Update protobuf dep version

Fix #939

.clang-format

@@ -34,7 +34,7 @@ BreakBeforeBinaryOperators: false
 BreakBeforeBraces: Custom
 BreakBeforeTernaryOperators: true
 BreakConstructorInitializersBeforeComma: true
-ColumnLimit:     80
+ColumnLimit:     120
 CommentPragmas:  '^ IWYU pragma:'
 ConstructorInitializerAllOnOneLineOrOnePerLine: true
 ConstructorInitializerIndentWidth: 4

.git-blame-ignore-revs

@@ -7,3 +7,4 @@
 # clang-format
 e41150248a97e4bdc1cf21b54650c4bb7c63928e
 2e542e7b0d94451a933c88778461cc8d3d7e6417
+d816ef54abd8e8e979b9c795bdb657a8d18f5e95

.githooks/pre-commit

@@ -4,7 +4,7 @@ exec 1>&2
 
 # paths to check and re-format
 sources="src unittests"
-formatter="clang-format -i"
+formatter="clang-format-11 -i"
 
 first=$(git diff $sources)
 find $sources -type f \( -name '*.cpp' -o -name '*.h' -o -name '*.ipp' \) -print0 | xargs -0 $formatter

.github/actions/build_clio/action.yml

@@ -0,0 +1,36 @@
+name: Build clio
+description: Build clio in build directory
+inputs:
+  conan_profile:
+    description: Conan profile name
+    required: true
+    default: default
+  conan_cache_hit:
+    description: Whether conan cache has been downloaded
+    required: true
+runs:
+  using: composite
+  steps:
+    - name: Get number of threads on mac
+      id: mac_threads
+      if: ${{ runner.os == 'macOS' }}
+      shell: bash
+      run: echo "num=$(($(sysctl -n hw.logicalcpu) - 2))" >> $GITHUB_OUTPUT
+
+    - name: Get number of threads on Linux
+      id: linux_threads
+      if: ${{ runner.os == 'Linux' }}
+      shell: bash
+      run: echo "num=$(($(nproc) - 2))" >> $GITHUB_OUTPUT
+
+    - name: Build Clio
+      shell: bash
+      env:
+        BUILD_OPTION: "${{ inputs.conan_cache_hit == 'true' && 'missing' || '' }}"
+      run: |
+        mkdir -p build
+        cd build
+        threads_num=${{ steps.mac_threads.outputs.num || steps.linux_threads.outputs.num }}
+        conan install .. -of . -b $BUILD_OPTION -s build_type=Release -o clio:tests=True --profile ${{ inputs.conan_profile }}
+        cmake -DCMAKE_TOOLCHAIN_FILE:FILEPATH=build/generators/conan_toolchain.cmake -DCMAKE_BUILD_TYPE=Release .. -G Ninja
+        cmake --build . --parallel $threads_num

.github/actions/clang_format/action.yml

@@ -1,3 +1,5 @@
+name: Check format
+description: Check format using clang-format-11
 runs:
   using: composite
   steps:

.github/actions/git_common_ancestor/action.yml

@@ -0,0 +1,14 @@
+name: Git common ancestor
+description: Find the closest common commit
+outputs:
+  commit:
+    description: Hash of commit
+    value: ${{ steps.find_common_ancestor.outputs.commit }}
+runs:
+  using: composite
+  steps:
+    - name: Find common git ancestor
+      id: find_common_ancestor
+      shell: bash
+      run: |
+        echo "commit=$(git merge-base --fork-point origin/develop)" >> $GITHUB_OUTPUT

.github/actions/restore_cache/action.yml

@@ -0,0 +1,50 @@
+name: Restore cache
+description: Find and restores conan and ccache cache
+inputs:
+  conan_dir:
+    description: Path to .conan directory
+    required: true
+  ccache_dir:
+    description: Path to .ccache directory
+    required: true
+outputs:
+  conan_hash:
+    description: Hash to use as a part of conan cache key
+    value: ${{ steps.conan_hash.outputs.hash }}
+  conan_cache_hit:
+    description: True if conan cache has been downloaded
+    value: ${{ steps.conan_cache.outputs.cache-hit }}
+  ccache_cache_hit:
+    description: True if ccache cache has been downloaded
+    value: ${{ steps.ccache_cache.outputs.cache-hit }}
+runs:
+  using: composite
+  steps:
+    - name: Find common commit
+      id: git_common_ancestor
+      uses: ./.github/actions/git_common_ancestor
+
+    - name: Calculate conan hash
+      id: conan_hash
+      shell: bash
+      run: |
+        conan info . -j info.json
+        packages_info=$(cat info.json | jq '.[] | "\(.display_name): \(.id)"' | grep -v 'clio')
+        echo "$packages_info"
+        hash=$(echo "$packages_info" | shasum -a 256 | cut -d ' ' -f 1)
+        rm info.json
+        echo "hash=$hash" >> $GITHUB_OUTPUT
+
+    - name: Restore conan cache
+      uses: actions/cache/restore@v3
+      id: conan_cache
+      with:
+        path: ${{ inputs.conan_dir }}/data
+        key: clio-conan_data-${{ runner.os }}-develop-${{ steps.conan_hash.outputs.hash }}
+
+    - name: Restore ccache cache
+      uses: actions/cache/restore@v3
+      id: ccache_cache
+      with:
+        path: ${{ inputs.ccache_dir }}
+        key: clio-ccache-${{ runner.os }}-develop-${{ steps.git_common_ancestor.outputs.commit }}

.github/actions/save_cache/action.yml

@@ -0,0 +1,46 @@
+name: Save cache
+description: Save conan and ccache cache for develop branch
+inputs:
+  conan_dir:
+    description: Path to .conan directory
+    required: true
+  conan_hash:
+    description: Hash to use as a part of conan cache key
+    required: true
+  conan_cache_hit:
+    description: Whether conan cache has been downloaded
+    required: true
+  ccache_dir:
+    description: Path to .ccache directory
+    required: true
+  ccache_cache_hit:
+    description: Whether conan cache has been downloaded
+    required: true
+runs:
+  using: composite
+  steps:
+    - name: Find common commit
+      id: git_common_ancestor
+      uses: ./.github/actions/git_common_ancestor
+
+    - name: Cleanup conan directory from extra data
+      if: ${{ inputs.conan_cache_hit != 'true' }}
+      shell: bash
+      run: |
+        conan remove "*" -s -b -f
+
+    - name: Save conan cache
+      if: ${{ inputs.conan_cache_hit != 'true' }}
+      uses: actions/cache/save@v3
+      with:
+        path: ${{ inputs.conan_dir }}/data
+        key: clio-conan_data-${{ runner.os }}-develop-${{ inputs.conan_hash }}
+
+    - name: Save ccache cache
+      if: ${{ inputs.ccache_cache_hit != 'true' }}
+      uses: actions/cache/save@v3
+      with:
+        path: ${{ inputs.ccache_dir }}
+        key: clio-ccache-${{ runner.os }}-develop-${{ steps.git_common_ancestor.outputs.commit }}
+
+

.github/actions/setup_conan/action.yml

@@ -0,0 +1,55 @@
+name: Setup conan
+description: Setup conan profile and artifactory
+outputs:
+  conan_profile:
+    description: Created conan profile name
+    value: ${{ steps.conan_export_output.outputs.conan_profile }}
+runs:
+  using: composite
+  steps:
+    - name: On mac
+      if: ${{ runner.os == 'macOS' }}
+      shell: bash
+      env:
+        CONAN_PROFILE: clio_clang_14
+      id: conan_setup_mac
+      run: |
+        echo "Creating $CONAN_PROFILE conan profile";
+        clang_path="$(brew --prefix llvm@14)/bin/clang"
+        clang_cxx_path="$(brew --prefix llvm@14)/bin/clang++"
+        conan profile new $CONAN_PROFILE --detect --force
+        conan profile update settings.compiler=clang $CONAN_PROFILE
+        conan profile update settings.compiler.version=14 $CONAN_PROFILE
+        conan profile update settings.compiler.cppstd=20 $CONAN_PROFILE
+        conan profile update "conf.tools.build:compiler_executables={\"c\": \"$clang_path\", \"cpp\": \"$clang_cxx_path\"}" $CONAN_PROFILE
+        conan profile update env.CC="$clang_path" $CONAN_PROFILE
+        conan profile update env.CXX="$clang_cxx_path" $CONAN_PROFILE
+        echo "created_conan_profile=$CONAN_PROFILE" >> $GITHUB_OUTPUT
+
+    - name: On linux
+      if: ${{ runner.os == 'Linux' }}
+      shell: bash
+      id: conan_setup_linux
+      run: |
+        conan profile new default --detect
+        conan profile update settings.compiler.cppstd=20 default
+        conan profile update settings.compiler.libcxx=libstdc++11 default
+        echo "created_conan_profile=default" >> $GITHUB_OUTPUT
+
+    - name: Export output variable
+      shell: bash
+      id: conan_export_output
+      run: |
+        echo "conan_profile=${{ steps.conan_setup_mac.outputs.created_conan_profile || steps.conan_setup_linux.outputs.created_conan_profile }}" >> $GITHUB_OUTPUT
+
+    - name: Add conan-non-prod artifactory
+      shell: bash
+      run: |
+        if [[ -z $(conan remote list | grep conan-non-prod) ]]; then
+            echo "Adding conan-non-prod"
+            conan remote add --insert 0 conan-non-prod http://18.143.149.228:8081/artifactory/api/conan/conan-non-prod
+        else
+            echo "Conan-non-prod is available"
+        fi
+
+

.github/workflows/build.yml

@@ -1,9 +1,9 @@
 name: Build Clio
 on:
   push:
-    branches: [master, release/*, develop, develop-next]
+    branches: [master, release/*, develop]
   pull_request:
-    branches: [master, release/*, develop, develop-next]
+    branches: [master, release/*, develop]
   workflow_dispatch:
 
 jobs:
@@ -13,183 +13,142 @@ jobs:
     steps:
       - uses: actions/checkout@v3
       - name: Run clang-format
-        uses: ./.github/actions/lint
+        uses: ./.github/actions/clang_format
 
-  build_clio:
-    name: Build Clio
-    runs-on: [self-hosted, Linux]
+  build_mac:
+    name: Build macOS
     needs: lint
-    strategy:
-      fail-fast: false
-      matrix:
-        type:
-          - suffix: deb
-            image: rippleci/clio-dpkg-builder:2022-09-17
-            script: dpkg
-          - suffix: rpm
-            image: rippleci/clio-rpm-builder:2022-09-17
-            script: rpm
-
-    container:
-      image: ${{ matrix.type.image }}
-      # options: --user 1001
+    runs-on: [self-hosted, macOS]
+    env:
+      CCACHE_DIR: ${{ github.workspace }}/.ccache
+      CONAN_USER_HOME: ${{ github.workspace }}
     steps:
       - uses: actions/checkout@v3
         with:
-          path: clio
-          fetch-depth: 0
+            fetch-depth: 0
 
-      - name: Clone Clio packaging repo
+      - name: Install packages
+        run: |
+          brew install llvm@14 pkg-config ninja bison cmake ccache jq
+
+      - name: Setup conan
+        uses: ./.github/actions/setup_conan
+        id: conan
+
+      - name: Restore cache
+        uses: ./.github/actions/restore_cache
+        id: restore_cache
+        with:
+          conan_dir: ${{ env.CONAN_USER_HOME }}/.conan
+          ccache_dir: ${{ env.CCACHE_DIR }}
+
+      - name: Build Clio
+        uses: ./.github/actions/build_clio
+        with:
+          conan_profile: ${{ steps.conan.outputs.conan_profile }}
+          conan_cache_hit: ${{ steps.restore_cache.outputs.conan_cache_hit }}
+
+      - name: Strip tests
+        run: strip build/clio_tests
+
+      - name: Upload clio_tests
+        uses: actions/upload-artifact@v3
+        with:
+          name: clio_tests_mac
+          path: build/clio_tests
+
+      - name: Save cache
+        uses: ./.github/actions/save_cache
+        with:
+          conan_dir: ${{ env.CONAN_USER_HOME }}/.conan
+          conan_hash: ${{ steps.restore_cache.outputs.conan_hash }}
+          conan_cache_hit: ${{ steps.restore_cache.outputs.conan_cache_hit }}
+          ccache_dir: ${{ env.CCACHE_DIR }}
+          ccache_cache_hit: ${{ steps.restore_cache.outputs.ccache_cache_hit }}
+
+  build_linux:
+    name: Build linux
+    needs: lint
+    runs-on: [self-hosted, Linux]
+    container:
+      image: conanio/gcc11:1.61.0
+      options: --user root
+    env:
+      CCACHE_DIR: /root/.ccache
+      CONAN_USER_HOME: /root/
+    steps:
+      - name: Get Clio
         uses: actions/checkout@v3
         with:
-          path: clio-packages
-          repository: XRPLF/clio-packages
-          ref: main
+          fetch-depth: 0
 
-      - name: Build
-        shell: bash
+      - name: Install packages
         run: |
-          export CLIO_ROOT=$(realpath clio)
-          if [ ${{ matrix.type.suffix }} == "rpm" ]; then
-            source /opt/rh/devtoolset-11/enable
-          fi
-          cmake -S clio-packages -B clio-packages/build -DCLIO_ROOT=$CLIO_ROOT
-          cmake --build clio-packages/build --parallel $(nproc)
-          cp ./clio-packages/build/clio-prefix/src/clio-build/clio_tests .
-          mv ./clio-packages/build/*.${{ matrix.type.suffix }} .
-      - name: Artifact packages
+          apt update -qq
+          apt install -y jq
+
+      - name: Install ccache
+        run: |
+          wget https://github.com/ccache/ccache/releases/download/v4.8.3/ccache-4.8.3-linux-x86_64.tar.xz
+          tar xf ./ccache-4.8.3-linux-x86_64.tar.xz
+          mv ./ccache-4.8.3-linux-x86_64/ccache /usr/bin/ccache
+
+      - name: Fix git permissions
+        run: git config --global --add safe.directory $PWD
+
+      - name: Setup conan
+        uses: ./.github/actions/setup_conan
+
+      - name: Restore cache
+        uses: ./.github/actions/restore_cache
+        id: restore_cache
+        with:
+          conan_dir: ${{ env.CONAN_USER_HOME }}/.conan
+          ccache_dir: ${{ env.CCACHE_DIR }}
+
+      - name: Build Clio
+        uses: ./.github/actions/build_clio
+        with:
+          conan_cache_hit: ${{ steps.restore_cache.outputs.conan_cache_hit }}
+
+      - name: Strip tests
+        run: strip build/clio_tests
+
+      - name: Upload clio_tests
         uses: actions/upload-artifact@v3
         with:
-          name: clio_${{ matrix.type.suffix }}_packages
-          path: ${{ github.workspace }}/*.${{ matrix.type.suffix }}
+          name: clio_tests_linux
+          path: build/clio_tests
 
-      - name: Artifact clio_tests
-        uses: actions/upload-artifact@v3
+      - name: Save cache
+        uses: ./.github/actions/save_cache
         with:
-          name: clio_tests-${{ matrix.type.suffix }}
-          path: ${{ github.workspace }}/clio_tests
-
-  build_dev:
-    name: Build on Mac/Clang14 and run tests
-    needs: lint
-    continue-on-error: false
-    runs-on: macos-12
+          conan_dir: ${{ env.CONAN_USER_HOME }}/.conan
+          conan_hash: ${{ steps.restore_cache.outputs.conan_hash }}
+          conan_cache_hit: ${{ steps.restore_cache.outputs.conan_cache_hit }}
+          ccache_dir: ${{ env.CCACHE_DIR }}
+          ccache_cache_hit: ${{ steps.restore_cache.outputs.ccache_cache_hit }}
 
+  test_mac:
+    needs: build_mac
+    runs-on: [self-hosted, macOS]
     steps:
-      - uses: actions/checkout@v3
+      - uses: actions/download-artifact@v3
         with:
-          path: clio
+          name: clio_tests_mac
+      - name: Run clio_tests
+        run: |
+          chmod +x ./clio_tests
+          ./clio_tests --gtest_filter="-BackendCassandraBaseTest*:BackendCassandraTest*:BackendCassandraFactoryTestWithDB*"
 
-      - name: Check Boost cache
-        id: boost
-        uses: actions/cache@v3
-        with:
-          path: boost
-          key: ${{ runner.os }}-boost
-
-      - name: Build boost
-        if: steps.boost.outputs.cache-hit != 'true'
-        run: |
-          curl -s -OJL "https://boostorg.jfrog.io/artifactory/main/release/1.77.0/source/boost_1_77_0.tar.gz"
-          tar zxf boost_1_77_0.tar.gz
-          mv boost_1_77_0 boost
-          cd boost
-          ./bootstrap.sh
-          ./b2 cxxflags="-std=c++14"
-      - name: install deps
-        run: |
-          brew install pkg-config protobuf openssl ninja cassandra-cpp-driver bison
-      - name: Build clio
-        run: |
-          export BOOST_ROOT=$(pwd)/boost
-          cd clio
-          cmake -B build
-          if ! cmake --build build -j$(nproc); then
-            echo '# 🔥🔥 MacOS AppleClang build failed!💥' >> $GITHUB_STEP_SUMMARY
-          fi
-      - name: Run Test
-        run: |
-          cd clio/build
-          ./clio_tests --gtest_filter="-Backend*"
-
-  test_clio:
-    name: Test Clio
-    runs-on: [self-hosted, Linux]
-    needs: build_clio
-    strategy:
-      fail-fast: false
-      matrix:
-        suffix: [rpm, deb]
+  test_linux:
+    needs: build_linux
+    runs-on: [self-hosted, x-heavy]
     steps:
-      - uses: actions/checkout@v3
-
-      - name: Get clio_tests artifact
-        uses: actions/download-artifact@v3
+      - uses: actions/download-artifact@v3
         with:
-          name: clio_tests-${{ matrix.suffix }}
-
-      - name: Run tests
-        timeout-minutes: 10
-        uses: ./.github/actions/test
-
-  code_coverage:
-    name: Build on Linux and code coverage
-    needs: lint
-    continue-on-error: false
-    runs-on: ubuntu-22.04
-
-    steps:
-      - uses: actions/checkout@v3
-        with:
-          path: clio
-
-      - name: Check Boost cache
-        id: boost
-        uses: actions/cache@v3
-        with:
-          path: boost
-          key: ${{ runner.os }}-boost
-
-      - name: Build boost
-        if: steps.boost.outputs.cache-hit != 'true'
+          name: clio_tests_linux
+      - name: Run clio_tests
         run: |
-          curl -s -OJL "https://boostorg.jfrog.io/artifactory/main/release/1.77.0/source/boost_1_77_0.tar.gz"
-          tar zxf boost_1_77_0.tar.gz
-          mv boost_1_77_0 boost
-          cd boost
-          ./bootstrap.sh
-          ./b2
-
-      - name: install deps
-        run: |
-          sudo apt-get -y install git pkg-config protobuf-compiler libprotobuf-dev libssl-dev wget build-essential doxygen bison flex autoconf clang-format gcovr
-
-      - name: Build clio
-        run: |
-          export BOOST_ROOT=$(pwd)/boost
-          cd clio
-          cmake -B build -DCODE_COVERAGE=on -DTEST_PARAMETER='--gtest_filter="-Backend*"'
-          if ! cmake --build build -j$(nproc); then
-            echo '# 🔥Ubuntu build🔥 failed!💥' >> $GITHUB_STEP_SUMMARY
-          fi
-          cd build
-          make clio_tests-ccov
-
-      - name: Code Coverage Summary Report
-        uses: irongut/CodeCoverageSummary@v1.2.0
-        with:
-          filename: clio/build/clio_tests-gcc-cov/out.xml
-          badge: true
-          output: both
-          format: markdown
-
-      - name: Save PR number and ccov report
-        run: |
-          mkdir -p ./UnitTestCoverage
-          echo ${{ github.event.number }} > ./UnitTestCoverage/NR
-          cp clio/build/clio_tests-gcc-cov/report.html ./UnitTestCoverage/report.html
-          cp code-coverage-results.md ./UnitTestCoverage/out.md
-      - uses: actions/upload-artifact@v2
-        with:
-          name: UnitTestCoverage
-          path: UnitTestCoverage/
+          chmod +x ./clio_tests
+          ./clio_tests --gtest_filter="-BackendCassandraBaseTest*:BackendCassandraTest*:BackendCassandraFactoryTestWithDB*"

.gitignore

@@ -1,6 +1,9 @@
 *clio*.log
-build*/
+/build*/
+.build
+.cache
 .vscode
 .python-version
+CMakeUserPresets.json
 config.json
 src/main/impl/Build.cpp

CMake/Ccache.cmake

@@ -0,0 +1,5 @@
+find_program (CCACHE_PATH "ccache")
+if (CCACHE_PATH)
+    set (CMAKE_CXX_COMPILER_LAUNCHER "${CCACHE_PATH}")
+    message (STATUS "Using ccache: ${CCACHE_PATH}")
+endif ()

CMake/CheckCompiler.cmake

@@ -0,0 +1,42 @@
+if (CMAKE_CXX_COMPILER_ID STREQUAL "Clang")
+  if (CMAKE_CXX_COMPILER_VERSION VERSION_LESS 14)
+    message (FATAL_ERROR "Clang 14+ required for building clio")
+  endif ()
+  set (is_clang TRUE)
+elseif (CMAKE_CXX_COMPILER_ID STREQUAL "AppleClang")
+  if (CMAKE_CXX_COMPILER_VERSION VERSION_LESS 14)
+    message (FATAL_ERROR "AppleClang 14+ required for building clio")
+  endif ()
+  set (is_appleclang TRUE)
+elseif (CMAKE_CXX_COMPILER_ID STREQUAL "GNU")
+  if (CMAKE_CXX_COMPILER_VERSION VERSION_LESS 11)
+    message (FATAL_ERROR "GCC 11+ required for building clio")
+  endif ()
+  set (is_gcc TRUE)
+else ()
+  message (FATAL_ERROR "Supported compilers: AppleClang 14+, Clang 14+, GCC 11+")
+endif ()
+
+if (san)
+  string (TOLOWER ${san} san)
+  set (SAN_FLAG "-fsanitize=${san}")
+  set (SAN_LIB "")
+  if (is_gcc)
+    if (san STREQUAL "address")
+      set (SAN_LIB "asan")
+    elseif (san STREQUAL "thread")
+      set (SAN_LIB "tsan")
+    elseif (san STREQUAL "memory")
+      set (SAN_LIB "msan")
+    elseif (san STREQUAL "undefined")
+      set (SAN_LIB "ubsan")
+    endif ()
+  endif ()
+  set (_saved_CRL ${CMAKE_REQUIRED_LIBRARIES})
+  set (CMAKE_REQUIRED_LIBRARIES "${SAN_FLAG};${SAN_LIB}")
+  CHECK_CXX_COMPILER_FLAG (${SAN_FLAG} COMPILER_SUPPORTS_SAN)
+  set (CMAKE_REQUIRED_LIBRARIES ${_saved_CRL})
+  if (NOT COMPILER_SUPPORTS_SAN)
+    message (FATAL_ERROR "${san} sanitizer does not seem to be supported by your compiler")
+  endif ()
+endif ()

CMake/ClioVersion.cmake

@@ -2,32 +2,38 @@
    write version to source
 #]===================================================================]
 
-find_package(Git REQUIRED)
+find_package (Git REQUIRED)
 
-set(GIT_COMMAND rev-parse --short HEAD)
-execute_process(COMMAND ${GIT_EXECUTABLE} ${GIT_COMMAND} OUTPUT_VARIABLE REV OUTPUT_STRIP_TRAILING_WHITESPACE)
+set (GIT_COMMAND rev-parse --short HEAD)
+execute_process (COMMAND ${GIT_EXECUTABLE} ${GIT_COMMAND} 
+  WORKING_DIRECTORY ${CMAKE_SOURCE_DIR}
+  OUTPUT_VARIABLE REV OUTPUT_STRIP_TRAILING_WHITESPACE)
 
-set(GIT_COMMAND branch --show-current)
-execute_process(COMMAND ${GIT_EXECUTABLE} ${GIT_COMMAND} OUTPUT_VARIABLE BRANCH OUTPUT_STRIP_TRAILING_WHITESPACE)
+set (GIT_COMMAND branch --show-current)
+execute_process (COMMAND ${GIT_EXECUTABLE} ${GIT_COMMAND} 
+  WORKING_DIRECTORY ${CMAKE_SOURCE_DIR}
+  OUTPUT_VARIABLE BRANCH OUTPUT_STRIP_TRAILING_WHITESPACE)
 
-if(BRANCH STREQUAL "")
-  set(BRANCH "dev")
-endif()
+if (BRANCH STREQUAL "")
+  set (BRANCH "dev")
+endif ()
 
-if(NOT (BRANCH MATCHES master OR BRANCH MATCHES release/*)) # for develop and any other branch name YYYYMMDDHMS-<branch>-<git-ref>
-  execute_process(COMMAND date +%Y%m%d%H%M%S OUTPUT_VARIABLE DATE OUTPUT_STRIP_TRAILING_WHITESPACE)
-  set(VERSION "${DATE}-${BRANCH}-${REV}")
-else()
-  set(GIT_COMMAND describe --tags)
-  execute_process(COMMAND ${GIT_EXECUTABLE} ${GIT_COMMAND} OUTPUT_VARIABLE TAG_VERSION OUTPUT_STRIP_TRAILING_WHITESPACE)
-  set(VERSION "${TAG_VERSION}-${REV}")
-endif()
+if (NOT (BRANCH MATCHES master OR BRANCH MATCHES release/*)) # for develop and any other branch name YYYYMMDDHMS-<branch>-<git-rev>
+  execute_process (COMMAND date +%Y%m%d%H%M%S OUTPUT_VARIABLE DATE OUTPUT_STRIP_TRAILING_WHITESPACE)
+  set (VERSION "${DATE}-${BRANCH}-${REV}")
+else ()
+  set (GIT_COMMAND describe --tags)
+  execute_process (COMMAND ${GIT_EXECUTABLE} ${GIT_COMMAND} 
+    WORKING_DIRECTORY ${CMAKE_SOURCE_DIR}
+    OUTPUT_VARIABLE TAG_VERSION OUTPUT_STRIP_TRAILING_WHITESPACE)
+  set (VERSION "${TAG_VERSION}-${REV}")
+endif ()
 
-if(CMAKE_BUILD_TYPE MATCHES Debug)
-  set(VERSION "${VERSION}+DEBUG")
-endif()
+if (CMAKE_BUILD_TYPE MATCHES Debug)
+  set (VERSION "${VERSION}+DEBUG")
+endif ()
 
-message(STATUS "Build version: ${VERSION}")
-set(clio_version "${VERSION}")
+message (STATUS "Build version: ${VERSION}")
+set (clio_version "${VERSION}")
 
-configure_file(CMake/Build.cpp.in ${CMAKE_SOURCE_DIR}/src/main/impl/Build.cpp)
+configure_file (CMake/Build.cpp.in ${CMAKE_SOURCE_DIR}/src/main/impl/Build.cpp)

CMake/Coverage.cmake

@@ -1,91 +1,91 @@
-# call add_converage(module_name) to add coverage targets for the given module
-function(add_converage module)
-  if("${CMAKE_C_COMPILER_ID}" MATCHES "(Apple)?[Cc]lang"
+# call add_coverage(module_name) to add coverage targets for the given module
+function (add_coverage module)
+  if ("${CMAKE_C_COMPILER_ID}" MATCHES "(Apple)?[Cc]lang"
      OR "${CMAKE_CXX_COMPILER_ID}" MATCHES "(Apple)?[Cc]lang")
-    message("[Coverage] Building with llvm Code Coverage Tools")
+    message ("[Coverage] Building with llvm Code Coverage Tools")
     # Using llvm gcov ; llvm install by xcode
-    set(LLVM_COV_PATH /Library/Developer/CommandLineTools/usr/bin)
-    if(NOT EXISTS ${LLVM_COV_PATH}/llvm-cov)
-      message(FATAL_ERROR "llvm-cov not found! Aborting.")
-    endif()
+    set (LLVM_COV_PATH /Library/Developer/CommandLineTools/usr/bin)
+    if (NOT EXISTS ${LLVM_COV_PATH}/llvm-cov)
+      message (FATAL_ERROR "llvm-cov not found! Aborting.")
+    endif ()
 
     # set Flags
-    target_compile_options(${module} PRIVATE -fprofile-instr-generate
-                                             -fcoverage-mapping)
-    target_link_options(${module} PUBLIC -fprofile-instr-generate
-                        -fcoverage-mapping)
+    target_compile_options (${module} PRIVATE 
+      -fprofile-instr-generate
+      -fcoverage-mapping)
+    
+    target_link_options (${module} PUBLIC 
+      -fprofile-instr-generate
+      -fcoverage-mapping)
 
-    target_compile_options(clio PRIVATE -fprofile-instr-generate
-                                        -fcoverage-mapping)
-    target_link_options(clio PUBLIC -fprofile-instr-generate
-                        -fcoverage-mapping)
+    target_compile_options (clio PRIVATE 
+      -fprofile-instr-generate
+      -fcoverage-mapping)
+    
+    target_link_options (clio PUBLIC 
+      -fprofile-instr-generate
+      -fcoverage-mapping)
 
     # llvm-cov
-    add_custom_target(
-      ${module}-ccov-preprocessing
+    add_custom_target (${module}-ccov-preprocessing
       COMMAND LLVM_PROFILE_FILE=${module}.profraw $<TARGET_FILE:${module}>
       COMMAND ${LLVM_COV_PATH}/llvm-profdata merge -sparse ${module}.profraw -o
               ${module}.profdata
       DEPENDS ${module})
 
-    add_custom_target(
-      ${module}-ccov-show
+    add_custom_target (${module}-ccov-show
       COMMAND ${LLVM_COV_PATH}/llvm-cov show $<TARGET_FILE:${module}>
               -instr-profile=${module}.profdata -show-line-counts-or-regions
       DEPENDS ${module}-ccov-preprocessing)
 
     # add summary for CI parse
-    add_custom_target(
-      ${module}-ccov-report
+    add_custom_target (${module}-ccov-report
       COMMAND
         ${LLVM_COV_PATH}/llvm-cov report $<TARGET_FILE:${module}>
         -instr-profile=${module}.profdata
-        -ignore-filename-regex=".*_makefiles|.*unittests"
+        -ignore-filename-regex=".*_makefiles|.*unittests|.*_deps"
         -show-region-summary=false
       DEPENDS ${module}-ccov-preprocessing)
 
     # exclude libs and unittests self
-    add_custom_target(
-      ${module}-ccov
+    add_custom_target (${module}-ccov
       COMMAND
         ${LLVM_COV_PATH}/llvm-cov show $<TARGET_FILE:${module}>
         -instr-profile=${module}.profdata -show-line-counts-or-regions
         -output-dir=${module}-llvm-cov -format="html"
-        -ignore-filename-regex=".*_makefiles|.*unittests" > /dev/null 2>&1
+        -ignore-filename-regex=".*_makefiles|.*unittests|.*_deps" > /dev/null 2>&1
       DEPENDS ${module}-ccov-preprocessing)
 
-    add_custom_command(
+    add_custom_command (
       TARGET ${module}-ccov
       POST_BUILD
       COMMENT
         "Open ${module}-llvm-cov/index.html in your browser to view the coverage report."
     )
-  elseif("${CMAKE_C_COMPILER_ID}" MATCHES "GNU" OR "${CMAKE_CXX_COMPILER_ID}"
-                                                   MATCHES "GNU")
-    message("[Coverage] Building with Gcc Code Coverage Tools")
+  elseif ("${CMAKE_C_COMPILER_ID}" MATCHES "GNU" OR "${CMAKE_CXX_COMPILER_ID}" MATCHES "GNU")
+    message ("[Coverage] Building with Gcc Code Coverage Tools")
 
-    find_program(GCOV_PATH gcov)
-    if(NOT GCOV_PATH)
-      message(FATAL_ERROR "gcov not found! Aborting...")
-    endif() # NOT GCOV_PATH
-    find_program(GCOVR_PATH gcovr)
-    if(NOT GCOVR_PATH)
-      message(FATAL_ERROR "gcovr not found! Aborting...")
-    endif() # NOT GCOVR_PATH
+    find_program (GCOV_PATH gcov)
+    if (NOT GCOV_PATH)
+      message (FATAL_ERROR "gcov not found! Aborting...")
+    endif () # NOT GCOV_PATH
+    find_program (GCOVR_PATH gcovr)
+    if (NOT GCOVR_PATH)
+      message (FATAL_ERROR "gcovr not found! Aborting...")
+    endif () # NOT GCOVR_PATH
 
-    set(COV_OUTPUT_PATH ${module}-gcc-cov)
-    target_compile_options(${module} PRIVATE -fprofile-arcs -ftest-coverage
+    set (COV_OUTPUT_PATH ${module}-gcc-cov)
+    target_compile_options (${module} PRIVATE -fprofile-arcs -ftest-coverage
                                              -fPIC)
-    target_link_libraries(${module} PRIVATE gcov)
+    target_link_libraries (${module} PRIVATE gcov)
 
-    target_compile_options(clio PRIVATE -fprofile-arcs -ftest-coverage
+    target_compile_options (clio PRIVATE -fprofile-arcs -ftest-coverage
                                              -fPIC)
-    target_link_libraries(clio PRIVATE gcov)
+    target_link_libraries (clio PRIVATE gcov)
     # this target is used for CI as well generate the summary out.xml will send
     # to github action to generate markdown, we can paste it to comments or
     # readme
-    add_custom_target(
-      ${module}-ccov
+    add_custom_target (${module}-ccov
       COMMAND ${module} ${TEST_PARAMETER}
       COMMAND rm -rf ${COV_OUTPUT_PATH}
       COMMAND mkdir ${COV_OUTPUT_PATH}
@@ -102,8 +102,7 @@ function(add_converage module)
       COMMENT "Running gcovr to produce Cobertura code coverage report.")
 
     # generate the detail report
-    add_custom_target(
-      ${module}-ccov-report
+    add_custom_target (${module}-ccov-report
       COMMAND ${module} ${TEST_PARAMETER}
       COMMAND rm -rf ${COV_OUTPUT_PATH}
       COMMAND mkdir ${COV_OUTPUT_PATH}
@@ -114,13 +113,13 @@ function(add_converage module)
         --exclude='${PROJECT_BINARY_DIR}/'
       WORKING_DIRECTORY ${PROJECT_BINARY_DIR}
       COMMENT "Running gcovr to produce Cobertura code coverage report.")
-    add_custom_command(
+    add_custom_command (
       TARGET ${module}-ccov-report
       POST_BUILD
       COMMENT
         "Open ${COV_OUTPUT_PATH}/index.html in your browser to view the coverage report."
     )
-  else()
-    message(FATAL_ERROR "Complier not support yet")
-  endif()
-endfunction()
+  else ()
+    message (FATAL_ERROR "Complier not support yet")
+  endif ()
+endfunction ()

CMake/Docs.cmake

@@ -0,0 +1,11 @@
+find_package (Doxygen REQUIRED)
+
+set (DOXYGEN_IN ${CMAKE_CURRENT_SOURCE_DIR}/Doxyfile)
+set (DOXYGEN_OUT ${CMAKE_CURRENT_BINARY_DIR}/Doxyfile)
+
+configure_file (${DOXYGEN_IN} ${DOXYGEN_OUT} @ONLY)
+add_custom_target (docs
+    COMMAND ${DOXYGEN_EXECUTABLE} ${DOXYGEN_OUT}
+    WORKING_DIRECTORY ${CMAKE_CURRENT_BINARY_DIR}
+    COMMENT "Generating API documentation with Doxygen"
+    VERBATIM)

CMake/Settings.cmake

@@ -0,0 +1,45 @@
+set(COMPILER_FLAGS
+  -Wall
+  -Wcast-align
+  -Wdouble-promotion
+  -Wextra
+  -Werror
+  -Wformat=2
+  -Wimplicit-fallthrough
+  -Wmisleading-indentation
+  -Wno-narrowing
+  -Wno-deprecated-declarations
+  -Wno-dangling-else
+  -Wno-unused-but-set-variable
+  -Wnon-virtual-dtor
+  -Wnull-dereference
+  -Wold-style-cast
+  -pedantic
+  -Wpedantic
+  -Wunused
+)
+
+if (is_gcc AND NOT lint)
+  list(APPEND COMPILER_FLAGS
+    -Wduplicated-branches
+    -Wduplicated-cond
+    -Wlogical-op
+    -Wuseless-cast
+  )
+endif ()
+
+if (is_clang)
+  list(APPEND COMPILER_FLAGS
+    -Wshadow # gcc is to aggressive with shadowing https://gcc.gnu.org/bugzilla/show_bug.cgi?id=78147
+  )
+endif ()
+
+if (is_appleclang)
+  list(APPEND COMPILER_FLAGS
+    -Wreorder-init-list
+  )
+endif ()
+
+# See https://github.com/cpp-best-practices/cppbestpractices/blob/master/02-Use_the_Tools_Available.md#gcc--clang for the flags description
+
+target_compile_options (clio PUBLIC ${COMPILER_FLAGS})

CMake/SourceLocation.cmake

@@ -0,0 +1,11 @@
+include (CheckIncludeFileCXX)
+
+check_include_file_cxx ("source_location" SOURCE_LOCATION_AVAILABLE)
+if (SOURCE_LOCATION_AVAILABLE) 
+    target_compile_definitions (clio PUBLIC "HAS_SOURCE_LOCATION")
+endif ()
+
+check_include_file_cxx ("experimental/source_location" EXPERIMENTAL_SOURCE_LOCATION_AVAILABLE)
+if (EXPERIMENTAL_SOURCE_LOCATION_AVAILABLE)     
+    target_compile_definitions (clio PUBLIC "HAS_EXPERIMENTAL_SOURCE_LOCATION")
+endif ()

CMake/deps/Boost.cmake

@@ -1,6 +1,11 @@
-set(Boost_USE_STATIC_LIBS ON)
-set(Boost_USE_STATIC_RUNTIME ON)
+set (Boost_USE_STATIC_LIBS ON)
+set (Boost_USE_STATIC_RUNTIME ON)
 
-find_package(Boost 1.75 COMPONENTS filesystem log_setup log thread system REQUIRED)
-
-target_link_libraries(clio PUBLIC ${Boost_LIBRARIES})
+find_package (Boost 1.82 REQUIRED
+  COMPONENTS
+    program_options
+    coroutine
+    system
+    log
+    log_setup
+)

CMake/deps/OpenSSL.cmake

@@ -0,0 +1,5 @@
+find_package (OpenSSL 1.1.1 REQUIRED)
+
+set_target_properties (OpenSSL::SSL PROPERTIES
+  INTERFACE_COMPILE_DEFINITIONS OPENSSL_NO_SSL2
+)

CMake/deps/Remove-bitset-operator.patch

@@ -1,24 +0,0 @@
-From 5cd9d09d960fa489a0c4379880cd7615b1c16e55 Mon Sep 17 00:00:00 2001
-From: CJ Cobb <ccobb@ripple.com>
-Date: Wed, 10 Aug 2022 12:30:01 -0400
-Subject: [PATCH] Remove bitset operator !=
-
----
- src/ripple/protocol/Feature.h | 1 -
- 1 file changed, 1 deletion(-)
-
-diff --git a/src/ripple/protocol/Feature.h b/src/ripple/protocol/Feature.h
-index b3ecb099b..6424be411 100644
---- a/src/ripple/protocol/Feature.h
-+++ b/src/ripple/protocol/Feature.h
-@@ -126,7 +126,6 @@ class FeatureBitset : private std::bitset<detail::numFeatures>
- public:
-     using base::bitset;
-     using base::operator==;
--    using base::operator!=;
- 
-     using base::all;
-     using base::any;
--- 
-2.32.0
-

CMake/deps/SourceLocation.cmake

@@ -1,11 +0,0 @@
-include(CheckIncludeFileCXX)
-
-check_include_file_cxx("source_location" SOURCE_LOCATION_AVAILABLE)
-if(SOURCE_LOCATION_AVAILABLE) 
-    target_compile_definitions(clio PUBLIC "HAS_SOURCE_LOCATION")
-endif()
-
-check_include_file_cxx("experimental/source_location" EXPERIMENTAL_SOURCE_LOCATION_AVAILABLE)
-if(EXPERIMENTAL_SOURCE_LOCATION_AVAILABLE)     
-    target_compile_definitions(clio PUBLIC "HAS_EXPERIMENTAL_SOURCE_LOCATION")
-endif()

CMake/deps/Threads.cmake

@@ -0,0 +1,2 @@
+set (THREADS_PREFER_PTHREAD_FLAG ON)
+find_package (Threads)

CMake/deps/cassandra.cmake

@@ -1,151 +1 @@
-find_package(ZLIB REQUIRED)
-
-find_library(cassandra NAMES cassandra)
-if(NOT cassandra)
-    message("System installed Cassandra cpp driver not found. Will build")
-    find_library(zlib NAMES zlib1g-dev zlib-devel zlib z)
-    if(NOT zlib)
-        message("zlib not found. will build")
-        add_library(zlib STATIC IMPORTED GLOBAL)
-        ExternalProject_Add(zlib_src
-            PREFIX ${nih_cache_path}
-            GIT_REPOSITORY https://github.com/madler/zlib.git
-            GIT_TAG v1.2.12
-            INSTALL_COMMAND ""
-            BUILD_BYPRODUCTS <BINARY_DIR>/${CMAKE_STATIC_LIBRARY_PREFIX}z.a
-            )
-        ExternalProject_Get_Property (zlib_src SOURCE_DIR)
-        ExternalProject_Get_Property (zlib_src BINARY_DIR)
-        set (zlib_src_SOURCE_DIR "${SOURCE_DIR}")
-        file (MAKE_DIRECTORY ${zlib_src_SOURCE_DIR}/include)
-        set_target_properties (zlib PROPERTIES
-            IMPORTED_LOCATION
-            ${BINARY_DIR}/${CMAKE_STATIC_LIBRARY_PREFIX}z.a
-            INTERFACE_INCLUDE_DIRECTORIES
-            ${SOURCE_DIR}/include)
-        add_dependencies(zlib zlib_src)
-        file(TO_CMAKE_PATH "${zlib_src_SOURCE_DIR}" zlib_src_SOURCE_DIR)
-    endif()
-    find_library(krb5 NAMES krb5-dev libkrb5-dev)
-    if(NOT krb5)
-        message("krb5 not found. will build")
-        add_library(krb5 STATIC IMPORTED GLOBAL)
-        ExternalProject_Add(krb5_src
-            PREFIX ${nih_cache_path}
-            GIT_REPOSITORY https://github.com/krb5/krb5.git
-            GIT_TAG krb5-1.20
-            UPDATE_COMMAND ""
-            CONFIGURE_COMMAND autoreconf src && CFLAGS=-fcommon ./src/configure --enable-static --disable-shared
-            BUILD_IN_SOURCE 1
-            BUILD_COMMAND make
-            INSTALL_COMMAND ""
-            BUILD_BYPRODUCTS <SOURCE_DIR>/lib/${CMAKE_STATIC_LIBRARY_PREFIX}krb5.a
-            )
-        message(${ep_lib_prefix}/krb5.a)
-        message(${CMAKE_STATIC_LIBRARY_PREFIX}krb5.a)
-        ExternalProject_Get_Property (krb5_src SOURCE_DIR)
-        ExternalProject_Get_Property (krb5_src BINARY_DIR)
-        set (krb5_src_SOURCE_DIR "${SOURCE_DIR}")
-        file (MAKE_DIRECTORY ${krb5_src_SOURCE_DIR}/include)
-        set_target_properties (krb5 PROPERTIES
-            IMPORTED_LOCATION
-	    ${SOURCE_DIR}/lib/${CMAKE_STATIC_LIBRARY_PREFIX}krb5.a
-            INTERFACE_INCLUDE_DIRECTORIES
-            ${SOURCE_DIR}/include)
-        add_dependencies(krb5 krb5_src)
-        file(TO_CMAKE_PATH "${krb5_src_SOURCE_DIR}" krb5_src_SOURCE_DIR)
-    endif()
-
-
-    find_library(libuv1 NAMES uv1 libuv1 liubuv1-dev libuv1:amd64)
-
-
-    if(NOT libuv1)
-        message("libuv1 not found, will build")
-        add_library(libuv1 STATIC IMPORTED GLOBAL)
-        ExternalProject_Add(libuv_src
-            PREFIX ${nih_cache_path}
-            GIT_REPOSITORY https://github.com/libuv/libuv.git
-            GIT_TAG v1.44.1
-            INSTALL_COMMAND ""
-            BUILD_BYPRODUCTS <BINARY_DIR>/${CMAKE_STATIC_LIBRARY_PREFIX}uv_a.a
-            )
-
-        ExternalProject_Get_Property (libuv_src SOURCE_DIR)
-        ExternalProject_Get_Property (libuv_src BINARY_DIR)
-        set (libuv_src_SOURCE_DIR "${SOURCE_DIR}")
-        file (MAKE_DIRECTORY ${libuv_src_SOURCE_DIR}/include)
-
-        set_target_properties (libuv1 PROPERTIES
-            IMPORTED_LOCATION
-            ${BINARY_DIR}/${CMAKE_STATIC_LIBRARY_PREFIX}uv_a.a
-            INTERFACE_INCLUDE_DIRECTORIES
-            ${SOURCE_DIR}/include)
-        add_dependencies(libuv1 libuv_src)
-
-        file(TO_CMAKE_PATH "${libuv_src_SOURCE_DIR}" libuv_src_SOURCE_DIR)
-    endif()
-    add_library (cassandra STATIC IMPORTED GLOBAL)
-    ExternalProject_Add(cassandra_src
-        PREFIX ${nih_cache_path}
-        GIT_REPOSITORY https://github.com/datastax/cpp-driver.git
-        GIT_TAG 2.16.2
-        CMAKE_ARGS
-        -DLIBUV_ROOT_DIR=${BINARY_DIR}
-        -DLIBUV_INCLUDE_DIR=${SOURCE_DIR}/include
-        -DCASS_BUILD_STATIC=ON
-        -DCASS_BUILD_SHARED=OFF
-        INSTALL_COMMAND ""
-        BUILD_BYPRODUCTS <BINARY_DIR>/${CMAKE_STATIC_LIBRARY_PREFIX}cassandra_static.a
-        )
-
-    ExternalProject_Get_Property (cassandra_src SOURCE_DIR)
-    ExternalProject_Get_Property (cassandra_src BINARY_DIR)
-    set (cassandra_src_SOURCE_DIR "${SOURCE_DIR}")
-    file (MAKE_DIRECTORY ${cassandra_src_SOURCE_DIR}/include)
-
-    set_target_properties (cassandra PROPERTIES
-        IMPORTED_LOCATION
-        ${BINARY_DIR}/${CMAKE_STATIC_LIBRARY_PREFIX}cassandra_static.a
-        INTERFACE_INCLUDE_DIRECTORIES
-        ${SOURCE_DIR}/include)
-    message("cass dirs")
-    message(${BINARY_DIR})
-    message(${SOURCE_DIR})
-    message(${BINARY_DIR}/${CMAKE_STATIC_LIBRARY_PREFIX}cassandra_static.a)
-    add_dependencies(cassandra cassandra_src)
-
-    if(NOT libuv1)
-        ExternalProject_Add_StepDependencies(cassandra_src build libuv1)
-        target_link_libraries(cassandra INTERFACE libuv1)
-    else()
-        target_link_libraries(cassandra INTERFACE ${libuv1})
-    endif()
-    if(NOT krb5)
-
-        ExternalProject_Add_StepDependencies(cassandra_src build krb5)
-        target_link_libraries(cassandra INTERFACE krb5)
-    else()
-        target_link_libraries(cassandra INTERFACE ${krb5})
-    endif()
-
-    if(NOT zlib)
-        ExternalProject_Add_StepDependencies(cassandra_src build zlib)
-        target_link_libraries(cassandra INTERFACE zlib)
-    else()
-        target_link_libraries(cassandra INTERFACE ${zlib})
-    endif()
-	set(OPENSSL_USE_STATIC_LIBS TRUE)
-	find_package(OpenSSL REQUIRED)
-	target_link_libraries(cassandra INTERFACE OpenSSL::SSL)
-
-    file(TO_CMAKE_PATH "${cassandra_src_SOURCE_DIR}" cassandra_src_SOURCE_DIR)
-    target_link_libraries(clio PUBLIC cassandra)
-else()
-    message("Found system installed cassandra cpp driver")
-    message(${cassandra})
-
-    find_path(cassandra_includes NAMES cassandra.h REQUIRED)
-    target_link_libraries (clio PUBLIC ${cassandra})
-    target_include_directories(clio INTERFACE ${cassandra_includes})
-endif()
+find_package (cassandra-cpp-driver REQUIRED)

CMake/deps/gtest.cmake

@@ -1,20 +1,4 @@
-FetchContent_Declare(
-  googletest
-  URL https://github.com/google/googletest/archive/609281088cfefc76f9d0ce82e1ff6c30cc3591e5.zip
-)
+find_package (GTest REQUIRED)
 
-FetchContent_GetProperties(googletest)
-
-if(NOT googletest_POPULATED)
-  FetchContent_Populate(googletest)
-  add_subdirectory(${googletest_SOURCE_DIR} ${googletest_BINARY_DIR} EXCLUDE_FROM_ALL)
-endif()
-
-target_link_libraries(clio_tests PUBLIC clio gmock_main)
-target_include_directories(clio_tests PRIVATE unittests)
-
-enable_testing()
-
-include(GoogleTest)
-
-gtest_discover_tests(clio_tests)
+enable_testing ()
+include (GoogleTest)

CMake/deps/libfmt.cmake

@@ -0,0 +1 @@
+find_package (fmt REQUIRED)

CMake/deps/libxrpl.cmake

@@ -0,0 +1 @@
+find_package (xrpl REQUIRED)

CMake/deps/rippled.cmake

@@ -1,20 +0,0 @@
-set(RIPPLED_REPO "https://github.com/ripple/rippled.git")
-set(RIPPLED_BRANCH "1.9.2")
-set(NIH_CACHE_ROOT "${CMAKE_CURRENT_BINARY_DIR}" CACHE INTERNAL "")
-set(patch_command ! grep operator!= src/ripple/protocol/Feature.h || git apply < ${CMAKE_CURRENT_SOURCE_DIR}/CMake/deps/Remove-bitset-operator.patch)
-message(STATUS "Cloning ${RIPPLED_REPO} branch ${RIPPLED_BRANCH}")
-FetchContent_Declare(rippled
-  GIT_REPOSITORY "${RIPPLED_REPO}"
-  GIT_TAG "${RIPPLED_BRANCH}"
-  GIT_SHALLOW ON
-  PATCH_COMMAND "${patch_command}"
-)
-
-FetchContent_GetProperties(rippled)
-if(NOT rippled_POPULATED)
-  FetchContent_Populate(rippled)
-  add_subdirectory(${rippled_SOURCE_DIR} ${rippled_BINARY_DIR} EXCLUDE_FROM_ALL)
-endif()
-
-target_link_libraries(clio PUBLIC xrpl_core grpc_pbufs)
-target_include_directories(clio PUBLIC ${rippled_SOURCE_DIR}/src ) # TODO: Seems like this shouldn't be needed?

CMake/install/install.cmake

@@ -1,16 +1,14 @@
-set(CLIO_INSTALL_DIR "/opt/clio")
-set(CMAKE_INSTALL_PREFIX ${CLIO_INSTALL_DIR})
+set (CLIO_INSTALL_DIR "/opt/clio")
+set (CMAKE_INSTALL_PREFIX ${CLIO_INSTALL_DIR})
 
-install(TARGETS clio_server DESTINATION bin)
-# install(TARGETS clio_tests DESTINATION bin) # NOTE: Do we want to install the tests?
+install (TARGETS clio_server DESTINATION bin)
 
-#install(FILES example-config.json DESTINATION etc RENAME config.json)
-file(READ example-config.json config)
-string(REGEX REPLACE "./clio_log" "/var/log/clio/" config "${config}")
-file(WRITE ${CMAKE_BINARY_DIR}/install-config.json "${config}")
-install(FILES ${CMAKE_BINARY_DIR}/install-config.json DESTINATION etc RENAME config.json)
+file (READ example-config.json config)
+string (REGEX REPLACE "./clio_log" "/var/log/clio/" config "${config}")
+file (WRITE ${CMAKE_BINARY_DIR}/install-config.json "${config}")
+install (FILES ${CMAKE_BINARY_DIR}/install-config.json DESTINATION etc RENAME config.json)
 
-configure_file("${CMAKE_SOURCE_DIR}/CMake/install/clio.service.in" "${CMAKE_BINARY_DIR}/clio.service")
+configure_file ("${CMAKE_SOURCE_DIR}/CMake/install/clio.service.in" "${CMAKE_BINARY_DIR}/clio.service")
 
-install(FILES "${CMAKE_BINARY_DIR}/clio.service" DESTINATION /lib/systemd/system)
+install (FILES "${CMAKE_BINARY_DIR}/clio.service" DESTINATION /lib/systemd/system)
 

CMake/settings.cmake

@@ -1,6 +0,0 @@
-target_compile_options(clio  
-  PUBLIC -Wall
-         -Werror 
-         -Wno-narrowing
-         -Wno-deprecated-declarations 
-         -Wno-dangling-else)

CMakeLists.txt

@@ -1,125 +1,272 @@
 cmake_minimum_required(VERSION 3.16.3)
-
 project(clio)
 
-if(CMAKE_CXX_COMPILER_VERSION VERSION_LESS 11)
-  message(FATAL_ERROR "GCC 11+ required for building clio")
-endif()
+# ========================================================================== #
+#                                   Options                                  #
+# ========================================================================== #
+option (verbose   "Verbose build"                                     FALSE)
+option (tests     "Build tests"                                       FALSE)
+option (docs      "Generate doxygen docs"                             FALSE)
+option (coverage  "Build test coverage report"                        FALSE)
+option (packaging "Create distribution packages"                      FALSE)
+# ========================================================================== #
+set (san "" CACHE STRING "Add sanitizer instrumentation")
+set (CMAKE_EXPORT_COMPILE_COMMANDS TRUE)
+set_property (CACHE san PROPERTY STRINGS ";undefined;memory;address;thread")
+# ========================================================================== #
 
-option(BUILD_TESTS "Build tests" TRUE)
+# Include required modules
+include (CMake/Ccache.cmake)
+include (CheckCXXCompilerFlag)
 
-option(VERBOSE "Verbose build" TRUE)
-if(VERBOSE)
-  set(CMAKE_VERBOSE_MAKEFILE TRUE)
-  set(FETCHCONTENT_QUIET FALSE CACHE STRING "Verbose FetchContent()")
-endif()
+if (verbose)
+  set (CMAKE_VERBOSE_MAKEFILE TRUE)
+endif ()
 
+if (packaging)
+  add_definitions (-DPKG=1)
+endif ()
 
-if(PACKAGING)
-  add_definitions(-DPKG=1)
-endif()
+add_library (clio)
 
-add_library(clio)
-target_compile_features(clio PUBLIC cxx_std_20)
-target_include_directories(clio PUBLIC src)
+# Clio tweaks and checks
+include (CMake/CheckCompiler.cmake)
+include (CMake/Settings.cmake)
+include (CMake/ClioVersion.cmake)
+include (CMake/SourceLocation.cmake)
 
-include(FetchContent)
-include(ExternalProject)
-include(CMake/settings.cmake)
-include(CMake/ClioVersion.cmake)
-include(CMake/deps/rippled.cmake)
-include(CMake/deps/Boost.cmake)
-include(CMake/deps/cassandra.cmake)
-include(CMake/deps/SourceLocation.cmake)
+# Clio deps
+include (CMake/deps/libxrpl.cmake)
+include (CMake/deps/Boost.cmake)
+include (CMake/deps/OpenSSL.cmake)
+include (CMake/deps/Threads.cmake)
+include (CMake/deps/libfmt.cmake)
+include (CMake/deps/cassandra.cmake)
 
-target_sources(clio PRIVATE
+# TODO: Include directory will be wrong when installed.
+target_include_directories (clio PUBLIC src)
+target_compile_features (clio PUBLIC cxx_std_20)
+
+target_link_libraries (clio
+  PUBLIC Boost::boost
+  PUBLIC Boost::coroutine
+  PUBLIC Boost::program_options
+  PUBLIC Boost::system
+  PUBLIC Boost::log
+  PUBLIC Boost::log_setup
+  PUBLIC cassandra-cpp-driver::cassandra-cpp-driver
+  PUBLIC fmt::fmt
+  PUBLIC OpenSSL::Crypto
+  PUBLIC OpenSSL::SSL
+  PUBLIC xrpl::libxrpl
+
+  INTERFACE Threads::Threads
+)
+
+if (is_gcc)
+  # FIXME: needed on gcc for now
+  target_compile_definitions (clio PUBLIC BOOST_ASIO_DISABLE_CONCEPTS)
+endif ()
+
+target_sources (clio PRIVATE
   ## Main
   src/main/impl/Build.cpp
   ## Backend
-  src/backend/BackendInterface.cpp
-  src/backend/CassandraBackend.cpp
-  src/backend/SimpleCache.cpp
+  src/data/BackendInterface.cpp
+  src/data/LedgerCache.cpp
+  src/data/cassandra/impl/Future.cpp
+  src/data/cassandra/impl/Cluster.cpp
+  src/data/cassandra/impl/Batch.cpp
+  src/data/cassandra/impl/Result.cpp
+  src/data/cassandra/impl/Tuple.cpp
+  src/data/cassandra/impl/SslContext.cpp
+  src/data/cassandra/Handle.cpp
+  src/data/cassandra/SettingsProvider.cpp
   ## ETL
-  src/etl/ETLSource.cpp
-  src/etl/ProbingETLSource.cpp
+  src/etl/Source.cpp
+  src/etl/ProbingSource.cpp
   src/etl/NFTHelpers.cpp
-  src/etl/ReportingETL.cpp
-  ## Subscriptions
-  src/subscriptions/SubscriptionManager.cpp
+  src/etl/ETLService.cpp
+  src/etl/LoadBalancer.cpp
+  src/etl/impl/ForwardCache.cpp
+  ## Feed
+  src/feed/SubscriptionManager.cpp
+  ## Web
+  src/web/IntervalSweepHandler.cpp
   ## RPC
   src/rpc/Errors.cpp
-  src/rpc/RPC.cpp
+  src/rpc/Factories.cpp
   src/rpc/RPCHelpers.cpp
   src/rpc/Counters.cpp
   src/rpc/WorkQueue.cpp
-  ## NextGen RPC
   src/rpc/common/Specs.cpp
   src/rpc/common/Validators.cpp
-  ## RPC Methods
-  # Account
+  src/rpc/common/MetaProcessors.cpp
+  src/rpc/common/impl/APIVersionParser.cpp
+  src/rpc/common/impl/HandlerProvider.cpp
+  ## RPC handlers
   src/rpc/handlers/AccountChannels.cpp
   src/rpc/handlers/AccountCurrencies.cpp
   src/rpc/handlers/AccountInfo.cpp
   src/rpc/handlers/AccountLines.cpp
-  src/rpc/handlers/AccountOffers.cpp
+  src/rpc/handlers/AccountNFTs.cpp
   src/rpc/handlers/AccountObjects.cpp
+  src/rpc/handlers/AccountOffers.cpp
+  src/rpc/handlers/AccountTx.cpp
+  src/rpc/handlers/BookChanges.cpp
+  src/rpc/handlers/BookOffers.cpp
+  src/rpc/handlers/DepositAuthorized.cpp
   src/rpc/handlers/GatewayBalances.cpp
-  src/rpc/handlers/NoRippleCheck.cpp
-  # NFT
-  src/rpc/handlers/NFTHistory.cpp
-  src/rpc/handlers/NFTInfo.cpp
-  src/rpc/handlers/NFTOffers.cpp
-  # Ledger
   src/rpc/handlers/Ledger.cpp
   src/rpc/handlers/LedgerData.cpp
   src/rpc/handlers/LedgerEntry.cpp
   src/rpc/handlers/LedgerRange.cpp
-  # Transaction
-  src/rpc/handlers/Tx.cpp
-  src/rpc/handlers/TransactionEntry.cpp
-  src/rpc/handlers/AccountTx.cpp
-  # Dex
-  src/rpc/handlers/BookChanges.cpp
-  src/rpc/handlers/BookOffers.cpp
-  # Payment Channel
-  src/rpc/handlers/ChannelAuthorize.cpp
-  src/rpc/handlers/ChannelVerify.cpp
-  # Subscribe
-  src/rpc/handlers/Subscribe.cpp
-  # Server
-  src/rpc/handlers/ServerInfo.cpp
-  # Utilities
+  src/rpc/handlers/NFTBuyOffers.cpp
+  src/rpc/handlers/NFTHistory.cpp
+  src/rpc/handlers/NFTInfo.cpp
+  src/rpc/handlers/NFTOffersCommon.cpp
+  src/rpc/handlers/NFTSellOffers.cpp
+  src/rpc/handlers/NoRippleCheck.cpp
   src/rpc/handlers/Random.cpp
-  src/config/Config.cpp
-  src/log/Logger.cpp
+  src/rpc/handlers/TransactionEntry.cpp
+  src/rpc/handlers/Tx.cpp
+  ## Util
+  src/util/config/Config.cpp
+  src/util/log/Logger.cpp
   src/util/Taggable.cpp)
 
-add_executable(clio_server src/main/main.cpp)
-target_link_libraries(clio_server PUBLIC clio)
+# Clio server
+add_executable (clio_server src/main/Main.cpp)
+target_link_libraries (clio_server PRIVATE clio)
+target_link_options(clio_server
+  PRIVATE
+    $<$<AND:$<NOT:$<BOOL:${APPLE}>>,$<NOT:$<BOOL:${san}>>>:-static-libstdc++ -static-libgcc>
+)
 
-if(BUILD_TESTS)
-  set(TEST_TARGET clio_tests)
-  add_executable(${TEST_TARGET}
+# Unittesting
+if (tests)
+  set (TEST_TARGET clio_tests)
+  add_executable (${TEST_TARGET}
+    # Common
+    unittests/Main.cpp
     unittests/Playground.cpp
-    unittests/Backend.cpp
-    unittests/Logger.cpp
-    unittests/Config.cpp
-    unittests/ProfilerTest.cpp
-    unittests/DOSGuard.cpp
+    unittests/LoggerTests.cpp
+    unittests/ConfigTests.cpp
+    unittests/ProfilerTests.cpp
+    unittests/JsonUtilTests.cpp
+    unittests/DOSGuardTests.cpp
+    unittests/SubscriptionTests.cpp
+    unittests/SubscriptionManagerTests.cpp
+    unittests/util/TestObject.cpp
+    unittests/util/StringUtils.cpp
+    # ETL
+    unittests/etl/ExtractionDataPipeTests.cpp
+    unittests/etl/ExtractorTests.cpp
+    unittests/etl/TransformerTests.cpp
+    unittests/etl/CacheLoaderTests.cpp
+    unittests/etl/AmendmentBlockHandlerTests.cpp
+    # RPC
     unittests/rpc/ErrorTests.cpp
     unittests/rpc/BaseTests.cpp
+    unittests/rpc/RPCHelpersTests.cpp
+    unittests/rpc/CountersTests.cpp
+    unittests/rpc/AdminVerificationTests.cpp
+    unittests/rpc/APIVersionTests.cpp
+    unittests/rpc/ForwardingProxyTests.cpp
+    unittests/rpc/WorkQueueTests.cpp
+    unittests/rpc/AmendmentsTests.cpp
+    unittests/rpc/JsonBoolTests.cpp
+    ## RPC handlers
+    unittests/rpc/handlers/DefaultProcessorTests.cpp
     unittests/rpc/handlers/TestHandlerTests.cpp
-    unittests/rpc/handlers/DefaultProcessorTests.cpp)
-  include(CMake/deps/gtest.cmake)
+    unittests/rpc/handlers/AccountCurrenciesTests.cpp
+    unittests/rpc/handlers/AccountLinesTests.cpp
+    unittests/rpc/handlers/AccountTxTests.cpp
+    unittests/rpc/handlers/AccountOffersTests.cpp
+    unittests/rpc/handlers/AccountInfoTests.cpp
+    unittests/rpc/handlers/AccountChannelsTests.cpp
+    unittests/rpc/handlers/AccountNFTsTests.cpp
+    unittests/rpc/handlers/BookOffersTests.cpp
+    unittests/rpc/handlers/DepositAuthorizedTests.cpp
+    unittests/rpc/handlers/GatewayBalancesTests.cpp
+    unittests/rpc/handlers/TxTests.cpp
+    unittests/rpc/handlers/TransactionEntryTests.cpp
+    unittests/rpc/handlers/LedgerEntryTests.cpp
+    unittests/rpc/handlers/LedgerRangeTests.cpp
+    unittests/rpc/handlers/NoRippleCheckTests.cpp
+    unittests/rpc/handlers/ServerInfoTests.cpp
+    unittests/rpc/handlers/PingTests.cpp
+    unittests/rpc/handlers/RandomTests.cpp
+    unittests/rpc/handlers/NFTInfoTests.cpp
+    unittests/rpc/handlers/NFTBuyOffersTests.cpp
+    unittests/rpc/handlers/NFTSellOffersTests.cpp
+    unittests/rpc/handlers/NFTHistoryTests.cpp
+    unittests/rpc/handlers/SubscribeTests.cpp
+    unittests/rpc/handlers/UnsubscribeTests.cpp
+    unittests/rpc/handlers/LedgerDataTests.cpp
+    unittests/rpc/handlers/AccountObjectsTests.cpp
+    unittests/rpc/handlers/BookChangesTests.cpp
+    unittests/rpc/handlers/LedgerTests.cpp
+    unittests/rpc/handlers/VersionHandlerTests.cpp
+    # Backend
+    unittests/data/BackendFactoryTests.cpp
+    unittests/data/cassandra/BaseTests.cpp
+    unittests/data/cassandra/BackendTests.cpp
+    unittests/data/cassandra/RetryPolicyTests.cpp
+    unittests/data/cassandra/SettingsProviderTests.cpp
+    unittests/data/cassandra/ExecutionStrategyTests.cpp
+    unittests/data/cassandra/AsyncExecutorTests.cpp
+    # Webserver
+    unittests/web/ServerTests.cpp
+    unittests/web/RPCServerHandlerTests.cpp
+    unittests/web/WhitelistHandlerTests.cpp
+    unittests/web/SweepHandlerTests.cpp)
 
-  # if CODE_COVERAGE enable, add clio_test-ccov
-  if(CODE_COVERAGE)
-    include(CMake/coverage.cmake)
-    add_converage(${TEST_TARGET})
-  endif()
-endif()
+  include (CMake/deps/gtest.cmake)
 
-include(CMake/install/install.cmake)
-if(PACKAGING)
-    include(CMake/packaging.cmake)
-endif()
+  # See https://github.com/google/googletest/issues/3475
+  gtest_discover_tests (clio_tests DISCOVERY_TIMEOUT 10)
+
+  # Fix for dwarf5 bug on ci
+  target_compile_options (clio PUBLIC -gdwarf-4)
+
+  target_compile_definitions (${TEST_TARGET} PUBLIC UNITTEST_BUILD)
+  target_include_directories (${TEST_TARGET} PRIVATE unittests)
+  target_link_libraries (${TEST_TARGET} PUBLIC clio gtest::gtest)
+
+  # Generate `clio_tests-ccov` if coverage is enabled
+  # Note: use `make clio_tests-ccov` to generate report
+  if (coverage)
+    target_compile_definitions(${TEST_TARGET} PRIVATE COVERAGE_ENABLED)
+    include (CMake/Coverage.cmake)
+    add_coverage (${TEST_TARGET})
+  endif ()
+endif ()
+
+# Enable selected sanitizer if enabled via `san`
+if (san)
+  target_compile_options (clio
+    PUBLIC
+      # Sanitizers recommend minimum of -O1 for reasonable performance
+      $<$<CONFIG:Debug>:-O1>
+      ${SAN_FLAG}
+      -fno-omit-frame-pointer)
+  target_compile_definitions (clio
+    PUBLIC
+      $<$<STREQUAL:${san},address>:SANITIZER=ASAN>
+      $<$<STREQUAL:${san},thread>:SANITIZER=TSAN>
+      $<$<STREQUAL:${san},memory>:SANITIZER=MSAN>
+      $<$<STREQUAL:${san},undefined>:SANITIZER=UBSAN>)
+  target_link_libraries (clio INTERFACE ${SAN_FLAG} ${SAN_LIB})
+endif ()
+
+# Generate `docs` target for doxygen documentation if enabled
+# Note: use `make docs` to generate the documentation
+if (docs)
+  include (CMake/Docs.cmake)
+endif ()
+
+include (CMake/install/install.cmake)
+if (packaging)
+  include (CMake/packaging.cmake) # This file exists only in build runner
+endif ()

CONTRIBUTING.md

@@ -3,18 +3,18 @@ Thank you for your interest in contributing to the `clio` project 🙏
 
 To contribute, please:
 1. Fork the repository under your own user.
-2. Create a new branch on which to write your changes.
+2. Create a new branch on which to commit/push your changes.
 3. Write and test your code.
 4. Ensure that your code compiles with the provided build engine and update the provided build engine as part of your PR where needed and where appropriate.
 5. Where applicable, write test cases for your code and include those in `unittests`.
 6. Ensure your code passes automated checks (e.g. clang-format)
-7. Squash your commits (i.e. rebase) into as few commits as is reasonable to describe your changes at a high level (typically a single commit for a small change.). See below for more details.
+7. Squash your commits (i.e. rebase) into as few commits as is reasonable to describe your changes at a high level (typically a single commit for a small change). See below for more details.
 8. Open a PR to the main repository onto the _develop_ branch, and follow the provided template.
 
-> **Note:** Please make sure you read the [Style guide](#style-guide).
+> **Note:** Please read the [Style guide](#style-guide).
 
 ## Install git hooks
-Please make sure to run the following command in order to use git hooks that are helpful for `clio` development.
+Please run the following command in order to use git hooks that are helpful for `clio` development.
 
 ``` bash
 git config --local core.hooksPath .githooks
@@ -22,10 +22,10 @@ git config --local core.hooksPath .githooks
 
 ## Git commands
 This sections offers a detailed look at the git commands you will need to use to get your PR submitted. 
-Please note that there are more than one way to do this and these commands are only  provided for your convenience.
+Please note that there are more than one way to do this and these commands are provided for your convenience.
 At this point it's assumed that you have already finished working on your feature/bug.
 
-> **Important:** Before you issue any of the commands below, please hit the `Sync fork` button and make sure your fork's `develop` branch is up to date with the main `clio` repository.
+> **Important:** Before you issue any of the commands below, please hit the `Sync fork` button and make sure your fork's `develop` branch is up-to-date with the main `clio` repository.
 
 ``` bash
 # Create a backup of your branch
@@ -37,16 +37,16 @@ git pull origin develop
 git checkout <your feature branch>
 git rebase -i develop
 ```
-For each commit in the list other than the first one please select `s` to squash.
-After this is done you will have the opportunity to write a message for the squashed commit.
+For each commit in the list other than the first one, enter `s` to squash.
+After this is done, you will have the opportunity to write a message for the squashed commit.
 
-> **Hint:** Please use **imperative mood** commit message capitalizing the first word of the subject.
+> **Hint:** Please use **imperative mood** in the commit message, and capitalize the first word.
 
 ``` bash
 # You should now have a single commit on top of a commit in `develop`
 git log
 ```
-> **Todo:** In case there are merge conflicts, please resolve them now
+> **Note:** If there are merge conflicts, please resolve them now.
 
 ``` bash
 # Use the same commit message as you did above
@@ -54,16 +54,16 @@ git commit -m 'Your message'
 git rebase --continue
 ```
 
-> **Important:** If you have no GPG keys setup please follow [this tutorial](https://docs.github.com/en/authentication/managing-commit-signature-verification/adding-a-gpg-key-to-your-github-account)
+> **Important:** If you have no GPG keys set up, please follow [this tutorial](https://docs.github.com/en/authentication/managing-commit-signature-verification/adding-a-gpg-key-to-your-github-account)
 
 ``` bash
-# Sign the commit with your GPG key and finally push your changes to the repo
+# Sign the commit with your GPG key, and push your changes
 git commit --amend -S
 git push --force
 ```
 
 ## Fixing issues found during code review
-While your code is in review it's possible that some changes will be requested by the reviewer.
+While your code is in review, it's possible that some changes will be requested by reviewer(s).
 This section describes the process of adding your fixes.
 
 We assume that you already made the required changes on your feature branch.
@@ -72,25 +72,26 @@ We assume that you already made the required changes on your feature branch.
 # Add the changed code
 git add <paths to add>
 
-# Add a folded commit message (so you can squash them later)
+# Add a [FOLD] commit message (so you remember to squash it later)
 # while also signing it with your GPG key
 git commit -S -m "[FOLD] Your commit message"
 
 # And finally push your changes
 git push
 ```
-## After code review
-Last but not least, when your PR is approved you still have to `Squash and merge` your code. 
-Luckily there is a button for that towards the bottom of the PR's page on github.
 
-> **Important:** Please leave the automatically generated link to PR in the subject line **and** in the description field please add `"Fixes #ISSUE_ID"` (replacing `ISSUE_ID` with yours).
+## After code review
+When your PR is approved and ready to merge, use `Squash and merge`.
+The button for that is near the bottom of the PR's page on GitHub.
+
+> **Important:** Please leave the automatically-generated mention/link to the PR in the subject line **and** in the description field add `"Fix #ISSUE_ID"` (replacing `ISSUE_ID` with yours) if the PR fixes an issue.
 > **Note:** See [issues](https://github.com/XRPLF/clio/issues) to find the `ISSUE_ID` for the feature/bug you were working on.
 
 # Style guide
-This is a non-exhaustive list of recommended style guidelines. These are not always strictly enforced and serve as a way to keep the codebase coherent rather than a set of _thou shalt not_ commandments.
+This is a non-exhaustive list of recommended style guidelines. These are not always strictly enforced and serve as a way to keep the codebase coherent.
 
 ## Formatting
-All code must conform to `clang-format` version 10, unless the result would be unreasonably difficult to read or maintain.
+Code must conform to `clang-format` version 10, unless the result would be unreasonably difficult to read or maintain.
 To change your code to conform use `clang-format -i <your changed files>`.
 
 ## Avoid
@@ -114,7 +115,7 @@ To change your code to conform use `clang-format -i <your changed files>`.
 Maintainers are ecosystem participants with elevated access to the repository. They are able to push new code, make decisions on when a release should be made, etc.
 
 ## Code Review
-PRs must be reviewed by at least one of the maintainers.
+A PR must be reviewed and approved by at least one of the maintainers before it can be merged.
 
 ## Adding and Removing
 New maintainers can be proposed by two existing maintainers, subject to a vote by a quorum of the existing maintainers. A minimum of 50% support and a 50% participation is required. In the event of a tie vote, the addition of the new maintainer will be rejected.
@@ -123,12 +124,11 @@ Existing maintainers can resign, or be subject to a vote for removal at the behe
 
 ## Existing Maintainers
 
-* [cjcobb23](https://github.com/cjcobb23) (Ripple)
-* [legleux](https://github.com/legleux) (Ripple)
-* [undertome](https://github.com/undertome) (Ripple)
+* [cindyyan317](https://github.com/cindyyan317) (Ripple)
 * [godexsoft](https://github.com/godexsoft) (Ripple)
-* [officialfrancismendoza](https://github.com/officialfrancismendoza) (Ripple)
+* [legleux](https://github.com/legleux) (Ripple)
 
 ## Honorable ex-Maintainers
 
+* [cjcobb23](https://github.com/cjcobb23) (ex-Ripple)
 * [natenichols](https://github.com/natenichols) (ex-Ripple)

Doxyfile

@@ -1,3 +1,16 @@
 PROJECT_NAME           = "Clio"
-INPUT                  = src
-RECURSIVE              = YES
+INPUT                  = ../src ../unittests
+EXCLUDE_PATTERNS       = *Test*.cpp *Test*.h
+RECURSIVE              = YES
+HAVE_DOT               = YES
+
+QUIET                  = YES
+WARNINGS               = NO
+WARN_NO_PARAMDOC       = NO
+WARN_IF_INCOMPLETE_DOC = NO
+WARN_IF_UNDOCUMENTED   = NO
+
+GENERATE_LATEX         = NO
+GENERATE_HTML          = YES
+
+SORT_MEMBERS_CTORS_1ST = YES

README.md

@@ -1,51 +1,102 @@
 # Clio
-Clio is an XRP Ledger API server. Clio is optimized for RPC calls, over WebSocket or JSON-RPC. Validated
-historical ledger and transaction data are stored in a more space-efficient format,
+
+Clio is an XRP Ledger API server. Clio is optimized for RPC calls, over WebSocket or JSON-RPC. 
+Validated historical ledger and transaction data are stored in a more space-efficient format,
 using up to 4 times less space than rippled. Clio can be configured to store data in Apache Cassandra or ScyllaDB,
-allowing for scalable read throughput. Multiple Clio nodes can share
-access to the same dataset, allowing for a highly available cluster of Clio nodes,
-without the need for redundant data storage or computation.
+allowing for scalable read throughput. Multiple Clio nodes can share access to the same dataset, 
+allowing for a highly available cluster of Clio nodes, without the need for redundant data storage or computation.
 
 Clio offers the full rippled API, with the caveat that Clio by default only returns validated data.
 This means that `ledger_index` defaults to `validated` instead of `current` for all requests.
 Other non-validated data is also not returned, such as information about queued transactions.
-For requests that require access to the p2p network, such as `fee` or `submit`, Clio automatically forwards the request to a rippled node and propagates the response back to the client. To access non-validated data for *any* request, simply add `ledger_index: "current"` to the request, and Clio will forward the request to rippled.
+For requests that require access to the p2p network, such as `fee` or `submit`, Clio automatically forwards the request to a rippled node and propagates the response back to the client. 
+To access non-validated data for *any* request, simply add `ledger_index: "current"` to the request, and Clio will forward the request to rippled.
 
 Clio does not connect to the peer-to-peer network. Instead, Clio extracts data from a group of specified rippled nodes. Running Clio requires access to at least one rippled node
 from which data can be extracted. The rippled node does not need to be running on the same machine as Clio.
 
-
 ## Requirements
 1. Access to a Cassandra cluster or ScyllaDB cluster. Can be local or remote.
-
 2. Access to one or more rippled nodes. Can be local or remote.
 
 ## Building
 
-Clio is built with CMake. Clio requires at least GCC-11/clang-14.0.0 (C++20), and Boost 1.75.0.
+Clio is built with CMake and uses Conan for managing dependencies. 
+It is written in C++20 and therefore requires a modern compiler.
 
-Use these instructions to build a Clio executable from the source. These instructions were tested on Ubuntu 20.04 LTS.
+## Prerequisites 
 
-```sh
-# Install dependencies
-  sudo apt-get -y install git pkg-config protobuf-compiler libprotobuf-dev libssl-dev wget build-essential bison flex autoconf cmake clang-format
-# Install gcovr to run code coverage
-  sudo apt-get -y install gcovr
+### Minimum Requirements
 
-# Compile Boost
-  wget -O $HOME/boost_1_75_0.tar.gz https://boostorg.jfrog.io/artifactory/main/release/1.75.0/source/boost_1_75_0.tar.gz
-  tar xvzf $HOME/boost_1_75_0.tar.gz
-  cd $HOME/boost_1_75_0
-  ./bootstrap.sh
-  ./b2 -j$(nproc)
-  echo "export BOOST_ROOT=$HOME/boost_1_75_0" >> $HOME/.profile && source $HOME/.profile
+- [Python 3.7](https://www.python.org/downloads/)
+- [Conan 1.55](https://conan.io/downloads.html)
+- [CMake 3.16](https://cmake.org/download/)
+- [**Optional**] [GCovr](https://gcc.gnu.org/onlinedocs/gcc/Gcov.html) (needed for code coverage generation)
 
-# Clone the Clio Git repository & build Clio
-  cd $HOME
-  git clone https://github.com/XRPLF/clio.git
-  cd $HOME/clio
-  cmake -B build && cmake --build build --parallel $(nproc)
+| Compiler    | Version |
+|-------------|---------|
+| GCC         | 11      |
+| Clang       | 14      |
+| Apple Clang | 14.0.3  |
+
+### Conan configuration
+
+Clio does not require anything but default settings in your (`~/.conan/profiles/default`) Conan profile. It's best to have no extra flags specified. 
+> Mac example:
 ```
+[settings]
+os=Macos
+os_build=Macos
+arch=armv8
+arch_build=armv8
+compiler=apple-clang
+compiler.version=14
+compiler.libcxx=libc++
+build_type=Release
+compiler.cppstd=20
+```
+> Linux example:
+```
+[settings]
+os=Linux
+os_build=Linux
+arch=x86_64
+arch_build=x86_64
+compiler=gcc
+compiler.version=11
+compiler.libcxx=libstdc++11
+build_type=Release
+compiler.cppstd=20
+```
+
+### Artifactory
+
+1. Make sure artifactory is setup with Conan
+```sh
+conan remote add --insert 0 conan-non-prod http://18.143.149.228:8081/artifactory/api/conan/conan-non-prod
+```
+Now you should be able to download prebuilt `xrpl` package on some platforms.
+
+2. Remove old packages you may have cached: 
+```sh 
+conan remove -f xrpl
+conan remove -f cassandra-cpp-driver
+```
+
+## Building Clio
+
+Navigate to Clio's root directory and perform
+```sh
+mkdir build && cd build
+conan install .. --output-folder . --build missing --settings build_type=Release -o tests=True
+cmake -DCMAKE_TOOLCHAIN_FILE:FILEPATH=build/generators/conan_toolchain.cmake -DCMAKE_BUILD_TYPE=Release ..
+cmake --build . --parallel 8 # or without the number if you feel extra adventurous
+```
+If all goes well, `conan install` will find required packages and `cmake` will do the rest. you should end up with `clio_server` and `clio_tests` in the `build` directory (the current directory).
+
+> **Tip:** You can omit the `-o tests=True` in `conan install` command above if you don't want to build `clio_tests`.
+
+> **Tip:** To generate a Code Coverage report, include `-o coverage=True` in the `conan install` command above, along with `-o tests=True` to enable tests. After running the `cmake` commands, execute `make clio_tests-ccov`. The coverage report will be found at `clio_tests-llvm-cov/index.html`.
 
 ## Running
 ```sh
@@ -96,12 +147,12 @@ The parameters `ssl_cert_file` and `ssl_key_file` can also be added to the top l
 An example of how to specify `ssl_cert_file` and `ssl_key_file` in the config:
 
 ```json
-"server":{
+"server": {
     "ip": "0.0.0.0",
     "port": 51233
 },
-"ssl_cert_file" : "/full/path/to/cert.file",
-"ssl_key_file" : "/full/path/to/key.file"
+"ssl_cert_file": "/full/path/to/cert.file",
+"ssl_key_file": "/full/path/to/key.file"
 ```
 
 Once your config files are ready, start rippled and Clio. It doesn't matter which you
@@ -172,6 +223,20 @@ which can cause high latencies. A possible alternative to this is to just deploy
 a database in each region, and the Clio nodes in each region use their region's database.
 This is effectively two systems.
 
+Clio supports API versioning as [described here](https://xrpl.org/request-formatting.html#api-versioning).
+It's possible to configure `minimum`, `maximum` and `default` version like so:
+```json
+"api_version": {
+    "min": 1,
+    "max": 2,
+    "default": 1
+}
+```
+All of the above are optional.
+Clio will fallback to hardcoded defaults when not specified in the config file or configured values are outside
+of the minimum and maximum supported versions hardcoded in `src/rpc/common/APIVersion.h`.
+> **Note:** See `example-config.json` for more details. 
+
 ## Developing against `rippled` in standalone mode
 
 If you wish you develop against a `rippled` instance running in standalone

cloud-example-config.json

@@ -1,38 +1,43 @@
+/*
+ * This is an example configuration file. Please do not use without modifying to suit your needs.
+ */
 {
-    "database":
-    {
-        "type":"cassandra",
-        "cassandra":
-        {
-            "secure_connect_bundle":"[path/to/zip. ignore if using contact_points]",
-            "contact_points":"[ip. ignore if using secure_connect_bundle]",
-            "port":"[port. ignore if using_secure_connect_bundle]",
-            "keyspace":"clio",
-            "username":"[username, if any]",
-            "password":"[password, if any]",
-            "max_requests_outstanding":25000,
-            "threads":8
+    "database": {
+        "type": "cassandra",
+        "cassandra": {
+            // This option can be used to setup a secure connect bundle connection
+            "secure_connect_bundle": "[path/to/zip. ignore if using contact_points]",
+            // The following options are used only if using contact_points
+            "contact_points": "[ip. ignore if using secure_connect_bundle]",
+            "port": "[port. ignore if using_secure_connect_bundle]",
+            // Authentication settings
+            "username": "[username, if any]",
+            "password": "[password, if any]",
+            // Other common settings
+            "keyspace": "clio",
+            "max_write_requests_outstanding": 25000,
+            "max_read_requests_outstanding": 30000,
+            "threads": 8
         }
     },
-    "etl_sources":
-    [
+    "etl_sources": [
         {
-            "ip":"[rippled ip]",
-            "ws_port":"6006",
-            "grpc_port":"50051"
+            "ip": "[rippled ip]",
+            "ws_port": "6006",
+            "grpc_port": "50051"
         }
     ],
-    "dos_guard":
-    {
-        "whitelist":["127.0.0.1"]
+    "dos_guard": {
+        "whitelist": [
+            "127.0.0.1"
+        ]
     },
-    "server":{
-        "ip":"0.0.0.0",
-        "port":8080
+    "server": {
+        "ip": "0.0.0.0",
+        "port": 8080
     },
-    "log_level":"debug",
-    "log_file":"./clio.log",
-    "online_delete":0,
-    "extractor_threads":8,
-    "read_only":false
+    "log_level": "debug",
+    "log_file": "./clio.log",
+    "extractor_threads": 8,
+    "read_only": false
 }

conanfile.py

@@ -0,0 +1,88 @@
+from conan import ConanFile
+from conan.tools.cmake import CMake, CMakeToolchain, cmake_layout
+import re
+
+class Clio(ConanFile):
+    name = 'clio'
+    license = 'ISC'
+    author = 'Alex Kremer <akremer@ripple.com>, John Freeman <jfreeman@ripple.com>'
+    url = 'https://github.com/xrplf/clio'
+    description = 'Clio RPC server'
+    settings = 'os', 'compiler', 'build_type', 'arch'
+    options = {
+        'fPIC': [True, False],
+        'verbose': [True, False],   
+        'tests': [True, False],     # build unit tests; create `clio_tests` binary
+        'docs': [True, False],      # doxygen API docs; create custom target 'docs'
+        'packaging': [True, False], # create distribution packages
+        'coverage': [True, False],  # build for test coverage report; create custom target `clio_tests-ccov`
+    }
+
+    requires = [
+        'boost/1.82.0',
+        'cassandra-cpp-driver/2.16.2',
+        'fmt/10.0.0',
+        'protobuf/3.21.12',
+        'grpc/1.50.1',
+        'openssl/1.1.1u',
+        'xrpl/1.12.0',
+    ]
+
+    default_options = {
+        'fPIC': True,
+        'verbose': False,
+        'tests': False,
+        'packaging': False,
+        'coverage': False,
+        'docs': False,
+        
+        'xrpl/*:tests': False,
+        'cassandra-cpp-driver/*:shared': False,
+        'date/*:header_only': True,
+        'grpc/*:shared': False,
+        'grpc/*:secure': True,
+        'libpq/*:shared': False,
+        'lz4/*:shared': False,
+        'openssl/*:shared': False,
+        'protobuf/*:shared': False,
+        'protobuf/*:with_zlib': True,
+        'snappy/*:shared': False,
+        'gtest/*:no_main': True,
+    }
+
+    exports_sources = (
+        'CMakeLists.txt', 'CMake/*', 'src/*'
+    )
+
+    def requirements(self):
+        if self.options.tests:
+            self.requires('gtest/1.13.0')
+
+    def configure(self):
+        if self.settings.compiler == 'apple-clang':
+            self.options['boost'].visibility = 'global'
+
+    def layout(self):
+        cmake_layout(self)
+        # Fix this setting to follow the default introduced in Conan 1.48 
+        # to align with our build instructions.
+        self.folders.generators = 'build/generators'
+
+    generators = 'CMakeDeps'
+    def generate(self):
+        tc = CMakeToolchain(self)
+        tc.variables['verbose'] = self.options.verbose
+        tc.variables['tests'] = self.options.tests
+        tc.variables['coverage'] = self.options.coverage
+        tc.variables['docs'] = self.options.docs
+        tc.variables['packaging'] = self.options.packaging
+        tc.generate()
+
+    def build(self):
+        cmake = CMake(self)
+        cmake.configure()
+        cmake.build()
+
+    def package(self):
+        cmake = CMake(self)
+        cmake.install()

example-config.json

@@ -1,3 +1,6 @@
+/*
+ * This is an example configuration file. Please do not use without modifying to suit your needs.
+ */
 {
     "database": {
         "type": "cassandra",
@@ -9,7 +12,18 @@
             "table_prefix": "",
             "max_write_requests_outstanding": 25000,
             "max_read_requests_outstanding": 30000,
-            "threads": 8
+            "threads": 8,
+            //
+            // Advanced options. USE AT OWN RISK:
+            // ---
+            "core_connections_per_host": 1 // Defaults to 1
+            //
+            // Below options will use defaults from cassandra driver if left unspecified.
+            // See https://docs.datastax.com/en/developer/cpp-driver/2.17/api/struct.CassCluster/ for details.
+            // 
+            // "queue_size_io": 2
+            //
+            // ---
         }
     },
     "etl_sources": [
@@ -20,21 +34,24 @@
         }
     ],
     "dos_guard": {
+        // Comma-separated list of IPs to exclude from rate limiting
         "whitelist": [
             "127.0.0.1"
-        ], // comma-separated list of ips to exclude from rate limiting
-        /* The below values are the default values and are only specified here
-         * for documentation purposes. The rate limiter currently limits
-         * connections and bandwidth per ip. The rate limiter looks at the raw
-         * ip of a client connection, and so requests routed through a load
-         * balancer will all have the same ip and be treated as a single client
-         */
-        "max_fetches": 1000000, // max bytes per ip per sweep interval
-        "max_connections": 20, // max connections per ip
-        "max_requests": 20, // max connections per ip
-        "sweep_interval": 1 // time in seconds before resetting bytes per ip count
+        ],
+        //
+        // The below values are the default values and are only specified here
+        // for documentation purposes. The rate limiter currently limits
+        // connections and bandwidth per IP. The rate limiter looks at the raw
+        // IP of a client connection, and so requests routed through a load
+        // balancer will all have the same IP and be treated as a single client.
+        //
+        "max_fetches": 1000000, // Max bytes per IP per sweep interval
+        "max_connections": 20, // Max connections per IP
+        "max_requests": 20, // Max connections per IP per sweep interval
+        "sweep_interval": 1 // Time in seconds before resetting max_fetches and max_requests
     },
     "cache": {
+        // Comma-separated list of peer nodes that Clio can use to download cache from at startup
         "peers": [
             {
                 "ip": "127.0.0.1",
@@ -45,11 +62,12 @@
     "server": {
         "ip": "0.0.0.0",
         "port": 51233,
-        /* Max number of requests to queue up before rejecting further requests.
-        * Defaults to 0, which disables the limit
-        */
+        // Max number of requests to queue up before rejecting further requests.
+        // Defaults to 0, which disables the limit.
         "max_queue_size": 500
     },
+    // Overrides log level on a per logging channel.
+    // Defaults to global "log_level" for each unspecified channel.
     "log_channels": [
         {
             "channel": "Backend",
@@ -77,17 +95,24 @@
         }
     ],
     "log_level": "info",
-    "log_format": "%TimeStamp% (%SourceLocation%) [%ThreadID%] %Channel%:%Severity% %Message%", // This is the default format
+    // Log format (this is the default format)
+    "log_format": "%TimeStamp% (%SourceLocation%) [%ThreadID%] %Channel%:%Severity% %Message%",
     "log_to_console": true,
-    "log_directory": "./clio_log",
+    // Clio logs to file in the specified directory only if "log_directory" is set
+    // "log_directory": "./clio_log",
     "log_rotation_size": 2048,
     "log_directory_max_size": 51200,
     "log_rotation_hour_interval": 12,
     "log_tag_style": "uint",
     "extractor_threads": 8,
     "read_only": false,
-    //"start_sequence": [integer] the ledger index to start from,
-    //"finish_sequence": [integer] the ledger index to finish at,
-    //"ssl_cert_file" : "/full/path/to/cert.file",
-    //"ssl_key_file" : "/full/path/to/key.file"
+    // "start_sequence": [integer] the ledger index to start from,
+    // "finish_sequence": [integer] the ledger index to finish at,
+    // "ssl_cert_file" : "/full/path/to/cert.file",
+    // "ssl_key_file" : "/full/path/to/key.file"
+    "api_version": {
+        "min": 1, // Minimum API version supported (could be 1 or 2)
+        "max": 2, // Maximum API version supported (could be 1 or 2, but >= min)
+        "default": 1 // Clio behaves the same as rippled by default
+    }
 }

external/cassandra/conanfile.py

@@ -0,0 +1,87 @@
+from conan import ConanFile, tools
+from conan.tools.cmake import CMake, CMakeToolchain
+
+class Cassandra(ConanFile):
+    name = 'cassandra-cpp-driver'
+    version = '2.16.2'
+    license = 'Apache-2.0'
+    url = 'https://github.com/conan-io/conan-center-index'
+    homepage = 'https://docs.datastax.com/en/developer/cpp-driver/'
+    description = 'Cassandra C++ Driver'
+    topics = ('conan', 'cassandra', 'driver')
+
+    settings = 'os', 'arch', 'compiler', 'build_type'
+    options = {
+        'shared': [True, False],
+        'fPIC': [True, False],
+        'install_header_in_subdir': [True, False],
+        'use_atomic': [None, 'boost', 'std'],
+        'with_openssl': [True, False],
+        'with_zlib': [True, False],
+        'with_kerberos': [True, False],
+        'use_timerfd': [True, False],
+    }
+    default_options = {
+        'shared': False,
+        'fPIC': True,
+        'install_header_in_subdir': False,
+        'use_atomic': None,
+        'with_openssl': True,
+        'with_zlib': True,
+        'with_kerberos': False,
+        'use_timerfd': True,
+    }
+
+    def requirements(self):
+        self.requires('libuv/1.44.1')
+        self.requires('http_parser/2.9.4')
+        if self.options.with_openssl:
+            self.requires('openssl/1.1.1q')
+        if self.options.with_zlib:
+            self.requires('minizip/1.2.12')
+            self.requires('zlib/1.2.13')
+        if self.options.use_atomic == 'boost':
+            self.requires('boost/1.79.0')
+
+    exports_sources = ['CMakeLists.txt']
+
+    def config_options(self):
+        if self.settings.os == 'Windows':
+            del self.options.fPIC
+
+    def configure(self):
+        self.options['libuv'].shared = self.options.shared
+
+    def generate(self):
+        tc = CMakeToolchain(self)
+        if self.settings.get_safe('compiler.cppstd') == '20':
+            tc.blocks['cppstd'].values['cppstd'] = 17
+        tc.variables['CASS_BUILD_STATIC'] = not self.options.shared
+        tc.variables['CASS_USE_STATIC_LIBS'] = not self.options.shared
+        tc.variables['CASS_BUILD_SHARED'] = self.options.shared
+        tc.variables['LIBUV_ROOT_DIR'] = self.deps_cpp_info['libuv'].rootpath
+        if self.options.with_openssl:
+            tc.variables['OPENSSL_ROOT_DIR'] = self.deps_cpp_info['openssl'].rootpath
+        tc.generate()
+
+    def source(self):
+        tools.files.get(self, f'https://github.com/datastax/cpp-driver/archive/refs/tags/{self.version}.tar.gz', strip_root=True)
+
+    def build(self):
+        cmake = CMake(self)
+        cmake.configure()
+        cmake.build()
+
+    def package(self):
+        cmake = CMake(self)
+        cmake.install()
+
+    def package_info(self):
+        if self.options.shared:
+            self.cpp_info.libs = ['cassandra']
+        else:
+            self.cpp_info.libs = ['cassandra_static']
+        if self.settings.os == 'Windows':
+            self.cpp_info.libs.extend(['iphlpapi', 'psapi', 'wsock32', 'crypt32', 'ws2_32', 'userenv'])
+            if not self.options.shared:
+                self.cpp_info.defines = ['CASS_STATIC']

src/backend/BackendInterface.h

@@ -1,663 +0,0 @@
-//------------------------------------------------------------------------------
-/*
-    This file is part of clio: https://github.com/XRPLF/clio
-    Copyright (c) 2022, the clio developers.
-
-    Permission to use, copy, modify, and distribute this software for any
-    purpose with or without fee is hereby granted, provided that the above
-    copyright notice and this permission notice appear in all copies.
-
-    THE  SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL WARRANTIES
-    WITH  REGARD  TO  THIS  SOFTWARE  INCLUDING  ALL  IMPLIED  WARRANTIES  OF
-    MERCHANTABILITY  AND  FITNESS. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR
-    ANY  SPECIAL,  DIRECT,  INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES
-    WHATSOEVER  RESULTING  FROM  LOSS  OF USE, DATA OR PROFITS, WHETHER IN AN
-    ACTION  OF  CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF
-    OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE.
-*/
-//==============================================================================
-
-#pragma once
-
-#include <ripple/ledger/ReadView.h>
-#include <backend/DBHelpers.h>
-#include <backend/SimpleCache.h>
-#include <backend/Types.h>
-#include <config/Config.h>
-#include <log/Logger.h>
-
-#include <boost/asio/spawn.hpp>
-#include <boost/json.hpp>
-
-#include <thread>
-#include <type_traits>
-
-namespace Backend {
-
-/**
- * @brief Throws an error when database read time limit is exceeded.
- *
- * This class is throws an error when read time limit is exceeded but
- * is also paired with a separate class to retry the connection.
- */
-class DatabaseTimeout : public std::exception
-{
-public:
-    const char*
-    what() const throw() override
-    {
-        return "Database read timed out. Please retry the request";
-    }
-};
-
-/**
- * @brief Separate class that reattempts connection after time limit.
- *
- * @tparam F Represents a class of handlers for Cassandra database.
- * @param func Instance of Cassandra database handler class.
- * @param waitMs Is the arbitrary time limit of 500ms.
- * @return auto
- */
-template <class F>
-auto
-retryOnTimeout(F func, size_t waitMs = 500)
-{
-    static clio::Logger log{"Backend"};
-
-    while (true)
-    {
-        try
-        {
-            return func();
-        }
-        catch (DatabaseTimeout& t)
-        {
-            log.error()
-                << "Database request timed out. Sleeping and retrying ... ";
-            std::this_thread::sleep_for(std::chrono::milliseconds(waitMs));
-        }
-    }
-}
-
-/**
- * @brief Passes in serialized handlers in an asynchronous fashion.
- *
- * Note that the synchronous auto passes handlers critical to supporting
- * the Clio backend. The coroutine types are checked if same/different.
- *
- * @tparam F Represents a class of handlers for Cassandra database.
- * @param f R-value instance of Cassandra handler class.
- * @return auto
- */
-template <class F>
-auto
-synchronous(F&& f)
-{
-    /** @brief Serialized handlers and their execution.
-     *
-     * The ctx class is converted into a serialized handler, also named
-     * ctx, and is used to pass a stream of data into the method.
-     */
-    boost::asio::io_context ctx;
-    boost::asio::io_context::strand strand(ctx);
-    std::optional<boost::asio::io_context::work> work;
-
-    /*! @brief Place the ctx within the vector of serialized handlers. */
-    work.emplace(ctx);
-
-    /**
-     * @brief If/else statements regarding coroutine type matching.
-     *
-     * R is the currently executing coroutine that is about to get passed.
-     * If corountine types do not match, the current one's type is stored.
-     */
-    using R = typename std::result_of<F(boost::asio::yield_context&)>::type;
-    if constexpr (!std::is_same<R, void>::value)
-    {
-        /**
-         * @brief When the coroutine type is the same
-         *
-         * The spawn function enables programs to implement asynchronous logic
-         * in a synchronous manner. res stores the instance of the currently
-         * executing coroutine, yield. The different type is returned.
-         */
-        R res;
-        boost::asio::spawn(
-            strand, [&f, &work, &res](boost::asio::yield_context yield) {
-                res = f(yield);
-                work.reset();
-            });
-
-        ctx.run();
-        return res;
-    }
-    else
-    {
-        /*! @brief When the corutine type is different, run as normal. */
-        boost::asio::spawn(
-            strand, [&f, &work](boost::asio::yield_context yield) {
-                f(yield);
-                work.reset();
-            });
-
-        ctx.run();
-    }
-}
-
-/**
- * @brief Reestablishes synchronous connection on timeout.
- *
- * @tparam Represents a class of handlers for Cassandra database.
- * @param f R-value instance of Cassandra database handler class.
- * @return auto
- */
-template <class F>
-auto
-synchronousAndRetryOnTimeout(F&& f)
-{
-    return retryOnTimeout([&]() { return synchronous(f); });
-}
-
-/*! @brief Handles ledger and transaction backend data. */
-class BackendInterface
-{
-    /**
-     * @brief Shared mutexes and a cache for the interface.
-     *
-     * rngMutex is a shared mutex. Shared mutexes prevent shared data
-     * from being accessed by multiple threads and has two levels of
-     * access: shared and exclusive.
-     */
-protected:
-    mutable std::shared_mutex rngMtx_;
-    std::optional<LedgerRange> range;
-    SimpleCache cache_;
-
-    /**
-     * @brief Public read methods
-     *
-     * All of these reads methods can throw DatabaseTimeout. When writing
-     * code in an RPC handler, this exception does not need to be caught:
-     * when an RPC results in a timeout, an error is returned to the client.
-     */
-
-public:
-    BackendInterface(clio::Config const& config)
-    {
-    }
-    virtual ~BackendInterface()
-    {
-    }
-
-    /*! @brief LEDGER METHODS */
-public:
-    /**
-     * @brief Cache that holds states of the ledger
-     *
-     * const version holds the original cache state; the other tracks
-     * historical changes.
-     *
-     * @return SimpleCache const&
-     */
-    SimpleCache const&
-    cache() const
-    {
-        return cache_;
-    }
-
-    SimpleCache&
-    cache()
-    {
-        return cache_;
-    }
-
-    /*! @brief Fetches a specific ledger by sequence number. */
-    virtual std::optional<ripple::LedgerInfo>
-    fetchLedgerBySequence(
-        std::uint32_t const sequence,
-        boost::asio::yield_context& yield) const = 0;
-
-    /*! @brief Fetches a specific ledger by hash. */
-    virtual std::optional<ripple::LedgerInfo>
-    fetchLedgerByHash(
-        ripple::uint256 const& hash,
-        boost::asio::yield_context& yield) const = 0;
-
-    /*! @brief Fetches the latest ledger sequence. */
-    virtual std::optional<std::uint32_t>
-    fetchLatestLedgerSequence(boost::asio::yield_context& yield) const = 0;
-
-    /*! @brief Fetches the current ledger range while locking that process */
-    std::optional<LedgerRange>
-    fetchLedgerRange() const
-    {
-        std::shared_lock lck(rngMtx_);
-        return range;
-    }
-
-    /**
-     * @brief Updates the range of sequences to be tracked.
-     *
-     * Function that continues updating the range sliding window or creates
-     * a new sliding window once the maxSequence limit has been reached.
-     *
-     * @param newMax Unsigned 32-bit integer representing new max of range.
-     */
-    void
-    updateRange(uint32_t newMax)
-    {
-        std::scoped_lock lck(rngMtx_);
-        assert(!range || newMax >= range->maxSequence);
-        if (!range)
-            range = {newMax, newMax};
-        else
-            range->maxSequence = newMax;
-    }
-
-    /**
-     * @brief Returns the fees for specific transactions.
-     *
-     * @param seq Unsigned 32-bit integer reprsenting sequence.
-     * @param yield The currently executing coroutine.
-     * @return std::optional<ripple::Fees>
-     */
-    std::optional<ripple::Fees>
-    fetchFees(std::uint32_t const seq, boost::asio::yield_context& yield) const;
-
-    /*! @brief TRANSACTION METHODS */
-    /**
-     * @brief Fetches a specific transaction.
-     *
-     * @param hash Unsigned 256-bit integer representing hash.
-     * @param yield The currently executing coroutine.
-     * @return std::optional<TransactionAndMetadata>
-     */
-    virtual std::optional<TransactionAndMetadata>
-    fetchTransaction(
-        ripple::uint256 const& hash,
-        boost::asio::yield_context& yield) const = 0;
-
-    /**
-     * @brief Fetches multiple transactions.
-     *
-     * @param hashes Unsigned integer value representing a hash.
-     * @param yield The currently executing coroutine.
-     * @return std::vector<TransactionAndMetadata>
-     */
-    virtual std::vector<TransactionAndMetadata>
-    fetchTransactions(
-        std::vector<ripple::uint256> const& hashes,
-        boost::asio::yield_context& yield) const = 0;
-
-    /**
-     * @brief Fetches all transactions for a specific account
-     *
-     * @param account A specific XRPL Account, speciifed by unique type
-     * accountID.
-     * @param limit Paging limit for how many transactions can be returned per
-     * page.
-     * @param forward Boolean whether paging happens forwards or backwards.
-     * @param cursor Important metadata returned every time paging occurs.
-     * @param yield Currently executing coroutine.
-     * @return TransactionsAndCursor
-     */
-    virtual TransactionsAndCursor
-    fetchAccountTransactions(
-        ripple::AccountID const& account,
-        std::uint32_t const limit,
-        bool forward,
-        std::optional<TransactionsCursor> const& cursor,
-        boost::asio::yield_context& yield) const = 0;
-
-    /**
-     * @brief Fetches all transactions from a specific ledger.
-     *
-     * @param ledgerSequence Unsigned 32-bit integer for latest total
-     * transactions.
-     * @param yield Currently executing coroutine.
-     * @return std::vector<TransactionAndMetadata>
-     */
-    virtual std::vector<TransactionAndMetadata>
-    fetchAllTransactionsInLedger(
-        std::uint32_t const ledgerSequence,
-        boost::asio::yield_context& yield) const = 0;
-
-    /**
-     * @brief Fetches all transaction hashes from a specific ledger.
-     *
-     * @param ledgerSequence Standard unsigned integer.
-     * @param yield Currently executing coroutine.
-     * @return std::vector<ripple::uint256>
-     */
-    virtual std::vector<ripple::uint256>
-    fetchAllTransactionHashesInLedger(
-        std::uint32_t const ledgerSequence,
-        boost::asio::yield_context& yield) const = 0;
-
-    /*! @brief NFT methods */
-    /**
-     * @brief Fetches a specific NFT
-     *
-     * @param tokenID Unsigned 256-bit integer.
-     * @param ledgerSequence Standard unsigned integer.
-     * @param yield Currently executing coroutine.
-     * @return std::optional<NFT>
-     */
-    virtual std::optional<NFT>
-    fetchNFT(
-        ripple::uint256 const& tokenID,
-        std::uint32_t const ledgerSequence,
-        boost::asio::yield_context& yield) const = 0;
-
-    /**
-     * @brief Fetches all transactions for a specific NFT.
-     *
-     * @param tokenID Unsigned 256-bit integer.
-     * @param limit Paging limit as to how many transactions return per page.
-     * @param forward Boolean whether paging happens forwards or backwards.
-     * @param cursorIn Represents transaction number and ledger sequence.
-     * @param yield Currently executing coroutine is passed in as input.
-     * @return TransactionsAndCursor
-     */
-    virtual TransactionsAndCursor
-    fetchNFTTransactions(
-        ripple::uint256 const& tokenID,
-        std::uint32_t const limit,
-        bool const forward,
-        std::optional<TransactionsCursor> const& cursorIn,
-        boost::asio::yield_context& yield) const = 0;
-
-    /*! @brief STATE DATA METHODS */
-    /**
-     * @brief Fetches a specific ledger object: vector of unsigned chars
-     *
-     * @param key Unsigned 256-bit integer.
-     * @param sequence Unsigned 32-bit integer.
-     * @param yield Currently executing coroutine.
-     * @return std::optional<Blob>
-     */
-    std::optional<Blob>
-    fetchLedgerObject(
-        ripple::uint256 const& key,
-        std::uint32_t const sequence,
-        boost::asio::yield_context& yield) const;
-
-    /**
-     * @brief Fetches all ledger objects: a vector of vectors of unsigned chars.
-     *
-     * @param keys Unsigned 256-bit integer.
-     * @param sequence Unsigned 32-bit integer.
-     * @param yield Currently executing coroutine.
-     * @return std::vector<Blob>
-     */
-    std::vector<Blob>
-    fetchLedgerObjects(
-        std::vector<ripple::uint256> const& keys,
-        std::uint32_t const sequence,
-        boost::asio::yield_context& yield) const;
-
-    /*! @brief Virtual function version of fetchLedgerObject */
-    virtual std::optional<Blob>
-    doFetchLedgerObject(
-        ripple::uint256 const& key,
-        std::uint32_t const sequence,
-        boost::asio::yield_context& yield) const = 0;
-
-    /*! @brief Virtual function version of fetchLedgerObjects */
-    virtual std::vector<Blob>
-    doFetchLedgerObjects(
-        std::vector<ripple::uint256> const& keys,
-        std::uint32_t const sequence,
-        boost::asio::yield_context& yield) const = 0;
-
-    /**
-     * @brief Returns the difference between ledgers: vector of objects
-     *
-     * Objects are made of a key value, vector of unsigned chars (blob),
-     * and a boolean detailing whether keys and blob match.
-     *
-     * @param ledgerSequence Standard unsigned integer.
-     * @param yield Currently executing coroutine.
-     * @return std::vector<LedgerObject>
-     */
-    virtual std::vector<LedgerObject>
-    fetchLedgerDiff(
-        std::uint32_t const ledgerSequence,
-        boost::asio::yield_context& yield) const = 0;
-
-    /**
-     * @brief Fetches a page of ledger objects, ordered by key/index.
-     *
-     * @param cursor Important metadata returned every time paging occurs.
-     * @param ledgerSequence Standard unsigned integer.
-     * @param limit Paging limit as to how many transactions returned per page.
-     * @param outOfOrder Boolean on whether ledger page is out of order.
-     * @param yield Currently executing coroutine.
-     * @return LedgerPage
-     */
-    LedgerPage
-    fetchLedgerPage(
-        std::optional<ripple::uint256> const& cursor,
-        std::uint32_t const ledgerSequence,
-        std::uint32_t const limit,
-        bool outOfOrder,
-        boost::asio::yield_context& yield) const;
-
-    /*! @brief Fetches successor object from key/index. */
-    std::optional<LedgerObject>
-    fetchSuccessorObject(
-        ripple::uint256 key,
-        std::uint32_t const ledgerSequence,
-        boost::asio::yield_context& yield) const;
-
-    /*! @brief Fetches successor key from key/index. */
-    std::optional<ripple::uint256>
-    fetchSuccessorKey(
-        ripple::uint256 key,
-        std::uint32_t const ledgerSequence,
-        boost::asio::yield_context& yield) const;
-
-    /*! @brief Virtual function version of fetchSuccessorKey. */
-    virtual std::optional<ripple::uint256>
-    doFetchSuccessorKey(
-        ripple::uint256 key,
-        std::uint32_t const ledgerSequence,
-        boost::asio::yield_context& yield) const = 0;
-
-    /**
-     * @brief Fetches book offers.
-     *
-     * @param book Unsigned 256-bit integer.
-     * @param ledgerSequence Standard unsigned integer.
-     * @param limit Pagaing limit as to how many transactions returned per page.
-     * @param cursor Important metadata returned every time paging occurs.
-     * @param yield Currently executing coroutine.
-     * @return BookOffersPage
-     */
-    BookOffersPage
-    fetchBookOffers(
-        ripple::uint256 const& book,
-        std::uint32_t const ledgerSequence,
-        std::uint32_t const limit,
-        std::optional<ripple::uint256> const& cursor,
-        boost::asio::yield_context& yield) const;
-
-    /**
-     * @brief Returns a ledger range
-     *
-     * Ledger range is a struct of min and max sequence numbers). Due to
-     * the use of [&], which denotes a special case of a lambda expression
-     * where values found outside the scope are passed by reference, wrt the
-     * currently executing coroutine.
-     *
-     * @return std::optional<LedgerRange>
-     */
-    std::optional<LedgerRange>
-    hardFetchLedgerRange() const
-    {
-        return synchronous([&](boost::asio::yield_context yield) {
-            return hardFetchLedgerRange(yield);
-        });
-    }
-
-    /*! @brief Virtual function equivalent of hardFetchLedgerRange. */
-    virtual std::optional<LedgerRange>
-    hardFetchLedgerRange(boost::asio::yield_context& yield) const = 0;
-
-    /*! @brief Fetches ledger range but doesn't throw timeout. Use with care. */
-    std::optional<LedgerRange>
-    hardFetchLedgerRangeNoThrow() const;
-    /*! @brief Fetches ledger range but doesn't throw timeout. Use with care. */
-    std::optional<LedgerRange>
-    hardFetchLedgerRangeNoThrow(boost::asio::yield_context& yield) const;
-
-    /**
-     * @brief Writes to a specific ledger.
-     *
-     * @param ledgerInfo Const on ledger information.
-     * @param ledgerHeader r-value string representing ledger header.
-     */
-    virtual void
-    writeLedger(
-        ripple::LedgerInfo const& ledgerInfo,
-        std::string&& ledgerHeader) = 0;
-
-    /**
-     * @brief Writes a new ledger object.
-     *
-     * The key and blob are r-value references and do NOT have memory addresses.
-     *
-     * @param key String represented as an r-value.
-     * @param seq Unsigned integer representing a sequence.
-     * @param blob r-value vector of unsigned characters (blob).
-     */
-    virtual void
-    writeLedgerObject(
-        std::string&& key,
-        std::uint32_t const seq,
-        std::string&& blob);
-
-    /**
-     * @brief Writes a new transaction.
-     *
-     * @param hash r-value reference. No memory address.
-     * @param seq Unsigned 32-bit integer.
-     * @param date Unsigned 32-bit integer.
-     * @param transaction r-value reference. No memory address.
-     * @param metadata r-value refrence. No memory address.
-     */
-    virtual void
-    writeTransaction(
-        std::string&& hash,
-        std::uint32_t const seq,
-        std::uint32_t const date,
-        std::string&& transaction,
-        std::string&& metadata) = 0;
-
-    /**
-     * @brief Write a new NFT.
-     *
-     * @param data Passed in as an r-value reference.
-     */
-    virtual void
-    writeNFTs(std::vector<NFTsData>&& data) = 0;
-
-    /**
-     * @brief Write a new set of account transactions.
-     *
-     * @param data Passed in as an r-value reference.
-     */
-    virtual void
-    writeAccountTransactions(std::vector<AccountTransactionsData>&& data) = 0;
-
-    /**
-     * @brief Write a new transaction for a specific NFT.
-     *
-     * @param data Passed in as an r-value reference.
-     */
-    virtual void
-    writeNFTTransactions(std::vector<NFTTransactionsData>&& data) = 0;
-
-    /**
-     * @brief Write a new successor.
-     *
-     * @param key Passed in as an r-value reference.
-     * @param seq Unsigned 32-bit integer.
-     * @param successor Passed in as an r-value reference.
-     */
-    virtual void
-    writeSuccessor(
-        std::string&& key,
-        std::uint32_t const seq,
-        std::string&& successor) = 0;
-
-    /*! @brief Tells database we will write data for a specific ledger. */
-    virtual void
-    startWrites() const = 0;
-
-    /**
-     * @brief Tells database we finished writing all data for a specific ledger.
-     *
-     * TODO: change the return value to represent different results:
-     * Committed, write conflict, errored, successful but not committed
-     *
-     * @param ledgerSequence Const unsigned 32-bit integer on ledger sequence.
-     * @return true
-     * @return false
-     */
-    bool
-    finishWrites(std::uint32_t const ledgerSequence);
-
-    /**
-     * @brief Selectively delets parts of the database.
-     *
-     * @param numLedgersToKeep Unsigned 32-bit integer on number of ledgers to
-     * keep.
-     * @param yield Currently executing coroutine.
-     * @return true
-     * @return false
-     */
-    virtual bool
-    doOnlineDelete(
-        std::uint32_t numLedgersToKeep,
-        boost::asio::yield_context& yield) const = 0;
-
-    /**
-     * @brief Opens the database
-     *
-     * Open the database. Set up all of the necessary objects and
-     * datastructures. After this call completes, the database is
-     * ready for use.
-     *
-     * @param readOnly Boolean whether ledger is read only.
-     */
-    virtual void
-    open(bool readOnly) = 0;
-
-    /*! @brief Closes the database, releasing any resources. */
-    virtual void
-    close(){};
-
-    virtual bool
-    isTooBusy() const = 0;
-
-private:
-    /**
-     * @brief Private helper method to write ledger object
-     *
-     * @param key r-value string representing key.
-     * @param seq Unsigned 32-bit integer representing sequence.
-     * @param blob r-value vector of unsigned chars.
-     */
-    virtual void
-    doWriteLedgerObject(
-        std::string&& key,
-        std::uint32_t const seq,
-        std::string&& blob) = 0;
-
-    virtual bool
-    doFinishWrites() = 0;
-};
-
-}  // namespace Backend
-using BackendInterface = Backend::BackendInterface;

src/backend/DBHelpers.h

@@ -1,201 +0,0 @@
-//------------------------------------------------------------------------------
-/*
-    This file is part of clio: https://github.com/XRPLF/clio
-    Copyright (c) 2022, the clio developers.
-
-    Permission to use, copy, modify, and distribute this software for any
-    purpose with or without fee is hereby granted, provided that the above
-    copyright notice and this permission notice appear in all copies.
-
-    THE  SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL WARRANTIES
-    WITH  REGARD  TO  THIS  SOFTWARE  INCLUDING  ALL  IMPLIED  WARRANTIES  OF
-    MERCHANTABILITY  AND  FITNESS. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR
-    ANY  SPECIAL,  DIRECT,  INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES
-    WHATSOEVER  RESULTING  FROM  LOSS  OF USE, DATA OR PROFITS, WHETHER IN AN
-    ACTION  OF  CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF
-    OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE.
-*/
-//==============================================================================
-
-#pragma once
-
-#include <ripple/basics/Log.h>
-#include <ripple/basics/StringUtilities.h>
-#include <ripple/ledger/ReadView.h>
-#include <ripple/protocol/SField.h>
-#include <ripple/protocol/STAccount.h>
-#include <ripple/protocol/TxMeta.h>
-
-#include <boost/container/flat_set.hpp>
-
-#include <backend/Types.h>
-
-/// Struct used to keep track of what to write to
-/// account_transactions/account_tx tables
-struct AccountTransactionsData
-{
-    boost::container::flat_set<ripple::AccountID> accounts;
-    std::uint32_t ledgerSequence;
-    std::uint32_t transactionIndex;
-    ripple::uint256 txHash;
-
-    AccountTransactionsData(
-        ripple::TxMeta& meta,
-        ripple::uint256 const& txHash,
-        beast::Journal& j)
-        : accounts(meta.getAffectedAccounts())
-        , ledgerSequence(meta.getLgrSeq())
-        , transactionIndex(meta.getIndex())
-        , txHash(txHash)
-    {
-    }
-
-    AccountTransactionsData() = default;
-};
-
-/// Represents a link from a tx to an NFT that was targeted/modified/created
-/// by it. Gets written to nf_token_transactions table and the like.
-struct NFTTransactionsData
-{
-    ripple::uint256 tokenID;
-    std::uint32_t ledgerSequence;
-    std::uint32_t transactionIndex;
-    ripple::uint256 txHash;
-
-    NFTTransactionsData(
-        ripple::uint256 const& tokenID,
-        ripple::TxMeta const& meta,
-        ripple::uint256 const& txHash)
-        : tokenID(tokenID)
-        , ledgerSequence(meta.getLgrSeq())
-        , transactionIndex(meta.getIndex())
-        , txHash(txHash)
-    {
-    }
-};
-
-/// Represents an NFT state at a particular ledger. Gets written to nf_tokens
-/// table and the like.
-struct NFTsData
-{
-    ripple::uint256 tokenID;
-    std::uint32_t ledgerSequence;
-
-    // The transaction index is only stored because we want to store only the
-    // final state of an NFT per ledger. Since we pull this from transactions
-    // we keep track of which tx index created this so we can de-duplicate, as
-    // it is possible for one ledger to have multiple txs that change the
-    // state of the same NFT.
-    std::uint32_t transactionIndex;
-    ripple::AccountID owner;
-    bool isBurned;
-
-    NFTsData(
-        ripple::uint256 const& tokenID,
-        ripple::AccountID const& owner,
-        ripple::TxMeta const& meta,
-        bool isBurned)
-        : tokenID(tokenID)
-        , ledgerSequence(meta.getLgrSeq())
-        , transactionIndex(meta.getIndex())
-        , owner(owner)
-        , isBurned(isBurned)
-    {
-    }
-};
-
-template <class T>
-inline bool
-isOffer(T const& object)
-{
-    short offer_bytes = (object[1] << 8) | object[2];
-    return offer_bytes == 0x006f;
-}
-
-template <class T>
-inline bool
-isOfferHex(T const& object)
-{
-    auto blob = ripple::strUnHex(4, object.begin(), object.begin() + 4);
-    if (blob)
-    {
-        short offer_bytes = ((*blob)[1] << 8) | (*blob)[2];
-        return offer_bytes == 0x006f;
-    }
-    return false;
-}
-
-template <class T>
-inline bool
-isDirNode(T const& object)
-{
-    short spaceKey = (object.data()[1] << 8) | object.data()[2];
-    return spaceKey == 0x0064;
-}
-
-template <class T, class R>
-inline bool
-isBookDir(T const& key, R const& object)
-{
-    if (!isDirNode(object))
-        return false;
-
-    ripple::STLedgerEntry const sle{
-        ripple::SerialIter{object.data(), object.size()}, key};
-    return !sle[~ripple::sfOwner].has_value();
-}
-
-template <class T>
-inline ripple::uint256
-getBook(T const& offer)
-{
-    ripple::SerialIter it{offer.data(), offer.size()};
-    ripple::SLE sle{it, {}};
-    ripple::uint256 book = sle.getFieldH256(ripple::sfBookDirectory);
-    return book;
-}
-
-template <class T>
-inline ripple::uint256
-getBookBase(T const& key)
-{
-    assert(key.size() == ripple::uint256::size());
-    ripple::uint256 ret;
-    for (size_t i = 0; i < 24; ++i)
-    {
-        ret.data()[i] = key.data()[i];
-    }
-    return ret;
-}
-
-inline ripple::LedgerInfo
-deserializeHeader(ripple::Slice data)
-{
-    ripple::SerialIter sit(data.data(), data.size());
-
-    ripple::LedgerInfo info;
-
-    info.seq = sit.get32();
-    info.drops = sit.get64();
-    info.parentHash = sit.get256();
-    info.txHash = sit.get256();
-    info.accountHash = sit.get256();
-    info.parentCloseTime =
-        ripple::NetClock::time_point{ripple::NetClock::duration{sit.get32()}};
-    info.closeTime =
-        ripple::NetClock::time_point{ripple::NetClock::duration{sit.get32()}};
-    info.closeTimeResolution = ripple::NetClock::duration{sit.get8()};
-    info.closeFlags = sit.get8();
-
-    info.hash = sit.get256();
-
-    return info;
-}
-
-inline std::string
-uint256ToString(ripple::uint256 const& uint)
-{
-    return {reinterpret_cast<const char*>(uint.data()), uint.size()};
-}
-
-static constexpr std::uint32_t rippleEpochStart = 946684800;

src/backend/SimpleCache.h

@@ -1,98 +0,0 @@
-//------------------------------------------------------------------------------
-/*
-    This file is part of clio: https://github.com/XRPLF/clio
-    Copyright (c) 2022, the clio developers.
-
-    Permission to use, copy, modify, and distribute this software for any
-    purpose with or without fee is hereby granted, provided that the above
-    copyright notice and this permission notice appear in all copies.
-
-    THE  SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL WARRANTIES
-    WITH  REGARD  TO  THIS  SOFTWARE  INCLUDING  ALL  IMPLIED  WARRANTIES  OF
-    MERCHANTABILITY  AND  FITNESS. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR
-    ANY  SPECIAL,  DIRECT,  INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES
-    WHATSOEVER  RESULTING  FROM  LOSS  OF USE, DATA OR PROFITS, WHETHER IN AN
-    ACTION  OF  CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF
-    OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE.
-*/
-//==============================================================================
-
-#pragma once
-
-#include <ripple/basics/base_uint.h>
-#include <ripple/basics/hardened_hash.h>
-#include <backend/Types.h>
-#include <map>
-#include <mutex>
-#include <shared_mutex>
-#include <utility>
-#include <vector>
-namespace Backend {
-class SimpleCache
-{
-    struct CacheEntry
-    {
-        uint32_t seq = 0;
-        Blob blob;
-    };
-
-    // counters for fetchLedgerObject(s) hit rate
-    mutable std::atomic_uint32_t objectReqCounter_;
-    mutable std::atomic_uint32_t objectHitCounter_;
-    // counters for fetchSuccessorKey hit rate
-    mutable std::atomic_uint32_t successorReqCounter_;
-    mutable std::atomic_uint32_t successorHitCounter_;
-
-    std::map<ripple::uint256, CacheEntry> map_;
-    mutable std::shared_mutex mtx_;
-    uint32_t latestSeq_ = 0;
-    std::atomic_bool full_ = false;
-    std::atomic_bool disabled_ = false;
-    // temporary set to prevent background thread from writing already deleted
-    // data. not used when cache is full
-    std::unordered_set<ripple::uint256, ripple::hardened_hash<>> deletes_;
-
-public:
-    // Update the cache with new ledger objects
-    // set isBackground to true when writing old data from a background thread
-    void
-    update(
-        std::vector<LedgerObject> const& blobs,
-        uint32_t seq,
-        bool isBackground = false);
-
-    std::optional<Blob>
-    get(ripple::uint256 const& key, uint32_t seq) const;
-
-    // always returns empty optional if isFull() is false
-    std::optional<LedgerObject>
-    getSuccessor(ripple::uint256 const& key, uint32_t seq) const;
-
-    // always returns empty optional if isFull() is false
-    std::optional<LedgerObject>
-    getPredecessor(ripple::uint256 const& key, uint32_t seq) const;
-
-    void
-    setDisabled();
-
-    void
-    setFull();
-
-    uint32_t
-    latestLedgerSequence() const;
-
-    // whether the cache has all data for the most recent ledger
-    bool
-    isFull() const;
-
-    size_t
-    size() const;
-
-    float
-    getObjectHitRate() const;
-
-    float
-    getSuccessorHitRate() const;
-};
-
-}  // namespace Backend

src/backend/Types.h

@@ -1,110 +0,0 @@
-//------------------------------------------------------------------------------
-/*
-    This file is part of clio: https://github.com/XRPLF/clio
-    Copyright (c) 2022, the clio developers.
-
-    Permission to use, copy, modify, and distribute this software for any
-    purpose with or without fee is hereby granted, provided that the above
-    copyright notice and this permission notice appear in all copies.
-
-    THE  SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL WARRANTIES
-    WITH  REGARD  TO  THIS  SOFTWARE  INCLUDING  ALL  IMPLIED  WARRANTIES  OF
-    MERCHANTABILITY  AND  FITNESS. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR
-    ANY  SPECIAL,  DIRECT,  INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES
-    WHATSOEVER  RESULTING  FROM  LOSS  OF USE, DATA OR PROFITS, WHETHER IN AN
-    ACTION  OF  CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF
-    OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE.
-*/
-//==============================================================================
-
-#pragma once
-
-#include <ripple/basics/base_uint.h>
-#include <ripple/protocol/AccountID.h>
-#include <optional>
-#include <string>
-#include <vector>
-
-namespace Backend {
-
-// *** return types
-
-using Blob = std::vector<unsigned char>;
-
-struct LedgerObject
-{
-    ripple::uint256 key;
-    Blob blob;
-    bool
-    operator==(const LedgerObject& other) const
-    {
-        return key == other.key && blob == other.blob;
-    }
-};
-
-struct LedgerPage
-{
-    std::vector<LedgerObject> objects;
-    std::optional<ripple::uint256> cursor;
-};
-struct BookOffersPage
-{
-    std::vector<LedgerObject> offers;
-    std::optional<ripple::uint256> cursor;
-};
-struct TransactionAndMetadata
-{
-    Blob transaction;
-    Blob metadata;
-    std::uint32_t ledgerSequence;
-    std::uint32_t date;
-    bool
-    operator==(const TransactionAndMetadata& other) const
-    {
-        return transaction == other.transaction && metadata == other.metadata &&
-            ledgerSequence == other.ledgerSequence && date == other.date;
-    }
-};
-
-struct TransactionsCursor
-{
-    std::uint32_t ledgerSequence;
-    std::uint32_t transactionIndex;
-};
-
-struct TransactionsAndCursor
-{
-    std::vector<TransactionAndMetadata> txns;
-    std::optional<TransactionsCursor> cursor;
-};
-
-struct NFT
-{
-    ripple::uint256 tokenID;
-    std::uint32_t ledgerSequence;
-    ripple::AccountID owner;
-    bool isBurned;
-
-    // clearly two tokens are the same if they have the same ID, but this
-    // struct stores the state of a given token at a given ledger sequence, so
-    // we also need to compare with ledgerSequence
-    bool
-    operator==(NFT const& other) const
-    {
-        return tokenID == other.tokenID &&
-            ledgerSequence == other.ledgerSequence;
-    }
-};
-
-struct LedgerRange
-{
-    std::uint32_t minSequence;
-    std::uint32_t maxSequence;
-};
-constexpr ripple::uint256 firstKey{
-    "0000000000000000000000000000000000000000000000000000000000000000"};
-constexpr ripple::uint256 lastKey{
-    "FFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFF"};
-constexpr ripple::uint256 hi192{
-    "0000000000000000000000000000000000000000000000001111111111111111"};
-}  // namespace Backend

src/data/BackendFactory.h

@@ -19,44 +19,50 @@
 
 #pragma once
 
-#include <backend/BackendInterface.h>
-#include <backend/CassandraBackend.h>
-#include <config/Config.h>
-#include <log/Logger.h>
+#include <data/BackendInterface.h>
+#include <data/CassandraBackend.h>
+#include <util/config/Config.h>
+#include <util/log/Logger.h>
 
 #include <boost/algorithm/string.hpp>
 
-namespace Backend {
-std::shared_ptr<BackendInterface>
-make_Backend(boost::asio::io_context& ioc, clio::Config const& config)
-{
-    static clio::Logger log{"Backend"};
-    log.info() << "Constructing BackendInterface";
+namespace data {
 
-    auto readOnly = config.valueOr("read_only", false);
-    auto type = config.value<std::string>("database.type");
+/**
+ * @brief A factory function that creates the backend based on a config.
+ *
+ * @param config The clio config to use
+ * @return A shared_ptr<BackendInterface> with the selected implementation
+ */
+std::shared_ptr<BackendInterface>
+make_Backend(util::Config const& config)
+{
+    static util::Logger log{"Backend"};
+    LOG(log.info()) << "Constructing BackendInterface";
+
+    auto const readOnly = config.valueOr("read_only", false);
+
+    auto const type = config.value<std::string>("database.type");
     std::shared_ptr<BackendInterface> backend = nullptr;
 
-    if (boost::iequals(type, "cassandra"))
+    // TODO: retire `cassandra-new` by next release after 2.0
+    if (boost::iequals(type, "cassandra") or boost::iequals(type, "cassandra-new"))
     {
         auto cfg = config.section("database." + type);
-        auto ttl = config.valueOr<uint32_t>("online_delete", 0) * 4;
-        backend = std::make_shared<CassandraBackend>(ioc, cfg, ttl);
+        backend = std::make_shared<data::cassandra::CassandraBackend>(data::cassandra::SettingsProvider{cfg}, readOnly);
     }
 
     if (!backend)
         throw std::runtime_error("Invalid database type");
 
-    backend->open(readOnly);
-    auto rng = backend->hardFetchLedgerRangeNoThrow();
+    auto const rng = backend->hardFetchLedgerRangeNoThrow();
     if (rng)
     {
         backend->updateRange(rng->minSequence);
         backend->updateRange(rng->maxSequence);
     }
 
-    log.info() << "Constructed BackendInterface Successfully";
-
+    LOG(log.info()) << "Constructed BackendInterface Successfully";
     return backend;
 }
-}  // namespace Backend
+}  // namespace data

src/data/BackendInterface.cpp

@@ -17,61 +17,40 @@
 */
 //==============================================================================
 
+#include <data/BackendInterface.h>
+#include <util/log/Logger.h>
+
 #include <ripple/protocol/Indexes.h>
 #include <ripple/protocol/STLedgerEntry.h>
-#include <backend/BackendInterface.h>
-#include <log/Logger.h>
-
-using namespace clio;
 
 // local to compilation unit loggers
 namespace {
-clio::Logger gLog{"Backend"};
+util::Logger gLog{"Backend"};
 }  // namespace
 
-namespace Backend {
+namespace data {
 bool
 BackendInterface::finishWrites(std::uint32_t const ledgerSequence)
 {
+    LOG(gLog.debug()) << "Want finish writes for " << ledgerSequence;
     auto commitRes = doFinishWrites();
     if (commitRes)
     {
+        LOG(gLog.debug()) << "Successfully commited. Updating range now to " << ledgerSequence;
         updateRange(ledgerSequence);
     }
     return commitRes;
 }
 void
-BackendInterface::writeLedgerObject(
-    std::string&& key,
-    std::uint32_t const seq,
-    std::string&& blob)
+BackendInterface::writeLedgerObject(std::string&& key, std::uint32_t const seq, std::string&& blob)
 {
     assert(key.size() == sizeof(ripple::uint256));
     doWriteLedgerObject(std::move(key), seq, std::move(blob));
 }
 
-std::optional<LedgerRange>
-BackendInterface::hardFetchLedgerRangeNoThrow(
-    boost::asio::yield_context& yield) const
-{
-    gLog.trace() << "called";
-    while (true)
-    {
-        try
-        {
-            return hardFetchLedgerRange(yield);
-        }
-        catch (DatabaseTimeout& t)
-        {
-            ;
-        }
-    }
-}
-
 std::optional<LedgerRange>
 BackendInterface::hardFetchLedgerRangeNoThrow() const
 {
-    gLog.trace() << "called";
     return retryOnTimeout([&]() { return hardFetchLedgerRange(); });
 }
 
@@ -80,22 +59,22 @@ std::optional<Blob>
 BackendInterface::fetchLedgerObject(
     ripple::uint256 const& key,
     std::uint32_t const sequence,
-    boost::asio::yield_context& yield) const
+    boost::asio::yield_context yield) const
 {
     auto obj = cache_.get(key, sequence);
     if (obj)
     {
-        gLog.trace() << "Cache hit - " << ripple::strHex(key);
+        LOG(gLog.trace()) << "Cache hit - " << ripple::strHex(key);
         return *obj;
     }
     else
     {
-        gLog.trace() << "Cache miss - " << ripple::strHex(key);
+        LOG(gLog.trace()) << "Cache miss - " << ripple::strHex(key);
         auto dbObj = doFetchLedgerObject(key, sequence, yield);
         if (!dbObj)
-            gLog.trace() << "Missed cache and missed in db";
+            LOG(gLog.trace()) << "Missed cache and missed in db";
         else
-            gLog.trace() << "Missed cache but found in db";
+            LOG(gLog.trace()) << "Missed cache but found in db";
         return dbObj;
     }
 }
@@ -104,7 +83,7 @@ std::vector<Blob>
 BackendInterface::fetchLedgerObjects(
     std::vector<ripple::uint256> const& keys,
     std::uint32_t const sequence,
-    boost::asio::yield_context& yield) const
+    boost::asio::yield_context yield) const
 {
     std::vector<Blob> results;
     results.resize(keys.size());
@@ -117,8 +96,7 @@ BackendInterface::fetchLedgerObjects(
         else
             misses.push_back(keys[i]);
     }
-    gLog.trace() << "Cache hits = " << keys.size() - misses.size()
-                 << " - cache misses = " << misses.size();
+    LOG(gLog.trace()) << "Cache hits = " << keys.size() - misses.size() << " - cache misses = " << misses.size();
 
     if (misses.size())
     {
@@ -140,13 +118,13 @@ std::optional<ripple::uint256>
 BackendInterface::fetchSuccessorKey(
     ripple::uint256 key,
     std::uint32_t const ledgerSequence,
-    boost::asio::yield_context& yield) const
+    boost::asio::yield_context yield) const
 {
     auto succ = cache_.getSuccessor(key, ledgerSequence);
     if (succ)
-        gLog.trace() << "Cache hit - " << ripple::strHex(key);
+        LOG(gLog.trace()) << "Cache hit - " << ripple::strHex(key);
     else
-        gLog.trace() << "Cache miss - " << ripple::strHex(key);
+        LOG(gLog.trace()) << "Cache miss - " << ripple::strHex(key);
     return succ ? succ->key : doFetchSuccessorKey(key, ledgerSequence, yield);
 }
 
@@ -154,7 +132,7 @@ std::optional<LedgerObject>
 BackendInterface::fetchSuccessorObject(
     ripple::uint256 key,
     std::uint32_t const ledgerSequence,
-    boost::asio::yield_context& yield) const
+    boost::asio::yield_context yield) const
 {
     auto succ = fetchSuccessorKey(key, ledgerSequence, yield);
     if (succ)
@@ -173,8 +151,7 @@ BackendInterface::fetchBookOffers(
     ripple::uint256 const& book,
     std::uint32_t const ledgerSequence,
     std::uint32_t const limit,
-    std::optional<ripple::uint256> const& cursor,
-    boost::asio::yield_context& yield) const
+    boost::asio::yield_context yield) const
 {
     // TODO try to speed this up. This can take a few seconds. The goal is
     // to get it down to a few hundred milliseconds.
@@ -182,10 +159,7 @@ BackendInterface::fetchBookOffers(
     const ripple::uint256 bookEnd = ripple::getQualityNext(book);
     ripple::uint256 uTipIndex = book;
     std::vector<ripple::uint256> keys;
-    auto getMillis = [](auto diff) {
-        return std::chrono::duration_cast<std::chrono::milliseconds>(diff)
-            .count();
-    };
+    auto getMillis = [](auto diff) { return std::chrono::duration_cast<std::chrono::milliseconds>(diff).count(); };
     auto begin = std::chrono::system_clock::now();
     std::uint32_t numSucc = 0;
     std::uint32_t numPages = 0;
@@ -200,29 +174,24 @@ BackendInterface::fetchBookOffers(
         succMillis += getMillis(mid2 - mid1);
         if (!offerDir || offerDir->key >= bookEnd)
         {
-            gLog.trace() << "offerDir.has_value() " << offerDir.has_value()
-                         << " breaking";
+            LOG(gLog.trace()) << "offerDir.has_value() " << offerDir.has_value() << " breaking";
             break;
         }
         uTipIndex = offerDir->key;
         while (keys.size() < limit)
         {
             ++numPages;
-            ripple::STLedgerEntry sle{
-                ripple::SerialIter{
-                    offerDir->blob.data(), offerDir->blob.size()},
-                offerDir->key};
+            ripple::STLedgerEntry sle{ripple::SerialIter{offerDir->blob.data(), offerDir->blob.size()}, offerDir->key};
             auto indexes = sle.getFieldV256(ripple::sfIndexes);
             keys.insert(keys.end(), indexes.begin(), indexes.end());
             auto next = sle.getFieldU64(ripple::sfIndexNext);
             if (!next)
             {
-                gLog.trace() << "Next is empty. breaking";
+                LOG(gLog.trace()) << "Next is empty. breaking";
                 break;
             }
             auto nextKey = ripple::keylet::page(uTipIndex, next);
-            auto nextDir =
-                fetchLedgerObject(nextKey.key, ledgerSequence, yield);
+            auto nextDir = fetchLedgerObject(nextKey.key, ledgerSequence, yield);
             assert(nextDir);
             offerDir->blob = *nextDir;
             offerDir->key = nextKey.key;
@@ -234,38 +203,56 @@ BackendInterface::fetchBookOffers(
     auto objs = fetchLedgerObjects(keys, ledgerSequence, yield);
     for (size_t i = 0; i < keys.size() && i < limit; ++i)
     {
-        gLog.trace() << "Key = " << ripple::strHex(keys[i])
-                     << " blob = " << ripple::strHex(objs[i])
-                     << " ledgerSequence = " << ledgerSequence;
+        LOG(gLog.trace()) << "Key = " << ripple::strHex(keys[i]) << " blob = " << ripple::strHex(objs[i])
+                          << " ledgerSequence = " << ledgerSequence;
         assert(objs[i].size());
         page.offers.push_back({keys[i], objs[i]});
     }
     auto end = std::chrono::system_clock::now();
-    gLog.debug() << "Fetching " << std::to_string(keys.size())
-                 << " offers took " << std::to_string(getMillis(mid - begin))
-                 << " milliseconds. Fetching next dir took "
-                 << std::to_string(succMillis)
-                 << " milliseonds. Fetched next dir " << std::to_string(numSucc)
-                 << " times"
-                 << " Fetching next page of dir took "
-                 << std::to_string(pageMillis) << " milliseconds"
-                 << ". num pages = " << std::to_string(numPages)
-                 << ". Fetching all objects took "
-                 << std::to_string(getMillis(end - mid))
-                 << " milliseconds. total time = "
-                 << std::to_string(getMillis(end - begin)) << " milliseconds"
-                 << " book = " << ripple::strHex(book);
+    LOG(gLog.debug()) << "Fetching " << std::to_string(keys.size()) << " offers took "
+                      << std::to_string(getMillis(mid - begin)) << " milliseconds. Fetching next dir took "
+                      << std::to_string(succMillis) << " milliseonds. Fetched next dir " << std::to_string(numSucc)
+                      << " times"
+                      << " Fetching next page of dir took " << std::to_string(pageMillis) << " milliseconds"
+                      << ". num pages = " << std::to_string(numPages) << ". Fetching all objects took "
+                      << std::to_string(getMillis(end - mid))
+                      << " milliseconds. total time = " << std::to_string(getMillis(end - begin)) << " milliseconds"
+                      << " book = " << ripple::strHex(book);
 
     return page;
 }
 
+std::optional<LedgerRange>
+BackendInterface::hardFetchLedgerRange() const
+{
+    return synchronous([this](auto yield) { return hardFetchLedgerRange(yield); });
+}
+
+std::optional<LedgerRange>
+BackendInterface::fetchLedgerRange() const
+{
+    std::shared_lock lck(rngMtx_);
+    return range;
+}
+
+void
+BackendInterface::updateRange(uint32_t newMax)
+{
+    std::scoped_lock lck(rngMtx_);
+    assert(!range || newMax >= range->maxSequence);
+    if (!range)
+        range = {newMax, newMax};
+    else
+        range->maxSequence = newMax;
+}
+
 LedgerPage
 BackendInterface::fetchLedgerPage(
     std::optional<ripple::uint256> const& cursor,
     std::uint32_t const ledgerSequence,
     std::uint32_t const limit,
     bool outOfOrder,
-    boost::asio::yield_context& yield) const
+    boost::asio::yield_context yield) const
 {
     LedgerPage page;
 
@@ -273,11 +260,8 @@ BackendInterface::fetchLedgerPage(
     bool reachedEnd = false;
     while (keys.size() < limit && !reachedEnd)
     {
-        ripple::uint256 const& curCursor = keys.size() ? keys.back()
-            : cursor                                   ? *cursor
-                                                       : firstKey;
-        std::uint32_t const seq =
-            outOfOrder ? range->maxSequence : ledgerSequence;
+        ripple::uint256 const& curCursor = keys.size() ? keys.back() : cursor ? *cursor : firstKey;
+        std::uint32_t const seq = outOfOrder ? range->maxSequence : ledgerSequence;
         auto succ = fetchSuccessorKey(curCursor, seq, yield);
         if (!succ)
             reachedEnd = true;
@@ -292,15 +276,14 @@ BackendInterface::fetchLedgerPage(
             page.objects.push_back({std::move(keys[i]), std::move(objects[i])});
         else if (!outOfOrder)
         {
-            gLog.error()
-                << "Deleted or non-existent object in successor table. key = "
-                << ripple::strHex(keys[i]) << " - seq = " << ledgerSequence;
+            LOG(gLog.error()) << "Deleted or non-existent object in successor table. key = " << ripple::strHex(keys[i])
+                              << " - seq = " << ledgerSequence;
             std::stringstream msg;
             for (size_t j = 0; j < objects.size(); ++j)
             {
                 msg << " - " << ripple::strHex(keys[j]);
             }
-            gLog.error() << msg.str();
+            LOG(gLog.error()) << msg.str();
         }
     }
     if (keys.size() && !reachedEnd)
@@ -310,9 +293,7 @@ BackendInterface::fetchLedgerPage(
 }
 
 std::optional<ripple::Fees>
-BackendInterface::fetchFees(
-    std::uint32_t const seq,
-    boost::asio::yield_context& yield) const
+BackendInterface::fetchFees(std::uint32_t const seq, boost::asio::yield_context yield) const
 {
     ripple::Fees fees;
 
@@ -321,7 +302,7 @@ BackendInterface::fetchFees(
 
     if (!bytes)
     {
-        gLog.error() << "Could not find fees";
+        LOG(gLog.error()) << "Could not find fees";
         return {};
     }
 
@@ -331,9 +312,6 @@ BackendInterface::fetchFees(
     if (sle.getFieldIndex(ripple::sfBaseFee) != -1)
         fees.base = sle.getFieldU64(ripple::sfBaseFee);
 
-    if (sle.getFieldIndex(ripple::sfReferenceFeeUnits) != -1)
-        fees.units = sle.getFieldU32(ripple::sfReferenceFeeUnits);
-
     if (sle.getFieldIndex(ripple::sfReserveBase) != -1)
         fees.reserve = sle.getFieldU32(ripple::sfReserveBase);
 
@@ -343,4 +321,4 @@ BackendInterface::fetchFees(
     return fees;
 }
 
-}  // namespace Backend
+}  // namespace data

src/data/BackendInterface.h

@@ -0,0 +1,568 @@
+//------------------------------------------------------------------------------
+/*
+    This file is part of clio: https://github.com/XRPLF/clio
+    Copyright (c) 2022, the clio developers.
+
+    Permission to use, copy, modify, and distribute this software for any
+    purpose with or without fee is hereby granted, provided that the above
+    copyright notice and this permission notice appear in all copies.
+
+    THE  SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL WARRANTIES
+    WITH  REGARD  TO  THIS  SOFTWARE  INCLUDING  ALL  IMPLIED  WARRANTIES  OF
+    MERCHANTABILITY  AND  FITNESS. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR
+    ANY  SPECIAL,  DIRECT,  INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES
+    WHATSOEVER  RESULTING  FROM  LOSS  OF USE, DATA OR PROFITS, WHETHER IN AN
+    ACTION  OF  CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF
+    OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE.
+*/
+//==============================================================================
+
+#pragma once
+
+#include <data/DBHelpers.h>
+#include <data/LedgerCache.h>
+#include <data/Types.h>
+#include <util/config/Config.h>
+#include <util/log/Logger.h>
+
+#include <ripple/protocol/Fees.h>
+#include <ripple/protocol/LedgerHeader.h>
+#include <boost/asio/spawn.hpp>
+#include <boost/json.hpp>
+
+#include <thread>
+#include <type_traits>
+
+namespace data {
+
+/**
+ * @brief Represents a database timeout error.
+ */
+class DatabaseTimeout : public std::exception
+{
+public:
+    const char*
+    what() const throw() override
+    {
+        return "Database read timed out. Please retry the request";
+    }
+};
+
+/**
+ * @brief A helper function that catches DatabaseTimout exceptions and retries indefinitely.
+ *
+ * @tparam FnType The type of function object to execute
+ * @param func The function object to execute
+ * @param waitMs Delay between retry attempts
+ * @return auto The same as the return type of func
+ */
+template <class FnType>
+auto
+retryOnTimeout(FnType func, size_t waitMs = 500)
+{
+    static util::Logger log{"Backend"};
+
+    while (true)
+    {
+        try
+        {
+            return func();
+        }
+        catch (DatabaseTimeout const&)
+        {
+            LOG(log.error()) << "Database request timed out. Sleeping and retrying ... ";
+            std::this_thread::sleep_for(std::chrono::milliseconds(waitMs));
+        }
+    }
+}
+
+/**
+ * @brief Synchronously executes the given function object inside a coroutine.
+ *
+ * @tparam FnType The type of function object to execute
+ * @param func The function object to execute
+ * @return auto The same as the return type of func
+ */
+template <class FnType>
+auto
+synchronous(FnType&& func)
+{
+    boost::asio::io_context ctx;
+
+    using R = typename boost::result_of<FnType(boost::asio::yield_context)>::type;
+    if constexpr (!std::is_same<R, void>::value)
+    {
+        R res;
+        boost::asio::spawn(
+            ctx, [_ = boost::asio::make_work_guard(ctx), &func, &res](auto yield) { res = func(yield); });
+
+        ctx.run();
+        return res;
+    }
+    else
+    {
+        boost::asio::spawn(ctx, [_ = boost::asio::make_work_guard(ctx), &func](auto yield) { func(yield); });
+        ctx.run();
+    }
+}
+
+/**
+ * @brief Synchronously execute the given function object and retry until no DatabaseTimeout is thrown.
+ *
+ * @tparam FnType The type of function object to execute
+ * @param func The function object to execute
+ * @return auto The same as the return type of func
+ */
+template <class FnType>
+auto
+synchronousAndRetryOnTimeout(FnType&& func)
+{
+    return retryOnTimeout([&]() { return synchronous(func); });
+}
+
+/**
+ * @brief The interface to the database used by Clio.
+ */
+class BackendInterface
+{
+protected:
+    mutable std::shared_mutex rngMtx_;
+    std::optional<LedgerRange> range;
+    LedgerCache cache_;
+
+public:
+    BackendInterface() = default;
+    virtual ~BackendInterface() = default;
+
+    // TODO: Remove this hack. Cache should not be exposed thru BackendInterface
+    /**
+     * @return Immutable cache
+     */
+    LedgerCache const&
+    cache() const
+    {
+        return cache_;
+    }
+
+    /**
+     * @return Mutable cache
+     */
+    LedgerCache&
+    cache()
+    {
+        return cache_;
+    }
+
+    /**
+     * @brief Fetches a specific ledger by sequence number.
+     *
+     * @param sequence The sequence number to fetch for
+     * @param yield The coroutine context
+     * @return The ripple::LedgerHeader if found; nullopt otherwise
+     */
+    virtual std::optional<ripple::LedgerHeader>
+    fetchLedgerBySequence(std::uint32_t const sequence, boost::asio::yield_context yield) const = 0;
+
+    /**
+     * @brief Fetches a specific ledger by hash.
+     *
+     * @param hash The hash to fetch for
+     * @param yield The coroutine context
+     * @return The ripple::LedgerHeader if found; nullopt otherwise
+     */
+    virtual std::optional<ripple::LedgerHeader>
+    fetchLedgerByHash(ripple::uint256 const& hash, boost::asio::yield_context yield) const = 0;
+
+    /**
+     * @brief Fetches the latest ledger sequence.
+     *
+     * @param yield The coroutine context
+     * @return Latest sequence wrapped in an optional if found; nullopt otherwise
+     */
+    virtual std::optional<std::uint32_t>
+    fetchLatestLedgerSequence(boost::asio::yield_context yield) const = 0;
+
+    /**
+     * @brief Fetch the current ledger range.
+     *
+     * @return The current ledger range if populated; nullopt otherwise
+     */
+    std::optional<LedgerRange>
+    fetchLedgerRange() const;
+
+    /**
+     * @brief Updates the range of sequences that are stored in the DB.
+     *
+     * @param newMax The new maximum sequence available
+     */
+    void
+    updateRange(uint32_t newMax);
+
+    /**
+     * @brief Fetch the fees from a specific ledger sequence.
+     *
+     * @param seq The sequence to fetch for
+     * @param yield The coroutine context
+     * @return ripple::Fees if fees are found; nullopt otherwise
+     */
+    std::optional<ripple::Fees>
+    fetchFees(std::uint32_t const seq, boost::asio::yield_context yield) const;
+
+    /**
+     * @brief Fetches a specific transaction.
+     *
+     * @param hash The hash of the transaction to fetch
+     * @param yield The coroutine context
+     * @return TransactionAndMetadata if transaction is found; nullopt otherwise
+     */
+    virtual std::optional<TransactionAndMetadata>
+    fetchTransaction(ripple::uint256 const& hash, boost::asio::yield_context yield) const = 0;
+
+    /**
+     * @brief Fetches multiple transactions.
+     *
+     * @param hashes A vector of hashes to fetch transactions for
+     * @param yield The coroutine context
+     * @return A vector of TransactionAndMetadata matching the given hashes
+     */
+    virtual std::vector<TransactionAndMetadata>
+    fetchTransactions(std::vector<ripple::uint256> const& hashes, boost::asio::yield_context yield) const = 0;
+
+    /**
+     * @brief Fetches all transactions for a specific account.
+     *
+     * @param account The account to fetch transactions for
+     * @param limit The maximum number of transactions per result page
+     * @param forward Whether to fetch the page forwards or backwards from the given cursor
+     * @param cursor The cursor to resume fetching from
+     * @param yield The coroutine context
+     * @return Results and a cursor to resume from
+     */
+    virtual TransactionsAndCursor
+    fetchAccountTransactions(
+        ripple::AccountID const& account,
+        std::uint32_t const limit,
+        bool forward,
+        std::optional<TransactionsCursor> const& cursor,
+        boost::asio::yield_context yield) const = 0;
+
+    /**
+     * @brief Fetches all transactions from a specific ledger.
+     *
+     * @param ledgerSequence The ledger sequence to fetch for
+     * @param yield The coroutine context
+     * @return Results as a vector of TransactionAndMetadata
+     */
+    virtual std::vector<TransactionAndMetadata>
+    fetchAllTransactionsInLedger(std::uint32_t const ledgerSequence, boost::asio::yield_context yield) const = 0;
+
+    /**
+     * @brief Fetches all transaction hashes from a specific ledger.
+     *
+     * @param ledgerSequence The ledger sequence to fetch for
+     * @param yield The coroutine context
+     * @return Hashes as ripple::uint256 in a vector
+     */
+    virtual std::vector<ripple::uint256>
+    fetchAllTransactionHashesInLedger(std::uint32_t const ledgerSequence, boost::asio::yield_context yield) const = 0;
+
+    /**
+     * @brief Fetches a specific NFT.
+     *
+     * @param tokenID The ID of the NFT
+     * @param ledgerSequence The ledger sequence to fetch for
+     * @param yield The coroutine context
+     * @return NFT object on success; nullopt otherwise
+     */
+    virtual std::optional<NFT>
+    fetchNFT(ripple::uint256 const& tokenID, std::uint32_t const ledgerSequence, boost::asio::yield_context yield)
+        const = 0;
+
+    /**
+     * @brief Fetches all transactions for a specific NFT.
+     *
+     * @param tokenID The ID of the NFT
+     * @param limit The maximum number of transactions per result page
+     * @param forward Whether to fetch the page forwards or backwards from the given cursor
+     * @param cursorIn The cursor to resume fetching from
+     * @param yield The coroutine context
+     * @return Results and a cursor to resume from
+     */
+    virtual TransactionsAndCursor
+    fetchNFTTransactions(
+        ripple::uint256 const& tokenID,
+        std::uint32_t const limit,
+        bool const forward,
+        std::optional<TransactionsCursor> const& cursorIn,
+        boost::asio::yield_context yield) const = 0;
+
+    /**
+     * @brief Fetches a specific ledger object.
+     *
+     * Currently the real fetch happens in doFetchLedgerObject and fetchLedgerObject attempts to fetch from Cache first
+     * and only calls out to the real DB if a cache miss ocurred.
+     *
+     * @param key The key of the object
+     * @param sequence The ledger sequence to fetch for
+     * @param yield The coroutine context
+     * @return The object as a Blob on success; nullopt otherwise
+     */
+    std::optional<Blob>
+    fetchLedgerObject(ripple::uint256 const& key, std::uint32_t const sequence, boost::asio::yield_context yield) const;
+
+    /**
+     * @brief Fetches all ledger objects by their keys.
+     *
+     * Currently the real fetch happens in doFetchLedgerObjects and fetchLedgerObjects attempts to fetch from Cache
+     * first and only calls out to the real DB for each of the keys that was not found in the cache.
+     *
+     * @param keys A vector with the keys of the objects to fetch
+     * @param sequence The ledger sequence to fetch for
+     * @param yield The coroutine context
+     * @return A vector of ledger objects as Blobs
+     */
+    std::vector<Blob>
+    fetchLedgerObjects(
+        std::vector<ripple::uint256> const& keys,
+        std::uint32_t const sequence,
+        boost::asio::yield_context yield) const;
+
+    /**
+     * @brief The database-specific implementation for fetching a ledger object.
+     *
+     * @param key The key to fetch for
+     * @param sequence The ledger sequence to fetch for
+     * @param yield The coroutine context
+     * @return The object as a Blob on success; nullopt otherwise
+     */
+    virtual std::optional<Blob>
+    doFetchLedgerObject(ripple::uint256 const& key, std::uint32_t const sequence, boost::asio::yield_context yield)
+        const = 0;
+
+    /**
+     * @brief The database-specific implementation for fetching ledger objects.
+     *
+     * @param keys The keys to fetch for
+     * @param sequence The ledger sequence to fetch for
+     * @param yield The coroutine context
+     * @return A vector of Blobs representing each fetched object
+     */
+    virtual std::vector<Blob>
+    doFetchLedgerObjects(
+        std::vector<ripple::uint256> const& keys,
+        std::uint32_t const sequence,
+        boost::asio::yield_context yield) const = 0;
+
+    /**
+     * @brief Returns the difference between ledgers.
+     *
+     * @param ledgerSequence The ledger sequence to fetch for
+     * @param yield The coroutine context
+     * @return A vector of LedgerObject representing the diff
+     */
+    virtual std::vector<LedgerObject>
+    fetchLedgerDiff(std::uint32_t const ledgerSequence, boost::asio::yield_context yield) const = 0;
+
+    /**
+     * @brief Fetches a page of ledger objects, ordered by key/index.
+     *
+     * @param cursor The cursor to resume fetching from
+     * @param ledgerSequence The ledger sequence to fetch for
+     * @param limit The maximum number of transactions per result page
+     * @param outOfOrder If set to true max available sequence is used instead of ledgerSequence
+     * @param yield The coroutine context
+     * @return The ledger page
+     */
+    LedgerPage
+    fetchLedgerPage(
+        std::optional<ripple::uint256> const& cursor,
+        std::uint32_t const ledgerSequence,
+        std::uint32_t const limit,
+        bool outOfOrder,
+        boost::asio::yield_context yield) const;
+
+    /**
+     * @brief Fetches the successor object.
+     *
+     * @param key The key to fetch for
+     * @param ledgerSequence The ledger sequence to fetch for
+     * @param yield The coroutine context
+     * @return The sucessor on success; nullopt otherwise
+     */
+    std::optional<LedgerObject>
+    fetchSuccessorObject(ripple::uint256 key, std::uint32_t const ledgerSequence, boost::asio::yield_context yield)
+        const;
+
+    /**
+     * @brief Fetches the successor key.
+     *
+     * Thea real fetch happens in doFetchSuccessorKey. This function will attempt to lookup the successor in the cache
+     * first and only if it's not found in the cache will it fetch from the actual DB.
+     *
+     * @param key The key to fetch for
+     * @param ledgerSequence The ledger sequence to fetch for
+     * @param yield The coroutine context
+     * @return The sucessor key on success; nullopt otherwise
+     */
+    std::optional<ripple::uint256>
+    fetchSuccessorKey(ripple::uint256 key, std::uint32_t const ledgerSequence, boost::asio::yield_context yield) const;
+
+    /**
+     * @brief Database-specific implementation of fetching the successor key
+     *
+     * @param key The key to fetch for
+     * @param ledgerSequence The ledger sequence to fetch for
+     * @param yield The coroutine context
+     * @return The sucessor on success; nullopt otherwise
+     */
+    virtual std::optional<ripple::uint256>
+    doFetchSuccessorKey(ripple::uint256 key, std::uint32_t const ledgerSequence, boost::asio::yield_context yield)
+        const = 0;
+
+    /**
+     * @brief Fetches book offers.
+     *
+     * @param book Unsigned 256-bit integer.
+     * @param ledgerSequence The ledger sequence to fetch for
+     * @param limit Pagaing limit as to how many transactions returned per page.
+     * @param yield The coroutine context
+     * @return The book offers page
+     */
+    BookOffersPage
+    fetchBookOffers(
+        ripple::uint256 const& book,
+        std::uint32_t const ledgerSequence,
+        std::uint32_t const limit,
+        boost::asio::yield_context yield) const;
+
+    /**
+     * @brief Synchronously fetches the ledger range from DB.
+     *
+     * This function just wraps hardFetchLedgerRange(boost::asio::yield_context) using synchronous(FnType&&).
+     *
+     * @return The ledger range if available; nullopt otherwise
+     */
+    std::optional<LedgerRange>
+    hardFetchLedgerRange() const;
+
+    /**
+     * @brief Fetches the ledger range from DB.
+     *
+     * @return The ledger range if available; nullopt otherwise
+     */
+    virtual std::optional<LedgerRange>
+    hardFetchLedgerRange(boost::asio::yield_context yield) const = 0;
+
+    /**
+     * @brief Fetches the ledger range from DB retrying until no DatabaseTimeout is thrown.
+     *
+     * @return The ledger range if available; nullopt otherwise
+     */
+    std::optional<LedgerRange>
+    hardFetchLedgerRangeNoThrow() const;
+
+    /**
+     * @brief Writes to a specific ledger.
+     *
+     * @param ledgerHeader Ledger header.
+     * @param blob r-value string serialization of ledger header.
+     */
+    virtual void
+    writeLedger(ripple::LedgerHeader const& ledgerHeader, std::string&& blob) = 0;
+
+    /**
+     * @brief Writes a new ledger object.
+     *
+     * @param key The key to write the ledger object under
+     * @param seq The ledger sequence to write for
+     * @param blob The data to write
+     */
+    virtual void
+    writeLedgerObject(std::string&& key, std::uint32_t const seq, std::string&& blob);
+
+    /**
+     * @brief Writes a new transaction.
+     *
+     * @param hash The hash of the transaction
+     * @param seq The ledger sequence to write for
+     * @param date The timestamp of the entry
+     * @param transaction The transaction data to write
+     * @param metadata The metadata to write
+     */
+    virtual void
+    writeTransaction(
+        std::string&& hash,
+        std::uint32_t const seq,
+        std::uint32_t const date,
+        std::string&& transaction,
+        std::string&& metadata) = 0;
+
+    /**
+     * @brief Writes NFTs to the database.
+     *
+     * @param data A vector of NFTsData objects representing the NFTs
+     */
+    virtual void
+    writeNFTs(std::vector<NFTsData>&& data) = 0;
+
+    /**
+     * @brief Write a new set of account transactions.
+     *
+     * @param data A vector of AccountTransactionsData objects representing the account transactions
+     */
+    virtual void
+    writeAccountTransactions(std::vector<AccountTransactionsData>&& data) = 0;
+
+    /**
+     * @brief Write NFTs transactions.
+     *
+     * @param data A vector of NFTTransactionsData objects
+     */
+    virtual void
+    writeNFTTransactions(std::vector<NFTTransactionsData>&& data) = 0;
+
+    /**
+     * @brief Write a new successor.
+     *
+     * @param key Key of the object that the passed successor will be the successor for
+     * @param seq The ledger sequence to write for
+     * @param successor The successor data to write
+     */
+    virtual void
+    writeSuccessor(std::string&& key, std::uint32_t const seq, std::string&& successor) = 0;
+
+    /**
+     * @brief Starts a write transaction with the DB. No-op for cassandra.
+     *
+     * Note: Can potentially be deprecated and removed.
+     */
+    virtual void
+    startWrites() const = 0;
+
+    /**
+     * @brief Tells database we finished writing all data for a specific ledger.
+     *
+     * Uses doFinishWrites to synchronize with the pending writes.
+     *
+     * @param ledgerSequence The ledger sequence to finish writing for
+     * @return true on success; false otherwise
+     */
+    bool
+    finishWrites(std::uint32_t const ledgerSequence);
+
+    /**
+     * @return true if database is overwhelmed; false otherwise
+     */
+    virtual bool
+    isTooBusy() const = 0;
+
+private:
+    virtual void
+    doWriteLedgerObject(std::string&& key, std::uint32_t const seq, std::string&& blob) = 0;
+
+    virtual bool
+    doFinishWrites() = 0;
+};
+
+}  // namespace data
+using BackendInterface = data::BackendInterface;

src/data/CassandraBackend.h

@@ -0,0 +1,829 @@
+//------------------------------------------------------------------------------
+/*
+    This file is part of clio: https://github.com/XRPLF/clio
+    Copyright (c) 2023, the clio developers.
+
+    Permission to use, copy, modify, and distribute this software for any
+    purpose with or without fee is hereby granted, provided that the above
+    copyright notice and this permission notice appear in all copies.
+
+    THE  SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL WARRANTIES
+    WITH  REGARD  TO  THIS  SOFTWARE  INCLUDING  ALL  IMPLIED  WARRANTIES  OF
+    MERCHANTABILITY  AND  FITNESS. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR
+    ANY  SPECIAL,  DIRECT,  INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES
+    WHATSOEVER  RESULTING  FROM  LOSS  OF USE, DATA OR PROFITS, WHETHER IN AN
+    ACTION  OF  CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF
+    OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE.
+*/
+//==============================================================================
+
+#pragma once
+
+#include <data/BackendInterface.h>
+#include <data/cassandra/Concepts.h>
+#include <data/cassandra/Handle.h>
+#include <data/cassandra/Schema.h>
+#include <data/cassandra/SettingsProvider.h>
+#include <data/cassandra/impl/ExecutionStrategy.h>
+#include <util/LedgerUtils.h>
+#include <util/Profiler.h>
+#include <util/log/Logger.h>
+
+#include <ripple/protocol/LedgerHeader.h>
+#include <ripple/protocol/nft.h>
+#include <boost/asio/spawn.hpp>
+
+namespace data::cassandra {
+
+/**
+ * @brief Implements @ref BackendInterface for Cassandra/ScyllaDB.
+ *
+ * Note: This is a safer and more correct rewrite of the original implementation of the backend.
+ *
+ * @tparam SettingsProviderType The settings provider type to use
+ * @tparam ExecutionStrategyType The execution strategy type to use
+ */
+template <SomeSettingsProvider SettingsProviderType, SomeExecutionStrategy ExecutionStrategyType>
+class BasicCassandraBackend : public BackendInterface
+{
+    util::Logger log_{"Backend"};
+
+    SettingsProviderType settingsProvider_;
+    Schema<SettingsProviderType> schema_;
+    Handle handle_;
+
+    // have to be mutable because BackendInterface constness :(
+    mutable ExecutionStrategyType executor_;
+
+    std::atomic_uint32_t ledgerSequence_ = 0u;
+
+public:
+    /**
+     * @brief Create a new cassandra/scylla backend instance.
+     *
+     * @param settingsProvider The settings provider to use
+     * @param readOnly Whether the database should be in readonly mode
+     */
+    BasicCassandraBackend(SettingsProviderType settingsProvider, bool readOnly)
+        : settingsProvider_{std::move(settingsProvider)}
+        , schema_{settingsProvider_}
+        , handle_{settingsProvider_.getSettings()}
+        , executor_{settingsProvider_.getSettings(), handle_}
+    {
+        if (auto const res = handle_.connect(); not res)
+            throw std::runtime_error("Could not connect to Cassandra: " + res.error());
+
+        if (not readOnly)
+        {
+            if (auto const res = handle_.execute(schema_.createKeyspace); not res)
+            {
+                // on datastax, creation of keyspaces can be configured to only be done thru the admin interface.
+                // this does not mean that the keyspace does not already exist tho.
+                if (res.error().code() != CASS_ERROR_SERVER_UNAUTHORIZED)
+                    throw std::runtime_error("Could not create keyspace: " + res.error());
+            }
+
+            if (auto const res = handle_.executeEach(schema_.createSchema); not res)
+                throw std::runtime_error("Could not create schema: " + res.error());
+        }
+
+        try
+        {
+            schema_.prepareStatements(handle_);
+        }
+        catch (std::runtime_error const& ex)
+        {
+            LOG(log_.error()) << "Failed to prepare the statements: " << ex.what() << "; readOnly: " << readOnly;
+            throw;
+        }
+
+        LOG(log_.info()) << "Created (revamped) CassandraBackend";
+    }
+
+    TransactionsAndCursor
+    fetchAccountTransactions(
+        ripple::AccountID const& account,
+        std::uint32_t const limit,
+        bool forward,
+        std::optional<TransactionsCursor> const& cursorIn,
+        boost::asio::yield_context yield) const override
+    {
+        auto rng = fetchLedgerRange();
+        if (!rng)
+            return {{}, {}};
+
+        Statement statement = [this, forward, &account]() {
+            if (forward)
+                return schema_->selectAccountTxForward.bind(account);
+            else
+                return schema_->selectAccountTx.bind(account);
+        }();
+
+        auto cursor = cursorIn;
+        if (cursor)
+        {
+            statement.bindAt(1, cursor->asTuple());
+            LOG(log_.debug()) << "account = " << ripple::strHex(account) << " tuple = " << cursor->ledgerSequence
+                              << cursor->transactionIndex;
+        }
+        else
+        {
+            auto const seq = forward ? rng->minSequence : rng->maxSequence;
+            auto const placeHolder = forward ? 0u : std::numeric_limits<std::uint32_t>::max();
+
+            statement.bindAt(1, std::make_tuple(placeHolder, placeHolder));
+            LOG(log_.debug()) << "account = " << ripple::strHex(account) << " idx = " << seq
+                              << " tuple = " << placeHolder;
+        }
+
+        // FIXME: Limit is a hack to support uint32_t properly for the time
+        // being. Should be removed later and schema updated to use proper
+        // types.
+        statement.bindAt(2, Limit{limit});
+        auto const res = executor_.read(yield, statement);
+        auto const& results = res.value();
+        if (not results.hasRows())
+        {
+            LOG(log_.debug()) << "No rows returned";
+            return {};
+        }
+
+        std::vector<ripple::uint256> hashes = {};
+        auto numRows = results.numRows();
+        LOG(log_.info()) << "num_rows = " << numRows;
+
+        for (auto [hash, data] : extract<ripple::uint256, std::tuple<uint32_t, uint32_t>>(results))
+        {
+            hashes.push_back(hash);
+            if (--numRows == 0)
+            {
+                LOG(log_.debug()) << "Setting cursor";
+                cursor = data;
+
+                // forward queries by ledger/tx sequence `>=`
+                // so we have to advance the index by one
+                if (forward)
+                    ++cursor->transactionIndex;
+            }
+        }
+
+        auto const txns = fetchTransactions(hashes, yield);
+        LOG(log_.debug()) << "Txns = " << txns.size();
+
+        if (txns.size() == limit)
+        {
+            LOG(log_.debug()) << "Returning cursor";
+            return {txns, cursor};
+        }
+
+        return {txns, {}};
+    }
+
+    bool
+    doFinishWrites() override
+    {
+        // wait for other threads to finish their writes
+        executor_.sync();
+
+        if (!range)
+        {
+            executor_.writeSync(schema_->updateLedgerRange, ledgerSequence_, false, ledgerSequence_);
+        }
+
+        if (not executeSyncUpdate(schema_->updateLedgerRange.bind(ledgerSequence_, true, ledgerSequence_ - 1)))
+        {
+            LOG(log_.warn()) << "Update failed for ledger " << ledgerSequence_;
+            return false;
+        }
+
+        LOG(log_.info()) << "Committed ledger " << ledgerSequence_;
+        return true;
+    }
+
+    void
+    writeLedger(ripple::LedgerHeader const& ledgerInfo, std::string&& blob) override
+    {
+        executor_.write(schema_->insertLedgerHeader, ledgerInfo.seq, std::move(blob));
+
+        executor_.write(schema_->insertLedgerHash, ledgerInfo.hash, ledgerInfo.seq);
+
+        ledgerSequence_ = ledgerInfo.seq;
+    }
+
+    std::optional<std::uint32_t>
+    fetchLatestLedgerSequence(boost::asio::yield_context yield) const override
+    {
+        if (auto const res = executor_.read(yield, schema_->selectLatestLedger); res)
+        {
+            if (auto const& result = res.value(); result)
+            {
+                if (auto const maybeValue = result.template get<uint32_t>(); maybeValue)
+                    return maybeValue;
+
+                LOG(log_.error()) << "Could not fetch latest ledger - no rows";
+                return std::nullopt;
+            }
+
+            LOG(log_.error()) << "Could not fetch latest ledger - no result";
+        }
+        else
+        {
+            LOG(log_.error()) << "Could not fetch latest ledger: " << res.error();
+        }
+
+        return std::nullopt;
+    }
+
+    std::optional<ripple::LedgerHeader>
+    fetchLedgerBySequence(std::uint32_t const sequence, boost::asio::yield_context yield) const override
+    {
+        auto const res = executor_.read(yield, schema_->selectLedgerBySeq, sequence);
+        if (res)
+        {
+            if (auto const& result = res.value(); result)
+            {
+                if (auto const maybeValue = result.template get<std::vector<unsigned char>>(); maybeValue)
+                {
+                    return util::deserializeHeader(ripple::makeSlice(*maybeValue));
+                }
+
+                LOG(log_.error()) << "Could not fetch ledger by sequence - no rows";
+                return std::nullopt;
+            }
+
+            LOG(log_.error()) << "Could not fetch ledger by sequence - no result";
+        }
+        else
+        {
+            LOG(log_.error()) << "Could not fetch ledger by sequence: " << res.error();
+        }
+
+        return std::nullopt;
+    }
+
+    std::optional<ripple::LedgerHeader>
+    fetchLedgerByHash(ripple::uint256 const& hash, boost::asio::yield_context yield) const override
+    {
+        if (auto const res = executor_.read(yield, schema_->selectLedgerByHash, hash); res)
+        {
+            if (auto const& result = res.value(); result)
+            {
+                if (auto const maybeValue = result.template get<uint32_t>(); maybeValue)
+                    return fetchLedgerBySequence(*maybeValue, yield);
+
+                LOG(log_.error()) << "Could not fetch ledger by hash - no rows";
+                return std::nullopt;
+            }
+
+            LOG(log_.error()) << "Could not fetch ledger by hash - no result";
+        }
+        else
+        {
+            LOG(log_.error()) << "Could not fetch ledger by hash: " << res.error();
+        }
+
+        return std::nullopt;
+    }
+
+    std::optional<LedgerRange>
+    hardFetchLedgerRange(boost::asio::yield_context yield) const override
+    {
+        if (auto const res = executor_.read(yield, schema_->selectLedgerRange); res)
+        {
+            auto const& results = res.value();
+            if (not results.hasRows())
+            {
+                LOG(log_.debug()) << "Could not fetch ledger range - no rows";
+                return std::nullopt;
+            }
+
+            // TODO: this is probably a good place to use user type in
+            // cassandra instead of having two rows with bool flag. or maybe at
+            // least use tuple<int, int>?
+            LedgerRange range;
+            std::size_t idx = 0;
+            for (auto [seq] : extract<uint32_t>(results))
+            {
+                if (idx == 0)
+                    range.maxSequence = range.minSequence = seq;
+                else if (idx == 1)
+                    range.maxSequence = seq;
+
+                ++idx;
+            }
+
+            if (range.minSequence > range.maxSequence)
+                std::swap(range.minSequence, range.maxSequence);
+
+            LOG(log_.debug()) << "After hardFetchLedgerRange range is " << range.minSequence << ":"
+                              << range.maxSequence;
+            return range;
+        }
+        else
+        {
+            LOG(log_.error()) << "Could not fetch ledger range: " << res.error();
+        }
+
+        return std::nullopt;
+    }
+
+    std::vector<TransactionAndMetadata>
+    fetchAllTransactionsInLedger(std::uint32_t const ledgerSequence, boost::asio::yield_context yield) const override
+    {
+        auto hashes = fetchAllTransactionHashesInLedger(ledgerSequence, yield);
+        return fetchTransactions(hashes, yield);
+    }
+
+    std::vector<ripple::uint256>
+    fetchAllTransactionHashesInLedger(std::uint32_t const ledgerSequence, boost::asio::yield_context yield)
+        const override
+    {
+        auto start = std::chrono::system_clock::now();
+        auto const res = executor_.read(yield, schema_->selectAllTransactionHashesInLedger, ledgerSequence);
+
+        if (not res)
+        {
+            LOG(log_.error()) << "Could not fetch all transaction hashes: " << res.error();
+            return {};
+        }
+
+        auto const& result = res.value();
+        if (not result.hasRows())
+        {
+            LOG(log_.error()) << "Could not fetch all transaction hashes - no rows; ledger = "
+                              << std::to_string(ledgerSequence);
+            return {};
+        }
+
+        std::vector<ripple::uint256> hashes;
+        for (auto [hash] : extract<ripple::uint256>(result))
+            hashes.push_back(std::move(hash));
+
+        auto end = std::chrono::system_clock::now();
+        LOG(log_.debug()) << "Fetched " << hashes.size() << " transaction hashes from Cassandra in "
+                          << std::chrono::duration_cast<std::chrono::milliseconds>(end - start).count()
+                          << " milliseconds";
+
+        return hashes;
+    }
+
+    std::optional<NFT>
+    fetchNFT(ripple::uint256 const& tokenID, std::uint32_t const ledgerSequence, boost::asio::yield_context yield)
+        const override
+    {
+        auto const res = executor_.read(yield, schema_->selectNFT, tokenID, ledgerSequence);
+        if (not res)
+            return std::nullopt;
+
+        if (auto const maybeRow = res->template get<uint32_t, ripple::AccountID, bool>(); maybeRow)
+        {
+            auto [seq, owner, isBurned] = *maybeRow;
+            auto result = std::make_optional<NFT>(tokenID, seq, owner, isBurned);
+
+            // now fetch URI. Usually we will have the URI even for burned NFTs,
+            // but if the first ledger on this clio included NFTokenBurn
+            // transactions we will not have the URIs for any of those tokens.
+            // In any other case not having the URI indicates something went
+            // wrong with our data.
+            //
+            // TODO - in the future would be great for any handlers that use
+            // this could inject a warning in this case (the case of not having
+            // a URI because it was burned in the first ledger) to indicate that
+            // even though we are returning a blank URI, the NFT might have had
+            // one.
+            auto uriRes = executor_.read(yield, schema_->selectNFTURI, tokenID, ledgerSequence);
+            if (uriRes)
+            {
+                if (auto const maybeUri = uriRes->template get<ripple::Blob>(); maybeUri)
+                    result->uri = *maybeUri;
+            }
+
+            return result;
+        }
+
+        LOG(log_.error()) << "Could not fetch NFT - no rows";
+        return std::nullopt;
+    }
+
+    TransactionsAndCursor
+    fetchNFTTransactions(
+        ripple::uint256 const& tokenID,
+        std::uint32_t const limit,
+        bool const forward,
+        std::optional<TransactionsCursor> const& cursorIn,
+        boost::asio::yield_context yield) const override
+    {
+        auto rng = fetchLedgerRange();
+        if (!rng)
+            return {{}, {}};
+
+        Statement statement = [this, forward, &tokenID]() {
+            if (forward)
+                return schema_->selectNFTTxForward.bind(tokenID);
+            else
+                return schema_->selectNFTTx.bind(tokenID);
+        }();
+
+        auto cursor = cursorIn;
+        if (cursor)
+        {
+            statement.bindAt(1, cursor->asTuple());
+            LOG(log_.debug()) << "token_id = " << ripple::strHex(tokenID) << " tuple = " << cursor->ledgerSequence
+                              << cursor->transactionIndex;
+        }
+        else
+        {
+            auto const seq = forward ? rng->minSequence : rng->maxSequence;
+            auto const placeHolder = forward ? 0 : std::numeric_limits<std::uint32_t>::max();
+
+            statement.bindAt(1, std::make_tuple(placeHolder, placeHolder));
+            LOG(log_.debug()) << "token_id = " << ripple::strHex(tokenID) << " idx = " << seq
+                              << " tuple = " << placeHolder;
+        }
+
+        statement.bindAt(2, Limit{limit});
+
+        auto const res = executor_.read(yield, statement);
+        auto const& results = res.value();
+        if (not results.hasRows())
+        {
+            LOG(log_.debug()) << "No rows returned";
+            return {};
+        }
+
+        std::vector<ripple::uint256> hashes = {};
+        auto numRows = results.numRows();
+        LOG(log_.info()) << "num_rows = " << numRows;
+
+        for (auto [hash, data] : extract<ripple::uint256, std::tuple<uint32_t, uint32_t>>(results))
+        {
+            hashes.push_back(hash);
+            if (--numRows == 0)
+            {
+                LOG(log_.debug()) << "Setting cursor";
+                cursor = data;
+
+                // forward queries by ledger/tx sequence `>=`
+                // so we have to advance the index by one
+                if (forward)
+                    ++cursor->transactionIndex;
+            }
+        }
+
+        auto const txns = fetchTransactions(hashes, yield);
+        LOG(log_.debug()) << "NFT Txns = " << txns.size();
+
+        if (txns.size() == limit)
+        {
+            LOG(log_.debug()) << "Returning cursor";
+            return {txns, cursor};
+        }
+
+        return {txns, {}};
+    }
+
+    std::optional<Blob>
+    doFetchLedgerObject(ripple::uint256 const& key, std::uint32_t const sequence, boost::asio::yield_context yield)
+        const override
+    {
+        LOG(log_.debug()) << "Fetching ledger object for seq " << sequence << ", key = " << ripple::to_string(key);
+        if (auto const res = executor_.read(yield, schema_->selectObject, key, sequence); res)
+        {
+            if (auto const result = res->template get<Blob>(); result)
+            {
+                if (result->size())
+                    return *result;
+            }
+            else
+            {
+                LOG(log_.debug()) << "Could not fetch ledger object - no rows";
+            }
+        }
+        else
+        {
+            LOG(log_.error()) << "Could not fetch ledger object: " << res.error();
+        }
+
+        return std::nullopt;
+    }
+
+    std::optional<TransactionAndMetadata>
+    fetchTransaction(ripple::uint256 const& hash, boost::asio::yield_context yield) const override
+    {
+        if (auto const res = executor_.read(yield, schema_->selectTransaction, hash); res)
+        {
+            if (auto const maybeValue = res->template get<Blob, Blob, uint32_t, uint32_t>(); maybeValue)
+            {
+                auto [transaction, meta, seq, date] = *maybeValue;
+                return std::make_optional<TransactionAndMetadata>(transaction, meta, seq, date);
+            }
+            else
+            {
+                LOG(log_.debug()) << "Could not fetch transaction - no rows";
+            }
+        }
+        else
+        {
+            LOG(log_.error()) << "Could not fetch transaction: " << res.error();
+        }
+
+        return std::nullopt;
+    }
+
+    std::optional<ripple::uint256>
+    doFetchSuccessorKey(ripple::uint256 key, std::uint32_t const ledgerSequence, boost::asio::yield_context yield)
+        const override
+    {
+        if (auto const res = executor_.read(yield, schema_->selectSuccessor, key, ledgerSequence); res)
+        {
+            if (auto const result = res->template get<ripple::uint256>(); result)
+            {
+                if (*result == lastKey)
+                    return std::nullopt;
+                return *result;
+            }
+            else
+            {
+                LOG(log_.debug()) << "Could not fetch successor - no rows";
+            }
+        }
+        else
+        {
+            LOG(log_.error()) << "Could not fetch successor: " << res.error();
+        }
+
+        return std::nullopt;
+    }
+
+    std::vector<TransactionAndMetadata>
+    fetchTransactions(std::vector<ripple::uint256> const& hashes, boost::asio::yield_context yield) const override
+    {
+        if (hashes.size() == 0)
+            return {};
+
+        auto const numHashes = hashes.size();
+        std::vector<TransactionAndMetadata> results;
+        results.reserve(numHashes);
+
+        std::vector<Statement> statements;
+        statements.reserve(numHashes);
+
+        auto const timeDiff = util::timed([this, yield, &results, &hashes, &statements]() {
+            // TODO: seems like a job for "hash IN (list of hashes)" instead?
+            std::transform(
+                std::cbegin(hashes), std::cend(hashes), std::back_inserter(statements), [this](auto const& hash) {
+                    return schema_->selectTransaction.bind(hash);
+                });
+
+            auto const entries = executor_.readEach(yield, statements);
+            std::transform(
+                std::cbegin(entries),
+                std::cend(entries),
+                std::back_inserter(results),
+                [](auto const& res) -> TransactionAndMetadata {
+                    if (auto const maybeRow = res.template get<Blob, Blob, uint32_t, uint32_t>(); maybeRow)
+                        return *maybeRow;
+                    else
+                        return {};
+                });
+        });
+
+        assert(numHashes == results.size());
+        LOG(log_.debug()) << "Fetched " << numHashes << " transactions from Cassandra in " << timeDiff
+                          << " milliseconds";
+        return results;
+    }
+
+    std::vector<Blob>
+    doFetchLedgerObjects(
+        std::vector<ripple::uint256> const& keys,
+        std::uint32_t const sequence,
+        boost::asio::yield_context yield) const override
+    {
+        if (keys.size() == 0)
+            return {};
+
+        auto const numKeys = keys.size();
+        LOG(log_.trace()) << "Fetching " << numKeys << " objects";
+
+        std::vector<Blob> results;
+        results.reserve(numKeys);
+
+        std::vector<Statement> statements;
+        statements.reserve(numKeys);
+
+        // TODO: seems like a job for "key IN (list of keys)" instead?
+        std::transform(
+            std::cbegin(keys), std::cend(keys), std::back_inserter(statements), [this, &sequence](auto const& key) {
+                return schema_->selectObject.bind(key, sequence);
+            });
+
+        auto const entries = executor_.readEach(yield, statements);
+        std::transform(
+            std::cbegin(entries), std::cend(entries), std::back_inserter(results), [](auto const& res) -> Blob {
+                if (auto const maybeValue = res.template get<Blob>(); maybeValue)
+                    return *maybeValue;
+                else
+                    return {};
+            });
+
+        LOG(log_.trace()) << "Fetched " << numKeys << " objects";
+        return results;
+    }
+
+    std::vector<LedgerObject>
+    fetchLedgerDiff(std::uint32_t const ledgerSequence, boost::asio::yield_context yield) const override
+    {
+        auto const [keys, timeDiff] = util::timed([this, &ledgerSequence, yield]() -> std::vector<ripple::uint256> {
+            auto const res = executor_.read(yield, schema_->selectDiff, ledgerSequence);
+            if (not res)
+            {
+                LOG(log_.error()) << "Could not fetch ledger diff: " << res.error() << "; ledger = " << ledgerSequence;
+                return {};
+            }
+
+            auto const& results = res.value();
+            if (not results)
+            {
+                LOG(log_.error()) << "Could not fetch ledger diff - no rows; ledger = " << ledgerSequence;
+                return {};
+            }
+
+            std::vector<ripple::uint256> resultKeys;
+            for (auto [key] : extract<ripple::uint256>(results))
+                resultKeys.push_back(key);
+
+            return resultKeys;
+        });
+
+        // one of the above errors must have happened
+        if (keys.empty())
+            return {};
+
+        LOG(log_.debug()) << "Fetched " << keys.size() << " diff hashes from Cassandra in " << timeDiff
+                          << " milliseconds";
+
+        auto const objs = fetchLedgerObjects(keys, ledgerSequence, yield);
+        std::vector<LedgerObject> results;
+        results.reserve(keys.size());
+
+        std::transform(
+            std::cbegin(keys),
+            std::cend(keys),
+            std::cbegin(objs),
+            std::back_inserter(results),
+            [](auto const& key, auto const& obj) {
+                return LedgerObject{key, obj};
+            });
+
+        return results;
+    }
+
+    void
+    doWriteLedgerObject(std::string&& key, std::uint32_t const seq, std::string&& blob) override
+    {
+        LOG(log_.trace()) << " Writing ledger object " << key.size() << ":" << seq << " [" << blob.size() << " bytes]";
+
+        if (range)
+            executor_.write(schema_->insertDiff, seq, key);
+
+        executor_.write(schema_->insertObject, std::move(key), seq, std::move(blob));
+    }
+
+    void
+    writeSuccessor(std::string&& key, std::uint32_t const seq, std::string&& successor) override
+    {
+        LOG(log_.trace()) << "Writing successor. key = " << key.size() << " bytes. "
+                          << " seq = " << std::to_string(seq) << " successor = " << successor.size() << " bytes.";
+        assert(key.size() != 0);
+        assert(successor.size() != 0);
+
+        executor_.write(schema_->insertSuccessor, std::move(key), seq, std::move(successor));
+    }
+
+    void
+    writeAccountTransactions(std::vector<AccountTransactionsData>&& data) override
+    {
+        std::vector<Statement> statements;
+        statements.reserve(data.size() * 10);  // assume 10 transactions avg
+
+        for (auto& record : data)
+        {
+            std::transform(
+                std::begin(record.accounts),
+                std::end(record.accounts),
+                std::back_inserter(statements),
+                [this, &record](auto&& account) {
+                    return schema_->insertAccountTx.bind(
+                        std::move(account),
+                        std::make_tuple(record.ledgerSequence, record.transactionIndex),
+                        record.txHash);
+                });
+        }
+
+        executor_.write(std::move(statements));
+    }
+
+    void
+    writeNFTTransactions(std::vector<NFTTransactionsData>&& data) override
+    {
+        std::vector<Statement> statements;
+        statements.reserve(data.size());
+
+        std::transform(std::cbegin(data), std::cend(data), std::back_inserter(statements), [this](auto const& record) {
+            return schema_->insertNFTTx.bind(
+                record.tokenID, std::make_tuple(record.ledgerSequence, record.transactionIndex), record.txHash);
+        });
+
+        executor_.write(std::move(statements));
+    }
+
+    void
+    writeTransaction(
+        std::string&& hash,
+        std::uint32_t const seq,
+        std::uint32_t const date,
+        std::string&& transaction,
+        std::string&& metadata) override
+    {
+        LOG(log_.trace()) << "Writing txn to cassandra";
+
+        executor_.write(schema_->insertLedgerTransaction, seq, hash);
+        executor_.write(
+            schema_->insertTransaction, std::move(hash), seq, date, std::move(transaction), std::move(metadata));
+    }
+
+    void
+    writeNFTs(std::vector<NFTsData>&& data) override
+    {
+        std::vector<Statement> statements;
+        statements.reserve(data.size() * 3);
+
+        for (NFTsData const& record : data)
+        {
+            statements.push_back(
+                schema_->insertNFT.bind(record.tokenID, record.ledgerSequence, record.owner, record.isBurned));
+
+            // If `uri` is set (and it can be set to an empty uri), we know this
+            // is a net-new NFT. That is, this NFT has not been seen before by
+            // us _OR_ it is in the extreme edge case of a re-minted NFT ID with
+            // the same NFT ID as an already-burned token. In this case, we need
+            // to record the URI and link to the issuer_nf_tokens table.
+            if (record.uri)
+            {
+                statements.push_back(schema_->insertIssuerNFT.bind(
+                    ripple::nft::getIssuer(record.tokenID),
+                    static_cast<uint32_t>(ripple::nft::getTaxon(record.tokenID)),
+                    record.tokenID));
+                statements.push_back(
+                    schema_->insertNFTURI.bind(record.tokenID, record.ledgerSequence, record.uri.value()));
+            }
+        }
+
+        executor_.write(std::move(statements));
+    }
+
+    void
+    startWrites() const override
+    {
+        // Note: no-op in original implementation too.
+        // probably was used in PG to start a transaction or smth.
+    }
+
+    bool
+    isTooBusy() const override
+    {
+        return executor_.isTooBusy();
+    }
+
+private:
+    bool
+    executeSyncUpdate(Statement statement)
+    {
+        auto const res = executor_.writeSync(statement);
+        auto maybeSuccess = res->template get<bool>();
+        if (not maybeSuccess)
+        {
+            LOG(log_.error()) << "executeSyncUpdate - error getting result - no row";
+            return false;
+        }
+
+        if (not maybeSuccess.value())
+        {
+            LOG(log_.warn()) << "Update failed. Checking if DB state is what we expect";
+
+            // error may indicate that another writer wrote something.
+            // in this case let's just compare the current state of things
+            // against what we were trying to write in the first place and
+            // use that as the source of truth for the result.
+            auto rng = hardFetchLedgerRangeNoThrow();
+            return rng && rng->maxSequence == ledgerSequence_;
+        }
+
+        return true;
+    }
+};
+
+using CassandraBackend = BasicCassandraBackend<SettingsProvider, detail::DefaultExecutionStrategy<>>;
+
+}  // namespace data::cassandra

src/data/DBHelpers.h

@@ -0,0 +1,253 @@
+//------------------------------------------------------------------------------
+/*
+    This file is part of clio: https://github.com/XRPLF/clio
+    Copyright (c) 2022, the clio developers.
+
+    Permission to use, copy, modify, and distribute this software for any
+    purpose with or without fee is hereby granted, provided that the above
+    copyright notice and this permission notice appear in all copies.
+
+    THE  SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL WARRANTIES
+    WITH  REGARD  TO  THIS  SOFTWARE  INCLUDING  ALL  IMPLIED  WARRANTIES  OF
+    MERCHANTABILITY  AND  FITNESS. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR
+    ANY  SPECIAL,  DIRECT,  INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES
+    WHATSOEVER  RESULTING  FROM  LOSS  OF USE, DATA OR PROFITS, WHETHER IN AN
+    ACTION  OF  CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF
+    OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE.
+*/
+//==============================================================================
+
+/** @file */
+#pragma once
+
+#include <ripple/basics/Log.h>
+#include <ripple/basics/StringUtilities.h>
+#include <ripple/protocol/SField.h>
+#include <ripple/protocol/STAccount.h>
+#include <ripple/protocol/TxMeta.h>
+
+#include <boost/container/flat_set.hpp>
+
+#include <data/Types.h>
+
+/**
+ * @brief Struct used to keep track of what to write to account_transactions/account_tx tables.
+ */
+struct AccountTransactionsData
+{
+    boost::container::flat_set<ripple::AccountID> accounts;
+    std::uint32_t ledgerSequence;
+    std::uint32_t transactionIndex;
+    ripple::uint256 txHash;
+
+    AccountTransactionsData(ripple::TxMeta& meta, ripple::uint256 const& txHash)
+        : accounts(meta.getAffectedAccounts())
+        , ledgerSequence(meta.getLgrSeq())
+        , transactionIndex(meta.getIndex())
+        , txHash(txHash)
+    {
+    }
+
+    AccountTransactionsData() = default;
+};
+
+/**
+ * @brief Represents a link from a tx to an NFT that was targeted/modified/created by it.
+ *
+ * Gets written to nf_token_transactions table and the like.
+ */
+struct NFTTransactionsData
+{
+    ripple::uint256 tokenID;
+    std::uint32_t ledgerSequence;
+    std::uint32_t transactionIndex;
+    ripple::uint256 txHash;
+
+    NFTTransactionsData(ripple::uint256 const& tokenID, ripple::TxMeta const& meta, ripple::uint256 const& txHash)
+        : tokenID(tokenID), ledgerSequence(meta.getLgrSeq()), transactionIndex(meta.getIndex()), txHash(txHash)
+    {
+    }
+};
+
+/**
+ * @brief Represents an NFT state at a particular ledger.
+ *
+ * Gets written to nf_tokens table and the like.
+ */
+struct NFTsData
+{
+    ripple::uint256 tokenID;
+    std::uint32_t ledgerSequence;
+
+    // The transaction index is only stored because we want to store only the
+    // final state of an NFT per ledger. Since we pull this from transactions
+    // we keep track of which tx index created this so we can de-duplicate, as
+    // it is possible for one ledger to have multiple txs that change the
+    // state of the same NFT. This field is not applicable when we are loading
+    // initial NFT state via ledger objects, since we do not have to tiebreak
+    // NFT state for a given ledger in that case.
+    std::optional<std::uint32_t> transactionIndex;
+    ripple::AccountID owner;
+    // We only set the uri if this is a mint tx, or if we are
+    // loading initial state from NFTokenPage objects. In other words,
+    // uri should only be set if the etl process believes this NFT hasn't
+    // been seen before in our local database. We do this so that we don't
+    // write to the the nf_token_uris table every
+    // time the same NFT changes hands. We also can infer if there is a URI
+    // that we need to write to the issuer_nf_tokens table.
+    std::optional<ripple::Blob> uri;
+    bool isBurned = false;
+
+    // This constructor is used when parsing an NFTokenMint tx.
+    // Unfortunately because of the extreme edge case of being able to
+    // re-mint an NFT with the same ID, we must explicitly record a null
+    // URI. For this reason, we _always_ write this field as a result of
+    // this tx.
+    NFTsData(
+        ripple::uint256 const& tokenID,
+        ripple::AccountID const& owner,
+        ripple::Blob const& uri,
+        ripple::TxMeta const& meta)
+        : tokenID(tokenID), ledgerSequence(meta.getLgrSeq()), transactionIndex(meta.getIndex()), owner(owner), uri(uri)
+    {
+    }
+
+    // This constructor is used when parsing an NFTokenBurn or
+    // NFTokenAcceptOffer tx
+    NFTsData(ripple::uint256 const& tokenID, ripple::AccountID const& owner, ripple::TxMeta const& meta, bool isBurned)
+        : tokenID(tokenID)
+        , ledgerSequence(meta.getLgrSeq())
+        , transactionIndex(meta.getIndex())
+        , owner(owner)
+        , isBurned(isBurned)
+    {
+    }
+
+    // This constructor is used when parsing an NFTokenPage directly from
+    // ledger state.
+    // Unfortunately because of the extreme edge case of being able to
+    // re-mint an NFT with the same ID, we must explicitly record a null
+    // URI. For this reason, we _always_ write this field as a result of
+    // this tx.
+    NFTsData(
+        ripple::uint256 const& tokenID,
+        std::uint32_t const ledgerSequence,
+        ripple::AccountID const& owner,
+        ripple::Blob const& uri)
+        : tokenID(tokenID), ledgerSequence(ledgerSequence), owner(owner), uri(uri)
+    {
+    }
+};
+
+/**
+ * @brief Check whether the supplied object is an offer.
+ *
+ * @param object The object to check
+ * @return true if the object is an offer; false otherwise
+ */
+template <class T>
+inline bool
+isOffer(T const& object)
+{
+    short offer_bytes = (object[1] << 8) | object[2];
+    return offer_bytes == 0x006f;
+}
+
+/**
+ * @brief Check whether the supplied hex represents an offer object.
+ *
+ * @param object The object to check
+ * @return true if the object is an offer; false otherwise
+ */
+template <class T>
+inline bool
+isOfferHex(T const& object)
+{
+    auto blob = ripple::strUnHex(4, object.begin(), object.begin() + 4);
+    if (blob)
+        return isOffer(*blob);
+    return false;
+}
+
+/**
+ * @brief Check whether the supplied object is a dir node.
+ *
+ * @param object The object to check
+ * @return true if the object is a dir node; false otherwise
+ */
+template <class T>
+inline bool
+isDirNode(T const& object)
+{
+    short spaceKey = (object.data()[1] << 8) | object.data()[2];
+    return spaceKey == 0x0064;
+}
+
+/**
+ * @brief Check whether the supplied object is a book dir.
+ *
+ * @param key The key into the object
+ * @param object The object to check
+ * @return true if the object is a book dir; false otherwise
+ */
+template <class T, class R>
+inline bool
+isBookDir(T const& key, R const& object)
+{
+    if (!isDirNode(object))
+        return false;
+
+    ripple::STLedgerEntry const sle{ripple::SerialIter{object.data(), object.size()}, key};
+    return !sle[~ripple::sfOwner].has_value();
+}
+
+/**
+ * @brief Get the book out of an offer object.
+ *
+ * @param offer The offer to get the book for
+ * @return Book as ripple::uint256
+ */
+template <class T>
+inline ripple::uint256
+getBook(T const& offer)
+{
+    ripple::SerialIter it{offer.data(), offer.size()};
+    ripple::SLE sle{it, {}};
+    ripple::uint256 book = sle.getFieldH256(ripple::sfBookDirectory);
+
+    return book;
+}
+
+/**
+ * @brief Get the book base.
+ *
+ * @param key The key to get the book base out of
+ * @return Book base as ripple::uint256
+ */
+template <class T>
+inline ripple::uint256
+getBookBase(T const& key)
+{
+    assert(key.size() == ripple::uint256::size());
+
+    ripple::uint256 ret;
+    for (size_t i = 0; i < 24; ++i)
+        ret.data()[i] = key.data()[i];
+
+    return ret;
+}
+
+/**
+ * @brief Stringify a ripple::uint256.
+ *
+ * @param input The input value
+ * @return The input value as a string
+ */
+inline std::string
+uint256ToString(ripple::uint256 const& input)
+{
+    return {reinterpret_cast<const char*>(input.data()), input.size()};
+}
+
+/** @brief The ripple epoch start timestamp. Midnight on 1st January 2000. */
+static constexpr std::uint32_t rippleEpochStart = 946684800;

src/data/LedgerCache.cpp

@@ -17,21 +17,19 @@
 */
 //==============================================================================
 
-#include <backend/SimpleCache.h>
-namespace Backend {
+#include <data/LedgerCache.h>
+
+namespace data {
 
 uint32_t
-SimpleCache::latestLedgerSequence() const
+LedgerCache::latestLedgerSequence() const
 {
     std::shared_lock lck{mtx_};
     return latestSeq_;
 }
 
 void
-SimpleCache::update(
-    std::vector<LedgerObject> const& objs,
-    uint32_t seq,
-    bool isBackground)
+LedgerCache::update(std::vector<LedgerObject> const& objs, uint32_t seq, bool isBackground)
 {
     if (disabled_)
         return;
@@ -67,11 +65,11 @@ SimpleCache::update(
 }
 
 std::optional<LedgerObject>
-SimpleCache::getSuccessor(ripple::uint256 const& key, uint32_t seq) const
+LedgerCache::getSuccessor(ripple::uint256 const& key, uint32_t seq) const
 {
     if (!full_)
         return {};
-    std::shared_lock{mtx_};
+    std::shared_lock lck{mtx_};
     successorReqCounter_++;
     if (seq != latestSeq_)
         return {};
@@ -83,7 +81,7 @@ SimpleCache::getSuccessor(ripple::uint256 const& key, uint32_t seq) const
 }
 
 std::optional<LedgerObject>
-SimpleCache::getPredecessor(ripple::uint256 const& key, uint32_t seq) const
+LedgerCache::getPredecessor(ripple::uint256 const& key, uint32_t seq) const
 {
     if (!full_)
         return {};
@@ -96,12 +94,13 @@ SimpleCache::getPredecessor(ripple::uint256 const& key, uint32_t seq) const
     --e;
     return {{e->first, e->second.blob}};
 }
+
 std::optional<Blob>
-SimpleCache::get(ripple::uint256 const& key, uint32_t seq) const
+LedgerCache::get(ripple::uint256 const& key, uint32_t seq) const
 {
+    std::shared_lock lck{mtx_};
     if (seq > latestSeq_)
         return {};
-    std::shared_lock lck{mtx_};
     objectReqCounter_++;
     auto e = map_.find(key);
     if (e == map_.end())
@@ -113,13 +112,13 @@ SimpleCache::get(ripple::uint256 const& key, uint32_t seq) const
 }
 
 void
-SimpleCache::setDisabled()
+LedgerCache::setDisabled()
 {
     disabled_ = true;
 }
 
 void
-SimpleCache::setFull()
+LedgerCache::setFull()
 {
     if (disabled_)
         return;
@@ -130,28 +129,32 @@ SimpleCache::setFull()
 }
 
 bool
-SimpleCache::isFull() const
+LedgerCache::isFull() const
 {
     return full_;
 }
+
 size_t
-SimpleCache::size() const
+LedgerCache::size() const
 {
     std::shared_lock lck{mtx_};
     return map_.size();
 }
+
 float
-SimpleCache::getObjectHitRate() const
+LedgerCache::getObjectHitRate() const
 {
     if (!objectReqCounter_)
         return 1;
-    return ((float)objectHitCounter_) / objectReqCounter_;
+    return static_cast<float>(objectHitCounter_) / objectReqCounter_;
 }
+
 float
-SimpleCache::getSuccessorHitRate() const
+LedgerCache::getSuccessorHitRate() const
 {
     if (!successorReqCounter_)
         return 1;
-    return ((float)successorHitCounter_) / successorReqCounter_;
+    return static_cast<float>(successorHitCounter_) / successorReqCounter_;
 }
-}  // namespace Backend
+
+}  // namespace data

src/data/LedgerCache.h

@@ -0,0 +1,154 @@
+//------------------------------------------------------------------------------
+/*
+    This file is part of clio: https://github.com/XRPLF/clio
+    Copyright (c) 2022, the clio developers.
+
+    Permission to use, copy, modify, and distribute this software for any
+    purpose with or without fee is hereby granted, provided that the above
+    copyright notice and this permission notice appear in all copies.
+
+    THE  SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL WARRANTIES
+    WITH  REGARD  TO  THIS  SOFTWARE  INCLUDING  ALL  IMPLIED  WARRANTIES  OF
+    MERCHANTABILITY  AND  FITNESS. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR
+    ANY  SPECIAL,  DIRECT,  INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES
+    WHATSOEVER  RESULTING  FROM  LOSS  OF USE, DATA OR PROFITS, WHETHER IN AN
+    ACTION  OF  CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF
+    OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE.
+*/
+//==============================================================================
+
+#pragma once
+
+#include <ripple/basics/base_uint.h>
+#include <ripple/basics/hardened_hash.h>
+#include <data/Types.h>
+#include <map>
+#include <mutex>
+#include <shared_mutex>
+#include <utility>
+#include <vector>
+
+namespace data {
+
+/**
+ * @brief Cache for an entire ledger.
+ */
+class LedgerCache
+{
+    struct CacheEntry
+    {
+        uint32_t seq = 0;
+        Blob blob;
+    };
+
+    // counters for fetchLedgerObject(s) hit rate
+    mutable std::atomic_uint32_t objectReqCounter_ = 0;
+    mutable std::atomic_uint32_t objectHitCounter_ = 0;
+
+    // counters for fetchSuccessorKey hit rate
+    mutable std::atomic_uint32_t successorReqCounter_ = 0;
+    mutable std::atomic_uint32_t successorHitCounter_ = 0;
+
+    std::map<ripple::uint256, CacheEntry> map_;
+
+    mutable std::shared_mutex mtx_;
+    uint32_t latestSeq_ = 0;
+    std::atomic_bool full_ = false;
+    std::atomic_bool disabled_ = false;
+
+    // temporary set to prevent background thread from writing already deleted data. not used when cache is full
+    std::unordered_set<ripple::uint256, ripple::hardened_hash<>> deletes_;
+
+public:
+    /**
+     * @brief Update the cache with new ledger objects.
+     *
+     * @param blobs The ledger objects to update cache with
+     * @param seq The sequence to update cache for
+     * @param isBackground Should be set to true when writing old data from a background thread
+     */
+    void
+    update(std::vector<LedgerObject> const& blobs, uint32_t seq, bool isBackground = false);
+
+    /**
+     * @brief Fetch a cached object by its key and sequence number.
+     *
+     * @param key The key to fetch for
+     * @param seq The sequence to fetch for
+     * @return If found in cache, will return the cached Blob; otherwise nullopt is returned
+     */
+    std::optional<Blob>
+    get(ripple::uint256 const& key, uint32_t seq) const;
+
+    /**
+     * @brief Gets a cached successor.
+     *
+     * Note: This function always returns std::nullopt when @ref isFull() returns false.
+     *
+     * @param key The key to fetch for
+     * @param seq The sequence to fetch for
+     * @return If found in cache, will return the cached successor; otherwise nullopt is returned
+     */
+    std::optional<LedgerObject>
+    getSuccessor(ripple::uint256 const& key, uint32_t seq) const;
+
+    /**
+     * @brief Gets a cached predcessor.
+     *
+     * Note: This function always returns std::nullopt when @ref isFull() returns false.
+     *
+     * @param key The key to fetch for
+     * @param seq The sequence to fetch for
+     * @return If found in cache, will return the cached predcessor; otherwise nullopt is returned
+     */
+    std::optional<LedgerObject>
+    getPredecessor(ripple::uint256 const& key, uint32_t seq) const;
+
+    /**
+     * @brief Disables the cache.
+     */
+    void
+    setDisabled();
+
+    /**
+     * @brief Sets the full flag to true.
+     *
+     * This is used when cache loaded in its entirety at startup of the application. This can be either loaded from DB,
+     * populated together with initial ledger download (on first run) or downloaded from a peer node (specified in
+     * config).
+     */
+    void
+    setFull();
+
+    /**
+     * @return The latest ledger sequence for which cache is available.
+     */
+    uint32_t
+    latestLedgerSequence() const;
+
+    /**
+     * @return true if the cache has all data for the most recent ledger; false otherwise
+     */
+    bool
+    isFull() const;
+
+    /**
+     * @return The total size of the cache.
+     */
+    size_t
+    size() const;
+
+    /**
+     * @return A number representing the success rate of hitting an object in the cache versus missing it.
+     */
+    float
+    getObjectHitRate() const;
+
+    /**
+     * @return A number representing the success rate of hitting a successor in the cache versus missing it.
+     */
+    float
+    getSuccessorHitRate() const;
+};
+
+}  // namespace data

src/data/README.md

@@ -1,4 +1,5 @@
-# Clio Backend
+# Backend
+
 ## Background
 The backend of Clio is responsible for handling the proper reading and writing of past ledger data from and to a given database. As of right now, Cassandra and ScyllaDB are the only supported databases that are production-ready. Support for database types can be easily extended by creating new implementations which implements the virtual methods of `BackendInterface.h`. Then, use the Factory Object Design Pattern to simply add logic statements to `BackendFactory.h` that return the new database interface for a specific `type` in Clio's configuration file. 
 
@@ -130,3 +131,91 @@ In each new ledger version with sequence `n`, a ledger object `v` can either be
 	 1. Being **created**, add two new records of `seq=n` with one being `e` pointing to `v`, and `v` pointing to `w` (Linked List insertion operation).
 	 2. Being **modified**, do nothing.
 	 3. Being **deleted**, add a record of `seq=n` with `e` pointing to `v`'s `next` value (Linked List deletion operation).
+
+### NFT data model
+In `rippled` NFTs are stored in NFTokenPage ledger objects. This object is
+implemented to save ledger space and has the property that it gives us O(1)
+lookup time for an NFT, assuming we know who owns the NFT at a particular
+ledger. However, if we do not know who owns the NFT at a specific ledger
+height we have no alternative in rippled other than scanning the entire
+ledger. Because of this tradeoff, clio implements a special NFT indexing data
+structure that allows clio users to query NFTs quickly, while keeping
+rippled's space-saving optimizations.
+
+#### `nf_tokens`
+```
+CREATE TABLE clio.nf_tokens (
+	token_id blob,         # The NFT's ID
+	sequence bigint,       # Sequence of ledger version
+	owner blob,            # The account ID of the owner of this NFT at this ledger
+	is_burned boolean,     # True if token was burned in this ledger
+	PRIMARY KEY (token_id, sequence)
+) WITH CLUSTERING ORDER BY (sequence DESC) ...
+```
+This table indexes NFT IDs with their owner at a given ledger. So
+```
+SELECT * FROM nf_tokens
+WHERE token_id = N AND seq <= Y
+ORDER BY seq DESC LIMIT 1;
+```
+will give you the owner of token N at ledger Y and whether it was burned. If
+the token is burned, the owner field indicates the account that owned the
+token at the time it was burned; it does not indicate the person who burned
+the token, necessarily. If you need to determine who burned the token you can
+use the `nft_history` API, which will give you the NFTokenBurn transaction
+that burned this token, along with the account that submitted that
+transaction.
+
+#### `issuer_nf_tokens_v2`
+```
+CREATE TABLE clio.issuer_nf_tokens_v2 (
+	issuer blob,       # The NFT issuer's account ID
+	taxon bigint,      # The NFT's token taxon
+	token_id blob,     # The NFT's ID
+	PRIMARY KEY (issuer, taxon, token_id)
+) WITH CLUSTERING ORDER BY (taxon ASC, token_id ASC) ...
+```
+This table indexes token IDs against their issuer and issuer/taxon
+combination. This is useful for determining all the NFTs a specific account
+issued, or all the NFTs a specific account issued with a specific taxon. It is
+not useful to know all the NFTs with a given taxon while excluding issuer, since the
+meaning of a taxon is left to an issuer.
+
+#### `nf_token_uris`
+```
+CREATE TABLE clio.nf_token_uris (
+	token_id blob,    # The NFT's ID
+	sequence bigint,  # Sequence of ledger version
+	uri blob,         # The NFT's URI
+	PRIMARY KEY (token_id, sequence)
+) WITH CLUSTERING ORDER BY (sequence DESC) ...
+```
+This table is used to store an NFT's URI. Without storing this here, we would
+need to traverse the NFT owner's entire set of NFTs to find the URI, again due
+to the way that NFTs are stored in rippled. Furthermore, instead of storing
+this in the `nf_tokens` table, we store it here to save space. A given NFT
+will have only one entry in this table (see caveat below), written to this
+table as soon as clio sees the NFTokenMint transaction, or when clio loads an
+NFTokenPage from the initial ledger it downloaded. However, the `nf_tokens`
+table is written to every time an NFT changes ownership, or if it is burned.
+
+Given this, why do we have to store the sequence? Unfortunately there is an
+extreme edge case where a given NFT ID can be burned, and then re-minted with
+a different URI. This is extremely unlikely, and might be fixed in a future
+version to rippled, but just in case we can handle that edge case by allowing
+a given NFT ID to have a new URI assigned in this case, without removing the
+prior URI.
+
+#### `nf_token_transactions`
+```
+CREATE TABLE clio.nf_token_transactions (
+	token_id blob,                  # The NFT's ID
+	seq_idx tuple<bigint, bigint>,  # Tuple of (ledger_index, transaction_index)
+	hash blob,                      # Hash of the transaction
+	PRIMARY KEY (token_id, seq_idx)
+) WITH CLUSTERING ORDER BY (seq_idx DESC) ...
+```
+This table is the NFT equivalent of `account_tx`. It's motivated by the exact
+same reasons and serves the analogous purpose here. It drives the
+`nft_history` API.
+

src/data/Types.h

@@ -0,0 +1,193 @@
+//------------------------------------------------------------------------------
+/*
+    This file is part of clio: https://github.com/XRPLF/clio
+    Copyright (c) 2022, the clio developers.
+
+    Permission to use, copy, modify, and distribute this software for any
+    purpose with or without fee is hereby granted, provided that the above
+    copyright notice and this permission notice appear in all copies.
+
+    THE  SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL WARRANTIES
+    WITH  REGARD  TO  THIS  SOFTWARE  INCLUDING  ALL  IMPLIED  WARRANTIES  OF
+    MERCHANTABILITY  AND  FITNESS. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR
+    ANY  SPECIAL,  DIRECT,  INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES
+    WHATSOEVER  RESULTING  FROM  LOSS  OF USE, DATA OR PROFITS, WHETHER IN AN
+    ACTION  OF  CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF
+    OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE.
+*/
+//==============================================================================
+
+#pragma once
+
+#include <ripple/basics/base_uint.h>
+#include <ripple/protocol/AccountID.h>
+
+#include <optional>
+#include <string>
+#include <vector>
+
+namespace data {
+
+using Blob = std::vector<unsigned char>;
+
+/**
+ * @brief Represents an object in the ledger.
+ */
+struct LedgerObject
+{
+    ripple::uint256 key;
+    Blob blob;
+
+    bool
+    operator==(const LedgerObject& other) const
+    {
+        return key == other.key && blob == other.blob;
+    }
+};
+
+/**
+ * @brief Represents a page of LedgerObjects.
+ */
+struct LedgerPage
+{
+    std::vector<LedgerObject> objects;
+    std::optional<ripple::uint256> cursor;
+};
+
+/**
+ * @brief Represents a page of book offer objects.
+ */
+struct BookOffersPage
+{
+    std::vector<LedgerObject> offers;
+    std::optional<ripple::uint256> cursor;
+};
+
+/**
+ * @brief Represents a transaction and its metadata bundled together.
+ */
+struct TransactionAndMetadata
+{
+    Blob transaction;
+    Blob metadata;
+    std::uint32_t ledgerSequence = 0;
+    std::uint32_t date = 0;
+
+    TransactionAndMetadata() = default;
+    TransactionAndMetadata(
+        Blob const& transaction,
+        Blob const& metadata,
+        std::uint32_t ledgerSequence,
+        std::uint32_t date)
+        : transaction{transaction}, metadata{metadata}, ledgerSequence{ledgerSequence}, date{date}
+    {
+    }
+
+    TransactionAndMetadata(std::tuple<Blob, Blob, std::uint32_t, std::uint32_t> data)
+        : transaction{std::get<0>(data)}
+        , metadata{std::get<1>(data)}
+        , ledgerSequence{std::get<2>(data)}
+        , date{std::get<3>(data)}
+    {
+    }
+
+    bool
+    operator==(const TransactionAndMetadata& other) const
+    {
+        return transaction == other.transaction && metadata == other.metadata &&
+            ledgerSequence == other.ledgerSequence && date == other.date;
+    }
+};
+
+/**
+ * @brief Represents a cursor into the transactions table.
+ */
+struct TransactionsCursor
+{
+    std::uint32_t ledgerSequence = 0;
+    std::uint32_t transactionIndex = 0;
+
+    TransactionsCursor() = default;
+    TransactionsCursor(std::uint32_t ledgerSequence, std::uint32_t transactionIndex)
+        : ledgerSequence{ledgerSequence}, transactionIndex{transactionIndex}
+    {
+    }
+
+    TransactionsCursor(std::tuple<std::uint32_t, std::uint32_t> data)
+        : ledgerSequence{std::get<0>(data)}, transactionIndex{std::get<1>(data)}
+    {
+    }
+
+    TransactionsCursor(TransactionsCursor const&) = default;
+
+    TransactionsCursor&
+    operator=(TransactionsCursor const&) = default;
+
+    bool
+    operator==(TransactionsCursor const& other) const = default;
+
+    [[nodiscard]] std::tuple<std::uint32_t, std::uint32_t>
+    asTuple() const
+    {
+        return std::make_tuple(ledgerSequence, transactionIndex);
+    }
+};
+
+/**
+ * @brief Represests a bundle of transactions with metadata and a cursor to the next page.
+ */
+struct TransactionsAndCursor
+{
+    std::vector<TransactionAndMetadata> txns;
+    std::optional<TransactionsCursor> cursor;
+};
+
+/**
+ * @brief Represents a NFToken.
+ */
+struct NFT
+{
+    ripple::uint256 tokenID;
+    std::uint32_t ledgerSequence;
+    ripple::AccountID owner;
+    Blob uri;
+    bool isBurned;
+
+    NFT() = default;
+    NFT(ripple::uint256 const& tokenID,
+        std::uint32_t ledgerSequence,
+        ripple::AccountID const& owner,
+        Blob const& uri,
+        bool isBurned)
+        : tokenID{tokenID}, ledgerSequence{ledgerSequence}, owner{owner}, uri{uri}, isBurned{isBurned}
+    {
+    }
+
+    NFT(ripple::uint256 const& tokenID, std::uint32_t ledgerSequence, ripple::AccountID const& owner, bool isBurned)
+        : NFT(tokenID, ledgerSequence, owner, {}, isBurned)
+    {
+    }
+
+    // clearly two tokens are the same if they have the same ID, but this struct stores the state of a given token at a
+    // given ledger sequence, so we also need to compare with ledgerSequence.
+    bool
+    operator==(NFT const& other) const
+    {
+        return tokenID == other.tokenID && ledgerSequence == other.ledgerSequence;
+    }
+};
+
+/**
+ * @brief Stores a range of sequences as a min and max pair.
+ */
+struct LedgerRange
+{
+    std::uint32_t minSequence = 0;
+    std::uint32_t maxSequence = 0;
+};
+
+constexpr ripple::uint256 firstKey{"0000000000000000000000000000000000000000000000000000000000000000"};
+constexpr ripple::uint256 lastKey{"FFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFF"};
+constexpr ripple::uint256 hi192{"0000000000000000000000000000000000000000000000001111111111111111"};
+
+}  // namespace data

src/data/cassandra/Concepts.h

@@ -0,0 +1,88 @@
+//------------------------------------------------------------------------------
+/*
+    This file is part of clio: https://github.com/XRPLF/clio
+    Copyright (c) 2023, the clio developers.
+
+    Permission to use, copy, modify, and distribute this software for any
+    purpose with or without fee is hereby granted, provided that the above
+    copyright notice and this permission notice appear in all copies.
+
+    THE  SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL WARRANTIES
+    WITH  REGARD  TO  THIS  SOFTWARE  INCLUDING  ALL  IMPLIED  WARRANTIES  OF
+    MERCHANTABILITY  AND  FITNESS. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR
+    ANY  SPECIAL,  DIRECT,  INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES
+    WHATSOEVER  RESULTING  FROM  LOSS  OF USE, DATA OR PROFITS, WHETHER IN AN
+    ACTION  OF  CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF
+    OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE.
+*/
+//==============================================================================
+
+#pragma once
+
+#include <data/cassandra/Types.h>
+
+#include <boost/asio/spawn.hpp>
+
+#include <chrono>
+#include <concepts>
+#include <optional>
+#include <string>
+
+namespace data::cassandra {
+
+/**
+ * @brief The requirements of a settings provider.
+ */
+// clang-format off
+template <typename T>
+concept SomeSettingsProvider = requires(T a) {
+    { a.getSettings() } -> std::same_as<Settings>;
+    { a.getKeyspace() } -> std::same_as<std::string>;
+    { a.getTablePrefix() } -> std::same_as<std::optional<std::string>>;
+    { a.getReplicationFactor() } -> std::same_as<uint16_t>;
+    { a.getTtl() } -> std::same_as<uint16_t>;
+};
+// clang-format on
+
+/**
+ * @brief The requirements of an execution strategy.
+ */
+// clang-format off
+template <typename T>
+concept SomeExecutionStrategy = requires(
+    T a, 
+    Settings settings, 
+    Handle handle, 
+    Statement statement, 
+    std::vector<Statement> statements,
+    PreparedStatement prepared,
+    boost::asio::yield_context token
+) {
+    { T(settings, handle) };
+    { a.sync() } -> std::same_as<void>;
+    { a.isTooBusy() } -> std::same_as<bool>;
+    { a.writeSync(statement) } -> std::same_as<ResultOrError>;
+    { a.writeSync(prepared) } -> std::same_as<ResultOrError>;
+    { a.write(prepared) } -> std::same_as<void>;
+    { a.write(std::move(statements)) } -> std::same_as<void>;
+    { a.read(token, prepared) } -> std::same_as<ResultOrError>;
+    { a.read(token, statement) } -> std::same_as<ResultOrError>;
+    { a.read(token, statements) } -> std::same_as<ResultOrError>;
+    { a.readEach(token, statements) } -> std::same_as<std::vector<Result>>;
+};
+// clang-format on
+
+/**
+ * @brief The requirements of a retry policy.
+ */
+// clang-format off
+template <typename T>
+concept SomeRetryPolicy = requires(T a, boost::asio::io_context ioc, CassandraError err, uint32_t attempt) {
+    { T(ioc) };
+    { a.shouldRetry(err) } -> std::same_as<bool>;
+    { a.retry([](){}) } -> std::same_as<void>;
+    { a.calculateDelay(attempt) } -> std::same_as<std::chrono::milliseconds>;
+};
+// clang-format on
+
+}  // namespace data::cassandra

src/data/cassandra/Error.h

@@ -0,0 +1,111 @@
+//------------------------------------------------------------------------------
+/*
+    This file is part of clio: https://github.com/XRPLF/clio
+    Copyright (c) 2023, the clio developers.
+
+    Permission to use, copy, modify, and distribute this software for any
+    purpose with or without fee is hereby granted, provided that the above
+    copyright notice and this permission notice appear in all copies.
+
+    THE  SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL WARRANTIES
+    WITH  REGARD  TO  THIS  SOFTWARE  INCLUDING  ALL  IMPLIED  WARRANTIES  OF
+    MERCHANTABILITY  AND  FITNESS. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR
+    ANY  SPECIAL,  DIRECT,  INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES
+    WHATSOEVER  RESULTING  FROM  LOSS  OF USE, DATA OR PROFITS, WHETHER IN AN
+    ACTION  OF  CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF
+    OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE.
+*/
+//==============================================================================
+
+#pragma once
+
+#include <cassandra.h>
+
+#include <string>
+
+namespace data::cassandra {
+
+/**
+ * @brief A simple container for both error message and error code.
+ */
+class CassandraError
+{
+    std::string message_;
+    uint32_t code_;
+
+public:
+    CassandraError() = default;  // default constructible required by Expected
+    CassandraError(std::string message, uint32_t code) : message_{message}, code_{code}
+    {
+    }
+
+    template <typename T>
+    friend std::string
+    operator+(T const& lhs, CassandraError const& rhs) requires std::is_convertible_v<T, std::string>
+    {
+        return lhs + rhs.message();
+    }
+
+    template <typename T>
+    friend bool
+    operator==(T const& lhs, CassandraError const& rhs) requires std::is_convertible_v<T, std::string>
+    {
+        return lhs == rhs.message();
+    }
+
+    template <std::integral T>
+    friend bool
+    operator==(T const& lhs, CassandraError const& rhs)
+    {
+        return lhs == rhs.code();
+    }
+
+    friend std::ostream&
+    operator<<(std::ostream& os, CassandraError const& err)
+    {
+        os << err.message();
+        return os;
+    }
+
+    /**
+     * @return The final error message as a std::string
+     */
+    std::string
+    message() const
+    {
+        return message_;
+    }
+
+    /**
+     * @return The error code
+     */
+    uint32_t
+    code() const
+    {
+        return code_;
+    }
+
+    /**
+     * @return true if the wrapped error is considered a timeout; false otherwise
+     */
+    bool
+    isTimeout() const
+    {
+        if (code_ == CASS_ERROR_LIB_NO_HOSTS_AVAILABLE or code_ == CASS_ERROR_LIB_REQUEST_TIMED_OUT or
+            code_ == CASS_ERROR_SERVER_UNAVAILABLE or code_ == CASS_ERROR_SERVER_OVERLOADED or
+            code_ == CASS_ERROR_SERVER_READ_TIMEOUT)
+            return true;
+        return false;
+    }
+
+    /**
+     * @return true if the wrapped error is an invalid query; false otherwise
+     */
+    bool
+    isInvalidQuery() const
+    {
+        return code_ == CASS_ERROR_SERVER_INVALID_QUERY;
+    }
+};
+
+}  // namespace data::cassandra

src/data/cassandra/Handle.cpp

@@ -0,0 +1,155 @@
+//------------------------------------------------------------------------------
+/*
+    This file is part of clio: https://github.com/XRPLF/clio
+    Copyright (c) 2023, the clio developers.
+
+    Permission to use, copy, modify, and distribute this software for any
+    purpose with or without fee is hereby granted, provided that the above
+    copyright notice and this permission notice appear in all copies.
+
+    THE  SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL WARRANTIES
+    WITH  REGARD  TO  THIS  SOFTWARE  INCLUDING  ALL  IMPLIED  WARRANTIES  OF
+    MERCHANTABILITY  AND  FITNESS. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR
+    ANY  SPECIAL,  DIRECT,  INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES
+    WHATSOEVER  RESULTING  FROM  LOSS  OF USE, DATA OR PROFITS, WHETHER IN AN
+    ACTION  OF  CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF
+    OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE.
+*/
+//==============================================================================
+
+#include <data/cassandra/Handle.h>
+
+namespace data::cassandra {
+
+Handle::Handle(Settings clusterSettings) : cluster_{clusterSettings}
+{
+}
+
+Handle::Handle(std::string_view contactPoints) : Handle{Settings::defaultSettings().withContactPoints(contactPoints)}
+{
+}
+
+Handle::~Handle()
+{
+    [[maybe_unused]] auto _ = disconnect();  // attempt to disconnect
+}
+
+Handle::FutureType
+Handle::asyncConnect() const
+{
+    return cass_session_connect(session_, cluster_);
+}
+
+Handle::MaybeErrorType
+Handle::connect() const
+{
+    return asyncConnect().await();
+}
+
+Handle::FutureType
+Handle::asyncConnect(std::string_view keyspace) const
+{
+    return cass_session_connect_keyspace(session_, cluster_, keyspace.data());
+}
+
+Handle::MaybeErrorType
+Handle::connect(std::string_view keyspace) const
+{
+    return asyncConnect(keyspace).await();
+}
+
+Handle::FutureType
+Handle::asyncDisconnect() const
+{
+    return cass_session_close(session_);
+}
+
+Handle::MaybeErrorType
+Handle::disconnect() const
+{
+    return asyncDisconnect().await();
+}
+
+Handle::FutureType
+Handle::asyncReconnect(std::string_view keyspace) const
+{
+    if (auto rc = asyncDisconnect().await(); not rc)  // sync
+        throw std::logic_error("Reconnect to keyspace '" + std::string{keyspace} + "' failed: " + rc.error());
+    return asyncConnect(keyspace);
+}
+
+Handle::MaybeErrorType
+Handle::reconnect(std::string_view keyspace) const
+{
+    return asyncReconnect(keyspace).await();
+}
+
+std::vector<Handle::FutureType>
+Handle::asyncExecuteEach(std::vector<Statement> const& statements) const
+{
+    std::vector<Handle::FutureType> futures;
+    for (auto const& statement : statements)
+        futures.push_back(cass_session_execute(session_, statement));
+    return futures;
+}
+
+Handle::MaybeErrorType
+Handle::executeEach(std::vector<Statement> const& statements) const
+{
+    for (auto futures = asyncExecuteEach(statements); auto const& future : futures)
+    {
+        if (auto const rc = future.await(); not rc)
+            return rc;
+    }
+
+    return {};
+}
+
+Handle::FutureType
+Handle::asyncExecute(Statement const& statement) const
+{
+    return cass_session_execute(session_, statement);
+}
+
+Handle::FutureWithCallbackType
+Handle::asyncExecute(Statement const& statement, std::function<void(Handle::ResultOrErrorType)>&& cb) const
+{
+    return Handle::FutureWithCallbackType{cass_session_execute(session_, statement), std::move(cb)};
+}
+
+Handle::ResultOrErrorType
+Handle::execute(Statement const& statement) const
+{
+    return asyncExecute(statement).get();
+}
+
+Handle::FutureType
+Handle::asyncExecute(std::vector<Statement> const& statements) const
+{
+    return cass_session_execute_batch(session_, Batch{statements});
+}
+
+Handle::MaybeErrorType
+Handle::execute(std::vector<Statement> const& statements) const
+{
+    return asyncExecute(statements).await();
+}
+
+Handle::FutureWithCallbackType
+Handle::asyncExecute(std::vector<Statement> const& statements, std::function<void(Handle::ResultOrErrorType)>&& cb)
+    const
+{
+    return Handle::FutureWithCallbackType{cass_session_execute_batch(session_, Batch{statements}), std::move(cb)};
+}
+
+Handle::PreparedStatementType
+Handle::prepare(std::string_view query) const
+{
+    Handle::FutureType future = cass_session_prepare(session_, query.data());
+    if (auto const rc = future.await(); rc)
+        return cass_future_get_prepared(future);
+    else
+        throw std::runtime_error(rc.error().message());
+}
+
+}  // namespace data::cassandra

src/data/cassandra/Handle.h

@@ -0,0 +1,332 @@
+//------------------------------------------------------------------------------
+/*
+    This file is part of clio: https://github.com/XRPLF/clio
+    Copyright (c) 2023, the clio developers.
+
+    Permission to use, copy, modify, and distribute this software for any
+    purpose with or without fee is hereby granted, provided that the above
+    copyright notice and this permission notice appear in all copies.
+
+    THE  SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL WARRANTIES
+    WITH  REGARD  TO  THIS  SOFTWARE  INCLUDING  ALL  IMPLIED  WARRANTIES  OF
+    MERCHANTABILITY  AND  FITNESS. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR
+    ANY  SPECIAL,  DIRECT,  INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES
+    WHATSOEVER  RESULTING  FROM  LOSS  OF USE, DATA OR PROFITS, WHETHER IN AN
+    ACTION  OF  CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF
+    OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE.
+*/
+//==============================================================================
+
+#pragma once
+
+#include <data/cassandra/Error.h>
+#include <data/cassandra/Types.h>
+#include <data/cassandra/impl/Batch.h>
+#include <data/cassandra/impl/Cluster.h>
+#include <data/cassandra/impl/Future.h>
+#include <data/cassandra/impl/ManagedObject.h>
+#include <data/cassandra/impl/Result.h>
+#include <data/cassandra/impl/Session.h>
+#include <data/cassandra/impl/Statement.h>
+#include <util/Expected.h>
+
+#include <cassandra.h>
+
+#include <chrono>
+#include <compare>
+#include <iterator>
+#include <vector>
+
+namespace data::cassandra {
+
+/**
+ * @brief Represents a handle to the cassandra database cluster
+ */
+class Handle
+{
+    detail::Cluster cluster_;
+    detail::Session session_;
+
+public:
+    using ResultOrErrorType = ResultOrError;
+    using MaybeErrorType = MaybeError;
+    using FutureWithCallbackType = FutureWithCallback;
+    using FutureType = Future;
+    using StatementType = Statement;
+    using PreparedStatementType = PreparedStatement;
+    using ResultType = Result;
+
+    /**
+     * @brief Construct a new handle from a @ref detail::Settings object.
+     *
+     * @param clusterSettings The settings to use
+     */
+    explicit Handle(Settings clusterSettings = Settings::defaultSettings());
+
+    /**
+     * @brief Construct a new handle with default settings and only by setting the contact points.
+     *
+     * @param contactPoints The contact points to use instead of settings
+     */
+    explicit Handle(std::string_view contactPoints);
+
+    /**
+     * @brief Disconnects gracefully if possible.
+     */
+    ~Handle();
+
+    /**
+     * @brief Move is supported.
+     */
+    Handle(Handle&&) = default;
+
+    /**
+     * @brief Connect to the cluster asynchronously.
+     *
+     * @return A future
+     */
+    [[nodiscard]] FutureType
+    asyncConnect() const;
+
+    /**
+     * @brief Synchonous version of the above.
+     *
+     * See @ref asyncConnect() const for how this works.
+     *
+     * @return Possibly an error
+     */
+    [[nodiscard]] MaybeErrorType
+    connect() const;
+
+    /**
+     * @brief Connect to the the specified keyspace asynchronously.
+     *
+     * @param keyspace The keyspace to use
+     * @return A future
+     */
+    [[nodiscard]] FutureType
+    asyncConnect(std::string_view keyspace) const;
+
+    /**
+     * @brief Synchonous version of the above.
+     *
+     * See @ref asyncConnect(std::string_view) const for how this works.
+     *
+     * @param keyspace The keyspace to use
+     * @return Possibly an error
+     */
+    [[nodiscard]] MaybeErrorType
+    connect(std::string_view keyspace) const;
+
+    /**
+     * @brief Disconnect from the cluster asynchronously.
+     *
+     * @return A future
+     */
+    [[nodiscard]] FutureType
+    asyncDisconnect() const;
+
+    /**
+     * @brief Synchonous version of the above.
+     *
+     * See @ref asyncDisconnect() const for how this works.
+     *
+     * @return Possibly an error
+     */
+    [[maybe_unused]] MaybeErrorType
+    disconnect() const;
+
+    /**
+     * @brief Reconnect to the the specified keyspace asynchronously.
+     *
+     * @param keyspace The keyspace to use
+     * @return A future
+     */
+    [[nodiscard]] FutureType
+    asyncReconnect(std::string_view keyspace) const;
+
+    /**
+     * @brief Synchonous version of the above.
+     *
+     * See @ref asyncReconnect(std::string_view) const for how this works.
+     *
+     * @param keyspace The keyspace to use
+     * @return Possibly an error
+     */
+    [[nodiscard]] MaybeErrorType
+    reconnect(std::string_view keyspace) const;
+
+    /**
+     * @brief Execute a simple query with optional args asynchronously.
+     *
+     * @param query The query to execute
+     * @param args The arguments to bind for execution
+     * @return A future
+     */
+    template <typename... Args>
+    [[nodiscard]] FutureType
+    asyncExecute(std::string_view query, Args&&... args) const
+    {
+        auto statement = StatementType{query, std::forward<Args>(args)...};
+        return cass_session_execute(session_, statement);
+    }
+
+    /**
+     * @brief Synchonous version of the above.
+     *
+     * See asyncExecute(std::string_view, Args&&...) const for how this works.
+     *
+     * @param query The query to execute
+     * @param args The arguments to bind for execution
+     * @return The result or an error
+     */
+    template <typename... Args>
+    [[maybe_unused]] ResultOrErrorType
+    execute(std::string_view query, Args&&... args) const
+    {
+        return asyncExecute<Args...>(query, std::forward<Args>(args)...).get();
+    }
+
+    /**
+     * @brief Execute each of the statements asynchronously.
+     *
+     * Batched version is not always the right option.
+     * Especially since it only supports INSERT, UPDATE and DELETE statements.
+     * This can be used as an alternative when statements need to execute in bulk.
+     *
+     * @param statements The statements to execute
+     * @return A vector of future objects
+     */
+    [[nodiscard]] std::vector<FutureType>
+    asyncExecuteEach(std::vector<StatementType> const& statements) const;
+
+    /**
+     * @brief Synchonous version of the above.
+     *
+     * See @ref asyncExecuteEach(std::vector<StatementType> const&) const for how this works.
+     *
+     * @param statements The statements to execute
+     * @return Possibly an error
+     */
+    [[maybe_unused]] MaybeErrorType
+    executeEach(std::vector<StatementType> const& statements) const;
+
+    /**
+     * @brief Execute a prepared statement with optional args asynchronously.
+     *
+     * @param statement The prepared statement to execute
+     * @param args The arguments to bind for execution
+     * @return A future
+     */
+    template <typename... Args>
+    [[nodiscard]] FutureType
+    asyncExecute(PreparedStatementType const& statement, Args&&... args) const
+    {
+        auto bound = statement.bind<Args...>(std::forward<Args>(args)...);
+        return cass_session_execute(session_, bound);
+    }
+
+    /**
+     * @brief Synchonous version of the above.
+     *
+     * See asyncExecute(std::vector<StatementType> const&, Args&&...) const for how this works.
+     *
+     * @param statement The prepared statement to bind and execute
+     * @param args The arguments to bind for execution
+     * @return The result or an error
+     */
+    template <typename... Args>
+    [[maybe_unused]] ResultOrErrorType
+    execute(PreparedStatementType const& statement, Args&&... args) const
+    {
+        return asyncExecute<Args...>(statement, std::forward<Args>(args)...).get();
+    }
+
+    /**
+     * @brief Execute one (bound or simple) statements asynchronously.
+     *
+     * @param statement The statement to execute
+     * @return A future
+     */
+    [[nodiscard]] FutureType
+    asyncExecute(StatementType const& statement) const;
+
+    /**
+     * @brief Execute one (bound or simple) statements asynchronously with a callback.
+     *
+     * @param statement The statement to execute
+     * @param cb The callback to execute when data is ready
+     * @return A future that holds onto the callback provided
+     */
+    [[nodiscard]] FutureWithCallbackType
+    asyncExecute(StatementType const& statement, std::function<void(ResultOrErrorType)>&& cb) const;
+
+    /**
+     * @brief Synchonous version of the above.
+     *
+     * See @ref asyncExecute(StatementType const&) const for how this works.
+     *
+     * @param statement The statement to execute
+     * @return The result or an error
+     */
+    [[maybe_unused]] ResultOrErrorType
+    execute(StatementType const& statement) const;
+
+    /**
+     * @brief Execute a batch of (bound or simple) statements asynchronously.
+     *
+     * @param statements The statements to execute
+     * @return A future
+     */
+    [[nodiscard]] FutureType
+    asyncExecute(std::vector<StatementType> const& statements) const;
+
+    /**
+     * @brief Synchonous version of the above.
+     *
+     * See @ref asyncExecute(std::vector<StatementType> const&) const for how this works.
+     *
+     * @param statements The statements to execute
+     * @return Possibly an error
+     */
+    [[maybe_unused]] MaybeErrorType
+    execute(std::vector<StatementType> const& statements) const;
+
+    /**
+     * @brief Execute a batch of (bound or simple) statements asynchronously with a completion callback.
+     *
+     * @param statements The statements to execute
+     * @param cb The callback to execute when data is ready
+     * @return A future that holds onto the callback provided
+     */
+    [[nodiscard]] FutureWithCallbackType
+    asyncExecute(std::vector<StatementType> const& statements, std::function<void(ResultOrErrorType)>&& cb) const;
+
+    /**
+     * @brief Prepare a statement.
+     *
+     * @param query
+     * @return A prepared statement
+     * @throws std::runtime_error with underlying error description on failure
+     */
+    [[nodiscard]] PreparedStatementType
+    prepare(std::string_view query) const;
+};
+
+/**
+ * @brief Extracts the results into series of std::tuple<Types...> by creating a simple wrapper with an STL input
+ * iterator inside.
+ *
+ * You can call .begin() and .end() in order to iterate as usual.
+ * This also means that you can use it in a range-based for or with some algorithms.
+ *
+ * @param result The result to iterate
+ */
+template <typename... Types>
+[[nodiscard]] detail::ResultExtractor<Types...>
+extract(Handle::ResultType const& result)
+{
+    return {result};
+}
+
+}  // namespace data::cassandra

src/data/cassandra/Schema.h

@@ -0,0 +1,662 @@
+//------------------------------------------------------------------------------
+/*
+    This file is part of clio: https://github.com/XRPLF/clio
+    Copyright (c) 2023, the clio developers.
+
+    Permission to use, copy, modify, and distribute this software for any
+    purpose with or without fee is hereby granted, provided that the above
+    copyright notice and this permission notice appear in all copies.
+
+    THE  SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL WARRANTIES
+    WITH  REGARD  TO  THIS  SOFTWARE  INCLUDING  ALL  IMPLIED  WARRANTIES  OF
+    MERCHANTABILITY  AND  FITNESS. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR
+    ANY  SPECIAL,  DIRECT,  INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES
+    WHATSOEVER  RESULTING  FROM  LOSS  OF USE, DATA OR PROFITS, WHETHER IN AN
+    ACTION  OF  CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF
+    OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE.
+*/
+//==============================================================================
+
+#pragma once
+
+#include <data/cassandra/Concepts.h>
+#include <data/cassandra/Handle.h>
+#include <data/cassandra/SettingsProvider.h>
+#include <data/cassandra/Types.h>
+#include <util/Expected.h>
+#include <util/config/Config.h>
+#include <util/log/Logger.h>
+
+#include <fmt/compile.h>
+
+namespace data::cassandra {
+
+template <SomeSettingsProvider SettingsProviderType>
+[[nodiscard]] std::string inline qualifiedTableName(SettingsProviderType const& provider, std::string_view name)
+{
+    return fmt::format("{}.{}{}", provider.getKeyspace(), provider.getTablePrefix().value_or(""), name);
+}
+
+/**
+ * @brief Manages the DB schema and provides access to prepared statements.
+ */
+template <SomeSettingsProvider SettingsProviderType>
+class Schema
+{
+    util::Logger log_{"Backend"};
+    std::reference_wrapper<SettingsProviderType const> settingsProvider_;
+
+public:
+    explicit Schema(SettingsProviderType const& settingsProvider) : settingsProvider_{std::cref(settingsProvider)}
+    {
+    }
+
+    std::string createKeyspace = [this]() {
+        return fmt::format(
+            R"(
+            CREATE KEYSPACE IF NOT EXISTS {} 
+              WITH replication = {{
+                     'class': 'SimpleStrategy',
+                     'replication_factor': '{}'
+                   }} 
+               AND durable_writes = true
+            )",
+            settingsProvider_.get().getKeyspace(),
+            settingsProvider_.get().getReplicationFactor());
+    }();
+
+    // =======================
+    // Schema creation queries
+    // =======================
+
+    std::vector<Statement> createSchema = [this]() {
+        std::vector<Statement> statements;
+
+        statements.emplace_back(fmt::format(
+            R"(
+           CREATE TABLE IF NOT EXISTS {}
+                  (      
+                         key blob, 
+                    sequence bigint, 
+                      object blob, 
+                     PRIMARY KEY (key, sequence) 
+                  ) 
+             WITH CLUSTERING ORDER BY (sequence DESC) 
+              AND default_time_to_live = {}
+            )",
+            qualifiedTableName(settingsProvider_.get(), "objects"),
+            settingsProvider_.get().getTtl()));
+
+        statements.emplace_back(fmt::format(
+            R"(
+           CREATE TABLE IF NOT EXISTS {}
+                  (     
+                        hash blob PRIMARY KEY, 
+             ledger_sequence bigint, 
+                        date bigint,
+                 transaction blob, 
+                    metadata blob 
+                  ) 
+             WITH default_time_to_live = {}
+            )",
+            qualifiedTableName(settingsProvider_.get(), "transactions"),
+            settingsProvider_.get().getTtl()));
+
+        statements.emplace_back(fmt::format(
+            R"(
+           CREATE TABLE IF NOT EXISTS {}
+                  (     
+             ledger_sequence bigint, 
+                        hash blob, 
+                     PRIMARY KEY (ledger_sequence, hash) 
+                  ) 
+             WITH default_time_to_live = {}
+            )",
+            qualifiedTableName(settingsProvider_.get(), "ledger_transactions"),
+            settingsProvider_.get().getTtl()));
+
+        statements.emplace_back(fmt::format(
+            R"(
+           CREATE TABLE IF NOT EXISTS {}
+                  (     
+                    key blob,
+                    seq bigint, 
+                   next blob, 
+                PRIMARY KEY (key, seq) 
+                  ) 
+             WITH default_time_to_live = {}
+            )",
+            qualifiedTableName(settingsProvider_.get(), "successor"),
+            settingsProvider_.get().getTtl()));
+
+        statements.emplace_back(fmt::format(
+            R"(
+           CREATE TABLE IF NOT EXISTS {}
+                  (     
+                    seq bigint, 
+                    key blob,
+                PRIMARY KEY (seq, key) 
+                  ) 
+             WITH default_time_to_live = {}
+            )",
+            qualifiedTableName(settingsProvider_.get(), "diff"),
+            settingsProvider_.get().getTtl()));
+
+        statements.emplace_back(fmt::format(
+            R"(
+           CREATE TABLE IF NOT EXISTS {}
+                  ( 
+                    account blob,    
+                    seq_idx tuple<bigint, bigint>, 
+                       hash blob,
+                    PRIMARY KEY (account, seq_idx) 
+                  ) 
+             WITH CLUSTERING ORDER BY (seq_idx DESC)
+              AND default_time_to_live = {}
+            )",
+            qualifiedTableName(settingsProvider_.get(), "account_tx"),
+            settingsProvider_.get().getTtl()));
+
+        statements.emplace_back(fmt::format(
+            R"(
+           CREATE TABLE IF NOT EXISTS {}
+                  ( 
+                    sequence bigint PRIMARY KEY,
+                      header blob
+                  ) 
+             WITH default_time_to_live = {}
+            )",
+            qualifiedTableName(settingsProvider_.get(), "ledgers"),
+            settingsProvider_.get().getTtl()));
+
+        statements.emplace_back(fmt::format(
+            R"(
+           CREATE TABLE IF NOT EXISTS {}
+                  ( 
+                    hash blob PRIMARY KEY,
+                sequence bigint
+                  ) 
+             WITH default_time_to_live = {}
+            )",
+            qualifiedTableName(settingsProvider_.get(), "ledger_hashes"),
+            settingsProvider_.get().getTtl()));
+
+        statements.emplace_back(fmt::format(
+            R"(
+           CREATE TABLE IF NOT EXISTS {}
+                  ( 
+                    is_latest boolean PRIMARY KEY,
+                     sequence bigint
+                  )
+            )",
+            qualifiedTableName(settingsProvider_.get(), "ledger_range")));
+
+        statements.emplace_back(fmt::format(
+            R"(
+           CREATE TABLE IF NOT EXISTS {}
+                  ( 
+                    token_id blob,    
+                    sequence bigint,
+                       owner blob,
+                   is_burned boolean,
+                     PRIMARY KEY (token_id, sequence) 
+                  ) 
+             WITH CLUSTERING ORDER BY (sequence DESC)
+              AND default_time_to_live = {}
+            )",
+            qualifiedTableName(settingsProvider_.get(), "nf_tokens"),
+            settingsProvider_.get().getTtl()));
+
+        statements.emplace_back(fmt::format(
+            R"(
+           CREATE TABLE IF NOT EXISTS {}
+                  ( 
+                      issuer blob,
+                       taxon bigint,
+                    token_id blob,
+                     PRIMARY KEY (issuer, taxon, token_id)
+                  ) 
+             WITH CLUSTERING ORDER BY (taxon ASC, token_id ASC)
+              AND default_time_to_live = {}
+            )",
+            qualifiedTableName(settingsProvider_.get(), "issuer_nf_tokens_v2"),
+            settingsProvider_.get().getTtl()));
+
+        statements.emplace_back(fmt::format(
+            R"(
+           CREATE TABLE IF NOT EXISTS {}
+                  ( 
+                    token_id blob,
+                    sequence bigint,
+                         uri blob,
+                     PRIMARY KEY (token_id, sequence)
+                  ) 
+             WITH CLUSTERING ORDER BY (sequence DESC)
+              AND default_time_to_live = {}
+            )",
+            qualifiedTableName(settingsProvider_.get(), "nf_token_uris"),
+            settingsProvider_.get().getTtl()));
+
+        statements.emplace_back(fmt::format(
+            R"(
+           CREATE TABLE IF NOT EXISTS {}
+                  ( 
+                    token_id blob,    
+                     seq_idx tuple<bigint, bigint>,
+                        hash blob,
+                     PRIMARY KEY (token_id, seq_idx) 
+                  ) 
+             WITH CLUSTERING ORDER BY (seq_idx DESC)
+              AND default_time_to_live = {}
+            )",
+            qualifiedTableName(settingsProvider_.get(), "nf_token_transactions"),
+            settingsProvider_.get().getTtl()));
+
+        return statements;
+    }();
+
+    /**
+     * @brief Prepared statements holder.
+     */
+    class Statements
+    {
+        std::reference_wrapper<SettingsProviderType const> settingsProvider_;
+        std::reference_wrapper<Handle const> handle_;
+
+    public:
+        Statements(SettingsProviderType const& settingsProvider, Handle const& handle)
+            : settingsProvider_{settingsProvider}, handle_{std::cref(handle)}
+        {
+        }
+
+        //
+        // Insert queries
+        //
+
+        PreparedStatement insertObject = [this]() {
+            return handle_.get().prepare(fmt::format(
+                R"(
+                INSERT INTO {} 
+                       (key, sequence, object)
+                VALUES (?, ?, ?)
+                )",
+                qualifiedTableName(settingsProvider_.get(), "objects")));
+        }();
+
+        PreparedStatement insertTransaction = [this]() {
+            return handle_.get().prepare(fmt::format(
+                R"(
+                INSERT INTO {} 
+                       (hash, ledger_sequence, date, transaction, metadata)
+                VALUES (?, ?, ?, ?, ?)
+                )",
+                qualifiedTableName(settingsProvider_.get(), "transactions")));
+        }();
+
+        PreparedStatement insertLedgerTransaction = [this]() {
+            return handle_.get().prepare(fmt::format(
+                R"(
+                INSERT INTO {} 
+                       (ledger_sequence, hash)
+                VALUES (?, ?)
+                )",
+                qualifiedTableName(settingsProvider_.get(), "ledger_transactions")));
+        }();
+
+        PreparedStatement insertSuccessor = [this]() {
+            return handle_.get().prepare(fmt::format(
+                R"(
+                INSERT INTO {} 
+                       (key, seq, next)
+                VALUES (?, ?, ?)
+                )",
+                qualifiedTableName(settingsProvider_.get(), "successor")));
+        }();
+
+        PreparedStatement insertDiff = [this]() {
+            return handle_.get().prepare(fmt::format(
+                R"(
+                INSERT INTO {} 
+                       (seq, key)
+                VALUES (?, ?)
+                )",
+                qualifiedTableName(settingsProvider_.get(), "diff")));
+        }();
+
+        PreparedStatement insertAccountTx = [this]() {
+            return handle_.get().prepare(fmt::format(
+                R"(
+                INSERT INTO {} 
+                       (account, seq_idx, hash)
+                VALUES (?, ?, ?)
+                )",
+                qualifiedTableName(settingsProvider_.get(), "account_tx")));
+        }();
+
+        PreparedStatement insertNFT = [this]() {
+            return handle_.get().prepare(fmt::format(
+                R"(
+                INSERT INTO {} 
+                       (token_id, sequence, owner, is_burned)
+                VALUES (?, ?, ?, ?)
+                )",
+                qualifiedTableName(settingsProvider_.get(), "nf_tokens")));
+        }();
+
+        PreparedStatement insertIssuerNFT = [this]() {
+            return handle_.get().prepare(fmt::format(
+                R"(
+                INSERT INTO {} 
+                       (issuer, taxon, token_id)
+                VALUES (?, ?, ?)
+                )",
+                qualifiedTableName(settingsProvider_.get(), "issuer_nf_tokens_v2")));
+        }();
+
+        PreparedStatement insertNFTURI = [this]() {
+            return handle_.get().prepare(fmt::format(
+                R"(
+                INSERT INTO {} 
+                       (token_id, sequence, uri)
+                VALUES (?, ?, ?)
+                )",
+                qualifiedTableName(settingsProvider_.get(), "nf_token_uris")));
+        }();
+
+        PreparedStatement insertNFTTx = [this]() {
+            return handle_.get().prepare(fmt::format(
+                R"(
+                INSERT INTO {} 
+                       (token_id, seq_idx, hash)
+                VALUES (?, ?, ?)
+                )",
+                qualifiedTableName(settingsProvider_.get(), "nf_token_transactions")));
+        }();
+
+        PreparedStatement insertLedgerHeader = [this]() {
+            return handle_.get().prepare(fmt::format(
+                R"(
+                INSERT INTO {} 
+                       (sequence, header)
+                VALUES (?, ?)
+                )",
+                qualifiedTableName(settingsProvider_.get(), "ledgers")));
+        }();
+
+        PreparedStatement insertLedgerHash = [this]() {
+            return handle_.get().prepare(fmt::format(
+                R"(
+                INSERT INTO {} 
+                       (hash, sequence)
+                VALUES (?, ?)
+                )",
+                qualifiedTableName(settingsProvider_.get(), "ledger_hashes")));
+        }();
+
+        //
+        // Update (and "delete") queries
+        //
+
+        PreparedStatement updateLedgerRange = [this]() {
+            return handle_.get().prepare(fmt::format(
+                R"(
+                UPDATE {} 
+                   SET sequence = ?
+                 WHERE is_latest = ? 
+                    IF sequence IN (?, null)
+                )",
+                qualifiedTableName(settingsProvider_.get(), "ledger_range")));
+        }();
+
+        PreparedStatement deleteLedgerRange = [this]() {
+            return handle_.get().prepare(fmt::format(
+                R"(
+                UPDATE {} 
+                   SET sequence = ?
+                 WHERE is_latest = false
+                )",
+                qualifiedTableName(settingsProvider_.get(), "ledger_range")));
+        }();
+
+        //
+        // Select queries
+        //
+
+        PreparedStatement selectSuccessor = [this]() {
+            return handle_.get().prepare(fmt::format(
+                R"(
+                SELECT next 
+                  FROM {}               
+                 WHERE key = ?
+                   AND seq <= ?
+              ORDER BY seq DESC 
+                 LIMIT 1
+                )",
+                qualifiedTableName(settingsProvider_.get(), "successor")));
+        }();
+
+        PreparedStatement selectDiff = [this]() {
+            return handle_.get().prepare(fmt::format(
+                R"(
+                SELECT key 
+                  FROM {}
+                 WHERE seq = ?
+                )",
+                qualifiedTableName(settingsProvider_.get(), "diff")));
+        }();
+
+        PreparedStatement selectObject = [this]() {
+            return handle_.get().prepare(fmt::format(
+                R"(
+                SELECT object, sequence 
+                  FROM {}               
+                 WHERE key = ?
+                   AND sequence <= ?
+              ORDER BY sequence DESC 
+                 LIMIT 1
+                )",
+                qualifiedTableName(settingsProvider_.get(), "objects")));
+        }();
+
+        PreparedStatement selectTransaction = [this]() {
+            return handle_.get().prepare(fmt::format(
+                R"(
+                SELECT transaction, metadata, ledger_sequence, date 
+                  FROM {}
+                 WHERE hash = ?
+                )",
+                qualifiedTableName(settingsProvider_.get(), "transactions")));
+        }();
+
+        PreparedStatement selectAllTransactionHashesInLedger = [this]() {
+            return handle_.get().prepare(fmt::format(
+                R"(
+                SELECT hash 
+                  FROM {}               
+                 WHERE ledger_sequence = ?               
+                )",
+                qualifiedTableName(settingsProvider_.get(), "ledger_transactions")));
+        }();
+
+        PreparedStatement selectLedgerPageKeys = [this]() {
+            return handle_.get().prepare(fmt::format(
+                R"(
+                SELECT key 
+                  FROM {}               
+                 WHERE TOKEN(key) >= ?
+                   AND sequence <= ?
+         PER PARTITION LIMIT 1 
+                 LIMIT ?
+                 ALLOW FILTERING
+                )",
+                qualifiedTableName(settingsProvider_.get(), "objects")));
+        }();
+
+        PreparedStatement selectLedgerPage = [this]() {
+            return handle_.get().prepare(fmt::format(
+                R"(
+                SELECT object, key
+                  FROM {}
+                 WHERE TOKEN(key) >= ?
+                   AND sequence <= ?
+         PER PARTITION LIMIT 1
+                 LIMIT ?
+                 ALLOW FILTERING
+                )",
+                qualifiedTableName(settingsProvider_.get(), "objects")));
+        }();
+
+        PreparedStatement getToken = [this]() {
+            return handle_.get().prepare(fmt::format(
+                R"(
+                SELECT TOKEN(key) 
+                  FROM {}               
+                 WHERE key = ?               
+                 LIMIT 1
+                )",
+                qualifiedTableName(settingsProvider_.get(), "objects")));
+        }();
+
+        PreparedStatement selectAccountTx = [this]() {
+            return handle_.get().prepare(fmt::format(
+                R"(
+                SELECT hash, seq_idx 
+                  FROM {}               
+                 WHERE account = ?
+                   AND seq_idx < ?
+                 LIMIT ?
+                )",
+                qualifiedTableName(settingsProvider_.get(), "account_tx")));
+        }();
+
+        PreparedStatement selectAccountTxForward = [this]() {
+            return handle_.get().prepare(fmt::format(
+                R"(
+                SELECT hash, seq_idx 
+                  FROM {}               
+                 WHERE account = ?
+                   AND seq_idx > ?
+              ORDER BY seq_idx ASC 
+                 LIMIT ?
+                )",
+                qualifiedTableName(settingsProvider_.get(), "account_tx")));
+        }();
+
+        PreparedStatement selectNFT = [this]() {
+            return handle_.get().prepare(fmt::format(
+                R"(
+                SELECT sequence, owner, is_burned
+                  FROM {}    
+                 WHERE token_id = ?
+                   AND sequence <= ?
+              ORDER BY sequence DESC
+                 LIMIT 1
+                )",
+                qualifiedTableName(settingsProvider_.get(), "nf_tokens")));
+        }();
+
+        PreparedStatement selectNFTURI = [this]() {
+            return handle_.get().prepare(fmt::format(
+                R"(
+                SELECT uri
+                  FROM {}    
+                 WHERE token_id = ?
+                   AND sequence <= ?
+              ORDER BY sequence DESC
+                 LIMIT 1
+                )",
+                qualifiedTableName(settingsProvider_.get(), "nf_token_uris")));
+        }();
+
+        PreparedStatement selectNFTTx = [this]() {
+            return handle_.get().prepare(fmt::format(
+                R"(
+                SELECT hash, seq_idx
+                  FROM {}    
+                 WHERE token_id = ?
+                   AND seq_idx < ?
+              ORDER BY seq_idx DESC
+                 LIMIT ?
+                )",
+                qualifiedTableName(settingsProvider_.get(), "nf_token_transactions")));
+        }();
+
+        PreparedStatement selectNFTTxForward = [this]() {
+            return handle_.get().prepare(fmt::format(
+                R"(
+                SELECT hash, seq_idx
+                  FROM {}    
+                 WHERE token_id = ?
+                   AND seq_idx >= ?
+              ORDER BY seq_idx ASC
+                 LIMIT ?
+                )",
+                qualifiedTableName(settingsProvider_.get(), "nf_token_transactions")));
+        }();
+
+        PreparedStatement selectLedgerByHash = [this]() {
+            return handle_.get().prepare(fmt::format(
+                R"(
+                SELECT sequence
+                  FROM {}
+                 WHERE hash = ?     
+                 LIMIT 1
+                )",
+                qualifiedTableName(settingsProvider_.get(), "ledger_hashes")));
+        }();
+
+        PreparedStatement selectLedgerBySeq = [this]() {
+            return handle_.get().prepare(fmt::format(
+                R"(
+                SELECT header
+                  FROM {}
+                 WHERE sequence = ?
+                )",
+                qualifiedTableName(settingsProvider_.get(), "ledgers")));
+        }();
+
+        PreparedStatement selectLatestLedger = [this]() {
+            return handle_.get().prepare(fmt::format(
+                R"(
+                SELECT sequence
+                  FROM {}    
+                 WHERE is_latest = true               
+                )",
+                qualifiedTableName(settingsProvider_.get(), "ledger_range")));
+        }();
+
+        PreparedStatement selectLedgerRange = [this]() {
+            return handle_.get().prepare(fmt::format(
+                R"(
+                SELECT sequence
+                  FROM {}
+                )",
+                qualifiedTableName(settingsProvider_.get(), "ledger_range")));
+        }();
+    };
+
+    /**
+     * @brief Recreates the prepared statements.
+     */
+    void
+    prepareStatements(Handle const& handle)
+    {
+        LOG(log_.info()) << "Preparing cassandra statements";
+        statements_ = std::make_unique<Statements>(settingsProvider_, handle);
+        LOG(log_.info()) << "Finished preparing statements";
+    }
+
+    /**
+     * @brief Provides access to statements.
+     */
+    std::unique_ptr<Statements> const&
+    operator->() const
+    {
+        return statements_;
+    }
+
+private:
+    std::unique_ptr<Statements> statements_{nullptr};
+};
+
+}  // namespace data::cassandra

src/data/cassandra/SettingsProvider.cpp

@@ -0,0 +1,139 @@
+//------------------------------------------------------------------------------
+/*
+    This file is part of clio: https://github.com/XRPLF/clio
+    Copyright (c) 2023, the clio developers.
+
+    Permission to use, copy, modify, and distribute this software for any
+    purpose with or without fee is hereby granted, provided that the above
+    copyright notice and this permission notice appear in all copies.
+
+    THE  SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL WARRANTIES
+    WITH  REGARD  TO  THIS  SOFTWARE  INCLUDING  ALL  IMPLIED  WARRANTIES  OF
+    MERCHANTABILITY  AND  FITNESS. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR
+    ANY  SPECIAL,  DIRECT,  INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES
+    WHATSOEVER  RESULTING  FROM  LOSS  OF USE, DATA OR PROFITS, WHETHER IN AN
+    ACTION  OF  CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF
+    OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE.
+*/
+//==============================================================================
+
+#include <data/cassandra/SettingsProvider.h>
+#include <data/cassandra/impl/Cluster.h>
+#include <data/cassandra/impl/Statement.h>
+#include <util/config/Config.h>
+
+#include <boost/json.hpp>
+
+#include <fstream>
+#include <string>
+#include <thread>
+
+namespace data::cassandra {
+
+namespace detail {
+inline Settings::ContactPoints
+tag_invoke(boost::json::value_to_tag<Settings::ContactPoints>, boost::json::value const& value)
+{
+    if (not value.is_object())
+        throw std::runtime_error(
+            "Feed entire Cassandra section to parse "
+            "Settings::ContactPoints instead");
+
+    util::Config obj{value};
+    Settings::ContactPoints out;
+
+    out.contactPoints = obj.valueOrThrow<std::string>("contact_points", "`contact_points` must be a string");
+    out.port = obj.maybeValue<uint16_t>("port");
+
+    return out;
+}
+
+inline Settings::SecureConnectionBundle
+tag_invoke(boost::json::value_to_tag<Settings::SecureConnectionBundle>, boost::json::value const& value)
+{
+    if (not value.is_string())
+        throw std::runtime_error("`secure_connect_bundle` must be a string");
+    return Settings::SecureConnectionBundle{value.as_string().data()};
+}
+}  // namespace detail
+
+SettingsProvider::SettingsProvider(util::Config const& cfg, uint16_t ttl)
+    : config_{cfg}
+    , keyspace_{cfg.valueOr<std::string>("keyspace", "clio")}
+    , tablePrefix_{cfg.maybeValue<std::string>("table_prefix")}
+    , replicationFactor_{cfg.valueOr<uint16_t>("replication_factor", 3)}
+    , ttl_{ttl}
+    , settings_{parseSettings()}
+{
+}
+
+Settings
+SettingsProvider::getSettings() const
+{
+    return settings_;
+}
+
+std::optional<std::string>
+SettingsProvider::parseOptionalCertificate() const
+{
+    if (auto const certPath = config_.maybeValue<std::string>("certfile"); certPath)
+    {
+        auto const path = std::filesystem::path(*certPath);
+        std::ifstream fileStream(path.string(), std::ios::in);
+        if (!fileStream)
+        {
+            throw std::system_error(errno, std::generic_category(), "Opening certificate " + path.string());
+        }
+
+        std::string contents(std::istreambuf_iterator<char>{fileStream}, std::istreambuf_iterator<char>{});
+        if (fileStream.bad())
+        {
+            throw std::system_error(errno, std::generic_category(), "Reading certificate " + path.string());
+        }
+
+        return contents;
+    }
+
+    return std::nullopt;
+}
+
+Settings
+SettingsProvider::parseSettings() const
+{
+    auto settings = Settings::defaultSettings();
+    if (auto const bundle = config_.maybeValue<Settings::SecureConnectionBundle>("secure_connect_bundle"); bundle)
+    {
+        settings.connectionInfo = *bundle;
+    }
+    else
+    {
+        settings.connectionInfo =
+            config_.valueOrThrow<Settings::ContactPoints>("Missing contact_points in Cassandra config");
+    }
+
+    settings.threads = config_.valueOr<uint32_t>("threads", settings.threads);
+    settings.maxWriteRequestsOutstanding =
+        config_.valueOr<uint32_t>("max_write_requests_outstanding", settings.maxWriteRequestsOutstanding);
+    settings.maxReadRequestsOutstanding =
+        config_.valueOr<uint32_t>("max_read_requests_outstanding", settings.maxReadRequestsOutstanding);
+    settings.coreConnectionsPerHost =
+        config_.valueOr<uint32_t>("core_connections_per_host", settings.coreConnectionsPerHost);
+
+    settings.queueSizeIO = config_.maybeValue<uint32_t>("queue_size_io");
+
+    auto const connectTimeoutSecond = config_.maybeValue<uint32_t>("connect_timeout");
+    if (connectTimeoutSecond)
+        settings.connectionTimeout = std::chrono::milliseconds{*connectTimeoutSecond * 1000};
+
+    auto const requestTimeoutSecond = config_.maybeValue<uint32_t>("request_timeout");
+    if (requestTimeoutSecond)
+        settings.requestTimeout = std::chrono::milliseconds{*requestTimeoutSecond * 1000};
+
+    settings.certificate = parseOptionalCertificate();
+    settings.username = config_.maybeValue<std::string>("username");
+    settings.password = config_.maybeValue<std::string>("password");
+
+    return settings;
+}
+
+}  // namespace data::cassandra

src/data/cassandra/SettingsProvider.h

@@ -0,0 +1,102 @@
+//------------------------------------------------------------------------------
+/*
+    This file is part of clio: https://github.com/XRPLF/clio
+    Copyright (c) 2023, the clio developers.
+
+    Permission to use, copy, modify, and distribute this software for any
+    purpose with or without fee is hereby granted, provided that the above
+    copyright notice and this permission notice appear in all copies.
+
+    THE  SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL WARRANTIES
+    WITH  REGARD  TO  THIS  SOFTWARE  INCLUDING  ALL  IMPLIED  WARRANTIES  OF
+    MERCHANTABILITY  AND  FITNESS. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR
+    ANY  SPECIAL,  DIRECT,  INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES
+    WHATSOEVER  RESULTING  FROM  LOSS  OF USE, DATA OR PROFITS, WHETHER IN AN
+    ACTION  OF  CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF
+    OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE.
+*/
+//==============================================================================
+
+#pragma once
+
+#include <data/cassandra/Handle.h>
+#include <data/cassandra/Types.h>
+#include <util/Expected.h>
+#include <util/config/Config.h>
+#include <util/log/Logger.h>
+
+namespace data::cassandra {
+
+/**
+ * @brief Provides settings for @ref BasicCassandraBackend.
+ */
+class SettingsProvider
+{
+    util::Config config_;
+
+    std::string keyspace_;
+    std::optional<std::string> tablePrefix_;
+    uint16_t replicationFactor_;
+    uint16_t ttl_;
+    Settings settings_;
+
+public:
+    /**
+     * @brief Create a settings provider from the specified config.
+     *
+     * @param cfg The config of Clio to use
+     * @param ttl Time to live setting
+     */
+    explicit SettingsProvider(util::Config const& cfg, uint16_t ttl = 0);
+
+    /**
+     * @return The cluster settings
+     */
+    [[nodiscard]] Settings
+    getSettings() const;
+
+    /**
+     * @return The specified keyspace
+     */
+    [[nodiscard]] inline std::string
+    getKeyspace() const
+    {
+        return keyspace_;
+    }
+
+    /**
+     * @return The optional table prefix to use in all queries
+     */
+    [[nodiscard]] inline std::optional<std::string>
+    getTablePrefix() const
+    {
+        return tablePrefix_;
+    }
+
+    /**
+     * @return The replication factor
+     */
+    [[nodiscard]] inline uint16_t
+    getReplicationFactor() const
+    {
+        return replicationFactor_;
+    }
+
+    /**
+     * @return The default time to live to use in all `create` queries
+     */
+    [[nodiscard]] inline uint16_t
+    getTtl() const
+    {
+        return ttl_;
+    }
+
+private:
+    [[nodiscard]] std::optional<std::string>
+    parseOptionalCertificate() const;
+
+    [[nodiscard]] Settings
+    parseSettings() const;
+};
+
+}  // namespace data::cassandra

src/data/cassandra/Types.h

@@ -0,0 +1,67 @@
+//------------------------------------------------------------------------------
+/*
+    This file is part of clio: https://github.com/XRPLF/clio
+    Copyright (c) 2023, the clio developers.
+
+    Permission to use, copy, modify, and distribute this software for any
+    purpose with or without fee is hereby granted, provided that the above
+    copyright notice and this permission notice appear in all copies.
+
+    THE  SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL WARRANTIES
+    WITH  REGARD  TO  THIS  SOFTWARE  INCLUDING  ALL  IMPLIED  WARRANTIES  OF
+    MERCHANTABILITY  AND  FITNESS. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR
+    ANY  SPECIAL,  DIRECT,  INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES
+    WHATSOEVER  RESULTING  FROM  LOSS  OF USE, DATA OR PROFITS, WHETHER IN AN
+    ACTION  OF  CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF
+    OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE.
+*/
+//==============================================================================
+
+#pragma once
+
+#include <util/Expected.h>
+
+#include <string>
+
+namespace data::cassandra {
+
+namespace detail {
+struct Settings;
+class Session;
+class Cluster;
+struct Future;
+class FutureWithCallback;
+struct Result;
+class Statement;
+class PreparedStatement;
+struct Batch;
+}  // namespace detail
+
+using Settings = detail::Settings;
+using Future = detail::Future;
+using FutureWithCallback = detail::FutureWithCallback;
+using Result = detail::Result;
+using Statement = detail::Statement;
+using PreparedStatement = detail::PreparedStatement;
+using Batch = detail::Batch;
+
+/**
+ * @brief A strong type wrapper for int32_t
+ *
+ * This is unfortunately needed right now to support uint32_t properly
+ * because clio uses bigint (int64) everywhere except for when one need
+ * to specify LIMIT, which needs an int32 :-/
+ */
+struct Limit
+{
+    int32_t limit;
+};
+
+class Handle;
+class CassandraError;
+
+using MaybeError = util::Expected<void, CassandraError>;
+using ResultOrError = util::Expected<Result, CassandraError>;
+using Error = util::Unexpected<CassandraError>;
+
+}  // namespace data::cassandra

src/data/cassandra/impl/AsyncExecutor.h

@@ -0,0 +1,119 @@
+//------------------------------------------------------------------------------
+/*
+    This file is part of clio: https://github.com/XRPLF/clio
+    Copyright (c) 2023, the clio developers.
+
+    Permission to use, copy, modify, and distribute this software for any
+    purpose with or without fee is hereby granted, provided that the above
+    copyright notice and this permission notice appear in all copies.
+
+    THE  SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL WARRANTIES
+    WITH  REGARD  TO  THIS  SOFTWARE  INCLUDING  ALL  IMPLIED  WARRANTIES  OF
+    MERCHANTABILITY  AND  FITNESS. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR
+    ANY  SPECIAL,  DIRECT,  INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES
+    WHATSOEVER  RESULTING  FROM  LOSS  OF USE, DATA OR PROFITS, WHETHER IN AN
+    ACTION  OF  CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF
+    OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE.
+*/
+//==============================================================================
+
+#pragma once
+
+#include <data/cassandra/Concepts.h>
+#include <data/cassandra/Handle.h>
+#include <data/cassandra/Types.h>
+#include <data/cassandra/impl/RetryPolicy.h>
+#include <util/Expected.h>
+#include <util/log/Logger.h>
+
+#include <boost/asio.hpp>
+
+#include <functional>
+#include <memory>
+
+namespace data::cassandra::detail {
+
+/**
+ * @brief A query executor with a changable retry policy
+ *
+ * Note: this is a bit of an anti-pattern and should be done differently
+ * eventually.
+ *
+ * Currently it's basically a saner implementation of the previous design that
+ * was used in production without much issue but was using raw new/delete and
+ * could leak easily. This version is slightly better but the overall design is
+ * flawed and should be reworked.
+ */
+template <
+    typename StatementType,
+    typename HandleType = Handle,
+    SomeRetryPolicy RetryPolicyType = ExponentialBackoffRetryPolicy>
+class AsyncExecutor : public std::enable_shared_from_this<AsyncExecutor<StatementType, HandleType, RetryPolicyType>>
+{
+    using FutureWithCallbackType = typename HandleType::FutureWithCallbackType;
+    using CallbackType = std::function<void(typename HandleType::ResultOrErrorType)>;
+
+    util::Logger log_{"Backend"};
+
+    StatementType data_;
+    RetryPolicyType retryPolicy_;
+    CallbackType onComplete_;
+
+    // does not exist during initial construction, hence optional
+    std::optional<FutureWithCallbackType> future_;
+    std::mutex mtx_;
+
+public:
+    /**
+     * @brief Create a new instance of the AsyncExecutor and execute it.
+     */
+    static void
+    run(boost::asio::io_context& ioc, HandleType const& handle, StatementType&& data, CallbackType&& onComplete)
+    {
+        // this is a helper that allows us to use std::make_shared below
+        struct EnableMakeShared : public AsyncExecutor<StatementType, HandleType, RetryPolicyType>
+        {
+            EnableMakeShared(boost::asio::io_context& ioc, StatementType&& data, CallbackType&& onComplete)
+                : AsyncExecutor(ioc, std::move(data), std::move(onComplete))
+            {
+            }
+        };
+
+        auto ptr = std::make_shared<EnableMakeShared>(ioc, std::move(data), std::move(onComplete));
+        ptr->execute(handle);
+    }
+
+private:
+    AsyncExecutor(boost::asio::io_context& ioc, StatementType&& data, CallbackType&& onComplete)
+        : data_{std::move(data)}, retryPolicy_{ioc}, onComplete_{std::move(onComplete)}
+    {
+    }
+
+    void
+    execute(HandleType const& handle)
+    {
+        auto self = this->shared_from_this();
+
+        // lifetime is extended by capturing self ptr
+        auto handler = [this, &handle, self](auto&& res) mutable {
+            if (res)
+            {
+                onComplete_(std::move(res));
+            }
+            else
+            {
+                if (retryPolicy_.shouldRetry(res.error()))
+                    retryPolicy_.retry([self, &handle]() { self->execute(handle); });
+                else
+                    onComplete_(std::move(res));  // report error
+            }
+
+            self = nullptr;  // explicitly decrement refcount
+        };
+
+        std::scoped_lock lck{mtx_};
+        future_.emplace(handle.asyncExecute(data_, std::move(handler)));
+    }
+};
+
+}  // namespace data::cassandra::detail

src/data/cassandra/impl/Batch.cpp

@@ -0,0 +1,55 @@
+//------------------------------------------------------------------------------
+/*
+    This file is part of clio: https://github.com/XRPLF/clio
+    Copyright (c) 2023, the clio developers.
+
+    Permission to use, copy, modify, and distribute this software for any
+    purpose with or without fee is hereby granted, provided that the above
+    copyright notice and this permission notice appear in all copies.
+
+    THE  SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL WARRANTIES
+    WITH  REGARD  TO  THIS  SOFTWARE  INCLUDING  ALL  IMPLIED  WARRANTIES  OF
+    MERCHANTABILITY  AND  FITNESS. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR
+    ANY  SPECIAL,  DIRECT,  INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES
+    WHATSOEVER  RESULTING  FROM  LOSS  OF USE, DATA OR PROFITS, WHETHER IN AN
+    ACTION  OF  CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF
+    OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE.
+*/
+//==============================================================================
+
+#include <data/cassandra/Error.h>
+#include <data/cassandra/impl/Batch.h>
+#include <data/cassandra/impl/Statement.h>
+#include <util/Expected.h>
+
+#include <exception>
+#include <vector>
+
+namespace {
+static constexpr auto batchDeleter = [](CassBatch* ptr) { cass_batch_free(ptr); };
+};
+
+namespace data::cassandra::detail {
+
+// TODO: Use an appropriate value instead of CASS_BATCH_TYPE_LOGGED for different use cases
+Batch::Batch(std::vector<Statement> const& statements)
+    : ManagedObject{cass_batch_new(CASS_BATCH_TYPE_LOGGED), batchDeleter}
+{
+    cass_batch_set_is_idempotent(*this, cass_true);
+
+    for (auto const& statement : statements)
+        if (auto const res = add(statement); not res)
+            throw std::runtime_error("Failed to add statement to batch: " + res.error());
+}
+
+MaybeError
+Batch::add(Statement const& statement)
+{
+    if (auto const rc = cass_batch_add_statement(*this, statement); rc != CASS_OK)
+    {
+        return Error{CassandraError{cass_error_desc(rc), rc}};
+    }
+    return {};
+}
+
+}  // namespace data::cassandra::detail

src/data/cassandra/impl/Batch.h

@@ -0,0 +1,37 @@
+//------------------------------------------------------------------------------
+/*
+    This file is part of clio: https://github.com/XRPLF/clio
+    Copyright (c) 2023, the clio developers.
+
+    Permission to use, copy, modify, and distribute this software for any
+    purpose with or without fee is hereby granted, provided that the above
+    copyright notice and this permission notice appear in all copies.
+
+    THE  SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL WARRANTIES
+    WITH  REGARD  TO  THIS  SOFTWARE  INCLUDING  ALL  IMPLIED  WARRANTIES  OF
+    MERCHANTABILITY  AND  FITNESS. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR
+    ANY  SPECIAL,  DIRECT,  INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES
+    WHATSOEVER  RESULTING  FROM  LOSS  OF USE, DATA OR PROFITS, WHETHER IN AN
+    ACTION  OF  CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF
+    OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE.
+*/
+//==============================================================================
+
+#pragma once
+
+#include <data/cassandra/Types.h>
+#include <data/cassandra/impl/ManagedObject.h>
+
+#include <cassandra.h>
+
+namespace data::cassandra::detail {
+
+struct Batch : public ManagedObject<CassBatch>
+{
+    Batch(std::vector<Statement> const& statements);
+
+    MaybeError
+    add(Statement const& statement);
+};
+
+}  // namespace data::cassandra::detail

src/data/cassandra/impl/Cluster.cpp

@@ -0,0 +1,153 @@
+//------------------------------------------------------------------------------
+/*
+    This file is part of clio: https://github.com/XRPLF/clio
+    Copyright (c) 2023, the clio developers.
+
+    Permission to use, copy, modify, and distribute this software for any
+    purpose with or without fee is hereby granted, provided that the above
+    copyright notice and this permission notice appear in all copies.
+
+    THE  SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL WARRANTIES
+    WITH  REGARD  TO  THIS  SOFTWARE  INCLUDING  ALL  IMPLIED  WARRANTIES  OF
+    MERCHANTABILITY  AND  FITNESS. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR
+    ANY  SPECIAL,  DIRECT,  INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES
+    WHATSOEVER  RESULTING  FROM  LOSS  OF USE, DATA OR PROFITS, WHETHER IN AN
+    ACTION  OF  CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF
+    OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE.
+*/
+//==============================================================================
+
+#include <data/cassandra/impl/Cluster.h>
+#include <data/cassandra/impl/SslContext.h>
+#include <data/cassandra/impl/Statement.h>
+#include <util/Expected.h>
+
+#include <fmt/core.h>
+
+#include <exception>
+#include <vector>
+
+namespace {
+static constexpr auto clusterDeleter = [](CassCluster* ptr) { cass_cluster_free(ptr); };
+
+template <class... Ts>
+struct overloadSet : Ts...
+{
+    using Ts::operator()...;
+};
+
+// explicit deduction guide (not needed as of C++20, but clang be clang)
+template <class... Ts>
+overloadSet(Ts...) -> overloadSet<Ts...>;
+};  // namespace
+
+namespace data::cassandra::detail {
+
+Cluster::Cluster(Settings const& settings) : ManagedObject{cass_cluster_new(), clusterDeleter}
+{
+    using std::to_string;
+
+    cass_cluster_set_token_aware_routing(*this, cass_true);
+    if (auto const rc = cass_cluster_set_protocol_version(*this, CASS_PROTOCOL_VERSION_V4); rc != CASS_OK)
+    {
+        throw std::runtime_error(
+            fmt::format("Error setting cassandra protocol version to v4: {}", cass_error_desc(rc)));
+    }
+
+    if (auto const rc = cass_cluster_set_num_threads_io(*this, settings.threads); rc != CASS_OK)
+    {
+        throw std::runtime_error(
+            fmt::format("Error setting cassandra io threads to {}: {}", settings.threads, cass_error_desc(rc)));
+    }
+
+    cass_log_set_level(settings.enableLog ? CASS_LOG_TRACE : CASS_LOG_DISABLED);
+    cass_cluster_set_connect_timeout(*this, settings.connectionTimeout.count());
+    cass_cluster_set_request_timeout(*this, settings.requestTimeout.count());
+
+    if (auto const rc = cass_cluster_set_core_connections_per_host(*this, settings.coreConnectionsPerHost);
+        rc != CASS_OK)
+    {
+        throw std::runtime_error(fmt::format("Could not set core connections per host: {}", cass_error_desc(rc)));
+    }
+
+    auto const queueSize =
+        settings.queueSizeIO.value_or(settings.maxWriteRequestsOutstanding + settings.maxReadRequestsOutstanding);
+    if (auto const rc = cass_cluster_set_queue_size_io(*this, queueSize); rc != CASS_OK)
+    {
+        throw std::runtime_error(fmt::format("Could not set queue size for IO per host: {}", cass_error_desc(rc)));
+    }
+
+    setupConnection(settings);
+    setupCertificate(settings);
+    setupCredentials(settings);
+
+    LOG(log_.info()) << "Threads: " << settings.threads;
+    LOG(log_.info()) << "Core connections per host: " << settings.coreConnectionsPerHost;
+    LOG(log_.info()) << "IO queue size: " << queueSize;
+}
+
+void
+Cluster::setupConnection(Settings const& settings)
+{
+    std::visit(
+        overloadSet{
+            [this](Settings::ContactPoints const& points) { setupContactPoints(points); },
+            [this](Settings::SecureConnectionBundle const& bundle) { setupSecureBundle(bundle); }},
+        settings.connectionInfo);
+}
+
+void
+Cluster::setupContactPoints(Settings::ContactPoints const& points)
+{
+    using std::to_string;
+    auto throwErrorIfNeeded = [](CassError rc, std::string const& label, std::string const& value) {
+        if (rc != CASS_OK)
+            throw std::runtime_error(
+                fmt::format("Cassandra: Error setting {} [{}]: {}", label, value, cass_error_desc(rc)));
+    };
+
+    {
+        LOG(log_.debug()) << "Attempt connection using contact points: " << points.contactPoints;
+        auto const rc = cass_cluster_set_contact_points(*this, points.contactPoints.data());
+        throwErrorIfNeeded(rc, "contact_points", points.contactPoints);
+    }
+
+    if (points.port)
+    {
+        auto const rc = cass_cluster_set_port(*this, points.port.value());
+        throwErrorIfNeeded(rc, "port", to_string(points.port.value()));
+    }
+}
+
+void
+Cluster::setupSecureBundle(Settings::SecureConnectionBundle const& bundle)
+{
+    LOG(log_.debug()) << "Attempt connection using secure bundle";
+    if (auto const rc = cass_cluster_set_cloud_secure_connection_bundle(*this, bundle.bundle.data()); rc != CASS_OK)
+    {
+        throw std::runtime_error("Failed to connect using secure connection bundle " + bundle.bundle);
+    }
+}
+
+void
+Cluster::setupCertificate(Settings const& settings)
+{
+    if (not settings.certificate)
+        return;
+
+    LOG(log_.debug()) << "Configure SSL context";
+    SslContext context = SslContext(*settings.certificate);
+    cass_cluster_set_ssl(*this, context);
+}
+
+void
+Cluster::setupCredentials(Settings const& settings)
+{
+    if (not settings.username || not settings.password)
+        return;
+
+    LOG(log_.debug()) << "Set credentials; username: " << settings.username.value();
+    cass_cluster_set_credentials(*this, settings.username.value().c_str(), settings.password.value().c_str());
+}
+
+}  // namespace data::cassandra::detail

src/data/cassandra/impl/Cluster.h

@@ -0,0 +1,141 @@
+//------------------------------------------------------------------------------
+/*
+    This file is part of clio: https://github.com/XRPLF/clio
+    Copyright (c) 2023, the clio developers.
+
+    Permission to use, copy, modify, and distribute this software for any
+    purpose with or without fee is hereby granted, provided that the above
+    copyright notice and this permission notice appear in all copies.
+
+    THE  SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL WARRANTIES
+    WITH  REGARD  TO  THIS  SOFTWARE  INCLUDING  ALL  IMPLIED  WARRANTIES  OF
+    MERCHANTABILITY  AND  FITNESS. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR
+    ANY  SPECIAL,  DIRECT,  INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES
+    WHATSOEVER  RESULTING  FROM  LOSS  OF USE, DATA OR PROFITS, WHETHER IN AN
+    ACTION  OF  CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF
+    OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE.
+*/
+//==============================================================================
+
+#pragma once
+
+#include <data/cassandra/impl/ManagedObject.h>
+#include <util/log/Logger.h>
+
+#include <cassandra.h>
+
+#include <chrono>
+#include <optional>
+#include <string>
+#include <string_view>
+#include <thread>
+#include <variant>
+
+namespace data::cassandra::detail {
+
+// TODO: move Settings to public interface, not detail
+
+/**
+ * @brief Bundles all cassandra settings in one place.
+ */
+struct Settings
+{
+    /**
+     * @brief Represents the configuration of contact points for cassandra.
+     */
+    struct ContactPoints
+    {
+        std::string contactPoints = "127.0.0.1";  // defaults to localhost
+        std::optional<uint16_t> port = {};
+    };
+
+    /**
+     * @brief Represents the configuration of a secure connection bundle.
+     */
+    struct SecureConnectionBundle
+    {
+        std::string bundle;  // no meaningful default
+    };
+
+    /** @brief Enables or disables cassandra driver logger */
+    bool enableLog = false;
+
+    /** @brief Connect timeout specified in milliseconds */
+    std::chrono::milliseconds connectionTimeout = std::chrono::milliseconds{10000};
+
+    /** @brief Request timeout specified in milliseconds */
+    std::chrono::milliseconds requestTimeout = std::chrono::milliseconds{0};  // no timeout at all
+
+    /** @brief Connection information; either ContactPoints or SecureConnectionBundle */
+    std::variant<ContactPoints, SecureConnectionBundle> connectionInfo = ContactPoints{};
+
+    /** @brief The number of threads for the driver to pool */
+    uint32_t threads = std::thread::hardware_concurrency();
+
+    /** @brief The maximum number of outstanding write requests at any given moment */
+    uint32_t maxWriteRequestsOutstanding = 10'000u;
+
+    /** @brief The maximum number of outstanding read requests at any given moment */
+    uint32_t maxReadRequestsOutstanding = 100'000u;
+
+    /** @brief The number of connection per host to always have active */
+    uint32_t coreConnectionsPerHost = 1u;
+
+    /** @brief Size of the IO queue */
+    std::optional<uint32_t> queueSizeIO;
+
+    /** @brief SSL certificate */
+    std::optional<std::string> certificate;  // ssl context
+
+    /** @brief Username/login */
+    std::optional<std::string> username;
+
+    /** @brief Password to match the `username` */
+    std::optional<std::string> password;
+
+    /**
+     * @brief Creates a new Settings object as a copy of the current one with overridden contact points.
+     */
+    Settings
+    withContactPoints(std::string_view contactPoints)
+    {
+        auto tmp = *this;
+        tmp.connectionInfo = ContactPoints{std::string{contactPoints}};
+        return tmp;
+    }
+
+    /**
+     * @brief Returns the default settings.
+     */
+    static Settings
+    defaultSettings()
+    {
+        return Settings();
+    }
+};
+
+class Cluster : public ManagedObject<CassCluster>
+{
+    util::Logger log_{"Backend"};
+
+public:
+    Cluster(Settings const& settings);
+
+private:
+    void
+    setupConnection(Settings const& settings);
+
+    void
+    setupContactPoints(Settings::ContactPoints const& points);
+
+    void
+    setupSecureBundle(Settings::SecureConnectionBundle const& bundle);
+
+    void
+    setupCertificate(Settings const& settings);
+
+    void
+    setupCredentials(Settings const& settings);
+};
+
+}  // namespace data::cassandra::detail

src/data/cassandra/impl/ExecutionStrategy.h

@@ -0,0 +1,442 @@
+//------------------------------------------------------------------------------
+/*
+    This file is part of clio: https://github.com/XRPLF/clio
+    Copyright (c) 2023, the clio developers.
+
+    Permission to use, copy, modify, and distribute this software for any
+    purpose with or without fee is hereby granted, provided that the above
+    copyright notice and this permission notice appear in all copies.
+
+    THE  SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL WARRANTIES
+    WITH  REGARD  TO  THIS  SOFTWARE  INCLUDING  ALL  IMPLIED  WARRANTIES  OF
+    MERCHANTABILITY  AND  FITNESS. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR
+    ANY  SPECIAL,  DIRECT,  INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES
+    WHATSOEVER  RESULTING  FROM  LOSS  OF USE, DATA OR PROFITS, WHETHER IN AN
+    ACTION  OF  CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF
+    OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE.
+*/
+//==============================================================================
+
+#pragma once
+
+#include <data/cassandra/Handle.h>
+#include <data/cassandra/Types.h>
+#include <data/cassandra/impl/AsyncExecutor.h>
+#include <util/Expected.h>
+#include <util/log/Logger.h>
+
+#include <boost/asio.hpp>
+#include <boost/asio/spawn.hpp>
+
+#include <atomic>
+#include <condition_variable>
+#include <functional>
+#include <memory>
+#include <mutex>
+#include <optional>
+#include <thread>
+
+namespace data::cassandra::detail {
+
+// TODO: this could probably be also moved out of detail and into the main cassandra namespace.
+
+/**
+ * @brief Implements async and sync querying against the cassandra DB with support for throttling.
+ *
+ * Note: A lot of the code that uses yield is repeated below.
+ * This is ok for now because we are hopefully going to be getting rid of it entirely later on.
+ */
+template <typename HandleType = Handle>
+class DefaultExecutionStrategy
+{
+    util::Logger log_{"Backend"};
+
+    std::uint32_t maxWriteRequestsOutstanding_;
+    std::atomic_uint32_t numWriteRequestsOutstanding_ = 0;
+
+    std::uint32_t maxReadRequestsOutstanding_;
+    std::atomic_uint32_t numReadRequestsOutstanding_ = 0;
+
+    std::mutex throttleMutex_;
+    std::condition_variable throttleCv_;
+
+    std::mutex syncMutex_;
+    std::condition_variable syncCv_;
+
+    boost::asio::io_context ioc_;
+    std::optional<boost::asio::io_service::work> work_;
+
+    std::reference_wrapper<HandleType const> handle_;
+    std::thread thread_;
+
+public:
+    using ResultOrErrorType = typename HandleType::ResultOrErrorType;
+    using StatementType = typename HandleType::StatementType;
+    using PreparedStatementType = typename HandleType::PreparedStatementType;
+    using FutureType = typename HandleType::FutureType;
+    using FutureWithCallbackType = typename HandleType::FutureWithCallbackType;
+    using ResultType = typename HandleType::ResultType;
+    using CompletionTokenType = boost::asio::yield_context;
+
+    /**
+     * @param settings The settings to use
+     * @param handle A handle to the cassandra database
+     */
+    DefaultExecutionStrategy(Settings const& settings, HandleType const& handle)
+        : maxWriteRequestsOutstanding_{settings.maxWriteRequestsOutstanding}
+        , maxReadRequestsOutstanding_{settings.maxReadRequestsOutstanding}
+        , work_{ioc_}
+        , handle_{std::cref(handle)}
+        , thread_{[this]() { ioc_.run(); }}
+    {
+        LOG(log_.info()) << "Max write requests outstanding is " << maxWriteRequestsOutstanding_
+                         << "; Max read requests outstanding is " << maxReadRequestsOutstanding_;
+    }
+
+    ~DefaultExecutionStrategy()
+    {
+        work_.reset();
+        ioc_.stop();
+        thread_.join();
+    }
+
+    /**
+     * @brief Wait for all async writes to finish before unblocking.
+     */
+    void
+    sync()
+    {
+        LOG(log_.debug()) << "Waiting to sync all writes...";
+        std::unique_lock<std::mutex> lck(syncMutex_);
+        syncCv_.wait(lck, [this]() { return finishedAllWriteRequests(); });
+        LOG(log_.debug()) << "Sync done.";
+    }
+
+    /**
+     * @return true if outstanding read requests allowance is exhausted; false otherwise
+     */
+    bool
+    isTooBusy() const
+    {
+        return numReadRequestsOutstanding_ >= maxReadRequestsOutstanding_;
+    }
+
+    /**
+     * @brief Blocking query execution used for writing data.
+     *
+     * Retries forever sleeping for 5 milliseconds between attempts.
+     */
+    ResultOrErrorType
+    writeSync(StatementType const& statement)
+    {
+        while (true)
+        {
+            if (auto res = handle_.get().execute(statement); res)
+            {
+                return res;
+            }
+            else
+            {
+                LOG(log_.warn()) << "Cassandra sync write error, retrying: " << res.error();
+                std::this_thread::sleep_for(std::chrono::milliseconds(5));
+            }
+        }
+    }
+
+    /**
+     * @brief Blocking query execution used for writing data.
+     *
+     * Retries forever sleeping for 5 milliseconds between attempts.
+     */
+    template <typename... Args>
+    ResultOrErrorType
+    writeSync(PreparedStatementType const& preparedStatement, Args&&... args)
+    {
+        return writeSync(preparedStatement.bind(std::forward<Args>(args)...));
+    }
+
+    /**
+     * @brief Non-blocking query execution used for writing data.
+     *
+     * Retries forever with retry policy specified by @ref AsyncExecutor
+     *
+     * @param preparedStatement Statement to prepare and execute
+     * @param args Args to bind to the prepared statement
+     * @throw DatabaseTimeout on timeout
+     */
+    template <typename... Args>
+    void
+    write(PreparedStatementType const& preparedStatement, Args&&... args)
+    {
+        auto statement = preparedStatement.bind(std::forward<Args>(args)...);
+        incrementOutstandingRequestCount();
+
+        // Note: lifetime is controlled by std::shared_from_this internally
+        AsyncExecutor<std::decay_t<decltype(statement)>, HandleType>::run(
+            ioc_, handle_, std::move(statement), [this](auto const&) { decrementOutstandingRequestCount(); });
+    }
+
+    /**
+     * @brief Non-blocking batched query execution used for writing data.
+     *
+     * Retries forever with retry policy specified by @ref AsyncExecutor.
+     *
+     * @param statements Vector of statements to execute as a batch
+     * @throw DatabaseTimeout on timeout
+     */
+    void
+    write(std::vector<StatementType>&& statements)
+    {
+        if (statements.empty())
+            return;
+
+        incrementOutstandingRequestCount();
+
+        // Note: lifetime is controlled by std::shared_from_this internally
+        AsyncExecutor<std::decay_t<decltype(statements)>, HandleType>::run(
+            ioc_, handle_, std::move(statements), [this](auto const&) { decrementOutstandingRequestCount(); });
+    }
+
+    /**
+     * @brief Coroutine-based query execution used for reading data.
+     *
+     * Retries forever until successful or throws an exception on timeout.
+     *
+     * @param token Completion token (yield_context)
+     * @param preparedStatement Statement to prepare and execute
+     * @param args Args to bind to the prepared statement
+     * @throw DatabaseTimeout on timeout
+     * @return ResultType or error wrapped in Expected
+     */
+    template <typename... Args>
+    [[maybe_unused]] ResultOrErrorType
+    read(CompletionTokenType token, PreparedStatementType const& preparedStatement, Args&&... args)
+    {
+        return read(token, preparedStatement.bind(std::forward<Args>(args)...));
+    }
+
+    /**
+     * @brief Coroutine-based query execution used for reading data.
+     *
+     * Retries forever until successful or throws an exception on timeout.
+     *
+     * @param token Completion token (yield_context)
+     * @param statements Statements to execute in a batch
+     * @throw DatabaseTimeout on timeout
+     * @return ResultType or error wrapped in Expected
+     */
+    [[maybe_unused]] ResultOrErrorType
+    read(CompletionTokenType token, std::vector<StatementType> const& statements)
+    {
+        auto const numStatements = statements.size();
+        std::optional<FutureWithCallbackType> future;
+
+        // todo: perhaps use policy instead
+        while (true)
+        {
+            numReadRequestsOutstanding_ += numStatements;
+
+            auto init = [this, &statements, &future]<typename Self>(Self& self) {
+                auto sself = std::make_shared<Self>(std::move(self));
+
+                future.emplace(handle_.get().asyncExecute(statements, [sself](auto&& res) mutable {
+                    boost::asio::post(
+                        boost::asio::get_associated_executor(*sself),
+                        [sself, res = std::move(res)]() mutable { sself->complete(std::move(res)); });
+                }));
+            };
+
+            auto res = boost::asio::async_compose<CompletionTokenType, void(ResultOrErrorType)>(
+                init, token, boost::asio::get_associated_executor(token));
+            numReadRequestsOutstanding_ -= numStatements;
+
+            if (res)
+            {
+                return res;
+            }
+            else
+            {
+                LOG(log_.error()) << "Failed batch read in coroutine: " << res.error();
+                throwErrorIfNeeded(res.error());
+            }
+        }
+    }
+
+    /**
+     * @brief Coroutine-based query execution used for reading data.
+     *
+     * Retries forever until successful or throws an exception on timeout.
+     *
+     * @param token Completion token (yield_context)
+     * @param statement Statement to execute
+     * @throw DatabaseTimeout on timeout
+     * @return ResultType or error wrapped in Expected
+     */
+    [[maybe_unused]] ResultOrErrorType
+    read(CompletionTokenType token, StatementType const& statement)
+    {
+        std::optional<FutureWithCallbackType> future;
+
+        // todo: perhaps use policy instead
+        while (true)
+        {
+            ++numReadRequestsOutstanding_;
+            auto init = [this, &statement, &future]<typename Self>(Self& self) {
+                auto sself = std::make_shared<Self>(std::move(self));
+
+                future.emplace(handle_.get().asyncExecute(statement, [sself](auto&& res) mutable {
+                    boost::asio::post(
+                        boost::asio::get_associated_executor(*sself),
+                        [sself, res = std::move(res)]() mutable { sself->complete(std::move(res)); });
+                }));
+            };
+
+            auto res = boost::asio::async_compose<CompletionTokenType, void(ResultOrErrorType)>(
+                init, token, boost::asio::get_associated_executor(token));
+            --numReadRequestsOutstanding_;
+
+            if (res)
+            {
+                return res;
+            }
+            else
+            {
+                LOG(log_.error()) << "Failed read in coroutine: " << res.error();
+                throwErrorIfNeeded(res.error());
+            }
+        }
+    }
+
+    /**
+     * @brief Coroutine-based query execution used for reading data.
+     *
+     * Attempts to execute each statement. On any error the whole vector will be
+     * discarded and exception will be thrown.
+     *
+     * @param token Completion token (yield_context)
+     * @param statements Statements to execute
+     * @throw DatabaseTimeout on db error
+     * @return Vector of results
+     */
+    std::vector<ResultType>
+    readEach(CompletionTokenType token, std::vector<StatementType> const& statements)
+    {
+        std::atomic_bool hadError = false;
+        std::atomic_int numOutstanding = statements.size();
+        numReadRequestsOutstanding_ += statements.size();
+
+        auto futures = std::vector<FutureWithCallbackType>{};
+        futures.reserve(numOutstanding);
+
+        auto init = [this, &statements, &futures, &hadError, &numOutstanding]<typename Self>(Self& self) {
+            auto sself = std::make_shared<Self>(std::move(self));
+            auto executionHandler = [&hadError, &numOutstanding, sself](auto const& res) mutable {
+                if (not res)
+                    hadError = true;
+
+                // when all async operations complete unblock the result
+                if (--numOutstanding == 0)
+                    boost::asio::post(
+                        boost::asio::get_associated_executor(*sself), [sself]() mutable { sself->complete(); });
+            };
+
+            std::transform(
+                std::cbegin(statements),
+                std::cend(statements),
+                std::back_inserter(futures),
+                [this, &executionHandler](auto const& statement) {
+                    return handle_.get().asyncExecute(statement, executionHandler);
+                });
+        };
+
+        boost::asio::async_compose<CompletionTokenType, void()>(
+            init, token, boost::asio::get_associated_executor(token));
+        numReadRequestsOutstanding_ -= statements.size();
+
+        if (hadError)
+            throw DatabaseTimeout{};
+
+        std::vector<ResultType> results;
+        results.reserve(futures.size());
+
+        // it's safe to call blocking get on futures here as we already waited for the coroutine to resume above.
+        std::transform(
+            std::make_move_iterator(std::begin(futures)),
+            std::make_move_iterator(std::end(futures)),
+            std::back_inserter(results),
+            [](auto&& future) {
+                auto entry = future.get();
+                auto&& res = entry.value();
+                return std::move(res);
+            });
+
+        assert(futures.size() == statements.size());
+        assert(results.size() == statements.size());
+        return results;
+    }
+
+private:
+    void
+    incrementOutstandingRequestCount()
+    {
+        {
+            std::unique_lock<std::mutex> lck(throttleMutex_);
+            if (!canAddWriteRequest())
+            {
+                LOG(log_.trace()) << "Max outstanding requests reached. "
+                                  << "Waiting for other requests to finish";
+                throttleCv_.wait(lck, [this]() { return canAddWriteRequest(); });
+            }
+        }
+        ++numWriteRequestsOutstanding_;
+    }
+
+    void
+    decrementOutstandingRequestCount()
+    {
+        // sanity check
+        if (numWriteRequestsOutstanding_ == 0)
+        {
+            assert(false);
+            throw std::runtime_error("decrementing num outstanding below 0");
+        }
+        size_t cur = (--numWriteRequestsOutstanding_);
+        {
+            // mutex lock required to prevent race condition around spurious
+            // wakeup
+            std::lock_guard lck(throttleMutex_);
+            throttleCv_.notify_one();
+        }
+        if (cur == 0)
+        {
+            // mutex lock required to prevent race condition around spurious
+            // wakeup
+            std::lock_guard lck(syncMutex_);
+            syncCv_.notify_one();
+        }
+    }
+
+    bool
+    canAddWriteRequest() const
+    {
+        return numWriteRequestsOutstanding_ < maxWriteRequestsOutstanding_;
+    }
+
+    bool
+    finishedAllWriteRequests() const
+    {
+        return numWriteRequestsOutstanding_ == 0;
+    }
+
+    void
+    throwErrorIfNeeded(CassandraError err) const
+    {
+        if (err.isTimeout())
+            throw DatabaseTimeout();
+
+        if (err.isInvalidQuery())
+            throw std::runtime_error("Invalid query");
+    }
+};
+
+}  // namespace data::cassandra::detail

src/data/cassandra/impl/Future.cpp

@@ -0,0 +1,105 @@
+//------------------------------------------------------------------------------
+/*
+    This file is part of clio: https://github.com/XRPLF/clio
+    Copyright (c) 2023, the clio developers.
+
+    Permission to use, copy, modify, and distribute this software for any
+    purpose with or without fee is hereby granted, provided that the above
+    copyright notice and this permission notice appear in all copies.
+
+    THE  SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL WARRANTIES
+    WITH  REGARD  TO  THIS  SOFTWARE  INCLUDING  ALL  IMPLIED  WARRANTIES  OF
+    MERCHANTABILITY  AND  FITNESS. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR
+    ANY  SPECIAL,  DIRECT,  INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES
+    WHATSOEVER  RESULTING  FROM  LOSS  OF USE, DATA OR PROFITS, WHETHER IN AN
+    ACTION  OF  CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF
+    OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE.
+*/
+//==============================================================================
+
+#include <data/cassandra/Error.h>
+#include <data/cassandra/impl/Future.h>
+#include <data/cassandra/impl/Result.h>
+
+#include <exception>
+#include <vector>
+
+namespace {
+static constexpr auto futureDeleter = [](CassFuture* ptr) { cass_future_free(ptr); };
+}  // namespace
+
+namespace data::cassandra::detail {
+
+/* implicit */ Future::Future(CassFuture* ptr) : ManagedObject{ptr, futureDeleter}
+{
+}
+
+MaybeError
+Future::await() const
+{
+    if (auto const rc = cass_future_error_code(*this); rc)
+    {
+        auto errMsg = [this](std::string const& label) {
+            char const* message;
+            std::size_t len;
+            cass_future_error_message(*this, &message, &len);
+            return label + ": " + std::string{message, len};
+        }(cass_error_desc(rc));
+        return Error{CassandraError{errMsg, rc}};
+    }
+    return {};
+}
+
+ResultOrError
+Future::get() const
+{
+    if (auto const rc = cass_future_error_code(*this); rc)
+    {
+        auto const errMsg = [this](std::string const& label) {
+            char const* message;
+            std::size_t len;
+            cass_future_error_message(*this, &message, &len);
+            return label + ": " + std::string{message, len};
+        }("future::get()");
+        return Error{CassandraError{errMsg, rc}};
+    }
+    else
+    {
+        return Result{cass_future_get_result(*this)};
+    }
+}
+
+void
+invokeHelper(CassFuture* ptr, void* cbPtr)
+{
+    // Note: can't use Future{ptr}.get() because double free will occur :/
+    // Note2: we are moving/copying it locally as a workaround for an issue we are seeing from asio recently.
+    // stackoverflow.com/questions/77004137/boost-asio-async-compose-gets-stuck-under-load
+    auto* cb = static_cast<FutureWithCallback::FnType*>(cbPtr);
+    auto local = std::make_unique<FutureWithCallback::FnType>(std::move(*cb));
+    if (auto const rc = cass_future_error_code(ptr); rc)
+    {
+        auto const errMsg = [&ptr](std::string const& label) {
+            char const* message;
+            std::size_t len;
+            cass_future_error_message(ptr, &message, &len);
+            return label + ": " + std::string{message, len};
+        }("invokeHelper");
+        (*local)(Error{CassandraError{errMsg, rc}});
+    }
+    else
+    {
+        (*local)(Result{cass_future_get_result(ptr)});
+    }
+}
+
+/* implicit */ FutureWithCallback::FutureWithCallback(CassFuture* ptr, FnType&& cb)
+    : Future{ptr}, cb_{std::make_unique<FnType>(std::move(cb))}
+{
+    // Instead of passing `this` as the userdata void*, we pass the address of
+    // the callback itself which will survive std::move of the
+    // FutureWithCallback parent. Not ideal but I have no better solution atm.
+    cass_future_set_callback(*this, &invokeHelper, cb_.get());
+}
+
+}  // namespace data::cassandra::detail

src/data/cassandra/impl/Future.h

@@ -0,0 +1,58 @@
+//------------------------------------------------------------------------------
+/*
+    This file is part of clio: https://github.com/XRPLF/clio
+    Copyright (c) 2023, the clio developers.
+
+    Permission to use, copy, modify, and distribute this software for any
+    purpose with or without fee is hereby granted, provided that the above
+    copyright notice and this permission notice appear in all copies.
+
+    THE  SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL WARRANTIES
+    WITH  REGARD  TO  THIS  SOFTWARE  INCLUDING  ALL  IMPLIED  WARRANTIES  OF
+    MERCHANTABILITY  AND  FITNESS. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR
+    ANY  SPECIAL,  DIRECT,  INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES
+    WHATSOEVER  RESULTING  FROM  LOSS  OF USE, DATA OR PROFITS, WHETHER IN AN
+    ACTION  OF  CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF
+    OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE.
+*/
+//==============================================================================
+
+#pragma once
+
+#include <data/cassandra/Types.h>
+#include <data/cassandra/impl/ManagedObject.h>
+
+#include <cassandra.h>
+
+namespace data::cassandra::detail {
+
+struct Future : public ManagedObject<CassFuture>
+{
+    /* implicit */ Future(CassFuture* ptr);
+
+    MaybeError
+    await() const;
+
+    ResultOrError
+    get() const;
+};
+
+void
+invokeHelper(CassFuture* ptr, void* self);
+
+class FutureWithCallback : public Future
+{
+public:
+    using FnType = std::function<void(ResultOrError)>;
+    using FnPtrType = std::unique_ptr<FnType>;
+
+    /* implicit */ FutureWithCallback(CassFuture* ptr, FnType&& cb);
+    FutureWithCallback(FutureWithCallback const&) = delete;
+    FutureWithCallback(FutureWithCallback&&) = default;
+
+private:
+    /** Wrapped in a unique_ptr so it can survive std::move :/ */
+    FnPtrType cb_;
+};
+
+}  // namespace data::cassandra::detail

src/data/cassandra/impl/ManagedObject.h

@@ -0,0 +1,47 @@
+//------------------------------------------------------------------------------
+/*
+    This file is part of clio: https://github.com/XRPLF/clio
+    Copyright (c) 2023, the clio developers.
+
+    Permission to use, copy, modify, and distribute this software for any
+    purpose with or without fee is hereby granted, provided that the above
+    copyright notice and this permission notice appear in all copies.
+
+    THE  SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL WARRANTIES
+    WITH  REGARD  TO  THIS  SOFTWARE  INCLUDING  ALL  IMPLIED  WARRANTIES  OF
+    MERCHANTABILITY  AND  FITNESS. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR
+    ANY  SPECIAL,  DIRECT,  INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES
+    WHATSOEVER  RESULTING  FROM  LOSS  OF USE, DATA OR PROFITS, WHETHER IN AN
+    ACTION  OF  CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF
+    OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE.
+*/
+//==============================================================================
+
+#pragma once
+
+#include <memory>
+
+namespace data::cassandra::detail {
+
+template <typename Managed>
+class ManagedObject
+{
+protected:
+    std::unique_ptr<Managed, void (*)(Managed*)> ptr_;
+
+public:
+    template <typename deleterCallable>
+    ManagedObject(Managed* rawPtr, deleterCallable deleter) : ptr_{rawPtr, deleter}
+    {
+        if (rawPtr == nullptr)
+            throw std::runtime_error("Could not create DB object - got nullptr");
+    }
+    ManagedObject(ManagedObject&&) = default;
+
+    operator Managed*() const
+    {
+        return ptr_.get();
+    }
+};
+
+}  // namespace data::cassandra::detail

src/data/cassandra/impl/Result.cpp

@@ -0,0 +1,69 @@
+//------------------------------------------------------------------------------
+/*
+    This file is part of clio: https://github.com/XRPLF/clio
+    Copyright (c) 2023, the clio developers.
+
+    Permission to use, copy, modify, and distribute this software for any
+    purpose with or without fee is hereby granted, provided that the above
+    copyright notice and this permission notice appear in all copies.
+
+    THE  SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL WARRANTIES
+    WITH  REGARD  TO  THIS  SOFTWARE  INCLUDING  ALL  IMPLIED  WARRANTIES  OF
+    MERCHANTABILITY  AND  FITNESS. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR
+    ANY  SPECIAL,  DIRECT,  INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES
+    WHATSOEVER  RESULTING  FROM  LOSS  OF USE, DATA OR PROFITS, WHETHER IN AN
+    ACTION  OF  CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF
+    OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE.
+*/
+//==============================================================================
+
+#include <data/cassandra/impl/Result.h>
+
+namespace {
+static constexpr auto resultDeleter = [](CassResult const* ptr) { cass_result_free(ptr); };
+static constexpr auto resultIteratorDeleter = [](CassIterator* ptr) { cass_iterator_free(ptr); };
+}  // namespace
+
+namespace data::cassandra::detail {
+
+/* implicit */ Result::Result(CassResult const* ptr) : ManagedObject{ptr, resultDeleter}
+{
+}
+
+[[nodiscard]] std::size_t
+Result::numRows() const
+{
+    return cass_result_row_count(*this);
+}
+
+[[nodiscard]] bool
+Result::hasRows() const
+{
+    return numRows() > 0;
+}
+
+/* implicit */ ResultIterator::ResultIterator(CassIterator* ptr)
+    : ManagedObject{ptr, resultIteratorDeleter}, hasMore_{cass_iterator_next(ptr)}
+{
+}
+
+[[nodiscard]] ResultIterator
+ResultIterator::fromResult(Result const& result)
+{
+    return {cass_iterator_from_result(result)};
+}
+
+[[maybe_unused]] bool
+ResultIterator::moveForward()
+{
+    hasMore_ = cass_iterator_next(*this);
+    return hasMore_;
+}
+
+[[nodiscard]] bool
+ResultIterator::hasMore() const
+{
+    return hasMore_;
+}
+
+}  // namespace data::cassandra::detail

src/data/cassandra/impl/Result.h

@@ -0,0 +1,257 @@
+//------------------------------------------------------------------------------
+/*
+    This file is part of clio: https://github.com/XRPLF/clio
+    Copyright (c) 2023, the clio developers.
+
+    Permission to use, copy, modify, and distribute this software for any
+    purpose with or without fee is hereby granted, provided that the above
+    copyright notice and this permission notice appear in all copies.
+
+    THE  SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL WARRANTIES
+    WITH  REGARD  TO  THIS  SOFTWARE  INCLUDING  ALL  IMPLIED  WARRANTIES  OF
+    MERCHANTABILITY  AND  FITNESS. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR
+    ANY  SPECIAL,  DIRECT,  INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES
+    WHATSOEVER  RESULTING  FROM  LOSS  OF USE, DATA OR PROFITS, WHETHER IN AN
+    ACTION  OF  CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF
+    OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE.
+*/
+//==============================================================================
+
+#pragma once
+
+#include <data/cassandra/impl/ManagedObject.h>
+#include <data/cassandra/impl/Tuple.h>
+#include <util/Expected.h>
+
+#include <ripple/basics/base_uint.h>
+#include <ripple/protocol/AccountID.h>
+#include <cassandra.h>
+
+#include <compare>
+#include <iterator>
+#include <tuple>
+
+namespace data::cassandra::detail {
+
+template <typename>
+static constexpr bool unsupported_v = false;
+
+template <typename Type>
+inline Type
+extractColumn(CassRow const* row, std::size_t idx)
+{
+    using std::to_string;
+    Type output;
+
+    auto throwErrorIfNeeded = [](CassError rc, std::string_view label) {
+        if (rc != CASS_OK)
+        {
+            auto const tag = '[' + std::string{label} + ']';
+            throw std::logic_error(tag + ": " + cass_error_desc(rc));
+        }
+    };
+
+    using DecayedType = std::decay_t<Type>;
+    using UintTupleType = std::tuple<uint32_t, uint32_t>;
+    using UCharVectorType = std::vector<unsigned char>;
+
+    if constexpr (std::is_same_v<DecayedType, ripple::uint256>)
+    {
+        cass_byte_t const* buf;
+        std::size_t bufSize;
+        auto const rc = cass_value_get_bytes(cass_row_get_column(row, idx), &buf, &bufSize);
+        throwErrorIfNeeded(rc, "Extract ripple::uint256");
+        output = ripple::uint256::fromVoid(buf);
+    }
+    else if constexpr (std::is_same_v<DecayedType, ripple::AccountID>)
+    {
+        cass_byte_t const* buf;
+        std::size_t bufSize;
+        auto const rc = cass_value_get_bytes(cass_row_get_column(row, idx), &buf, &bufSize);
+        throwErrorIfNeeded(rc, "Extract ripple::AccountID");
+        output = ripple::AccountID::fromVoid(buf);
+    }
+    else if constexpr (std::is_same_v<DecayedType, UCharVectorType>)
+    {
+        cass_byte_t const* buf;
+        std::size_t bufSize;
+        auto const rc = cass_value_get_bytes(cass_row_get_column(row, idx), &buf, &bufSize);
+        throwErrorIfNeeded(rc, "Extract vector<unsigned char>");
+        output = UCharVectorType{buf, buf + bufSize};
+    }
+    else if constexpr (std::is_same_v<DecayedType, UintTupleType>)
+    {
+        auto const* tuple = cass_row_get_column(row, idx);
+        output = TupleIterator::fromTuple(tuple).extract<uint32_t, uint32_t>();
+    }
+    else if constexpr (std::is_convertible_v<DecayedType, std::string>)
+    {
+        char const* value;
+        std::size_t len;
+        auto const rc = cass_value_get_string(cass_row_get_column(row, idx), &value, &len);
+        throwErrorIfNeeded(rc, "Extract string");
+        output = std::string{value, len};
+    }
+    else if constexpr (std::is_same_v<DecayedType, bool>)
+    {
+        cass_bool_t flag;
+        auto const rc = cass_value_get_bool(cass_row_get_column(row, idx), &flag);
+        throwErrorIfNeeded(rc, "Extract bool");
+        output = flag ? true : false;
+    }
+    // clio only uses bigint (int64_t) so we convert any incoming type
+    else if constexpr (std::is_convertible_v<DecayedType, int64_t>)
+    {
+        int64_t out;
+        auto const rc = cass_value_get_int64(cass_row_get_column(row, idx), &out);
+        throwErrorIfNeeded(rc, "Extract int64");
+        output = static_cast<DecayedType>(out);
+    }
+    else
+    {
+        // type not supported for extraction
+        static_assert(unsupported_v<DecayedType>);
+    }
+
+    return output;
+}
+
+struct Result : public ManagedObject<CassResult const>
+{
+    /* implicit */ Result(CassResult const* ptr);
+
+    [[nodiscard]] std::size_t
+    numRows() const;
+
+    [[nodiscard]] bool
+    hasRows() const;
+
+    template <typename... RowTypes>
+    std::optional<std::tuple<RowTypes...>>
+    get() const requires(std::tuple_size<std::tuple<RowTypes...>>{} > 1)
+    {
+        // row managed internally by cassandra driver, hence no ManagedObject.
+        auto const* row = cass_result_first_row(*this);
+        if (row == nullptr)
+            return std::nullopt;
+
+        std::size_t idx = 0;
+        auto advanceId = [&idx]() { return idx++; };
+
+        return std::make_optional<std::tuple<RowTypes...>>({extractColumn<RowTypes>(row, advanceId())...});
+    }
+
+    template <typename RowType>
+    std::optional<RowType>
+    get() const
+    {
+        // row managed internally by cassandra driver, hence no ManagedObject.
+        auto const* row = cass_result_first_row(*this);
+        if (row == nullptr)
+            return std::nullopt;
+        return std::make_optional<RowType>(extractColumn<RowType>(row, 0));
+    }
+};
+
+class ResultIterator : public ManagedObject<CassIterator>
+{
+    bool hasMore_ = false;
+
+public:
+    /* implicit */ ResultIterator(CassIterator* ptr);
+
+    [[nodiscard]] static ResultIterator
+    fromResult(Result const& result);
+
+    [[maybe_unused]] bool
+    moveForward();
+
+    [[nodiscard]] bool
+    hasMore() const;
+
+    template <typename... RowTypes>
+    std::tuple<RowTypes...>
+    extractCurrentRow() const
+    {
+        // note: row is invalidated on each iteration.
+        // managed internally by cassandra driver, hence no ManagedObject.
+        auto const* row = cass_iterator_get_row(*this);
+
+        std::size_t idx = 0;
+        auto advanceId = [&idx]() { return idx++; };
+
+        return {extractColumn<RowTypes>(row, advanceId())...};
+    }
+};
+
+template <typename... Types>
+class ResultExtractor
+{
+    std::reference_wrapper<Result const> ref_;
+
+public:
+    struct Sentinel
+    {
+    };
+
+    struct Iterator
+    {
+        using iterator_category = std::input_iterator_tag;
+        using difference_type = std::size_t;  // rows count
+        using value_type = std::tuple<Types...>;
+
+        /* implicit */ Iterator(ResultIterator iterator) : iterator_{std::move(iterator)}
+        {
+        }
+
+        Iterator(Iterator const&) = delete;
+        Iterator&
+        operator=(Iterator const&) = delete;
+
+        value_type
+        operator*() const
+        {
+            return iterator_.extractCurrentRow<Types...>();
+        }
+
+        value_type
+        operator->()
+        {
+            return iterator_.extractCurrentRow<Types...>();
+        }
+
+        Iterator&
+        operator++()
+        {
+            iterator_.moveForward();
+            return *this;
+        }
+
+        bool
+        operator==(Sentinel const&) const
+        {
+            return not iterator_.hasMore();
+        }
+
+    private:
+        ResultIterator iterator_;
+    };
+
+    ResultExtractor(Result const& result) : ref_{std::cref(result)}
+    {
+    }
+
+    Iterator
+    begin()
+    {
+        return ResultIterator::fromResult(ref_);
+    }
+
+    Sentinel
+    end()
+    {
+        return {};
+    }
+};
+
+}  // namespace data::cassandra::detail

src/data/cassandra/impl/RetryPolicy.h

@@ -0,0 +1,94 @@
+//------------------------------------------------------------------------------
+/*
+    This file is part of clio: https://github.com/XRPLF/clio
+    Copyright (c) 2023, the clio developers.
+
+    Permission to use, copy, modify, and distribute this software for any
+    purpose with or without fee is hereby granted, provided that the above
+    copyright notice and this permission notice appear in all copies.
+
+    THE  SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL WARRANTIES
+    WITH  REGARD  TO  THIS  SOFTWARE  INCLUDING  ALL  IMPLIED  WARRANTIES  OF
+    MERCHANTABILITY  AND  FITNESS. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR
+    ANY  SPECIAL,  DIRECT,  INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES
+    WHATSOEVER  RESULTING  FROM  LOSS  OF USE, DATA OR PROFITS, WHETHER IN AN
+    ACTION  OF  CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF
+    OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE.
+*/
+//==============================================================================
+
+#pragma once
+
+#include <data/cassandra/Handle.h>
+#include <data/cassandra/Types.h>
+#include <util/Expected.h>
+#include <util/log/Logger.h>
+
+#include <boost/asio.hpp>
+
+#include <algorithm>
+#include <chrono>
+#include <cmath>
+
+namespace data::cassandra::detail {
+
+/**
+ * @brief A retry policy that employs exponential backoff
+ */
+class ExponentialBackoffRetryPolicy
+{
+    util::Logger log_{"Backend"};
+
+    boost::asio::steady_timer timer_;
+    uint32_t attempt_ = 0u;
+
+public:
+    /**
+     * @brief Create a new retry policy instance with the io_context provided
+     */
+    ExponentialBackoffRetryPolicy(boost::asio::io_context& ioc) : timer_{boost::asio::make_strand(ioc)}
+    {
+    }
+
+    /**
+     * @brief Computes next retry delay and returns true unconditionally
+     *
+     * @param err The cassandra error that triggered the retry
+     */
+    [[nodiscard]] bool
+    shouldRetry([[maybe_unused]] CassandraError err)
+    {
+        auto const delay = calculateDelay(attempt_);
+        LOG(log_.error()) << "Cassandra write error: " << err << ", current retries " << attempt_ << ", retrying in "
+                          << delay.count() << " milliseconds";
+
+        return true;  // keep retrying forever
+    }
+
+    /**
+     * @brief Schedules next retry
+     *
+     * @param fn The callable to execute
+     */
+    template <typename Fn>
+    void
+    retry(Fn&& fn)
+    {
+        timer_.expires_after(calculateDelay(attempt_++));
+        timer_.async_wait([fn = std::forward<Fn>(fn)]([[maybe_unused]] const auto& err) {
+            // todo: deal with cancellation (thru err)
+            fn();
+        });
+    }
+
+    /**
+     * @brief Calculates the wait time before attempting another retry
+     */
+    std::chrono::milliseconds
+    calculateDelay(uint32_t attempt)
+    {
+        return std::chrono::milliseconds{lround(std::pow(2, std::min(10u, attempt)))};
+    }
+};
+
+}  // namespace data::cassandra::detail

src/data/cassandra/impl/Session.h

@@ -0,0 +1,38 @@
+//------------------------------------------------------------------------------
+/*
+    This file is part of clio: https://github.com/XRPLF/clio
+    Copyright (c) 2023, the clio developers.
+
+    Permission to use, copy, modify, and distribute this software for any
+    purpose with or without fee is hereby granted, provided that the above
+    copyright notice and this permission notice appear in all copies.
+
+    THE  SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL WARRANTIES
+    WITH  REGARD  TO  THIS  SOFTWARE  INCLUDING  ALL  IMPLIED  WARRANTIES  OF
+    MERCHANTABILITY  AND  FITNESS. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR
+    ANY  SPECIAL,  DIRECT,  INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES
+    WHATSOEVER  RESULTING  FROM  LOSS  OF USE, DATA OR PROFITS, WHETHER IN AN
+    ACTION  OF  CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF
+    OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE.
+*/
+//==============================================================================
+
+#pragma once
+
+#include <data/cassandra/impl/ManagedObject.h>
+
+#include <cassandra.h>
+
+namespace data::cassandra::detail {
+
+class Session : public ManagedObject<CassSession>
+{
+    static constexpr auto deleter = [](CassSession* ptr) { cass_session_free(ptr); };
+
+public:
+    Session() : ManagedObject{cass_session_new(), deleter}
+    {
+    }
+};
+
+}  // namespace data::cassandra::detail

src/data/cassandra/impl/SslContext.cpp

@@ -0,0 +1,37 @@
+//------------------------------------------------------------------------------
+/*
+    This file is part of clio: https://github.com/XRPLF/clio
+    Copyright (c) 2023, the clio developers.
+
+    Permission to use, copy, modify, and distribute this software for any
+    purpose with or without fee is hereby granted, provided that the above
+    copyright notice and this permission notice appear in all copies.
+
+    THE  SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL WARRANTIES
+    WITH  REGARD  TO  THIS  SOFTWARE  INCLUDING  ALL  IMPLIED  WARRANTIES  OF
+    MERCHANTABILITY  AND  FITNESS. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR
+    ANY  SPECIAL,  DIRECT,  INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES
+    WHATSOEVER  RESULTING  FROM  LOSS  OF USE, DATA OR PROFITS, WHETHER IN AN
+    ACTION  OF  CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF
+    OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE.
+*/
+//==============================================================================
+
+#include <data/cassandra/impl/SslContext.h>
+
+namespace {
+static constexpr auto contextDeleter = [](CassSsl* ptr) { cass_ssl_free(ptr); };
+}  // namespace
+
+namespace data::cassandra::detail {
+
+SslContext::SslContext(std::string const& certificate) : ManagedObject{cass_ssl_new(), contextDeleter}
+{
+    cass_ssl_set_verify_flags(*this, CASS_SSL_VERIFY_NONE);
+    if (auto const rc = cass_ssl_add_trusted_cert(*this, certificate.c_str()); rc != CASS_OK)
+    {
+        throw std::runtime_error(std::string{"Error setting Cassandra SSL Context: "} + cass_error_desc(rc));
+    }
+}
+
+}  // namespace data::cassandra::detail

src/data/cassandra/impl/SslContext.h

@@ -0,0 +1,35 @@
+//------------------------------------------------------------------------------
+/*
+    This file is part of clio: https://github.com/XRPLF/clio
+    Copyright (c) 2023, the clio developers.
+
+    Permission to use, copy, modify, and distribute this software for any
+    purpose with or without fee is hereby granted, provided that the above
+    copyright notice and this permission notice appear in all copies.
+
+    THE  SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL WARRANTIES
+    WITH  REGARD  TO  THIS  SOFTWARE  INCLUDING  ALL  IMPLIED  WARRANTIES  OF
+    MERCHANTABILITY  AND  FITNESS. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR
+    ANY  SPECIAL,  DIRECT,  INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES
+    WHATSOEVER  RESULTING  FROM  LOSS  OF USE, DATA OR PROFITS, WHETHER IN AN
+    ACTION  OF  CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF
+    OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE.
+*/
+//==============================================================================
+
+#pragma once
+
+#include <data/cassandra/impl/ManagedObject.h>
+
+#include <cassandra.h>
+
+#include <string>
+
+namespace data::cassandra::detail {
+
+struct SslContext : public ManagedObject<CassSsl>
+{
+    explicit SslContext(std::string const& certificate);
+};
+
+}  // namespace data::cassandra::detail

src/data/cassandra/impl/Statement.h

@@ -0,0 +1,186 @@
+//------------------------------------------------------------------------------
+/*
+    This file is part of clio: https://github.com/XRPLF/clio
+    Copyright (c) 2023, the clio developers.
+
+    Permission to use, copy, modify, and distribute this software for any
+    purpose with or without fee is hereby granted, provided that the above
+    copyright notice and this permission notice appear in all copies.
+
+    THE  SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL WARRANTIES
+    WITH  REGARD  TO  THIS  SOFTWARE  INCLUDING  ALL  IMPLIED  WARRANTIES  OF
+    MERCHANTABILITY  AND  FITNESS. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR
+    ANY  SPECIAL,  DIRECT,  INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES
+    WHATSOEVER  RESULTING  FROM  LOSS  OF USE, DATA OR PROFITS, WHETHER IN AN
+    ACTION  OF  CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF
+    OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE.
+*/
+//==============================================================================
+
+#pragma once
+
+#include <data/cassandra/Types.h>
+#include <data/cassandra/impl/ManagedObject.h>
+#include <data/cassandra/impl/Tuple.h>
+#include <util/Expected.h>
+
+#include <ripple/basics/base_uint.h>
+#include <ripple/protocol/STAccount.h>
+#include <cassandra.h>
+#include <fmt/core.h>
+
+#include <chrono>
+#include <compare>
+#include <iterator>
+
+namespace data::cassandra::detail {
+
+class Statement : public ManagedObject<CassStatement>
+{
+    static constexpr auto deleter = [](CassStatement* ptr) { cass_statement_free(ptr); };
+
+    template <typename>
+    static constexpr bool unsupported_v = false;
+
+public:
+    /**
+     * @brief Construct a new statement with optionally provided arguments.
+     *
+     * Note: it's up to the user to make sure the bound parameters match
+     * the format of the query (e.g. amount of '?' matches count of args).
+     */
+    template <typename... Args>
+    explicit Statement(std::string_view query, Args&&... args)
+        : ManagedObject{cass_statement_new(query.data(), sizeof...(args)), deleter}
+    {
+        cass_statement_set_consistency(*this, CASS_CONSISTENCY_QUORUM);
+        cass_statement_set_is_idempotent(*this, cass_true);
+        bind<Args...>(std::forward<Args>(args)...);
+    }
+
+    /* implicit */ Statement(CassStatement* ptr) : ManagedObject{ptr, deleter}
+    {
+        cass_statement_set_consistency(*this, CASS_CONSISTENCY_QUORUM);
+        cass_statement_set_is_idempotent(*this, cass_true);
+    }
+
+    Statement(Statement&&) = default;
+
+    /**
+     * @brief Binds the given arguments to the statement.
+     *
+     * @param args Arguments to bind
+     */
+    template <typename... Args>
+    void
+    bind(Args&&... args) const
+    {
+        std::size_t idx = 0;
+        (this->bindAt<Args>(idx++, std::forward<Args>(args)), ...);
+    }
+
+    /**
+     * @brief Binds an argument to a specific index.
+     *
+     * @param idx The index of the argument
+     * @param value The value to bind it to
+     */
+    template <typename Type>
+    void
+    bindAt(std::size_t const idx, Type&& value) const
+    {
+        using std::to_string;
+        auto throwErrorIfNeeded = [idx](CassError rc, std::string_view label) {
+            if (rc != CASS_OK)
+                throw std::logic_error(fmt::format("[{}] at idx {}: {}", label, idx, cass_error_desc(rc)));
+        };
+
+        auto bindBytes = [this, idx](auto const* data, size_t size) {
+            return cass_statement_bind_bytes(*this, idx, static_cast<cass_byte_t const*>(data), size);
+        };
+
+        using DecayedType = std::decay_t<Type>;
+        using UCharVectorType = std::vector<unsigned char>;
+        using UintTupleType = std::tuple<uint32_t, uint32_t>;
+
+        if constexpr (std::is_same_v<DecayedType, ripple::uint256>)
+        {
+            auto const rc = bindBytes(value.data(), value.size());
+            throwErrorIfNeeded(rc, "Bind ripple::uint256");
+        }
+        else if constexpr (std::is_same_v<DecayedType, ripple::AccountID>)
+        {
+            auto const rc = bindBytes(value.data(), value.size());
+            throwErrorIfNeeded(rc, "Bind ripple::AccountID");
+        }
+        else if constexpr (std::is_same_v<DecayedType, UCharVectorType>)
+        {
+            auto const rc = bindBytes(value.data(), value.size());
+            throwErrorIfNeeded(rc, "Bind vector<unsigned char>");
+        }
+        else if constexpr (std::is_convertible_v<DecayedType, std::string>)
+        {
+            // reinterpret_cast is needed here :'(
+            auto const rc = bindBytes(reinterpret_cast<unsigned char const*>(value.data()), value.size());
+            throwErrorIfNeeded(rc, "Bind string (as bytes)");
+        }
+        else if constexpr (std::is_same_v<DecayedType, UintTupleType>)
+        {
+            auto const rc = cass_statement_bind_tuple(*this, idx, Tuple{std::move(value)});
+            throwErrorIfNeeded(rc, "Bind tuple<uint32, uint32>");
+        }
+        else if constexpr (std::is_same_v<DecayedType, bool>)
+        {
+            auto const rc = cass_statement_bind_bool(*this, idx, value ? cass_true : cass_false);
+            throwErrorIfNeeded(rc, "Bind bool");
+        }
+        else if constexpr (std::is_same_v<DecayedType, Limit>)
+        {
+            auto const rc = cass_statement_bind_int32(*this, idx, value.limit);
+            throwErrorIfNeeded(rc, "Bind limit (int32)");
+        }
+        // clio only uses bigint (int64_t) so we convert any incoming type
+        else if constexpr (std::is_convertible_v<DecayedType, int64_t>)
+        {
+            auto const rc = cass_statement_bind_int64(*this, idx, value);
+            throwErrorIfNeeded(rc, "Bind int64");
+        }
+        else
+        {
+            // type not supported for binding
+            static_assert(unsupported_v<DecayedType>);
+        }
+    }
+};
+
+/**
+ * @brief Represents a prepared statement on the DB side.
+ *
+ * This is used to produce Statement objects that can be executed.
+ */
+class PreparedStatement : public ManagedObject<CassPrepared const>
+{
+    static constexpr auto deleter = [](CassPrepared const* ptr) { cass_prepared_free(ptr); };
+
+public:
+    /* implicit */ PreparedStatement(CassPrepared const* ptr) : ManagedObject{ptr, deleter}
+    {
+    }
+
+    /**
+     * @brief Bind the given arguments and produce a ready to execute Statement.
+     *
+     * @param args The arguments to bind
+     * @return A bound and ready to execute Statement object
+     */
+    template <typename... Args>
+    Statement
+    bind(Args&&... args) const
+    {
+        Statement statement = cass_prepared_bind(*this);
+        statement.bind<Args...>(std::forward<Args>(args)...);
+        return statement;
+    }
+};
+
+}  // namespace data::cassandra::detail

src/data/cassandra/impl/Tuple.cpp

@@ -0,0 +1,43 @@
+//------------------------------------------------------------------------------
+/*
+    This file is part of clio: https://github.com/XRPLF/clio
+    Copyright (c) 2023, the clio developers.
+
+    Permission to use, copy, modify, and distribute this software for any
+    purpose with or without fee is hereby granted, provided that the above
+    copyright notice and this permission notice appear in all copies.
+
+    THE  SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL WARRANTIES
+    WITH  REGARD  TO  THIS  SOFTWARE  INCLUDING  ALL  IMPLIED  WARRANTIES  OF
+    MERCHANTABILITY  AND  FITNESS. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR
+    ANY  SPECIAL,  DIRECT,  INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES
+    WHATSOEVER  RESULTING  FROM  LOSS  OF USE, DATA OR PROFITS, WHETHER IN AN
+    ACTION  OF  CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF
+    OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE.
+*/
+//==============================================================================
+
+#include <data/cassandra/impl/Tuple.h>
+
+namespace {
+static constexpr auto tupleDeleter = [](CassTuple* ptr) { cass_tuple_free(ptr); };
+static constexpr auto tupleIteratorDeleter = [](CassIterator* ptr) { cass_iterator_free(ptr); };
+}  // namespace
+
+namespace data::cassandra::detail {
+
+/* implicit */ Tuple::Tuple(CassTuple* ptr) : ManagedObject{ptr, tupleDeleter}
+{
+}
+
+/* implicit */ TupleIterator::TupleIterator(CassIterator* ptr) : ManagedObject{ptr, tupleIteratorDeleter}
+{
+}
+
+[[nodiscard]] TupleIterator
+TupleIterator::fromTuple(CassValue const* value)
+{
+    return {cass_iterator_from_tuple(value)};
+}
+
+}  // namespace data::cassandra::detail

src/data/cassandra/impl/Tuple.h

@@ -0,0 +1,149 @@
+//------------------------------------------------------------------------------
+/*
+    This file is part of clio: https://github.com/XRPLF/clio
+    Copyright (c) 2023, the clio developers.
+
+    Permission to use, copy, modify, and distribute this software for any
+    purpose with or without fee is hereby granted, provided that the above
+    copyright notice and this permission notice appear in all copies.
+
+    THE  SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL WARRANTIES
+    WITH  REGARD  TO  THIS  SOFTWARE  INCLUDING  ALL  IMPLIED  WARRANTIES  OF
+    MERCHANTABILITY  AND  FITNESS. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR
+    ANY  SPECIAL,  DIRECT,  INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES
+    WHATSOEVER  RESULTING  FROM  LOSS  OF USE, DATA OR PROFITS, WHETHER IN AN
+    ACTION  OF  CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF
+    OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE.
+*/
+//==============================================================================
+
+#pragma once
+
+#include <data/cassandra/impl/ManagedObject.h>
+
+#include <cassandra.h>
+
+#include <functional>
+#include <string>
+#include <string_view>
+#include <tuple>
+
+namespace data::cassandra::detail {
+
+class Tuple : public ManagedObject<CassTuple>
+{
+    static constexpr auto deleter = [](CassTuple* ptr) { cass_tuple_free(ptr); };
+
+    template <typename>
+    static constexpr bool unsupported_v = false;
+
+public:
+    /* implicit */ Tuple(CassTuple* ptr);
+
+    template <typename... Types>
+    explicit Tuple(std::tuple<Types...>&& value)
+        : ManagedObject{cass_tuple_new(std::tuple_size<std::tuple<Types...>>{}), deleter}
+    {
+        std::apply(std::bind_front(&Tuple::bind<Types...>, this), std::move(value));
+    }
+
+    template <typename... Args>
+    void
+    bind(Args&&... args) const
+    {
+        std::size_t idx = 0;
+        (this->bindAt<Args>(idx++, std::forward<Args>(args)), ...);
+    }
+
+    template <typename Type>
+    void
+    bindAt(std::size_t const idx, Type&& value) const
+    {
+        using std::to_string;
+        auto throwErrorIfNeeded = [idx](CassError rc, std::string_view label) {
+            if (rc != CASS_OK)
+            {
+                auto const tag = '[' + std::string{label} + ']';
+                throw std::logic_error(tag + " at idx " + to_string(idx) + ": " + cass_error_desc(rc));
+            }
+        };
+
+        using DecayedType = std::decay_t<Type>;
+
+        if constexpr (std::is_same_v<DecayedType, bool>)
+        {
+            auto const rc = cass_tuple_set_bool(*this, idx, value ? cass_true : cass_false);
+            throwErrorIfNeeded(rc, "Bind bool");
+        }
+        // clio only uses bigint (int64_t) so we convert any incoming type
+        else if constexpr (std::is_convertible_v<DecayedType, int64_t>)
+        {
+            auto const rc = cass_tuple_set_int64(*this, idx, value);
+            throwErrorIfNeeded(rc, "Bind int64");
+        }
+        else
+        {
+            // type not supported for binding
+            static_assert(unsupported_v<DecayedType>);
+        }
+    }
+};
+
+class TupleIterator : public ManagedObject<CassIterator>
+{
+    template <typename>
+    static constexpr bool unsupported_v = false;
+
+public:
+    /* implicit */ TupleIterator(CassIterator* ptr);
+
+    [[nodiscard]] static TupleIterator
+    fromTuple(CassValue const* value);
+
+    template <typename... Types>
+    [[nodiscard]] std::tuple<Types...>
+    extract() const
+    {
+        return {extractNext<Types>()...};
+    }
+
+private:
+    template <typename Type>
+    Type
+    extractNext() const
+    {
+        using std::to_string;
+        Type output;
+
+        if (not cass_iterator_next(*this))
+            throw std::logic_error("Could not extract next value from tuple iterator");
+
+        auto throwErrorIfNeeded = [](CassError rc, std::string_view label) {
+            if (rc != CASS_OK)
+            {
+                auto const tag = '[' + std::string{label} + ']';
+                throw std::logic_error(tag + ": " + cass_error_desc(rc));
+            }
+        };
+
+        using DecayedType = std::decay_t<Type>;
+
+        // clio only uses bigint (int64_t) so we convert any incoming type
+        if constexpr (std::is_convertible_v<DecayedType, int64_t>)
+        {
+            int64_t out;
+            auto const rc = cass_value_get_int64(cass_iterator_get_value(*this), &out);
+            throwErrorIfNeeded(rc, "Extract int64 from tuple");
+            output = static_cast<DecayedType>(out);
+        }
+        else
+        {
+            // type not supported for extraction
+            static_assert(unsupported_v<DecayedType>);
+        }
+
+        return output;
+    }
+};
+
+}  // namespace data::cassandra::detail

src/etl/ETLHelpers.h

@@ -17,6 +17,7 @@
 */
 //==============================================================================
 
+/** @file */
 #pragma once
 
 #include <ripple/basics/base_uint.h>
@@ -26,31 +27,38 @@
 #include <queue>
 #include <sstream>
 
-/// This datastructure is used to keep track of the sequence of the most recent
-/// ledger validated by the network. There are two methods that will wait until
-/// certain conditions are met. This datastructure is able to be "stopped". When
-/// the datastructure is stopped, any threads currently waiting are unblocked.
-/// Any later calls to methods of this datastructure will not wait. Once the
-/// datastructure is stopped, the datastructure remains stopped for the rest of
-/// its lifetime.
+namespace etl {
+/**
+ * @brief This datastructure is used to keep track of the sequence of the most recent ledger validated by the network.
+ *
+ * There are two methods that will wait until certain conditions are met. This datastructure is able to be "stopped".
+ * When the datastructure is stopped, any threads currently waiting are unblocked.
+ * Any later calls to methods of this datastructure will not wait. Once the datastructure is stopped, the datastructure
+ * remains stopped for the rest of its lifetime.
+ */
 class NetworkValidatedLedgers
 {
     // max sequence validated by network
     std::optional<uint32_t> max_;
 
     mutable std::mutex m_;
-
     std::condition_variable cv_;
 
 public:
+    /**
+     * @brief A factory function for NetworkValidatedLedgers.
+     */
     static std::shared_ptr<NetworkValidatedLedgers>
     make_ValidatedLedgers()
     {
         return std::make_shared<NetworkValidatedLedgers>();
     }
 
-    /// Notify the datastructure that idx has been validated by the network
-    /// @param idx sequence validated by network
+    /**
+     * @brief Notify the datastructure that idx has been validated by the network.
+     *
+     * @param idx Sequence validated by network
+     */
     void
     push(uint32_t idx)
     {
@@ -60,10 +68,13 @@ public:
         cv_.notify_all();
     }
 
-    /// Get most recently validated sequence. If no ledgers are known to have
-    /// been validated, this function waits until the next ledger is validated
-    /// @return sequence of most recently validated ledger. empty optional if
-    /// the datastructure has been stopped
+    /**
+     * @brief Get most recently validated sequence.
+     *
+     * If no ledgers are known to have been validated, this function waits until the next ledger is validated
+     *
+     * @return Sequence of most recently validated ledger. empty optional if the datastructure has been stopped
+     */
     std::optional<uint32_t>
     getMostRecent()
     {
@@ -72,19 +83,18 @@ public:
         return max_;
     }
 
-    /// Waits for the sequence to be validated by the network
-    /// @param sequence to wait for
-    /// @return true if sequence was validated, false otherwise
-    /// a return value of false means the datastructure has been stopped
+    /**
+     * @brief Waits for the sequence to be validated by the network.
+     *
+     * @param sequence The sequence to wait for
+     * @return true if sequence was validated, false otherwise a return value of false means the datastructure has been
+     * stopped
+     */
     bool
-    waitUntilValidatedByNetwork(
-        uint32_t sequence,
-        std::optional<uint32_t> maxWaitMs = {})
+    waitUntilValidatedByNetwork(uint32_t sequence, std::optional<uint32_t> maxWaitMs = {})
     {
         std::unique_lock lck(m_);
-        auto pred = [sequence, this]() -> bool {
-            return (max_ && sequence <= *max_);
-        };
+        auto pred = [sequence, this]() -> bool { return (max_ && sequence <= *max_); };
         if (maxWaitMs)
             cv_.wait_for(lck, std::chrono::milliseconds(*maxWaitMs));
         else
@@ -93,10 +103,13 @@ public:
     }
 };
 
-/// Generic thread-safe queue with an optional maximum size
-/// Note, we can't use a lockfree queue here, since we need the ability to wait
-/// for an element to be added or removed from the queue. These waits are
-/// blocking calls.
+// TODO: does the note make sense? lockfree queues provide the same blocking behaviour just without mutex, don't they?
+/**
+ * @brief Generic thread-safe queue with a max capacity.
+ *
+ * @note (original note) We can't use a lockfree queue here, since we need the ability to wait for an element to be
+ * added or removed from the queue. These waits are blocking calls.
+ */
 template <class T>
 class ThreadSafeQueue
 {
@@ -104,75 +117,96 @@ class ThreadSafeQueue
 
     mutable std::mutex m_;
     std::condition_variable cv_;
-    std::optional<uint32_t> maxSize_;
+    uint32_t maxSize_;
 
 public:
-    /// @param maxSize maximum size of the queue. Calls that would cause the
-    /// queue to exceed this size will block until free space is available
+    /**
+     * @brief Create an instance of the queue.
+     *
+     * @param maxSize maximum size of the queue. Calls that would cause the queue to exceed this size will block until
+     * free space is available.
+     */
     ThreadSafeQueue(uint32_t maxSize) : maxSize_(maxSize)
     {
     }
 
-    /// Create a queue with no maximum size
-    ThreadSafeQueue() = default;
-
-    /// @param elt element to push onto queue
-    /// if maxSize is set, this method will block until free space is available
+    /**
+     * @brief Push element onto the queue.
+     *
+     * Note: This method will block until free space is available.
+     *
+     * @param elt Element to push onto queue
+     */
     void
     push(T const& elt)
     {
         std::unique_lock lck(m_);
-        // if queue has a max size, wait until not full
-        if (maxSize_)
-            cv_.wait(lck, [this]() { return queue_.size() <= *maxSize_; });
+        cv_.wait(lck, [this]() { return queue_.size() <= maxSize_; });
         queue_.push(elt);
         cv_.notify_all();
     }
 
-    /// @param elt element to push onto queue. elt is moved from
-    /// if maxSize is set, this method will block until free space is available
+    /**
+     * @brief Push element onto the queue.
+     *
+     * Note: This method will block until free space is available
+     *
+     * @param elt Element to push onto queue. Ownership is transferred
+     */
     void
     push(T&& elt)
     {
         std::unique_lock lck(m_);
-        // if queue has a max size, wait until not full
-        if (maxSize_)
-            cv_.wait(lck, [this]() { return queue_.size() <= *maxSize_; });
+        cv_.wait(lck, [this]() { return queue_.size() <= maxSize_; });
         queue_.push(std::move(elt));
         cv_.notify_all();
     }
 
-    /// @return element popped from queue. Will block until queue is non-empty
+    /**
+     * @brief Pop element from the queue.
+     *
+     * Note: Will block until queue is non-empty.
+     *
+     * @return Element popped from queue
+     */
     T
     pop()
     {
         std::unique_lock lck(m_);
         cv_.wait(lck, [this]() { return !queue_.empty(); });
+
         T ret = std::move(queue_.front());
         queue_.pop();
-        // if queue has a max size, unblock any possible pushers
-        if (maxSize_)
-            cv_.notify_all();
+
+        cv_.notify_all();
         return ret;
     }
-    /// @return element popped from queue. Will block until queue is non-empty
+
+    /**
+     * @brief Attempt to pop an element.
+     *
+     * @return Element popped from queue or empty optional if queue was empty
+     */
     std::optional<T>
     tryPop()
     {
         std::scoped_lock lck(m_);
         if (queue_.empty())
             return {};
+
         T ret = std::move(queue_.front());
         queue_.pop();
-        // if queue has a max size, unblock any possible pushers
-        if (maxSize_)
-            cv_.notify_all();
+
+        cv_.notify_all();
         return ret;
     }
 };
 
-/// Parititions the uint256 keyspace into numMarkers partitions, each of equal
-/// size.
+/**
+ * @brief Parititions the uint256 keyspace into numMarkers partitions, each of equal size.
+ *
+ * @param numMarkers Total markers to partition for
+ */
 inline std::vector<ripple::uint256>
 getMarkers(size_t numMarkers)
 {
@@ -190,3 +224,4 @@ getMarkers(size_t numMarkers)
     }
     return markers;
 }
+}  // namespace etl

src/etl/ETLService.cpp

@@ -0,0 +1,282 @@
+//------------------------------------------------------------------------------
+/*
+    This file is part of clio: https://github.com/XRPLF/clio
+    Copyright (c) 2023, the clio developers.
+
+    Permission to use, copy, modify, and distribute this software for any
+    purpose with or without fee is hereby granted, provided that the above
+    copyright notice and this permission notice appear in all copies.
+
+    THE  SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL WARRANTIES
+    WITH  REGARD  TO  THIS  SOFTWARE  INCLUDING  ALL  IMPLIED  WARRANTIES  OF
+    MERCHANTABILITY  AND  FITNESS. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR
+    ANY  SPECIAL,  DIRECT,  INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES
+    WHATSOEVER  RESULTING  FROM  LOSS  OF USE, DATA OR PROFITS, WHETHER IN AN
+    ACTION  OF  CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF
+    OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE.
+*/
+//==============================================================================
+
+#include <etl/ETLService.h>
+
+#include <ripple/protocol/LedgerHeader.h>
+
+namespace etl {
+// Database must be populated when this starts
+std::optional<uint32_t>
+ETLService::runETLPipeline(uint32_t startSequence, uint32_t numExtractors)
+{
+    if (finishSequence_ && startSequence > *finishSequence_)
+        return {};
+
+    LOG(log_.debug()) << "Starting etl pipeline";
+    state_.isWriting = true;
+
+    auto rng = backend_->hardFetchLedgerRangeNoThrow();
+    if (!rng || rng->maxSequence < startSequence - 1)
+    {
+        assert(false);
+        throw std::runtime_error("runETLPipeline: parent ledger is null");
+    }
+
+    auto const begin = std::chrono::system_clock::now();
+    auto extractors = std::vector<std::unique_ptr<ExtractorType>>{};
+    auto pipe = DataPipeType{numExtractors, startSequence};
+
+    for (auto i = 0u; i < numExtractors; ++i)
+        extractors.push_back(std::make_unique<ExtractorType>(
+            pipe, networkValidatedLedgers_, ledgerFetcher_, startSequence + i, finishSequence_, state_));
+
+    auto transformer =
+        TransformerType{pipe, backend_, ledgerLoader_, ledgerPublisher_, amendmentBlockHandler_, startSequence, state_};
+    transformer.waitTillFinished();  // suspend current thread until exit condition is met
+    pipe.cleanup();                  // TODO: this should probably happen automatically using destructor
+
+    // wait for all of the extractors to stop
+    for (auto& t : extractors)
+        t->waitTillFinished();
+
+    auto const end = std::chrono::system_clock::now();
+    auto const lastPublishedSeq = ledgerPublisher_.getLastPublishedSequence();
+    LOG(log_.debug()) << "Extracted and wrote " << lastPublishedSeq.value_or(startSequence) - startSequence << " in "
+                      << ((end - begin).count()) / 1000000000.0;
+
+    state_.isWriting = false;
+
+    LOG(log_.debug()) << "Stopping etl pipeline";
+    return lastPublishedSeq;
+}
+
+// Main loop of ETL.
+// The software begins monitoring the ledgers that are validated by the nework.
+// The member networkValidatedLedgers_ keeps track of the sequences of ledgers validated by the network.
+// Whenever a ledger is validated by the network, the software looks for that ledger in the database. Once the ledger is
+// found in the database, the software publishes that ledger to the ledgers stream. If a network validated ledger is not
+// found in the database after a certain amount of time, then the software attempts to take over responsibility of the
+// ETL process, where it writes new ledgers to the database. The software will relinquish control of the ETL process if
+// it detects that another process has taken over ETL.
+void
+ETLService::monitor()
+{
+    auto rng = backend_->hardFetchLedgerRangeNoThrow();
+    if (!rng)
+    {
+        LOG(log_.info()) << "Database is empty. Will download a ledger from the network.";
+        std::optional<ripple::LedgerHeader> ledger;
+
+        try
+        {
+            if (startSequence_)
+            {
+                LOG(log_.info()) << "ledger sequence specified in config. "
+                                 << "Will begin ETL process starting with ledger " << *startSequence_;
+                ledger = ledgerLoader_.loadInitialLedger(*startSequence_);
+            }
+            else
+            {
+                LOG(log_.info()) << "Waiting for next ledger to be validated by network...";
+                std::optional<uint32_t> mostRecentValidated = networkValidatedLedgers_->getMostRecent();
+
+                if (mostRecentValidated)
+                {
+                    LOG(log_.info()) << "Ledger " << *mostRecentValidated << " has been validated. Downloading...";
+                    ledger = ledgerLoader_.loadInitialLedger(*mostRecentValidated);
+                }
+                else
+                {
+                    LOG(log_.info()) << "The wait for the next validated ledger has been aborted. Exiting monitor loop";
+                    return;
+                }
+            }
+        }
+        catch (std::runtime_error const& e)
+        {
+            LOG(log_.fatal()) << "Failed to load initial ledger: " << e.what();
+            return amendmentBlockHandler_.onAmendmentBlock();
+        }
+
+        if (ledger)
+        {
+            rng = backend_->hardFetchLedgerRangeNoThrow();
+        }
+        else
+        {
+            LOG(log_.error()) << "Failed to load initial ledger. Exiting monitor loop";
+            return;
+        }
+    }
+    else
+    {
+        if (startSequence_)
+            LOG(log_.warn()) << "start sequence specified but db is already populated";
+
+        LOG(log_.info()) << "Database already populated. Picking up from the tip of history";
+        cacheLoader_.load(rng->maxSequence);
+    }
+
+    assert(rng);
+    uint32_t nextSequence = rng->maxSequence + 1;
+
+    LOG(log_.debug()) << "Database is populated. "
+                      << "Starting monitor loop. sequence = " << nextSequence;
+
+    while (true)
+    {
+        nextSequence = publishNextSequence(nextSequence);
+    }
+}
+
+uint32_t
+ETLService::publishNextSequence(uint32_t nextSequence)
+{
+    if (auto rng = backend_->hardFetchLedgerRangeNoThrow(); rng && rng->maxSequence >= nextSequence)
+    {
+        ledgerPublisher_.publish(nextSequence, {});
+        ++nextSequence;
+    }
+    else if (networkValidatedLedgers_->waitUntilValidatedByNetwork(nextSequence, 1000))
+    {
+        LOG(log_.info()) << "Ledger with sequence = " << nextSequence << " has been validated by the network. "
+                         << "Attempting to find in database and publish";
+
+        // Attempt to take over responsibility of ETL writer after 10 failed
+        // attempts to publish the ledger. publishLedger() fails if the
+        // ledger that has been validated by the network is not found in the
+        // database after the specified number of attempts. publishLedger()
+        // waits one second between each attempt to read the ledger from the
+        // database
+        constexpr size_t timeoutSeconds = 10;
+        bool success = ledgerPublisher_.publish(nextSequence, timeoutSeconds);
+
+        if (!success)
+        {
+            LOG(log_.warn()) << "Failed to publish ledger with sequence = " << nextSequence << " . Beginning ETL";
+
+            // returns the most recent sequence published empty optional if no sequence was published
+            std::optional<uint32_t> lastPublished = runETLPipeline(nextSequence, extractorThreads_);
+            LOG(log_.info()) << "Aborting ETL. Falling back to publishing";
+
+            // if no ledger was published, don't increment nextSequence
+            if (lastPublished)
+                nextSequence = *lastPublished + 1;
+        }
+        else
+        {
+            ++nextSequence;
+        }
+    }
+    return nextSequence;
+}
+
+void
+ETLService::monitorReadOnly()
+{
+    LOG(log_.debug()) << "Starting reporting in strict read only mode";
+
+    const auto latestSequenceOpt = [this]() -> std::optional<uint32_t> {
+        auto rng = backend_->hardFetchLedgerRangeNoThrow();
+
+        if (!rng)
+        {
+            if (auto net = networkValidatedLedgers_->getMostRecent())
+                return *net;
+            else
+                return std::nullopt;
+        }
+        else
+        {
+            return rng->maxSequence;
+        }
+    }();
+
+    if (!latestSequenceOpt.has_value())
+    {
+        return;
+    }
+
+    uint32_t latestSequence = *latestSequenceOpt;
+
+    cacheLoader_.load(latestSequence);
+    latestSequence++;
+
+    while (true)
+    {
+        if (auto rng = backend_->hardFetchLedgerRangeNoThrow(); rng && rng->maxSequence >= latestSequence)
+        {
+            ledgerPublisher_.publish(latestSequence, {});
+            latestSequence = latestSequence + 1;
+        }
+        else
+        {
+            // if we can't, wait until it's validated by the network, or 1 second passes, whichever occurs first.
+            // Even if we don't hear from rippled, if ledgers are being written to the db, we publish them.
+            networkValidatedLedgers_->waitUntilValidatedByNetwork(latestSequence, 1000);
+        }
+    }
+}
+
+void
+ETLService::run()
+{
+    LOG(log_.info()) << "Starting reporting etl";
+    state_.isStopping = false;
+
+    doWork();
+}
+
+void
+ETLService::doWork()
+{
+    worker_ = std::thread([this]() {
+        beast::setCurrentThreadName("ETLService worker");
+
+        if (state_.isReadOnly)
+            monitorReadOnly();
+        else
+            monitor();
+    });
+}
+
+ETLService::ETLService(
+    util::Config const& config,
+    boost::asio::io_context& ioc,
+    std::shared_ptr<BackendInterface> backend,
+    std::shared_ptr<SubscriptionManagerType> subscriptions,
+    std::shared_ptr<LoadBalancerType> balancer,
+    std::shared_ptr<NetworkValidatedLedgersType> ledgers)
+    : backend_(backend)
+    , loadBalancer_(balancer)
+    , networkValidatedLedgers_(ledgers)
+    , cacheLoader_(config, ioc, backend, backend->cache())
+    , ledgerFetcher_(backend, balancer)
+    , ledgerLoader_(backend, balancer, ledgerFetcher_, state_)
+    , ledgerPublisher_(ioc, backend, subscriptions, state_)
+    , amendmentBlockHandler_(ioc, state_)
+{
+    startSequence_ = config.maybeValue<uint32_t>("start_sequence");
+    finishSequence_ = config.maybeValue<uint32_t>("finish_sequence");
+    state_.isReadOnly = config.valueOr("read_only", state_.isReadOnly);
+    extractorThreads_ = config.valueOr<uint32_t>("extractor_threads", extractorThreads_);
+    txnThreshold_ = config.valueOr<size_t>("txn_threshold", txnThreshold_);
+}
+}  // namespace etl

src/etl/ETLService.h

@@ -0,0 +1,285 @@
+//------------------------------------------------------------------------------
+/*
+    This file is part of clio: https://github.com/XRPLF/clio
+    Copyright (c) 2023, the clio developers.
+
+    Permission to use, copy, modify, and distribute this software for any
+    purpose with or without fee is hereby granted, provided that the above
+    copyright notice and this permission notice appear in all copies.
+
+    THE  SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL WARRANTIES
+    WITH  REGARD  TO  THIS  SOFTWARE  INCLUDING  ALL  IMPLIED  WARRANTIES  OF
+    MERCHANTABILITY  AND  FITNESS. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR
+    ANY  SPECIAL,  DIRECT,  INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES
+    WHATSOEVER  RESULTING  FROM  LOSS  OF USE, DATA OR PROFITS, WHETHER IN AN
+    ACTION  OF  CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF
+    OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE.
+*/
+//==============================================================================
+
+#pragma once
+
+#include <data/BackendInterface.h>
+#include <data/LedgerCache.h>
+#include <etl/LoadBalancer.h>
+#include <etl/Source.h>
+#include <etl/SystemState.h>
+#include <etl/impl/AmendmentBlock.h>
+#include <etl/impl/CacheLoader.h>
+#include <etl/impl/ExtractionDataPipe.h>
+#include <etl/impl/Extractor.h>
+#include <etl/impl/LedgerFetcher.h>
+#include <etl/impl/LedgerLoader.h>
+#include <etl/impl/LedgerPublisher.h>
+#include <etl/impl/Transformer.h>
+#include <feed/SubscriptionManager.h>
+#include <util/log/Logger.h>
+
+#include <ripple/proto/org/xrpl/rpc/v1/xrp_ledger.grpc.pb.h>
+#include <boost/asio/steady_timer.hpp>
+#include <grpcpp/grpcpp.h>
+
+#include <memory>
+
+struct AccountTransactionsData;
+struct NFTTransactionsData;
+struct NFTsData;
+namespace feed {
+class SubscriptionManager;
+}
+
+/**
+ * @brief This namespace contains everything to do with the ETL and ETL sources.
+ */
+namespace etl {
+
+/**
+ * @brief This class is responsible for continuously extracting data from a p2p node, and writing that data to the
+ * databases.
+ *
+ * Usually, multiple different processes share access to the same network accessible databases, in which case only one
+ * such process is performing ETL and writing to the database. The other processes simply monitor the database for new
+ * ledgers, and publish those ledgers to the various subscription streams. If a monitoring process determines that the
+ * ETL writer has failed (no new ledgers written for some time), the process will attempt to become the ETL writer.
+ *
+ * If there are multiple monitoring processes that try to become the ETL writer at the same time, one will win out, and
+ * the others will fall back to monitoring/publishing. In this sense, this class dynamically transitions from monitoring
+ * to writing and from writing to monitoring, based on the activity of other processes running on different machines.
+ */
+class ETLService
+{
+    // TODO: make these template parameters in ETLService
+    using SubscriptionManagerType = feed::SubscriptionManager;
+    using LoadBalancerType = LoadBalancer;
+    using NetworkValidatedLedgersType = NetworkValidatedLedgers;
+    using DataPipeType = etl::detail::ExtractionDataPipe<org::xrpl::rpc::v1::GetLedgerResponse>;
+    using CacheLoaderType = etl::detail::CacheLoader<data::LedgerCache>;
+    using LedgerFetcherType = etl::detail::LedgerFetcher<LoadBalancerType>;
+    using ExtractorType = etl::detail::Extractor<DataPipeType, NetworkValidatedLedgersType, LedgerFetcherType>;
+    using LedgerLoaderType = etl::detail::LedgerLoader<LoadBalancerType, LedgerFetcherType>;
+    using LedgerPublisherType = etl::detail::LedgerPublisher<SubscriptionManagerType>;
+    using AmendmentBlockHandlerType = etl::detail::AmendmentBlockHandler<>;
+    using TransformerType =
+        etl::detail::Transformer<DataPipeType, LedgerLoaderType, LedgerPublisherType, AmendmentBlockHandlerType>;
+
+    util::Logger log_{"ETL"};
+
+    std::shared_ptr<BackendInterface> backend_;
+    std::shared_ptr<LoadBalancerType> loadBalancer_;
+    std::shared_ptr<NetworkValidatedLedgersType> networkValidatedLedgers_;
+
+    std::uint32_t extractorThreads_ = 1;
+    std::thread worker_;
+
+    CacheLoaderType cacheLoader_;
+    LedgerFetcherType ledgerFetcher_;
+    LedgerLoaderType ledgerLoader_;
+    LedgerPublisherType ledgerPublisher_;
+    AmendmentBlockHandlerType amendmentBlockHandler_;
+
+    SystemState state_;
+
+    size_t numMarkers_ = 2;
+    std::optional<uint32_t> startSequence_;
+    std::optional<uint32_t> finishSequence_;
+    size_t txnThreshold_ = 0;
+
+public:
+    /**
+     * @brief Create an instance of ETLService.
+     *
+     * @param config The configuration to use
+     * @param ioc io context to run on
+     * @param backend BackendInterface implementation
+     * @param subscriptions Subscription manager
+     * @param balancer Load balancer to use
+     * @param ledgers The network validated ledgers datastructure
+     */
+    ETLService(
+        util::Config const& config,
+        boost::asio::io_context& ioc,
+        std::shared_ptr<BackendInterface> backend,
+        std::shared_ptr<SubscriptionManagerType> subscriptions,
+        std::shared_ptr<LoadBalancerType> balancer,
+        std::shared_ptr<NetworkValidatedLedgersType> ledgers);
+
+    /**
+     * @brief A factory function to spawn new ETLService instances.
+     *
+     * Creates and runs the ETL service.
+     *
+     * @param config The configuration to use
+     * @param ioc io context to run on
+     * @param backend BackendInterface implementation
+     * @param subscriptions Subscription manager
+     * @param balancer Load balancer to use
+     * @param ledgers The network validated ledgers datastructure
+     */
+    static std::shared_ptr<ETLService>
+    make_ETLService(
+        util::Config const& config,
+        boost::asio::io_context& ioc,
+        std::shared_ptr<BackendInterface> backend,
+        std::shared_ptr<SubscriptionManagerType> subscriptions,
+        std::shared_ptr<LoadBalancerType> balancer,
+        std::shared_ptr<NetworkValidatedLedgersType> ledgers)
+    {
+        auto etl = std::make_shared<ETLService>(config, ioc, backend, subscriptions, balancer, ledgers);
+        etl->run();
+
+        return etl;
+    }
+
+    /**
+     * @brief Stops components and joins worker thread.
+     */
+    ~ETLService()
+    {
+        LOG(log_.info()) << "onStop called";
+        LOG(log_.debug()) << "Stopping Reporting ETL";
+
+        state_.isStopping = true;
+        cacheLoader_.stop();
+
+        if (worker_.joinable())
+            worker_.join();
+
+        LOG(log_.debug()) << "Joined ETLService worker thread";
+    }
+
+    /**
+     * @brief Get time passed since last ledger close, in seconds.
+     */
+    std::uint32_t
+    lastCloseAgeSeconds() const
+    {
+        return ledgerPublisher_.lastCloseAgeSeconds();
+    }
+
+    /**
+     * @brief Check for the amendment blocked state.
+     *
+     * @return true if currently amendment blocked; false otherwise
+     */
+    bool
+    isAmendmentBlocked() const
+    {
+        return state_.isAmendmentBlocked;
+    }
+
+    /**
+     * @brief Get state of ETL as a JSON object
+     */
+    boost::json::object
+    getInfo() const
+    {
+        boost::json::object result;
+
+        result["etl_sources"] = loadBalancer_->toJson();
+        result["is_writer"] = state_.isWriting.load();
+        result["read_only"] = state_.isReadOnly;
+        auto last = ledgerPublisher_.getLastPublish();
+        if (last.time_since_epoch().count() != 0)
+            result["last_publish_age_seconds"] = std::to_string(ledgerPublisher_.lastPublishAgeSeconds());
+        return result;
+    }
+
+private:
+    /**
+     * @brief Run the ETL pipeline.
+     *
+     * Extracts ledgers and writes them to the database, until a write conflict occurs (or the server shuts down).
+     * @note database must already be populated when this function is called
+     *
+     * @param startSequence the first ledger to extract
+     * @param numExtractors number of extractors to use
+     * @return the last ledger written to the database, if any
+     */
+    std::optional<uint32_t>
+    runETLPipeline(uint32_t startSequence, uint32_t numExtractors);
+
+    /**
+     * @brief Monitor the network for newly validated ledgers.
+     *
+     * Also monitor the database to see if any process is writing those ledgers.
+     * This function is called when the application starts, and will only return when the application is shutting down.
+     * If the software detects the database is empty, this function will call loadInitialLedger(). If the software
+     * detects ledgers are not being written, this function calls runETLPipeline(). Otherwise, this function publishes
+     * ledgers as they are written to the database.
+     */
+    void
+    monitor();
+
+    /**
+     * @brief Monitor the network for newly validated ledgers and publish them to the ledgers stream
+     *
+     * @param nextSequence the ledger sequence to publish
+     * @return the next ledger sequence to publish
+     */
+    uint32_t
+    publishNextSequence(uint32_t nextSequence);
+
+    /**
+     * @brief Monitor the database for newly written ledgers.
+     *
+     * Similar to the monitor(), except this function will never call runETLPipeline() or loadInitialLedger().
+     * This function only publishes ledgers as they are written to the database.
+     */
+    void
+    monitorReadOnly();
+
+    /**
+     * @return true if stopping; false otherwise
+     */
+    bool
+    isStopping()
+    {
+        return state_.isStopping;
+    }
+
+    /**
+     * @brief Get the number of markers to use during the initial ledger download.
+     *
+     * This is equivelent to the degree of parallelism during the initial ledger download.
+     *
+     * @return the number of markers
+     */
+    std::uint32_t
+    getNumMarkers()
+    {
+        return numMarkers_;
+    }
+
+    /**
+     * @brief Start all components to run ETL service.
+     */
+    void
+    run();
+
+    /**
+     * @brief Spawn the worker thread and start monitoring.
+     */
+    void
+    doWork();
+};
+}  // namespace etl

src/etl/ETLSource.h

@@ -1,724 +0,0 @@
-//------------------------------------------------------------------------------
-/*
-    This file is part of clio: https://github.com/XRPLF/clio
-    Copyright (c) 2022, the clio developers.
-
-    Permission to use, copy, modify, and distribute this software for any
-    purpose with or without fee is hereby granted, provided that the above
-    copyright notice and this permission notice appear in all copies.
-
-    THE  SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL WARRANTIES
-    WITH  REGARD  TO  THIS  SOFTWARE  INCLUDING  ALL  IMPLIED  WARRANTIES  OF
-    MERCHANTABILITY  AND  FITNESS. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR
-    ANY  SPECIAL,  DIRECT,  INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES
-    WHATSOEVER  RESULTING  FROM  LOSS  OF USE, DATA OR PROFITS, WHETHER IN AN
-    ACTION  OF  CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF
-    OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE.
-*/
-//==============================================================================
-
-#pragma once
-
-#include <backend/BackendInterface.h>
-#include <config/Config.h>
-#include <etl/ETLHelpers.h>
-#include <log/Logger.h>
-#include <subscriptions/SubscriptionManager.h>
-
-#include "org/xrpl/rpc/v1/xrp_ledger.grpc.pb.h"
-#include <grpcpp/grpcpp.h>
-
-#include <boost/algorithm/string.hpp>
-#include <boost/asio.hpp>
-#include <boost/beast/core.hpp>
-#include <boost/beast/core/string.hpp>
-#include <boost/beast/ssl.hpp>
-#include <boost/beast/websocket.hpp>
-
-class ETLLoadBalancer;
-class ETLSource;
-class ProbingETLSource;
-class SubscriptionManager;
-
-/// This class manages a connection to a single ETL source. This is almost
-/// always a rippled node, but really could be another reporting node. This
-/// class subscribes to the ledgers and transactions_proposed streams of the
-/// associated rippled node, and keeps track of which ledgers the rippled node
-/// has. This class also has methods for extracting said ledgers. Lastly this
-/// class forwards transactions received on the transactions_proposed streams to
-/// any subscribers.
-class ForwardCache
-{
-    using response_type = std::optional<boost::json::object>;
-
-    clio::Logger log_{"ETL"};
-    mutable std::atomic_bool stopping_ = false;
-    mutable std::shared_mutex mtx_;
-    std::unordered_map<std::string, response_type> latestForwarded_;
-
-    boost::asio::io_context::strand strand_;
-    boost::asio::steady_timer timer_;
-    ETLSource const& source_;
-    std::uint32_t duration_ = 10;
-
-    void
-    clear();
-
-public:
-    ForwardCache(
-        clio::Config const& config,
-        boost::asio::io_context& ioc,
-        ETLSource const& source)
-        : strand_(ioc), timer_(strand_), source_(source)
-    {
-        if (config.contains("cache"))
-        {
-            auto commands =
-                config.arrayOrThrow("cache", "ETLSource cache must be array");
-
-            if (config.contains("cache_duration"))
-                duration_ = config.valueOrThrow<uint32_t>(
-                    "cache_duration",
-                    "ETLSource cache_duration must be a number");
-
-            for (auto const& command : commands)
-            {
-                auto key = command.valueOrThrow<std::string>(
-                    "ETLSource forward command must be array of strings");
-                latestForwarded_[key] = {};
-            }
-        }
-    }
-
-    // This is to be called every freshenDuration_ seconds.
-    // It will request information from this etlSource, and
-    // will populate the cache with the latest value. If the
-    // request fails, it will evict that value from the cache.
-    void
-    freshen();
-
-    std::optional<boost::json::object>
-    get(boost::json::object const& command) const;
-};
-
-class ETLSource
-{
-public:
-    virtual bool
-    isConnected() const = 0;
-
-    virtual boost::json::object
-    toJson() const = 0;
-
-    virtual void
-    run() = 0;
-
-    virtual void
-    pause() = 0;
-
-    virtual void
-    resume() = 0;
-
-    virtual std::string
-    toString() const = 0;
-
-    virtual bool
-    hasLedger(uint32_t sequence) const = 0;
-
-    virtual std::pair<grpc::Status, org::xrpl::rpc::v1::GetLedgerResponse>
-    fetchLedger(
-        uint32_t ledgerSequence,
-        bool getObjects = true,
-        bool getObjectNeighbors = false) = 0;
-
-    virtual bool
-    loadInitialLedger(
-        uint32_t sequence,
-        std::uint32_t numMarkers,
-        bool cacheOnly = false) = 0;
-
-    virtual std::optional<boost::json::object>
-    forwardToRippled(
-        boost::json::object const& request,
-        std::string const& clientIp,
-        boost::asio::yield_context& yield) const = 0;
-
-    virtual ~ETLSource()
-    {
-    }
-
-protected:
-    clio::Logger log_{"ETL"};
-
-private:
-    friend ForwardCache;
-    friend ProbingETLSource;
-
-    virtual std::optional<boost::json::object>
-    requestFromRippled(
-        boost::json::object const& request,
-        std::string const& clientIp,
-        boost::asio::yield_context& yield) const = 0;
-};
-
-struct ETLSourceHooks
-{
-    enum class Action { STOP, PROCEED };
-
-    std::function<Action(boost::beast::error_code)> onConnected;
-    std::function<Action(boost::beast::error_code)> onDisconnected;
-};
-
-template <class Derived>
-class ETLSourceImpl : public ETLSource
-{
-    std::string wsPort_;
-
-    std::string grpcPort_;
-
-    std::unique_ptr<org::xrpl::rpc::v1::XRPLedgerAPIService::Stub> stub_;
-
-    boost::asio::ip::tcp::resolver resolver_;
-
-    boost::beast::flat_buffer readBuffer_;
-
-    std::vector<std::pair<uint32_t, uint32_t>> validatedLedgers_;
-
-    std::string validatedLedgersRaw_{"N/A"};
-
-    std::shared_ptr<NetworkValidatedLedgers> networkValidatedLedgers_;
-
-    // beast::Journal journal_;
-
-    mutable std::mutex mtx_;
-
-    std::atomic_bool connected_{false};
-
-    // true if this ETL source is forwarding transactions received on the
-    // transactions_proposed stream. There are usually multiple ETL sources,
-    // so to avoid forwarding the same transaction multiple times, we only
-    // forward from one particular ETL source at a time.
-    std::atomic_bool forwardingStream_{false};
-
-    // The last time a message was received on the ledgers stream
-    std::chrono::system_clock::time_point lastMsgTime_;
-    mutable std::mutex lastMsgTimeMtx_;
-
-    std::shared_ptr<BackendInterface> backend_;
-    std::shared_ptr<SubscriptionManager> subscriptions_;
-    ETLLoadBalancer& balancer_;
-
-    ForwardCache forwardCache_;
-
-    std::optional<boost::json::object>
-    requestFromRippled(
-        boost::json::object const& request,
-        std::string const& clientIp,
-        boost::asio::yield_context& yield) const override;
-
-protected:
-    Derived&
-    derived()
-    {
-        return static_cast<Derived&>(*this);
-    }
-
-    std::string ip_;
-
-    size_t numFailures_ = 0;
-
-    boost::asio::io_context& ioc_;
-
-    // used for retrying connections
-    boost::asio::steady_timer timer_;
-
-    std::atomic_bool closing_{false};
-
-    std::atomic_bool paused_{false};
-
-    ETLSourceHooks hooks_;
-
-    void
-    run() override
-    {
-        log_.trace() << toString();
-
-        auto const host = ip_;
-        auto const port = wsPort_;
-
-        resolver_.async_resolve(host, port, [this](auto ec, auto results) {
-            onResolve(ec, results);
-        });
-    }
-
-public:
-    ~ETLSourceImpl()
-    {
-        derived().close(false);
-    }
-
-    bool
-    isConnected() const override
-    {
-        return connected_;
-    }
-
-    std::chrono::system_clock::time_point
-    getLastMsgTime() const
-    {
-        std::lock_guard lck(lastMsgTimeMtx_);
-        return lastMsgTime_;
-    }
-
-    void
-    setLastMsgTime()
-    {
-        std::lock_guard lck(lastMsgTimeMtx_);
-        lastMsgTime_ = std::chrono::system_clock::now();
-    }
-
-    /// Create ETL source without gRPC endpoint
-    /// Fetch ledger and load initial ledger will fail for this source
-    /// Primarly used in read-only mode, to monitor when ledgers are validated
-    ETLSourceImpl(
-        clio::Config const& config,
-        boost::asio::io_context& ioContext,
-        std::shared_ptr<BackendInterface> backend,
-        std::shared_ptr<SubscriptionManager> subscriptions,
-        std::shared_ptr<NetworkValidatedLedgers> networkValidatedLedgers,
-        ETLLoadBalancer& balancer,
-        ETLSourceHooks hooks)
-        : resolver_(boost::asio::make_strand(ioContext))
-        , networkValidatedLedgers_(networkValidatedLedgers)
-        , backend_(backend)
-        , subscriptions_(subscriptions)
-        , balancer_(balancer)
-        , forwardCache_(config, ioContext, *this)
-        , ioc_(ioContext)
-        , timer_(ioContext)
-        , hooks_(hooks)
-    {
-        ip_ = config.valueOr<std::string>("ip", {});
-        wsPort_ = config.valueOr<std::string>("ws_port", {});
-
-        if (auto value = config.maybeValue<std::string>("grpc_port"); value)
-        {
-            grpcPort_ = *value;
-            try
-            {
-                boost::asio::ip::tcp::endpoint endpoint{
-                    boost::asio::ip::make_address(ip_), std::stoi(grpcPort_)};
-                std::stringstream ss;
-                ss << endpoint;
-                grpc::ChannelArguments chArgs;
-                chArgs.SetMaxReceiveMessageSize(-1);
-                stub_ = org::xrpl::rpc::v1::XRPLedgerAPIService::NewStub(
-                    grpc::CreateCustomChannel(
-                        ss.str(), grpc::InsecureChannelCredentials(), chArgs));
-                log_.debug() << "Made stub for remote = " << toString();
-            }
-            catch (std::exception const& e)
-            {
-                log_.debug() << "Exception while creating stub = " << e.what()
-                             << " . Remote = " << toString();
-            }
-        }
-    }
-
-    /// @param sequence ledger sequence to check for
-    /// @return true if this source has the desired ledger
-    bool
-    hasLedger(uint32_t sequence) const override
-    {
-        std::lock_guard lck(mtx_);
-        for (auto& pair : validatedLedgers_)
-        {
-            if (sequence >= pair.first && sequence <= pair.second)
-            {
-                return true;
-            }
-            else if (sequence < pair.first)
-            {
-                // validatedLedgers_ is a sorted list of disjoint ranges
-                // if the sequence comes before this range, the sequence will
-                // come before all subsequent ranges
-                return false;
-            }
-        }
-        return false;
-    }
-
-    /// process the validated range received on the ledgers stream. set the
-    /// appropriate member variable
-    /// @param range validated range received on ledgers stream
-    void
-    setValidatedRange(std::string const& range)
-    {
-        std::vector<std::pair<uint32_t, uint32_t>> pairs;
-        std::vector<std::string> ranges;
-        boost::split(ranges, range, boost::is_any_of(","));
-        for (auto& pair : ranges)
-        {
-            std::vector<std::string> minAndMax;
-
-            boost::split(minAndMax, pair, boost::is_any_of("-"));
-
-            if (minAndMax.size() == 1)
-            {
-                uint32_t sequence = std::stoll(minAndMax[0]);
-                pairs.push_back(std::make_pair(sequence, sequence));
-            }
-            else
-            {
-                assert(minAndMax.size() == 2);
-                uint32_t min = std::stoll(minAndMax[0]);
-                uint32_t max = std::stoll(minAndMax[1]);
-                pairs.push_back(std::make_pair(min, max));
-            }
-        }
-        std::sort(pairs.begin(), pairs.end(), [](auto left, auto right) {
-            return left.first < right.first;
-        });
-
-        // we only hold the lock here, to avoid blocking while string processing
-        std::lock_guard lck(mtx_);
-        validatedLedgers_ = std::move(pairs);
-        validatedLedgersRaw_ = range;
-    }
-
-    /// @return the validated range of this source
-    /// @note this is only used by server_info
-    std::string
-    getValidatedRange() const
-    {
-        std::lock_guard lck(mtx_);
-        return validatedLedgersRaw_;
-    }
-
-    /// Fetch the specified ledger
-    /// @param ledgerSequence sequence of the ledger to fetch
-    /// @getObjects whether to get the account state diff between this ledger
-    /// and the prior one
-    /// @return the extracted data and the result status
-    std::pair<grpc::Status, org::xrpl::rpc::v1::GetLedgerResponse>
-    fetchLedger(
-        uint32_t ledgerSequence,
-        bool getObjects = true,
-        bool getObjectNeighbors = false) override;
-
-    std::string
-    toString() const override
-    {
-        return "{validated_ledger: " + getValidatedRange() + ", ip: " + ip_ +
-            ", web socket port: " + wsPort_ + ", grpc port: " + grpcPort_ + "}";
-    }
-
-    boost::json::object
-    toJson() const override
-    {
-        boost::json::object res;
-        res["validated_range"] = getValidatedRange();
-        res["is_connected"] = std::to_string(isConnected());
-        res["ip"] = ip_;
-        res["ws_port"] = wsPort_;
-        res["grpc_port"] = grpcPort_;
-        auto last = getLastMsgTime();
-        if (last.time_since_epoch().count() != 0)
-            res["last_msg_age_seconds"] = std::to_string(
-                std::chrono::duration_cast<std::chrono::seconds>(
-                    std::chrono::system_clock::now() - getLastMsgTime())
-                    .count());
-        return res;
-    }
-
-    /// Download a ledger in full
-    /// @param ledgerSequence sequence of the ledger to download
-    /// @param writeQueue queue to push downloaded ledger objects
-    /// @return true if the download was successful
-    bool
-    loadInitialLedger(
-        std::uint32_t ledgerSequence,
-        std::uint32_t numMarkers,
-        bool cacheOnly = false) override;
-
-    /// Attempt to reconnect to the ETL source
-    void
-    reconnect(boost::beast::error_code ec);
-
-    /// Pause the source effectively stopping it from trying to reconnect
-    void
-    pause() override
-    {
-        paused_ = true;
-        derived().close(false);
-    }
-
-    /// Resume the source allowing it to reconnect again
-    void
-    resume() override
-    {
-        paused_ = false;
-        derived().close(true);
-    }
-
-    /// Callback
-    void
-    onResolve(
-        boost::beast::error_code ec,
-        boost::asio::ip::tcp::resolver::results_type results);
-
-    /// Callback
-    virtual void
-    onConnect(
-        boost::beast::error_code ec,
-        boost::asio::ip::tcp::resolver::results_type::endpoint_type
-            endpoint) = 0;
-
-    /// Callback
-    void
-    onHandshake(boost::beast::error_code ec);
-
-    /// Callback
-    void
-    onWrite(boost::beast::error_code ec, size_t size);
-
-    /// Callback
-    void
-    onRead(boost::beast::error_code ec, size_t size);
-
-    /// Handle the most recently received message
-    /// @return true if the message was handled successfully. false on error
-    bool
-    handleMessage();
-
-    std::optional<boost::json::object>
-    forwardToRippled(
-        boost::json::object const& request,
-        std::string const& clientIp,
-        boost::asio::yield_context& yield) const override;
-};
-
-class PlainETLSource : public ETLSourceImpl<PlainETLSource>
-{
-    std::unique_ptr<boost::beast::websocket::stream<boost::beast::tcp_stream>>
-        ws_;
-
-public:
-    PlainETLSource(
-        clio::Config const& config,
-        boost::asio::io_context& ioc,
-        std::shared_ptr<BackendInterface> backend,
-        std::shared_ptr<SubscriptionManager> subscriptions,
-        std::shared_ptr<NetworkValidatedLedgers> nwvl,
-        ETLLoadBalancer& balancer,
-        ETLSourceHooks hooks)
-        : ETLSourceImpl(
-              config,
-              ioc,
-              backend,
-              subscriptions,
-              nwvl,
-              balancer,
-              std::move(hooks))
-        , ws_(std::make_unique<
-              boost::beast::websocket::stream<boost::beast::tcp_stream>>(
-              boost::asio::make_strand(ioc)))
-    {
-    }
-
-    void
-    onConnect(
-        boost::beast::error_code ec,
-        boost::asio::ip::tcp::resolver::results_type::endpoint_type endpoint)
-        override;
-
-    /// Close the websocket
-    /// @param startAgain whether to reconnect
-    void
-    close(bool startAgain);
-
-    boost::beast::websocket::stream<boost::beast::tcp_stream>&
-    ws()
-    {
-        return *ws_;
-    }
-};
-
-class SslETLSource : public ETLSourceImpl<SslETLSource>
-{
-    std::optional<std::reference_wrapper<boost::asio::ssl::context>> sslCtx_;
-
-    std::unique_ptr<boost::beast::websocket::stream<
-        boost::beast::ssl_stream<boost::beast::tcp_stream>>>
-        ws_;
-
-public:
-    SslETLSource(
-        clio::Config const& config,
-        boost::asio::io_context& ioc,
-        std::optional<std::reference_wrapper<boost::asio::ssl::context>> sslCtx,
-        std::shared_ptr<BackendInterface> backend,
-        std::shared_ptr<SubscriptionManager> subscriptions,
-        std::shared_ptr<NetworkValidatedLedgers> nwvl,
-        ETLLoadBalancer& balancer,
-        ETLSourceHooks hooks)
-        : ETLSourceImpl(
-              config,
-              ioc,
-              backend,
-              subscriptions,
-              nwvl,
-              balancer,
-              std::move(hooks))
-        , sslCtx_(sslCtx)
-        , ws_(std::make_unique<boost::beast::websocket::stream<
-                  boost::beast::ssl_stream<boost::beast::tcp_stream>>>(
-              boost::asio::make_strand(ioc_),
-              *sslCtx_))
-    {
-    }
-
-    void
-    onConnect(
-        boost::beast::error_code ec,
-        boost::asio::ip::tcp::resolver::results_type::endpoint_type endpoint)
-        override;
-
-    void
-    onSslHandshake(
-        boost::beast::error_code ec,
-        boost::asio::ip::tcp::resolver::results_type::endpoint_type endpoint);
-
-    /// Close the websocket
-    /// @param startAgain whether to reconnect
-    void
-    close(bool startAgain);
-
-    boost::beast::websocket::stream<
-        boost::beast::ssl_stream<boost::beast::tcp_stream>>&
-    ws()
-    {
-        return *ws_;
-    }
-};
-
-/// This class is used to manage connections to transaction processing processes
-/// This class spawns a listener for each etl source, which listens to messages
-/// on the ledgers stream (to keep track of which ledgers have been validated by
-/// the network, and the range of ledgers each etl source has). This class also
-/// allows requests for ledger data to be load balanced across all possible etl
-/// sources.
-class ETLLoadBalancer
-{
-private:
-    clio::Logger log_{"ETL"};
-    std::vector<std::unique_ptr<ETLSource>> sources_;
-    std::uint32_t downloadRanges_ = 16;
-
-public:
-    ETLLoadBalancer(
-        clio::Config const& config,
-        boost::asio::io_context& ioContext,
-        std::shared_ptr<BackendInterface> backend,
-        std::shared_ptr<SubscriptionManager> subscriptions,
-        std::shared_ptr<NetworkValidatedLedgers> nwvl);
-
-    static std::shared_ptr<ETLLoadBalancer>
-    make_ETLLoadBalancer(
-        clio::Config const& config,
-        boost::asio::io_context& ioc,
-        std::shared_ptr<BackendInterface> backend,
-        std::shared_ptr<SubscriptionManager> subscriptions,
-        std::shared_ptr<NetworkValidatedLedgers> validatedLedgers)
-    {
-        return std::make_shared<ETLLoadBalancer>(
-            config, ioc, backend, subscriptions, validatedLedgers);
-    }
-
-    ~ETLLoadBalancer()
-    {
-        sources_.clear();
-    }
-
-    /// Load the initial ledger, writing data to the queue
-    /// @param sequence sequence of ledger to download
-    void
-    loadInitialLedger(uint32_t sequence, bool cacheOnly = false);
-
-    /// Fetch data for a specific ledger. This function will continuously try
-    /// to fetch data for the specified ledger until the fetch succeeds, the
-    /// ledger is found in the database, or the server is shutting down.
-    /// @param ledgerSequence sequence of ledger to fetch data for
-    /// @param getObjects if true, fetch diff between specified ledger and
-    /// previous
-    /// @return the extracted data, if extraction was successful. If the ledger
-    /// was found in the database or the server is shutting down, the optional
-    /// will be empty
-    std::optional<org::xrpl::rpc::v1::GetLedgerResponse>
-    fetchLedger(
-        uint32_t ledgerSequence,
-        bool getObjects,
-        bool getObjectNeighbors);
-
-    /// Determine whether messages received on the transactions_proposed stream
-    /// should be forwarded to subscribing clients. The server subscribes to
-    /// transactions_proposed on multiple ETLSources, yet only forwards messages
-    /// from one source at any given time (to avoid sending duplicate messages
-    /// to clients).
-    /// @param in ETLSource in question
-    /// @return true if messages should be forwarded
-    bool
-    shouldPropagateTxnStream(ETLSource* in) const
-    {
-        for (auto& src : sources_)
-        {
-            assert(src);
-            // We pick the first ETLSource encountered that is connected
-            if (src->isConnected())
-            {
-                if (src.get() == in)
-                    return true;
-                else
-                    return false;
-            }
-        }
-
-        // If no sources connected, then this stream has not been forwarded
-        return true;
-    }
-
-    boost::json::value
-    toJson() const
-    {
-        boost::json::array ret;
-        for (auto& src : sources_)
-        {
-            ret.push_back(src->toJson());
-        }
-        return ret;
-    }
-
-    /// Forward a JSON RPC request to a randomly selected rippled node
-    /// @param request JSON-RPC request
-    /// @return response received from rippled node
-    std::optional<boost::json::object>
-    forwardToRippled(
-        boost::json::object const& request,
-        std::string const& clientIp,
-        boost::asio::yield_context& yield) const;
-
-private:
-    /// f is a function that takes an ETLSource as an argument and returns a
-    /// bool. Attempt to execute f for one randomly chosen ETLSource that has
-    /// the specified ledger. If f returns false, another randomly chosen
-    /// ETLSource is used. The process repeats until f returns true.
-    /// @param f function to execute. This function takes the ETL source as an
-    /// argument, and returns a bool.
-    /// @param ledgerSequence f is executed for each ETLSource that has this
-    /// ledger
-    /// @return true if f was eventually executed successfully. false if the
-    /// ledger was found in the database or the server is shutting down
-    template <class Func>
-    bool
-    execute(Func f, uint32_t ledgerSequence);
-};

src/etl/LoadBalancer.cpp

@@ -0,0 +1,242 @@
+//------------------------------------------------------------------------------
+/*
+    This file is part of clio: https://github.com/XRPLF/clio
+    Copyright (c) 2023, the clio developers.
+
+    Permission to use, copy, modify, and distribute this software for any
+    purpose with or without fee is hereby granted, provided that the above
+    copyright notice and this permission notice appear in all copies.
+
+    THE  SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL WARRANTIES
+    WITH  REGARD  TO  THIS  SOFTWARE  INCLUDING  ALL  IMPLIED  WARRANTIES  OF
+    MERCHANTABILITY  AND  FITNESS. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR
+    ANY  SPECIAL,  DIRECT,  INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES
+    WHATSOEVER  RESULTING  FROM  LOSS  OF USE, DATA OR PROFITS, WHETHER IN AN
+    ACTION  OF  CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF
+    OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE.
+*/
+//==============================================================================
+
+#include <data/DBHelpers.h>
+#include <etl/ETLService.h>
+#include <etl/NFTHelpers.h>
+#include <etl/ProbingSource.h>
+#include <etl/Source.h>
+#include <rpc/RPCHelpers.h>
+#include <util/Profiler.h>
+#include <util/log/Logger.h>
+
+#include <ripple/beast/net/IPEndpoint.h>
+#include <ripple/protocol/STLedgerEntry.h>
+#include <boost/asio/strand.hpp>
+#include <boost/beast/http.hpp>
+#include <boost/beast/ssl.hpp>
+#include <boost/json.hpp>
+#include <boost/json/src.hpp>
+
+#include <thread>
+
+using namespace util;
+
+namespace etl {
+
+std::unique_ptr<Source>
+LoadBalancer::make_Source(
+    Config const& config,
+    boost::asio::io_context& ioc,
+    std::shared_ptr<BackendInterface> backend,
+    std::shared_ptr<feed::SubscriptionManager> subscriptions,
+    std::shared_ptr<NetworkValidatedLedgers> validatedLedgers,
+    LoadBalancer& balancer)
+{
+    auto src = std::make_unique<ProbingSource>(config, ioc, backend, subscriptions, validatedLedgers, balancer);
+    src->run();
+
+    return src;
+}
+
+std::shared_ptr<LoadBalancer>
+LoadBalancer::make_LoadBalancer(
+    Config const& config,
+    boost::asio::io_context& ioc,
+    std::shared_ptr<BackendInterface> backend,
+    std::shared_ptr<feed::SubscriptionManager> subscriptions,
+    std::shared_ptr<NetworkValidatedLedgers> validatedLedgers)
+{
+    return std::make_shared<LoadBalancer>(config, ioc, backend, subscriptions, validatedLedgers);
+}
+
+LoadBalancer::LoadBalancer(
+    Config const& config,
+    boost::asio::io_context& ioc,
+    std::shared_ptr<BackendInterface> backend,
+    std::shared_ptr<feed::SubscriptionManager> subscriptions,
+    std::shared_ptr<NetworkValidatedLedgers> validatedLedgers)
+{
+    if (auto value = config.maybeValue<uint32_t>("num_markers"); value)
+        downloadRanges_ = std::clamp(*value, 1u, 256u);
+    else if (backend->fetchLedgerRange())
+        downloadRanges_ = 4;
+
+    for (auto const& entry : config.array("etl_sources"))
+    {
+        std::unique_ptr<Source> source = make_Source(entry, ioc, backend, subscriptions, validatedLedgers, *this);
+
+        sources_.push_back(std::move(source));
+        LOG(log_.info()) << "Added etl source - " << sources_.back()->toString();
+    }
+}
+
+LoadBalancer::~LoadBalancer()
+{
+    sources_.clear();
+}
+
+std::pair<std::vector<std::string>, bool>
+LoadBalancer::loadInitialLedger(uint32_t sequence, bool cacheOnly)
+{
+    std::vector<std::string> response;
+    auto const success = execute(
+        [this, &response, &sequence, cacheOnly](auto& source) {
+            auto [data, res] = source->loadInitialLedger(sequence, downloadRanges_, cacheOnly);
+
+            if (!res)
+                LOG(log_.error()) << "Failed to download initial ledger."
+                                  << " Sequence = " << sequence << " source = " << source->toString();
+            else
+                response = std::move(data);
+
+            return res;
+        },
+        sequence);
+    return {std::move(response), success};
+}
+
+LoadBalancer::OptionalGetLedgerResponseType
+LoadBalancer::fetchLedger(uint32_t ledgerSequence, bool getObjects, bool getObjectNeighbors)
+{
+    GetLedgerResponseType response;
+    bool success = execute(
+        [&response, ledgerSequence, getObjects, getObjectNeighbors, log = log_](auto& source) {
+            auto [status, data] = source->fetchLedger(ledgerSequence, getObjects, getObjectNeighbors);
+            response = std::move(data);
+            if (status.ok() && response.validated())
+            {
+                LOG(log.info()) << "Successfully fetched ledger = " << ledgerSequence
+                                << " from source = " << source->toString();
+                return true;
+            }
+            else
+            {
+                LOG(log.warn()) << "Could not fetch ledger " << ledgerSequence << ", Reply: " << response.DebugString()
+                                << ", error_code: " << status.error_code() << ", error_msg: " << status.error_message()
+                                << ", source = " << source->toString();
+                return false;
+            }
+        },
+        ledgerSequence);
+    if (success)
+        return response;
+    else
+        return {};
+}
+
+std::optional<boost::json::object>
+LoadBalancer::forwardToRippled(
+    boost::json::object const& request,
+    std::string const& clientIp,
+    boost::asio::yield_context yield) const
+{
+    srand(static_cast<unsigned>(time(0)));
+    auto sourceIdx = rand() % sources_.size();
+    auto numAttempts = 0u;
+
+    while (numAttempts < sources_.size())
+    {
+        if (auto res = sources_[sourceIdx]->forwardToRippled(request, clientIp, yield))
+            return res;
+
+        sourceIdx = (sourceIdx + 1) % sources_.size();
+        ++numAttempts;
+    }
+
+    return {};
+}
+
+bool
+LoadBalancer::shouldPropagateTxnStream(Source* in) const
+{
+    for (auto& src : sources_)
+    {
+        assert(src);
+
+        // We pick the first Source encountered that is connected
+        if (src->isConnected())
+            return *src == *in;
+    }
+
+    // If no sources connected, then this stream has not been forwarded
+    return true;
+}
+
+boost::json::value
+LoadBalancer::toJson() const
+{
+    boost::json::array ret;
+    for (auto& src : sources_)
+        ret.push_back(src->toJson());
+
+    return ret;
+}
+
+template <class Func>
+bool
+LoadBalancer::execute(Func f, uint32_t ledgerSequence)
+{
+    srand(static_cast<unsigned>(time(0)));
+    auto sourceIdx = rand() % sources_.size();
+    auto numAttempts = 0;
+
+    while (true)
+    {
+        auto& source = sources_[sourceIdx];
+
+        LOG(log_.debug()) << "Attempting to execute func. ledger sequence = " << ledgerSequence
+                          << " - source = " << source->toString();
+        // Originally, it was (source->hasLedger(ledgerSequence) || true)
+        /* Sometimes rippled has ledger but doesn't actually know. However,
+        but this does NOT happen in the normal case and is safe to remove
+        This || true is only needed when loading full history standalone */
+        if (source->hasLedger(ledgerSequence))
+        {
+            bool res = f(source);
+            if (res)
+            {
+                LOG(log_.debug()) << "Successfully executed func at source = " << source->toString()
+                                  << " - ledger sequence = " << ledgerSequence;
+                break;
+            }
+            else
+            {
+                LOG(log_.warn()) << "Failed to execute func at source = " << source->toString()
+                                 << " - ledger sequence = " << ledgerSequence;
+            }
+        }
+        else
+        {
+            LOG(log_.warn()) << "Ledger not present at source = " << source->toString()
+                             << " - ledger sequence = " << ledgerSequence;
+        }
+        sourceIdx = (sourceIdx + 1) % sources_.size();
+        numAttempts++;
+        if (numAttempts % sources_.size() == 0)
+        {
+            LOG(log_.info()) << "Ledger sequence " << ledgerSequence
+                             << " is not yet available from any configured sources. "
+                             << "Sleeping and trying again";
+            std::this_thread::sleep_for(std::chrono::seconds(2));
+        }
+    }
+    return true;
+}
+}  // namespace etl

src/etl/LoadBalancer.h

@@ -0,0 +1,189 @@
+//------------------------------------------------------------------------------
+/*
+    This file is part of clio: https://github.com/XRPLF/clio
+    Copyright (c) 2022, the clio developers.
+
+    Permission to use, copy, modify, and distribute this software for any
+    purpose with or without fee is hereby granted, provided that the above
+    copyright notice and this permission notice appear in all copies.
+
+    THE  SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL WARRANTIES
+    WITH  REGARD  TO  THIS  SOFTWARE  INCLUDING  ALL  IMPLIED  WARRANTIES  OF
+    MERCHANTABILITY  AND  FITNESS. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR
+    ANY  SPECIAL,  DIRECT,  INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES
+    WHATSOEVER  RESULTING  FROM  LOSS  OF USE, DATA OR PROFITS, WHETHER IN AN
+    ACTION  OF  CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF
+    OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE.
+*/
+//==============================================================================
+
+#pragma once
+
+#include <data/BackendInterface.h>
+#include <etl/ETLHelpers.h>
+#include <feed/SubscriptionManager.h>
+#include <util/config/Config.h>
+#include <util/log/Logger.h>
+
+#include <ripple/proto/org/xrpl/rpc/v1/xrp_ledger.grpc.pb.h>
+#include <boost/asio.hpp>
+#include <grpcpp/grpcpp.h>
+
+namespace etl {
+class Source;
+class ProbingSource;
+}  // namespace etl
+
+namespace feed {
+class SubscriptionManager;
+}  // namespace feed
+
+namespace etl {
+
+/**
+ * @brief This class is used to manage connections to transaction processing processes.
+ *
+ * This class spawns a listener for each etl source, which listens to messages on the ledgers stream (to keep track of
+ * which ledgers have been validated by the network, and the range of ledgers each etl source has). This class also
+ * allows requests for ledger data to be load balanced across all possible ETL sources.
+ */
+class LoadBalancer
+{
+public:
+    using RawLedgerObjectType = org::xrpl::rpc::v1::RawLedgerObject;
+    using GetLedgerResponseType = org::xrpl::rpc::v1::GetLedgerResponse;
+    using OptionalGetLedgerResponseType = std::optional<GetLedgerResponseType>;
+
+private:
+    util::Logger log_{"ETL"};
+    std::vector<std::unique_ptr<Source>> sources_;
+    std::uint32_t downloadRanges_ = 16; /*< The number of markers to use when downloading intial ledger */
+
+public:
+    /**
+     * @brief Create an instance of the load balancer.
+     *
+     * @param config The configuration to use
+     * @param ioc The io_context to run on
+     * @param backend BackendInterface implementation
+     * @param subscriptions Subscription manager
+     * @param validatedLedgers The network validated ledgers datastructure
+     */
+    LoadBalancer(
+        util::Config const& config,
+        boost::asio::io_context& ioc,
+        std::shared_ptr<BackendInterface> backend,
+        std::shared_ptr<feed::SubscriptionManager> subscriptions,
+        std::shared_ptr<NetworkValidatedLedgers> validatedLedgers);
+
+    /**
+     * @brief A factory function for the load balancer.
+     *
+     * @param config The configuration to use
+     * @param ioc The io_context to run on
+     * @param backend BackendInterface implementation
+     * @param subscriptions Subscription manager
+     * @param validatedLedgers The network validated ledgers datastructure
+     */
+    static std::shared_ptr<LoadBalancer>
+    make_LoadBalancer(
+        util::Config const& config,
+        boost::asio::io_context& ioc,
+        std::shared_ptr<BackendInterface> backend,
+        std::shared_ptr<feed::SubscriptionManager> subscriptions,
+        std::shared_ptr<NetworkValidatedLedgers> validatedLedgers);
+
+    /**
+     * @brief A factory function for the ETL source.
+     *
+     * @param config The configuration to use
+     * @param ioc The io_context to run on
+     * @param backend BackendInterface implementation
+     * @param subscriptions Subscription manager
+     * @param validatedLedgers The network validated ledgers datastructure
+     * @param balancer The load balancer
+     */
+    static std::unique_ptr<Source>
+    make_Source(
+        util::Config const& config,
+        boost::asio::io_context& ioc,
+        std::shared_ptr<BackendInterface> backend,
+        std::shared_ptr<feed::SubscriptionManager> subscriptions,
+        std::shared_ptr<NetworkValidatedLedgers> validatedLedgers,
+        LoadBalancer& balancer);
+
+    ~LoadBalancer();
+
+    /**
+     * @brief Load the initial ledger, writing data to the queue.
+     *
+     * @param sequence Sequence of ledger to download
+     * @param cacheOnly Whether to only write to cache and not to the DB; defaults to false
+     */
+    std::pair<std::vector<std::string>, bool>
+    loadInitialLedger(uint32_t sequence, bool cacheOnly = false);
+
+    /**
+     * @brief Fetch data for a specific ledger.
+     *
+     * This function will continuously try to fetch data for the specified ledger until the fetch succeeds, the ledger
+     * is found in the database, or the server is shutting down.
+     *
+     * @param ledgerSequence Sequence of the ledger to fetch
+     * @param getObjects Whether to get the account state diff between this ledger and the prior one
+     * @param getObjectNeighbors Whether to request object neighbors
+     * @return The extracted data, if extraction was successful. If the ledger was found in the database or the server
+     * is shutting down, the optional will be empty
+     */
+    OptionalGetLedgerResponseType
+    fetchLedger(uint32_t ledgerSequence, bool getObjects, bool getObjectNeighbors);
+
+    /**
+     * @brief Determine whether messages received on the transactions_proposed stream should be forwarded to subscribing
+     * clients.
+     *
+     * The server subscribes to transactions_proposed on multiple Sources, yet only forwards messages from one source at
+     * any given time (to avoid sending duplicate messages to clients).
+     *
+     * @param in Source in question
+     * @return true if messages should be forwarded
+     */
+    bool
+    shouldPropagateTxnStream(Source* in) const;
+
+    /**
+     * @return JSON representation of the state of this load balancer.
+     */
+    boost::json::value
+    toJson() const;
+
+    /**
+     * @brief Forward a JSON RPC request to a randomly selected rippled node.
+     *
+     * @param request JSON-RPC request to forward
+     * @param clientIp The IP address of the peer
+     * @param yield The coroutine context
+     * @return Response received from rippled node as JSON object on success; nullopt on failure
+     */
+    std::optional<boost::json::object>
+    forwardToRippled(boost::json::object const& request, std::string const& clientIp, boost::asio::yield_context yield)
+        const;
+
+private:
+    /**
+     * @brief Execute a function on a randomly selected source.
+     *
+     * @note f is a function that takes an Source as an argument and returns a bool.
+     * Attempt to execute f for one randomly chosen Source that has the specified ledger. If f returns false, another
+     * randomly chosen Source is used. The process repeats until f returns true.
+     *
+     * @param f Function to execute. This function takes the ETL source as an argument, and returns a bool
+     * @param ledgerSequence f is executed for each Source that has this ledger
+     * @return true if f was eventually executed successfully. false if the ledger was found in the database or the
+     * server is shutting down
+     */
+    template <class Func>
+    bool
+    execute(Func f, uint32_t ledgerSequence);
+};
+}  // namespace etl

src/etl/NFTHelpers.cpp

@@ -17,15 +17,17 @@
 */
 //==============================================================================
 
-#include <ripple/app/tx/impl/details/NFTokenUtils.h>
 #include <ripple/protocol/STBase.h>
 #include <ripple/protocol/STTx.h>
 #include <ripple/protocol/TxMeta.h>
 #include <vector>
 
-#include <backend/BackendInterface.h>
-#include <backend/DBHelpers.h>
-#include <backend/Types.h>
+#include <data/BackendInterface.h>
+#include <data/DBHelpers.h>
+#include <data/Types.h>
+#include <fmt/core.h>
+
+namespace etl {
 
 std::pair<std::vector<NFTTransactionsData>, std::optional<NFTsData>>
 getNFTokenMintData(ripple::TxMeta const& txMeta, ripple::STTx const& sttx)
@@ -45,25 +47,18 @@ getNFTokenMintData(ripple::TxMeta const& txMeta, ripple::STTx const& sttx)
 
     for (ripple::STObject const& node : txMeta.getNodes())
     {
-        if (node.getFieldU16(ripple::sfLedgerEntryType) !=
-            ripple::ltNFTOKEN_PAGE)
+        if (node.getFieldU16(ripple::sfLedgerEntryType) != ripple::ltNFTOKEN_PAGE)
             continue;
 
         if (!owner)
-            owner = ripple::AccountID::fromVoid(
-                node.getFieldH256(ripple::sfLedgerIndex).data());
+            owner = ripple::AccountID::fromVoid(node.getFieldH256(ripple::sfLedgerIndex).data());
 
         if (node.getFName() == ripple::sfCreatedNode)
         {
             ripple::STArray const& toAddNFTs =
-                node.peekAtField(ripple::sfNewFields)
-                    .downcast<ripple::STObject>()
-                    .getFieldArray(ripple::sfNFTokens);
+                node.peekAtField(ripple::sfNewFields).downcast<ripple::STObject>().getFieldArray(ripple::sfNFTokens);
             std::transform(
-                toAddNFTs.begin(),
-                toAddNFTs.end(),
-                std::back_inserter(finalIDs),
-                [](ripple::STObject const& nft) {
+                toAddNFTs.begin(), toAddNFTs.end(), std::back_inserter(finalIDs), [](ripple::STObject const& nft) {
                     return nft.getFieldH256(ripple::sfNFTokenID);
                 });
         }
@@ -81,102 +76,85 @@ getNFTokenMintData(ripple::TxMeta const& txMeta, ripple::STTx const& sttx)
             // as rippled outputs all fields in final fields even if they were
             // not changed.
             ripple::STObject const& previousFields =
-                node.peekAtField(ripple::sfPreviousFields)
-                    .downcast<ripple::STObject>();
+                node.peekAtField(ripple::sfPreviousFields).downcast<ripple::STObject>();
             if (!previousFields.isFieldPresent(ripple::sfNFTokens))
                 continue;
 
-            ripple::STArray const& toAddNFTs =
-                previousFields.getFieldArray(ripple::sfNFTokens);
+            ripple::STArray const& toAddNFTs = previousFields.getFieldArray(ripple::sfNFTokens);
             std::transform(
-                toAddNFTs.begin(),
-                toAddNFTs.end(),
-                std::back_inserter(prevIDs),
-                [](ripple::STObject const& nft) {
+                toAddNFTs.begin(), toAddNFTs.end(), std::back_inserter(prevIDs), [](ripple::STObject const& nft) {
                     return nft.getFieldH256(ripple::sfNFTokenID);
                 });
 
             ripple::STArray const& toAddFinalNFTs =
-                node.peekAtField(ripple::sfFinalFields)
-                    .downcast<ripple::STObject>()
-                    .getFieldArray(ripple::sfNFTokens);
+                node.peekAtField(ripple::sfFinalFields).downcast<ripple::STObject>().getFieldArray(ripple::sfNFTokens);
             std::transform(
                 toAddFinalNFTs.begin(),
                 toAddFinalNFTs.end(),
                 std::back_inserter(finalIDs),
-                [](ripple::STObject const& nft) {
-                    return nft.getFieldH256(ripple::sfNFTokenID);
-                });
+                [](ripple::STObject const& nft) { return nft.getFieldH256(ripple::sfNFTokenID); });
         }
     }
 
     std::sort(finalIDs.begin(), finalIDs.end());
     std::sort(prevIDs.begin(), prevIDs.end());
-    std::vector<ripple::uint256> tokenIDResult;
-    std::set_difference(
-        finalIDs.begin(),
-        finalIDs.end(),
-        prevIDs.begin(),
-        prevIDs.end(),
-        std::inserter(tokenIDResult, tokenIDResult.begin()));
-    if (tokenIDResult.size() == 1 && owner)
-        return {
-            {NFTTransactionsData(
-                tokenIDResult.front(), txMeta, sttx.getTransactionID())},
-            NFTsData(tokenIDResult.front(), *owner, txMeta, false)};
 
-    std::stringstream msg;
-    msg << " - unexpected NFTokenMint data in tx " << sttx.getTransactionID();
-    throw std::runtime_error(msg.str());
+    // Find the first NFT ID that doesn't match.  We're looking for an
+    // added NFT, so the one we want will be the mismatch in finalIDs.
+    auto const diff = std::mismatch(finalIDs.begin(), finalIDs.end(), prevIDs.begin(), prevIDs.end());
+
+    // There should always be a difference so the returned finalIDs
+    // iterator should never be end().  But better safe than sorry.
+    if (finalIDs.size() != prevIDs.size() + 1 || diff.first == finalIDs.end() || !owner)
+        throw std::runtime_error(
+            fmt::format(" - unexpected NFTokenMint data in tx {}", strHex(sttx.getTransactionID())));
+
+    return {
+        {NFTTransactionsData(*diff.first, txMeta, sttx.getTransactionID())},
+        NFTsData(*diff.first, *owner, sttx.getFieldVL(ripple::sfURI), txMeta)};
 }
 
 std::pair<std::vector<NFTTransactionsData>, std::optional<NFTsData>>
 getNFTokenBurnData(ripple::TxMeta const& txMeta, ripple::STTx const& sttx)
 {
     ripple::uint256 const tokenID = sttx.getFieldH256(ripple::sfNFTokenID);
-    std::vector<NFTTransactionsData> const txs = {
-        NFTTransactionsData(tokenID, txMeta, sttx.getTransactionID())};
+    std::vector<NFTTransactionsData> const txs = {NFTTransactionsData(tokenID, txMeta, sttx.getTransactionID())};
 
     // Determine who owned the token when it was burned by finding an
     // NFTokenPage that was deleted or modified that contains this
     // tokenID.
     for (ripple::STObject const& node : txMeta.getNodes())
     {
-        if (node.getFieldU16(ripple::sfLedgerEntryType) !=
-                ripple::ltNFTOKEN_PAGE ||
+        if (node.getFieldU16(ripple::sfLedgerEntryType) != ripple::ltNFTOKEN_PAGE ||
             node.getFName() == ripple::sfCreatedNode)
             continue;
 
         // NFT burn can result in an NFTokenPage being modified to no longer
         // include the target, or an NFTokenPage being deleted. If this is
         // modified, we want to look for the target in the fields prior to
-        // modification. If deleted, it's possible that the page was modified
-        // to remove the target NFT prior to the entire page being deleted. In
-        // this case, we need to look in the PreviousFields. Otherwise, the
-        // page was not modified prior to deleting and we need to look in the
-        // FinalFields.
+        // modification. If deleted, it's possible that the page was
+        // modified to remove the target NFT prior to the entire page being
+        // deleted. In this case, we need to look in the PreviousFields.
+        // Otherwise, the page was not modified prior to deleting and we
+        // need to look in the FinalFields.
         std::optional<ripple::STArray> prevNFTs;
 
         if (node.isFieldPresent(ripple::sfPreviousFields))
         {
             ripple::STObject const& previousFields =
-                node.peekAtField(ripple::sfPreviousFields)
-                    .downcast<ripple::STObject>();
+                node.peekAtField(ripple::sfPreviousFields).downcast<ripple::STObject>();
             if (previousFields.isFieldPresent(ripple::sfNFTokens))
                 prevNFTs = previousFields.getFieldArray(ripple::sfNFTokens);
         }
         else if (!prevNFTs && node.getFName() == ripple::sfDeletedNode)
-            prevNFTs = node.peekAtField(ripple::sfFinalFields)
-                           .downcast<ripple::STObject>()
-                           .getFieldArray(ripple::sfNFTokens);
+            prevNFTs =
+                node.peekAtField(ripple::sfFinalFields).downcast<ripple::STObject>().getFieldArray(ripple::sfNFTokens);
 
         if (!prevNFTs)
             continue;
 
-        auto const nft = std::find_if(
-            prevNFTs->begin(),
-            prevNFTs->end(),
-            [&tokenID](ripple::STObject const& candidate) {
+        auto const nft =
+            std::find_if(prevNFTs->begin(), prevNFTs->end(), [&tokenID](ripple::STObject const& candidate) {
                 return candidate.getFieldH256(ripple::sfNFTokenID) == tokenID;
             });
         if (nft != prevNFTs->end())
@@ -184,92 +162,74 @@ getNFTokenBurnData(ripple::TxMeta const& txMeta, ripple::STTx const& sttx)
                 txs,
                 NFTsData(
                     tokenID,
-                    ripple::AccountID::fromVoid(
-                        node.getFieldH256(ripple::sfLedgerIndex).data()),
+                    ripple::AccountID::fromVoid(node.getFieldH256(ripple::sfLedgerIndex).data()),
                     txMeta,
                     true));
     }
 
     std::stringstream msg;
-    msg << " - could not determine owner at burntime for tx "
-        << sttx.getTransactionID();
+    msg << " - could not determine owner at burntime for tx " << sttx.getTransactionID();
     throw std::runtime_error(msg.str());
 }
 
 std::pair<std::vector<NFTTransactionsData>, std::optional<NFTsData>>
-getNFTokenAcceptOfferData(
-    ripple::TxMeta const& txMeta,
-    ripple::STTx const& sttx)
+getNFTokenAcceptOfferData(ripple::TxMeta const& txMeta, ripple::STTx const& sttx)
 {
     // If we have the buy offer from this tx, we can determine the owner
     // more easily by just looking at the owner of the accepted NFTokenOffer
     // object.
     if (sttx.isFieldPresent(ripple::sfNFTokenBuyOffer))
     {
-        auto const affectedBuyOffer = std::find_if(
-            txMeta.getNodes().begin(),
-            txMeta.getNodes().end(),
-            [&sttx](ripple::STObject const& node) {
-                return node.getFieldH256(ripple::sfLedgerIndex) ==
-                    sttx.getFieldH256(ripple::sfNFTokenBuyOffer);
+        auto const affectedBuyOffer =
+            std::find_if(txMeta.getNodes().begin(), txMeta.getNodes().end(), [&sttx](ripple::STObject const& node) {
+                return node.getFieldH256(ripple::sfLedgerIndex) == sttx.getFieldH256(ripple::sfNFTokenBuyOffer);
             });
         if (affectedBuyOffer == txMeta.getNodes().end())
         {
             std::stringstream msg;
-            msg << " - unexpected NFTokenAcceptOffer data in tx "
-                << sttx.getTransactionID();
+            msg << " - unexpected NFTokenAcceptOffer data in tx " << sttx.getTransactionID();
             throw std::runtime_error(msg.str());
         }
 
-        ripple::uint256 const tokenID =
-            affectedBuyOffer->peekAtField(ripple::sfFinalFields)
-                .downcast<ripple::STObject>()
-                .getFieldH256(ripple::sfNFTokenID);
+        ripple::uint256 const tokenID = affectedBuyOffer->peekAtField(ripple::sfFinalFields)
+                                            .downcast<ripple::STObject>()
+                                            .getFieldH256(ripple::sfNFTokenID);
 
-        ripple::AccountID const owner =
-            affectedBuyOffer->peekAtField(ripple::sfFinalFields)
-                .downcast<ripple::STObject>()
-                .getAccountID(ripple::sfOwner);
+        ripple::AccountID const owner = affectedBuyOffer->peekAtField(ripple::sfFinalFields)
+                                            .downcast<ripple::STObject>()
+                                            .getAccountID(ripple::sfOwner);
         return {
-            {NFTTransactionsData(tokenID, txMeta, sttx.getTransactionID())},
-            NFTsData(tokenID, owner, txMeta, false)};
+            {NFTTransactionsData(tokenID, txMeta, sttx.getTransactionID())}, NFTsData(tokenID, owner, txMeta, false)};
     }
 
     // Otherwise we have to infer the new owner from the affected nodes.
-    auto const affectedSellOffer = std::find_if(
-        txMeta.getNodes().begin(),
-        txMeta.getNodes().end(),
-        [&sttx](ripple::STObject const& node) {
-            return node.getFieldH256(ripple::sfLedgerIndex) ==
-                sttx.getFieldH256(ripple::sfNFTokenSellOffer);
+    auto const affectedSellOffer =
+        std::find_if(txMeta.getNodes().begin(), txMeta.getNodes().end(), [&sttx](ripple::STObject const& node) {
+            return node.getFieldH256(ripple::sfLedgerIndex) == sttx.getFieldH256(ripple::sfNFTokenSellOffer);
         });
     if (affectedSellOffer == txMeta.getNodes().end())
     {
         std::stringstream msg;
-        msg << " - unexpected NFTokenAcceptOffer data in tx "
-            << sttx.getTransactionID();
+        msg << " - unexpected NFTokenAcceptOffer data in tx " << sttx.getTransactionID();
         throw std::runtime_error(msg.str());
     }
 
-    ripple::uint256 const tokenID =
-        affectedSellOffer->peekAtField(ripple::sfFinalFields)
-            .downcast<ripple::STObject>()
-            .getFieldH256(ripple::sfNFTokenID);
+    ripple::uint256 const tokenID = affectedSellOffer->peekAtField(ripple::sfFinalFields)
+                                        .downcast<ripple::STObject>()
+                                        .getFieldH256(ripple::sfNFTokenID);
 
-    ripple::AccountID const seller =
-        affectedSellOffer->peekAtField(ripple::sfFinalFields)
-            .downcast<ripple::STObject>()
-            .getAccountID(ripple::sfOwner);
+    ripple::AccountID const seller = affectedSellOffer->peekAtField(ripple::sfFinalFields)
+                                         .downcast<ripple::STObject>()
+                                         .getAccountID(ripple::sfOwner);
 
     for (ripple::STObject const& node : txMeta.getNodes())
     {
-        if (node.getFieldU16(ripple::sfLedgerEntryType) !=
-                ripple::ltNFTOKEN_PAGE ||
+        if (node.getFieldU16(ripple::sfLedgerEntryType) != ripple::ltNFTOKEN_PAGE ||
             node.getFName() == ripple::sfDeletedNode)
             continue;
 
-        ripple::AccountID const nodeOwner = ripple::AccountID::fromVoid(
-            node.getFieldH256(ripple::sfLedgerIndex).data());
+        ripple::AccountID const nodeOwner =
+            ripple::AccountID::fromVoid(node.getFieldH256(ripple::sfLedgerIndex).data());
         if (nodeOwner == seller)
             continue;
 
@@ -283,12 +243,9 @@ getNFTokenAcceptOfferData(
                 .getFieldArray(ripple::sfNFTokens);
         }();
 
-        auto const nft = std::find_if(
-            nfts.begin(),
-            nfts.end(),
-            [&tokenID](ripple::STObject const& candidate) {
-                return candidate.getFieldH256(ripple::sfNFTokenID) == tokenID;
-            });
+        auto const nft = std::find_if(nfts.begin(), nfts.end(), [&tokenID](ripple::STObject const& candidate) {
+            return candidate.getFieldH256(ripple::sfNFTokenID) == tokenID;
+        });
         if (nft != nfts.end())
             return {
                 {NFTTransactionsData(tokenID, txMeta, sttx.getTransactionID())},
@@ -296,8 +253,7 @@ getNFTokenAcceptOfferData(
     }
 
     std::stringstream msg;
-    msg << " - unexpected NFTokenAcceptOffer data in tx "
-        << sttx.getTransactionID();
+    msg << " - unexpected NFTokenAcceptOffer data in tx " << sttx.getTransactionID();
     throw std::runtime_error(msg.str());
 }
 
@@ -306,40 +262,28 @@ getNFTokenAcceptOfferData(
 // transaction using this feature. This transaction also never returns an
 // NFTsData because it does not change the state of an NFT itself.
 std::pair<std::vector<NFTTransactionsData>, std::optional<NFTsData>>
-getNFTokenCancelOfferData(
-    ripple::TxMeta const& txMeta,
-    ripple::STTx const& sttx)
+getNFTokenCancelOfferData(ripple::TxMeta const& txMeta, ripple::STTx const& sttx)
 {
     std::vector<NFTTransactionsData> txs;
     for (ripple::STObject const& node : txMeta.getNodes())
     {
-        if (node.getFieldU16(ripple::sfLedgerEntryType) !=
-            ripple::ltNFTOKEN_OFFER)
+        if (node.getFieldU16(ripple::sfLedgerEntryType) != ripple::ltNFTOKEN_OFFER)
             continue;
 
-        ripple::uint256 const tokenID = node.peekAtField(ripple::sfFinalFields)
-                                            .downcast<ripple::STObject>()
-                                            .getFieldH256(ripple::sfNFTokenID);
+        ripple::uint256 const tokenID =
+            node.peekAtField(ripple::sfFinalFields).downcast<ripple::STObject>().getFieldH256(ripple::sfNFTokenID);
         txs.emplace_back(tokenID, txMeta, sttx.getTransactionID());
     }
 
     // Deduplicate any transactions based on tokenID/txIdx combo. Can't just
     // use txIdx because in this case one tx can cancel offers for several
     // NFTs.
-    std::sort(
-        txs.begin(),
-        txs.end(),
-        [](NFTTransactionsData const& a, NFTTransactionsData const& b) {
-            return a.tokenID < b.tokenID &&
-                a.transactionIndex < b.transactionIndex;
-        });
-    auto last = std::unique(
-        txs.begin(),
-        txs.end(),
-        [](NFTTransactionsData const& a, NFTTransactionsData const& b) {
-            return a.tokenID == b.tokenID &&
-                a.transactionIndex == b.transactionIndex;
-        });
+    std::sort(txs.begin(), txs.end(), [](NFTTransactionsData const& a, NFTTransactionsData const& b) {
+        return a.tokenID < b.tokenID && a.transactionIndex < b.transactionIndex;
+    });
+    auto last = std::unique(txs.begin(), txs.end(), [](NFTTransactionsData const& a, NFTTransactionsData const& b) {
+        return a.tokenID == b.tokenID && a.transactionIndex == b.transactionIndex;
+    });
     txs.erase(last, txs.end());
     return {txs, {}};
 }
@@ -347,20 +291,13 @@ getNFTokenCancelOfferData(
 // This transaction never returns an NFTokensData because it does not
 // change the state of an NFT itself.
 std::pair<std::vector<NFTTransactionsData>, std::optional<NFTsData>>
-getNFTokenCreateOfferData(
-    ripple::TxMeta const& txMeta,
-    ripple::STTx const& sttx)
+getNFTokenCreateOfferData(ripple::TxMeta const& txMeta, ripple::STTx const& sttx)
 {
-    return {
-        {NFTTransactionsData(
-            sttx.getFieldH256(ripple::sfNFTokenID),
-            txMeta,
-            sttx.getTransactionID())},
-        {}};
+    return {{NFTTransactionsData(sttx.getFieldH256(ripple::sfNFTokenID), txMeta, sttx.getTransactionID())}, {}};
 }
 
 std::pair<std::vector<NFTTransactionsData>, std::optional<NFTsData>>
-getNFTData(ripple::TxMeta const& txMeta, ripple::STTx const& sttx)
+getNFTDataFromTx(ripple::TxMeta const& txMeta, ripple::STTx const& sttx)
 {
     if (txMeta.getResultTER() != ripple::tesSUCCESS)
         return {{}, {}};
@@ -386,3 +323,21 @@ getNFTData(ripple::TxMeta const& txMeta, ripple::STTx const& sttx)
             return {{}, {}};
     }
 }
+
+std::vector<NFTsData>
+getNFTDataFromObj(std::uint32_t const seq, std::string const& key, std::string const& blob)
+{
+    std::vector<NFTsData> nfts;
+    ripple::STLedgerEntry const sle =
+        ripple::STLedgerEntry(ripple::SerialIter{blob.data(), blob.size()}, ripple::uint256::fromVoid(key.data()));
+
+    if (sle.getFieldU16(ripple::sfLedgerEntryType) != ripple::ltNFTOKEN_PAGE)
+        return nfts;
+
+    auto const owner = ripple::AccountID::fromVoid(key.data());
+    for (ripple::STObject const& node : sle.getFieldArray(ripple::sfNFTokens))
+        nfts.emplace_back(node.getFieldH256(ripple::sfNFTokenID), seq, owner, node.getFieldVL(ripple::sfURI));
+
+    return nfts;
+}
+}  // namespace etl

src/etl/NFTHelpers.h

@@ -0,0 +1,51 @@
+//------------------------------------------------------------------------------
+/*
+    This file is part of clio: https://github.com/XRPLF/clio
+    Copyright (c) 2023, the clio developers.
+
+    Permission to use, copy, modify, and distribute this software for any
+    purpose with or without fee is hereby granted, provided that the above
+    copyright notice and this permission notice appear in all copies.
+
+    THE  SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL WARRANTIES
+    WITH  REGARD  TO  THIS  SOFTWARE  INCLUDING  ALL  IMPLIED  WARRANTIES  OF
+    MERCHANTABILITY  AND  FITNESS. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR
+    ANY  SPECIAL,  DIRECT,  INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES
+    WHATSOEVER  RESULTING  FROM  LOSS  OF USE, DATA OR PROFITS, WHETHER IN AN
+    ACTION  OF  CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF
+    OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE.
+*/
+//==============================================================================
+
+/** @file */
+#pragma once
+
+#include <data/DBHelpers.h>
+
+#include <ripple/protocol/STTx.h>
+#include <ripple/protocol/TxMeta.h>
+
+namespace etl {
+
+/**
+ * @brief Pull NFT data from TX via ETLService.
+ *
+ * @param txMeta Transaction metadata
+ * @param sttx The transaction
+ * @return NFT transactions data as a pair of transactions and optional NFTsData
+ */
+std::pair<std::vector<NFTTransactionsData>, std::optional<NFTsData>>
+getNFTDataFromTx(ripple::TxMeta const& txMeta, ripple::STTx const& sttx);
+
+/**
+ * @brief Pull NFT data from ledger object via loadInitialLedger.
+ *
+ * @param seq The ledger sequence to pull for
+ * @param key The owner key
+ * @param blob Object data as blob
+ * @return The NFT data as a vector
+ */
+std::vector<NFTsData>
+getNFTDataFromObj(std::uint32_t const seq, std::string const& key, std::string const& blob);
+
+}  // namespace etl

src/etl/ProbingSource.cpp

@@ -17,69 +17,54 @@
 */
 //==============================================================================
 
-#include <etl/ProbingETLSource.h>
-#include <log/Logger.h>
+#include <etl/ProbingSource.h>
 
-using namespace clio;
+namespace etl {
 
-ProbingETLSource::ProbingETLSource(
-    clio::Config const& config,
+ProbingSource::ProbingSource(
+    util::Config const& config,
     boost::asio::io_context& ioc,
     std::shared_ptr<BackendInterface> backend,
-    std::shared_ptr<SubscriptionManager> subscriptions,
+    std::shared_ptr<feed::SubscriptionManager> subscriptions,
     std::shared_ptr<NetworkValidatedLedgers> nwvl,
-    ETLLoadBalancer& balancer,
+    LoadBalancer& balancer,
     boost::asio::ssl::context sslCtx)
     : sslCtx_{std::move(sslCtx)}
-    , sslSrc_{make_shared<SslETLSource>(
-          config,
-          ioc,
-          std::ref(sslCtx_),
-          backend,
-          subscriptions,
-          nwvl,
-          balancer,
-          make_SSLHooks())}
-    , plainSrc_{make_shared<PlainETLSource>(
-          config,
-          ioc,
-          backend,
-          subscriptions,
-          nwvl,
-          balancer,
-          make_PlainHooks())}
+    , sslSrc_{make_shared<
+          SslSource>(config, ioc, std::ref(sslCtx_), backend, subscriptions, nwvl, balancer, make_SSLHooks())}
+    , plainSrc_{make_shared<PlainSource>(config, ioc, backend, subscriptions, nwvl, balancer, make_PlainHooks())}
 {
 }
 
 void
-ProbingETLSource::run()
+ProbingSource::run()
 {
     sslSrc_->run();
     plainSrc_->run();
 }
 
 void
-ProbingETLSource::pause()
+ProbingSource::pause()
 {
     sslSrc_->pause();
     plainSrc_->pause();
 }
 
 void
-ProbingETLSource::resume()
+ProbingSource::resume()
 {
     sslSrc_->resume();
     plainSrc_->resume();
 }
 
 bool
-ProbingETLSource::isConnected() const
+ProbingSource::isConnected() const
 {
     return currentSrc_ && currentSrc_->isConnected();
 }
 
 bool
-ProbingETLSource::hasLedger(uint32_t sequence) const
+ProbingSource::hasLedger(uint32_t sequence) const
 {
     if (!currentSrc_)
         return false;
@@ -87,7 +72,7 @@ ProbingETLSource::hasLedger(uint32_t sequence) const
 }
 
 boost::json::object
-ProbingETLSource::toJson() const
+ProbingSource::toJson() const
 {
     if (!currentSrc_)
     {
@@ -104,43 +89,42 @@ ProbingETLSource::toJson() const
 }
 
 std::string
-ProbingETLSource::toString() const
+ProbingSource::toString() const
 {
     if (!currentSrc_)
-        return "{probing... ws: " + plainSrc_->toString() +
-            ", wss: " + sslSrc_->toString() + "}";
+        return "{probing... ws: " + plainSrc_->toString() + ", wss: " + sslSrc_->toString() + "}";
     return currentSrc_->toString();
 }
 
-bool
-ProbingETLSource::loadInitialLedger(
-    std::uint32_t ledgerSequence,
-    std::uint32_t numMarkers,
-    bool cacheOnly)
+boost::uuids::uuid
+ProbingSource::token() const
 {
     if (!currentSrc_)
-        return false;
-    return currentSrc_->loadInitialLedger(
-        ledgerSequence, numMarkers, cacheOnly);
+        return boost::uuids::nil_uuid();
+    return currentSrc_->token();
 }
 
-std::pair<grpc::Status, org::xrpl::rpc::v1::GetLedgerResponse>
-ProbingETLSource::fetchLedger(
-    uint32_t ledgerSequence,
-    bool getObjects,
-    bool getObjectNeighbors)
+std::pair<std::vector<std::string>, bool>
+ProbingSource::loadInitialLedger(std::uint32_t sequence, std::uint32_t numMarkers, bool cacheOnly)
+{
+    if (!currentSrc_)
+        return {{}, false};
+    return currentSrc_->loadInitialLedger(sequence, numMarkers, cacheOnly);
+}
+
+std::pair<grpc::Status, ProbingSource::GetLedgerResponseType>
+ProbingSource::fetchLedger(uint32_t sequence, bool getObjects, bool getObjectNeighbors)
 {
     if (!currentSrc_)
         return {};
-    return currentSrc_->fetchLedger(
-        ledgerSequence, getObjects, getObjectNeighbors);
+    return currentSrc_->fetchLedger(sequence, getObjects, getObjectNeighbors);
 }
 
 std::optional<boost::json::object>
-ProbingETLSource::forwardToRippled(
+ProbingSource::forwardToRippled(
     boost::json::object const& request,
     std::string const& clientIp,
-    boost::asio::yield_context& yield) const
+    boost::asio::yield_context yield) const
 {
     if (!currentSrc_)
         return {};
@@ -148,72 +132,71 @@ ProbingETLSource::forwardToRippled(
 }
 
 std::optional<boost::json::object>
-ProbingETLSource::requestFromRippled(
+ProbingSource::requestFromRippled(
     boost::json::object const& request,
     std::string const& clientIp,
-    boost::asio::yield_context& yield) const
+    boost::asio::yield_context yield) const
 {
     if (!currentSrc_)
         return {};
     return currentSrc_->requestFromRippled(request, clientIp, yield);
 }
 
-ETLSourceHooks
-ProbingETLSource::make_SSLHooks() noexcept
+SourceHooks
+ProbingSource::make_SSLHooks() noexcept
 {
     return {// onConnected
             [this](auto ec) {
                 std::lock_guard lck(mtx_);
                 if (currentSrc_)
-                    return ETLSourceHooks::Action::STOP;
+                    return SourceHooks::Action::STOP;
 
                 if (!ec)
                 {
                     plainSrc_->pause();
                     currentSrc_ = sslSrc_;
-                    log_.info() << "Selected WSS as the main source: "
-                                << currentSrc_->toString();
+                    LOG(log_.info()) << "Selected WSS as the main source: " << currentSrc_->toString();
                 }
-                return ETLSourceHooks::Action::PROCEED;
+                return SourceHooks::Action::PROCEED;
             },
             // onDisconnected
-            [this](auto ec) {
+            [this](auto /* ec */) {
                 std::lock_guard lck(mtx_);
                 if (currentSrc_)
                 {
                     currentSrc_ = nullptr;
                     plainSrc_->resume();
                 }
-                return ETLSourceHooks::Action::STOP;
+                return SourceHooks::Action::STOP;
             }};
 }
 
-ETLSourceHooks
-ProbingETLSource::make_PlainHooks() noexcept
+SourceHooks
+ProbingSource::make_PlainHooks() noexcept
 {
     return {// onConnected
             [this](auto ec) {
                 std::lock_guard lck(mtx_);
                 if (currentSrc_)
-                    return ETLSourceHooks::Action::STOP;
+                    return SourceHooks::Action::STOP;
 
                 if (!ec)
                 {
                     sslSrc_->pause();
                     currentSrc_ = plainSrc_;
-                    log_.info() << "Selected Plain WS as the main source: "
-                                << currentSrc_->toString();
+                    LOG(log_.info()) << "Selected Plain WS as the main source: " << currentSrc_->toString();
                 }
-                return ETLSourceHooks::Action::PROCEED;
+                return SourceHooks::Action::PROCEED;
             },
             // onDisconnected
-            [this](auto ec) {
+            [this](auto /* ec */) {
                 std::lock_guard lck(mtx_);
                 if (currentSrc_)
                 {
                     currentSrc_ = nullptr;
                     sslSrc_->resume();
                 }
-                return ETLSourceHooks::Action::STOP;
+                return SourceHooks::Action::STOP;
             }};
-}
+};
+}  // namespace etl

src/etl/ProbingSource.h

@@ -19,6 +19,10 @@
 
 #pragma once
 
+#include <etl/Source.h>
+#include <util/config/Config.h>
+#include <util/log/Logger.h>
+
 #include <boost/asio.hpp>
 #include <boost/beast/core.hpp>
 #include <boost/beast/core/string.hpp>
@@ -27,36 +31,51 @@
 
 #include <mutex>
 
-#include <config/Config.h>
-#include <etl/ETLSource.h>
-#include <log/Logger.h>
+namespace etl {
 
-/// This ETLSource implementation attempts to connect over both secure websocket
-/// and plain websocket. First to connect pauses the other and the probing is
-/// considered done at this point. If however the connected source loses
-/// connection the probing is kickstarted again.
-class ProbingETLSource : public ETLSource
+/**
+ * @brief This Source implementation attempts to connect over both secure websocket and plain websocket.
+ *
+ * First to connect pauses the other and the probing is considered done at this point.
+ * If however the connected source loses connection the probing is kickstarted again.
+ */
+class ProbingSource : public Source
 {
-    clio::Logger log_{"ETL"};
+public:
+    // TODO: inject when unit tests will be written for ProbingSource
+    using GetLedgerResponseType = org::xrpl::rpc::v1::GetLedgerResponse;
+
+private:
+    util::Logger log_{"ETL"};
 
     std::mutex mtx_;
     boost::asio::ssl::context sslCtx_;
-    std::shared_ptr<ETLSource> sslSrc_;
-    std::shared_ptr<ETLSource> plainSrc_;
-    std::shared_ptr<ETLSource> currentSrc_;
+    std::shared_ptr<Source> sslSrc_;
+    std::shared_ptr<Source> plainSrc_;
+    std::shared_ptr<Source> currentSrc_;
 
 public:
-    ProbingETLSource(
-        clio::Config const& config,
+    /**
+     * @brief Create an instance of the probing source.
+     *
+     * @param config The configuration to use
+     * @param ioc io context to run on
+     * @param backend BackendInterface implementation
+     * @param subscriptions Subscription manager
+     * @param nwvl The network validated ledgers datastructure
+     * @param balancer Load balancer to use
+     * @param sslCtx The SSL context to use; defaults to tlsv12
+     */
+    ProbingSource(
+        util::Config const& config,
         boost::asio::io_context& ioc,
         std::shared_ptr<BackendInterface> backend,
-        std::shared_ptr<SubscriptionManager> subscriptions,
+        std::shared_ptr<feed::SubscriptionManager> subscriptions,
         std::shared_ptr<NetworkValidatedLedgers> nwvl,
-        ETLLoadBalancer& balancer,
-        boost::asio::ssl::context sslCtx = boost::asio::ssl::context{
-            boost::asio::ssl::context::tlsv12});
+        LoadBalancer& balancer,
+        boost::asio::ssl::context sslCtx = boost::asio::ssl::context{boost::asio::ssl::context::tlsv12});
 
-    ~ProbingETLSource() = default;
+    ~ProbingSource() = default;
 
     void
     run() override;
@@ -79,34 +98,30 @@ public:
     std::string
     toString() const override;
 
-    bool
-    loadInitialLedger(
-        std::uint32_t ledgerSequence,
-        std::uint32_t numMarkers,
-        bool cacheOnly = false) override;
+    std::pair<std::vector<std::string>, bool>
+    loadInitialLedger(std::uint32_t sequence, std::uint32_t numMarkers, bool cacheOnly = false) override;
 
-    std::pair<grpc::Status, org::xrpl::rpc::v1::GetLedgerResponse>
-    fetchLedger(
-        uint32_t ledgerSequence,
-        bool getObjects = true,
-        bool getObjectNeighbors = false) override;
+    std::pair<grpc::Status, GetLedgerResponseType>
+    fetchLedger(uint32_t sequence, bool getObjects = true, bool getObjectNeighbors = false) override;
 
     std::optional<boost::json::object>
-    forwardToRippled(
-        boost::json::object const& request,
-        std::string const& clientIp,
-        boost::asio::yield_context& yield) const override;
+    forwardToRippled(boost::json::object const& request, std::string const& clientIp, boost::asio::yield_context yield)
+        const override;
+
+    boost::uuids::uuid
+    token() const override;
 
 private:
     std::optional<boost::json::object>
     requestFromRippled(
         boost::json::object const& request,
         std::string const& clientIp,
-        boost::asio::yield_context& yield) const override;
+        boost::asio::yield_context yield) const override;
 
-    ETLSourceHooks
+    SourceHooks
     make_SSLHooks() noexcept;
 
-    ETLSourceHooks
+    SourceHooks
     make_PlainHooks() noexcept;
 };
+}  // namespace etl

src/etl/README.md

@@ -1,3 +1,5 @@
+# ETL subsystem
+
 A single clio node has one or more ETL sources, specified in the config
 file. clio will subscribe to the `ledgers` stream of each of the ETL
 sources. This stream sends a message whenever a new ledger is validated. Upon

src/etl/ReportingETL.h

@@ -1,415 +0,0 @@
-//------------------------------------------------------------------------------
-/*
-    This file is part of clio: https://github.com/XRPLF/clio
-    Copyright (c) 2022, the clio developers.
-
-    Permission to use, copy, modify, and distribute this software for any
-    purpose with or without fee is hereby granted, provided that the above
-    copyright notice and this permission notice appear in all copies.
-
-    THE  SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL WARRANTIES
-    WITH  REGARD  TO  THIS  SOFTWARE  INCLUDING  ALL  IMPLIED  WARRANTIES  OF
-    MERCHANTABILITY  AND  FITNESS. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR
-    ANY  SPECIAL,  DIRECT,  INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES
-    WHATSOEVER  RESULTING  FROM  LOSS  OF USE, DATA OR PROFITS, WHETHER IN AN
-    ACTION  OF  CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF
-    OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE.
-*/
-//==============================================================================
-
-#pragma once
-
-#include <ripple/ledger/ReadView.h>
-#include <boost/algorithm/string.hpp>
-#include <boost/beast/core.hpp>
-#include <boost/beast/core/string.hpp>
-#include <boost/beast/websocket.hpp>
-#include <backend/BackendInterface.h>
-#include <etl/ETLSource.h>
-#include <log/Logger.h>
-#include <subscriptions/SubscriptionManager.h>
-
-#include "org/xrpl/rpc/v1/xrp_ledger.grpc.pb.h"
-#include <grpcpp/grpcpp.h>
-
-#include <condition_variable>
-#include <mutex>
-#include <queue>
-
-#include <chrono>
-
-/**
- * Helper function for the ReportingETL, implemented in NFTHelpers.cpp, to
- * pull to-write data out of a transaction that relates to NFTs.
- */
-std::pair<std::vector<NFTTransactionsData>, std::optional<NFTsData>>
-getNFTData(ripple::TxMeta const& txMeta, ripple::STTx const& sttx);
-
-struct AccountTransactionsData;
-struct NFTTransactionsData;
-struct NFTsData;
-struct FormattedTransactionsData
-{
-    std::vector<AccountTransactionsData> accountTxData;
-    std::vector<NFTTransactionsData> nfTokenTxData;
-    std::vector<NFTsData> nfTokensData;
-};
-class SubscriptionManager;
-
-/**
- * This class is responsible for continuously extracting data from a
- * p2p node, and writing that data to the databases. Usually, multiple different
- * processes share access to the same network accessible databases, in which
- * case only one such process is performing ETL and writing to the database. The
- * other processes simply monitor the database for new ledgers, and publish
- * those ledgers to the various subscription streams. If a monitoring process
- * determines that the ETL writer has failed (no new ledgers written for some
- * time), the process will attempt to become the ETL writer. If there are
- * multiple monitoring processes that try to become the ETL writer at the same
- * time, one will win out, and the others will fall back to
- * monitoring/publishing. In this sense, this class dynamically transitions from
- * monitoring to writing and from writing to monitoring, based on the activity
- * of other processes running on different machines.
- */
-class ReportingETL
-{
-private:
-    clio::Logger log_{"ETL"};
-
-    std::shared_ptr<BackendInterface> backend_;
-    std::shared_ptr<SubscriptionManager> subscriptions_;
-    std::shared_ptr<ETLLoadBalancer> loadBalancer_;
-    std::optional<std::uint32_t> onlineDeleteInterval_;
-    std::uint32_t extractorThreads_ = 1;
-
-    enum class CacheLoadStyle { ASYNC, SYNC, NOT_AT_ALL };
-
-    CacheLoadStyle cacheLoadStyle_ = CacheLoadStyle::ASYNC;
-
-    // number of diffs to use to generate cursors to traverse the ledger in
-    // parallel during initial cache download
-    size_t numCacheDiffs_ = 32;
-    // number of markers to use at one time to traverse the ledger in parallel
-    // during initial cache download
-    size_t numCacheMarkers_ = 48;
-    // number of ledger objects to fetch concurrently per marker during cache
-    // download
-    size_t cachePageFetchSize_ = 512;
-    // thread responsible for syncing the cache on startup
-    std::thread cacheDownloader_;
-
-    struct ClioPeer
-    {
-        std::string ip;
-        int port;
-    };
-
-    std::vector<ClioPeer> clioPeers;
-
-    std::thread worker_;
-    boost::asio::io_context& ioContext_;
-
-    /// Strand to ensure that ledgers are published in order.
-    /// If ETL is started far behind the network, ledgers will be written and
-    /// published very rapidly. Monitoring processes will publish ledgers as
-    /// they are written. However, to publish a ledger, the monitoring process
-    /// needs to read all of the transactions for that ledger from the database.
-    /// Reading the transactions from the database requires network calls, which
-    /// can be slow. It is imperative however that the monitoring processes keep
-    /// up with the writer, else the monitoring processes will not be able to
-    /// detect if the writer failed. Therefore, publishing each ledger (which
-    /// includes reading all of the transactions from the database) is done from
-    /// the application wide asio io_service, and a strand is used to ensure
-    /// ledgers are published in order
-    boost::asio::io_context::strand publishStrand_;
-
-    /// Mechanism for communicating with ETL sources. ETLLoadBalancer wraps an
-    /// arbitrary number of ETL sources and load balances ETL requests across
-    /// those sources.
-
-    /// Mechanism for detecting when the network has validated a new ledger.
-    /// This class provides a way to wait for a specific ledger to be validated
-    std::shared_ptr<NetworkValidatedLedgers> networkValidatedLedgers_;
-
-    /// Whether the software is stopping
-    std::atomic_bool stopping_ = false;
-    /// Whether the software is performing online delete
-    // TODO this needs to live in the database, so diff servers can coordinate
-    // deletion
-    std::atomic_bool deleting_ = false;
-
-    /// This variable controls the number of GetLedgerData calls that will be
-    /// executed in parallel during the initial ledger download. GetLedgerData
-    /// allows clients to page through a ledger over many RPC calls.
-    /// GetLedgerData returns a marker that is used as an offset in a subsequent
-    /// call. If numMarkers_ is greater than 1, there will be multiple chains of
-    /// GetLedgerData calls iterating over different parts of the same ledger in
-    /// parallel. This can dramatically speed up the time to download the
-    /// initial ledger. However, a higher value for this member variable puts
-    /// more load on the ETL source.
-    size_t numMarkers_ = 2;
-
-    /// Whether the process is in strict read-only mode. In strict read-only
-    /// mode, the process will never attempt to become the ETL writer, and will
-    /// only publish ledgers as they are written to the database.
-    bool readOnly_ = false;
-
-    /// Whether the process is writing to the database. Used by server_info
-    std::atomic_bool writing_ = false;
-
-    /// Ledger sequence to start ETL from. If this is empty, ETL will start from
-    /// the next ledger validated by the network. If this is set, and the
-    /// database is already populated, an error is thrown.
-    std::optional<uint32_t> startSequence_;
-    std::optional<uint32_t> finishSequence_;
-
-    size_t txnThreshold_ = 0;
-
-    /// The time that the most recently published ledger was published. Used by
-    /// server_info
-    std::chrono::time_point<std::chrono::system_clock> lastPublish_;
-
-    mutable std::shared_mutex publishTimeMtx_;
-
-    void
-    setLastPublish()
-    {
-        std::scoped_lock lck(publishTimeMtx_);
-        lastPublish_ = std::chrono::system_clock::now();
-    }
-
-    /// The time that the most recently published ledger was closed.
-    std::chrono::time_point<ripple::NetClock> lastCloseTime_;
-
-    mutable std::shared_mutex closeTimeMtx_;
-
-    void
-    setLastClose(std::chrono::time_point<ripple::NetClock> lastCloseTime)
-    {
-        std::scoped_lock lck(closeTimeMtx_);
-        lastCloseTime_ = lastCloseTime;
-    }
-
-    /// Download a ledger with specified sequence in full, via GetLedgerData,
-    /// and write the data to the databases. This takes several minutes or
-    /// longer.
-    /// @param sequence the sequence of the ledger to download
-    /// @return The ledger downloaded, with a full transaction and account state
-    /// map
-    std::optional<ripple::LedgerInfo>
-    loadInitialLedger(uint32_t sequence);
-
-    /// Populates the cache by walking through the given ledger. Should only be
-    /// called once. The default behavior is to return immediately and populate
-    /// the cache in the background. This can be overridden via config
-    /// parameter, to populate synchronously, or not at all
-    void
-    loadCache(uint32_t seq);
-
-    void
-    loadCacheFromDb(uint32_t seq);
-
-    bool
-    loadCacheFromClioPeer(
-        uint32_t ledgerSequence,
-        std::string const& ip,
-        std::string const& port,
-        boost::asio::yield_context& yield);
-
-    /// Run ETL. Extracts ledgers and writes them to the database, until a
-    /// write conflict occurs (or the server shuts down).
-    /// @note database must already be populated when this function is
-    /// called
-    /// @param startSequence the first ledger to extract
-    /// @return the last ledger written to the database, if any
-    std::optional<uint32_t>
-    runETLPipeline(uint32_t startSequence, int offset);
-
-    /// Monitor the network for newly validated ledgers. Also monitor the
-    /// database to see if any process is writing those ledgers. This function
-    /// is called when the application starts, and will only return when the
-    /// application is shutting down. If the software detects the database is
-    /// empty, this function will call loadInitialLedger(). If the software
-    /// detects ledgers are not being written, this function calls
-    /// runETLPipeline(). Otherwise, this function publishes ledgers as they are
-    /// written to the database.
-    void
-    monitor();
-
-    /// Monitor the database for newly written ledgers.
-    /// Similar to the monitor(), except this function will never call
-    /// runETLPipeline() or loadInitialLedger(). This function only publishes
-    /// ledgers as they are written to the database.
-    void
-    monitorReadOnly();
-
-    /// Extract data for a particular ledger from an ETL source. This function
-    /// continously tries to extract the specified ledger (using all available
-    /// ETL sources) until the extraction succeeds, or the server shuts down.
-    /// @param sequence sequence of the ledger to extract
-    /// @return ledger header and transaction+metadata blobs. Empty optional
-    /// if the server is shutting down
-    std::optional<org::xrpl::rpc::v1::GetLedgerResponse>
-    fetchLedgerData(uint32_t sequence);
-
-    /// Extract data for a particular ledger from an ETL source. This function
-    /// continously tries to extract the specified ledger (using all available
-    /// ETL sources) until the extraction succeeds, or the server shuts down.
-    /// @param sequence sequence of the ledger to extract
-    /// @return ledger header, transaction+metadata blobs, and all ledger
-    /// objects created, modified or deleted between this ledger and the parent.
-    /// Empty optional if the server is shutting down
-    std::optional<org::xrpl::rpc::v1::GetLedgerResponse>
-    fetchLedgerDataAndDiff(uint32_t sequence);
-
-    /// Insert all of the extracted transactions into the ledger, returning
-    /// transactions related to accounts, transactions related to NFTs, and
-    /// NFTs themselves for later processsing.
-    /// @param ledger ledger to insert transactions into
-    /// @param data data extracted from an ETL source
-    /// @return struct that contains the neccessary info to write to the
-    /// account_transactions/account_tx and nft_token_transactions tables
-    /// (mostly transaction hashes, corresponding nodestore hashes and affected
-    /// accounts)
-    FormattedTransactionsData
-    insertTransactions(
-        ripple::LedgerInfo const& ledger,
-        org::xrpl::rpc::v1::GetLedgerResponse& data);
-
-    // TODO update this documentation
-    /// Build the next ledger using the previous ledger and the extracted data.
-    /// This function calls insertTransactions()
-    /// @note rawData should be data that corresponds to the ledger immediately
-    /// following parent
-    /// @param parent the previous ledger
-    /// @param rawData data extracted from an ETL source
-    /// @return the newly built ledger and data to write to the database
-    std::pair<ripple::LedgerInfo, bool>
-    buildNextLedger(org::xrpl::rpc::v1::GetLedgerResponse& rawData);
-
-    /// Attempt to read the specified ledger from the database, and then publish
-    /// that ledger to the ledgers stream.
-    /// @param ledgerSequence the sequence of the ledger to publish
-    /// @param maxAttempts the number of times to attempt to read the ledger
-    /// from the database. 1 attempt per second
-    /// @return whether the ledger was found in the database and published
-    bool
-    publishLedger(uint32_t ledgerSequence, std::optional<uint32_t> maxAttempts);
-
-    /// Publish the passed in ledger
-    /// @param ledger the ledger to publish
-    void
-    publishLedger(ripple::LedgerInfo const& lgrInfo);
-
-    bool
-    isStopping()
-    {
-        return stopping_;
-    }
-
-    /// Get the number of markers to use during the initial ledger download.
-    /// This is equivelent to the degree of parallelism during the initial
-    /// ledger download
-    /// @return the number of markers
-    std::uint32_t
-    getNumMarkers()
-    {
-        return numMarkers_;
-    }
-
-    /// start all of the necessary components and begin ETL
-    void
-    run()
-    {
-        log_.info() << "Starting reporting etl";
-        stopping_ = false;
-
-        doWork();
-    }
-
-    void
-    doWork();
-
-public:
-    ReportingETL(
-        clio::Config const& config,
-        boost::asio::io_context& ioc,
-        std::shared_ptr<BackendInterface> backend,
-        std::shared_ptr<SubscriptionManager> subscriptions,
-        std::shared_ptr<ETLLoadBalancer> balancer,
-        std::shared_ptr<NetworkValidatedLedgers> ledgers);
-
-    static std::shared_ptr<ReportingETL>
-    make_ReportingETL(
-        clio::Config const& config,
-        boost::asio::io_context& ioc,
-        std::shared_ptr<BackendInterface> backend,
-        std::shared_ptr<SubscriptionManager> subscriptions,
-        std::shared_ptr<ETLLoadBalancer> balancer,
-        std::shared_ptr<NetworkValidatedLedgers> ledgers)
-    {
-        auto etl = std::make_shared<ReportingETL>(
-            config, ioc, backend, subscriptions, balancer, ledgers);
-
-        etl->run();
-
-        return etl;
-    }
-
-    ~ReportingETL()
-    {
-        log_.info() << "onStop called";
-        log_.debug() << "Stopping Reporting ETL";
-        stopping_ = true;
-
-        if (worker_.joinable())
-            worker_.join();
-        if (cacheDownloader_.joinable())
-            cacheDownloader_.join();
-
-        log_.debug() << "Joined ReportingETL worker thread";
-    }
-
-    boost::json::object
-    getInfo() const
-    {
-        boost::json::object result;
-
-        result["etl_sources"] = loadBalancer_->toJson();
-        result["is_writer"] = writing_.load();
-        result["read_only"] = readOnly_;
-        auto last = getLastPublish();
-        if (last.time_since_epoch().count() != 0)
-            result["last_publish_age_seconds"] =
-                std::to_string(lastPublishAgeSeconds());
-        return result;
-    }
-
-    std::chrono::time_point<std::chrono::system_clock>
-    getLastPublish() const
-    {
-        std::shared_lock lck(publishTimeMtx_);
-        return lastPublish_;
-    }
-
-    std::uint32_t
-    lastPublishAgeSeconds() const
-    {
-        return std::chrono::duration_cast<std::chrono::seconds>(
-                   std::chrono::system_clock::now() - getLastPublish())
-            .count();
-    }
-
-    std::uint32_t
-    lastCloseAgeSeconds() const
-    {
-        std::shared_lock lck(closeTimeMtx_);
-        auto now = std::chrono::duration_cast<std::chrono::seconds>(
-                       std::chrono::system_clock::now().time_since_epoch())
-                       .count();
-        auto closeTime = lastCloseTime_.time_since_epoch().count();
-        if (now < (rippleEpochStart + closeTime))
-            return 0;
-        return now - (rippleEpochStart + closeTime);
-    }
-};

src/etl/Source.cpp

@@ -0,0 +1,194 @@
+//------------------------------------------------------------------------------
+/*
+    This file is part of clio: https://github.com/XRPLF/clio
+    Copyright (c) 2022, the clio developers.
+
+    Permission to use, copy, modify, and distribute this software for any
+    purpose with or without fee is hereby granted, provided that the above
+    copyright notice and this permission notice appear in all copies.
+
+    THE  SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL WARRANTIES
+    WITH  REGARD  TO  THIS  SOFTWARE  INCLUDING  ALL  IMPLIED  WARRANTIES  OF
+    MERCHANTABILITY  AND  FITNESS. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR
+    ANY  SPECIAL,  DIRECT,  INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES
+    WHATSOEVER  RESULTING  FROM  LOSS  OF USE, DATA OR PROFITS, WHETHER IN AN
+    ACTION  OF  CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF
+    OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE.
+*/
+//==============================================================================
+
+#include <data/DBHelpers.h>
+#include <etl/ETLService.h>
+#include <etl/LoadBalancer.h>
+#include <etl/ProbingSource.h>
+#include <etl/Source.h>
+#include <rpc/RPCHelpers.h>
+#include <util/Profiler.h>
+
+#include <ripple/beast/net/IPEndpoint.h>
+#include <ripple/protocol/STLedgerEntry.h>
+#include <boost/asio/strand.hpp>
+#include <boost/beast/http.hpp>
+#include <boost/beast/ssl.hpp>
+#include <boost/json.hpp>
+
+#include <thread>
+
+namespace etl {
+
+static boost::beast::websocket::stream_base::timeout
+make_TimeoutOption()
+{
+    return boost::beast::websocket::stream_base::timeout::suggested(boost::beast::role_type::client);
+}
+
+void
+PlainSource::close(bool startAgain)
+{
+    timer_.cancel();
+    boost::asio::post(strand_, [this, startAgain]() {
+        if (closing_)
+            return;
+
+        if (derived().ws().is_open())
+        {
+            // onStop() also calls close(). If the async_close is called twice,
+            // an assertion fails. Using closing_ makes sure async_close is only
+            // called once
+            closing_ = true;
+            derived().ws().async_close(boost::beast::websocket::close_code::normal, [this, startAgain](auto ec) {
+                if (ec)
+                {
+                    LOG(log_.error()) << "async_close: error code = " << ec << " - " << toString();
+                }
+                closing_ = false;
+                if (startAgain)
+                {
+                    ws_ = std::make_unique<StreamType>(strand_);
+                    run();
+                }
+            });
+        }
+        else if (startAgain)
+        {
+            ws_ = std::make_unique<StreamType>(strand_);
+            run();
+        }
+    });
+}
+
+void
+SslSource::close(bool startAgain)
+{
+    timer_.cancel();
+    boost::asio::post(strand_, [this, startAgain]() {
+        if (closing_)
+            return;
+
+        if (derived().ws().is_open())
+        {
+            // onStop() also calls close(). If the async_close is called twice, an assertion fails. Using closing_ makes
+            // sure async_close is only called once
+            closing_ = true;
+            derived().ws().async_close(boost::beast::websocket::close_code::normal, [this, startAgain](auto ec) {
+                if (ec)
+                {
+                    LOG(log_.error()) << "async_close: error code = " << ec << " - " << toString();
+                }
+                closing_ = false;
+                if (startAgain)
+                {
+                    ws_ = std::make_unique<StreamType>(strand_, *sslCtx_);
+                    run();
+                }
+            });
+        }
+        else if (startAgain)
+        {
+            ws_ = std::make_unique<StreamType>(strand_, *sslCtx_);
+            run();
+        }
+    });
+}
+
+void
+PlainSource::onConnect(
+    boost::beast::error_code ec,
+    boost::asio::ip::tcp::resolver::results_type::endpoint_type endpoint)
+{
+    if (ec)
+    {
+        // start over
+        reconnect(ec);
+    }
+    else
+    {
+        connected_ = true;
+        numFailures_ = 0;
+
+        // Websocket stream has it's own timeout system
+        boost::beast::get_lowest_layer(derived().ws()).expires_never();
+
+        derived().ws().set_option(make_TimeoutOption());
+        derived().ws().set_option(
+            boost::beast::websocket::stream_base::decorator([](boost::beast::websocket::request_type& req) {
+                req.set(boost::beast::http::field::user_agent, "clio-client");
+                req.set("X-User", "clio-client");
+            }));
+
+        // Update the host_ string. This will provide the value of the
+        // Host HTTP header during the WebSocket handshake.
+        // See https://tools.ietf.org/html/rfc7230#section-5.4
+        auto host = ip_ + ':' + std::to_string(endpoint.port());
+        derived().ws().async_handshake(host, "/", [this](auto ec) { onHandshake(ec); });
+    }
+}
+
+void
+SslSource::onConnect(boost::beast::error_code ec, boost::asio::ip::tcp::resolver::results_type::endpoint_type endpoint)
+{
+    if (ec)
+    {
+        // start over
+        reconnect(ec);
+    }
+    else
+    {
+        connected_ = true;
+        numFailures_ = 0;
+
+        // Websocket stream has it's own timeout system
+        boost::beast::get_lowest_layer(derived().ws()).expires_never();
+
+        derived().ws().set_option(make_TimeoutOption());
+        derived().ws().set_option(
+            boost::beast::websocket::stream_base::decorator([](boost::beast::websocket::request_type& req) {
+                req.set(boost::beast::http::field::user_agent, "clio-client");
+                req.set("X-User", "clio-client");
+            }));
+
+        // Update the host_ string. This will provide the value of the
+        // Host HTTP header during the WebSocket handshake.
+        // See https://tools.ietf.org/html/rfc7230#section-5.4
+        auto host = ip_ + ':' + std::to_string(endpoint.port());
+        ws().next_layer().async_handshake(
+            boost::asio::ssl::stream_base::client, [this, endpoint](auto ec) { onSslHandshake(ec, endpoint); });
+    }
+}
+
+void
+SslSource::onSslHandshake(
+    boost::beast::error_code ec,
+    boost::asio::ip::tcp::resolver::results_type::endpoint_type endpoint)
+{
+    if (ec)
+    {
+        reconnect(ec);
+    }
+    else
+    {
+        auto host = ip_ + ':' + std::to_string(endpoint.port());
+        ws().async_handshake(host, "/", [this](auto ec) { onHandshake(ec); });
+    }
+}
+}  // namespace etl

src/etl/Source.h

@@ -0,0 +1,989 @@
+//------------------------------------------------------------------------------
+/*
+    This file is part of clio: https://github.com/XRPLF/clio
+    Copyright (c) 2022, the clio developers.
+
+    Permission to use, copy, modify, and distribute this software for any
+    purpose with or without fee is hereby granted, provided that the above
+    copyright notice and this permission notice appear in all copies.
+
+    THE  SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL WARRANTIES
+    WITH  REGARD  TO  THIS  SOFTWARE  INCLUDING  ALL  IMPLIED  WARRANTIES  OF
+    MERCHANTABILITY  AND  FITNESS. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR
+    ANY  SPECIAL,  DIRECT,  INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES
+    WHATSOEVER  RESULTING  FROM  LOSS  OF USE, DATA OR PROFITS, WHETHER IN AN
+    ACTION  OF  CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF
+    OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE.
+*/
+//==============================================================================
+
+#pragma once
+
+#include <data/BackendInterface.h>
+#include <etl/ETLHelpers.h>
+#include <etl/LoadBalancer.h>
+#include <etl/impl/AsyncData.h>
+#include <etl/impl/ForwardCache.h>
+#include <feed/SubscriptionManager.h>
+#include <util/config/Config.h>
+#include <util/log/Logger.h>
+
+#include <ripple/proto/org/xrpl/rpc/v1/xrp_ledger.grpc.pb.h>
+#include <boost/algorithm/string.hpp>
+#include <boost/asio.hpp>
+#include <boost/beast/core.hpp>
+#include <boost/beast/core/string.hpp>
+#include <boost/beast/ssl.hpp>
+#include <boost/beast/websocket.hpp>
+#include <boost/uuid/uuid.hpp>
+#include <boost/uuid/uuid_generators.hpp>
+#include <grpcpp/grpcpp.h>
+
+class ProbingSource;
+namespace feed {
+class SubscriptionManager;
+}
+
+// TODO: we use Source so that we can store a vector of Sources
+// but we also use CRTP for implementation of the common logic - this is a bit strange because CRTP as used here is
+// supposed to be used instead of an abstract base.
+// Maybe we should rework this a bit. At this point there is not too much use in the CRTP implementation - we can move
+// things into the base class instead.
+
+namespace etl {
+
+/**
+ * @brief Base class for all ETL sources.
+ *
+ * Note: Since sources below are implemented via CRTP, it sort of makes no sense to have a virtual base class.
+ * We should consider using a vector of ProbingSources instead of vector of unique ptrs to this virtual base.
+ */
+class Source
+{
+public:
+    /** @return true if source is connected; false otherwise */
+    virtual bool
+    isConnected() const = 0;
+
+    /** @return JSON representation of the source */
+    virtual boost::json::object
+    toJson() const = 0;
+
+    /** @brief Runs the source */
+    virtual void
+    run() = 0;
+
+    /** @brief Request to pause the source (i.e. disconnect and do nothing) */
+    virtual void
+    pause() = 0;
+
+    /** @brief Reconnect and resume this source */
+    virtual void
+    resume() = 0;
+
+    /** @return String representation of the source (for debug) */
+    virtual std::string
+    toString() const = 0;
+
+    /**
+     * @brief Check if ledger is known by this source.
+     *
+     * @param sequence The ledger sequence to check
+     * @return true if ledger is in the range of this source; false otherwise
+     */
+    virtual bool
+    hasLedger(uint32_t sequence) const = 0;
+
+    /**
+     * @brief Fetch data for a specific ledger.
+     *
+     * This function will continuously try to fetch data for the specified ledger until the fetch succeeds, the ledger
+     * is found in the database, or the server is shutting down.
+     *
+     * @param sequence Sequence of the ledger to fetch
+     * @param getObjects Whether to get the account state diff between this ledger and the prior one; defaults to true
+     * @param getObjectNeighbors Whether to request object neighbors; defaults to false
+     * @return A std::pair of the response status and the response itself
+     */
+    virtual std::pair<grpc::Status, org::xrpl::rpc::v1::GetLedgerResponse>
+    fetchLedger(uint32_t sequence, bool getObjects = true, bool getObjectNeighbors = false) = 0;
+
+    /**
+     * @brief Download a ledger in full.
+     *
+     * @param sequence Sequence of the ledger to download
+     * @param numMarkers Number of markers to generate for async calls
+     * @param cacheOnly Only insert into cache, not the DB; defaults to false
+     * @return A std::pair of the data and a bool indicating whether the download was successfull
+     */
+    virtual std::pair<std::vector<std::string>, bool>
+    loadInitialLedger(uint32_t sequence, std::uint32_t numMarkers, bool cacheOnly = false) = 0;
+
+    /**
+     * @brief Forward a request to rippled.
+     *
+     * @param request The request to forward
+     * @param clientIp IP of the client forwarding this request
+     * @param yield The coroutine context
+     * @return Response wrapped in an optional on success; nullopt otherwise
+     */
+    virtual std::optional<boost::json::object>
+    forwardToRippled(boost::json::object const& request, std::string const& clientIp, boost::asio::yield_context yield)
+        const = 0;
+
+    /**
+     * @return A token that uniquely identifies this source instance.
+     */
+    virtual boost::uuids::uuid
+    token() const = 0;
+
+    virtual ~Source() = default;
+
+    /**
+     * @brief Comparison is done via comparing tokens provided by the token() function.
+     *
+     * @param other The other source to compare to
+     * @return true if sources are equal; false otherwise
+     */
+    bool
+    operator==(Source const& other) const
+    {
+        return token() == other.token();
+    }
+
+protected:
+    util::Logger log_{"ETL"};
+
+private:
+    friend etl::detail::ForwardCache;
+    friend ProbingSource;
+
+    virtual std::optional<boost::json::object>
+    requestFromRippled(
+        boost::json::object const& request,
+        std::string const& clientIp,
+        boost::asio::yield_context yield) const = 0;
+};
+
+/**
+ * @brief Hooks for source events such as connects and disconnects.
+ */
+struct SourceHooks
+{
+    enum class Action { STOP, PROCEED };
+
+    std::function<Action(boost::beast::error_code)> onConnected;
+    std::function<Action(boost::beast::error_code)> onDisconnected;
+};
+
+/**
+ * @brief Base implementation of shared source logic.
+ *
+ * @tparam Derived The derived class for CRTP
+ */
+template <class Derived>
+class SourceImpl : public Source
+{
+    std::string wsPort_;
+    std::string grpcPort_;
+
+    std::vector<std::pair<uint32_t, uint32_t>> validatedLedgers_;
+    std::string validatedLedgersRaw_{"N/A"};
+    std::shared_ptr<NetworkValidatedLedgers> networkValidatedLedgers_;
+
+    mutable std::mutex mtx_;
+
+    // true if this ETL source is forwarding transactions received on the transactions_proposed stream. There are
+    // usually multiple ETL sources, so to avoid forwarding the same transaction multiple times, we only forward from
+    // one particular ETL source at a time.
+    std::atomic_bool forwardingStream_{false};
+
+    std::chrono::system_clock::time_point lastMsgTime_;
+    mutable std::mutex lastMsgTimeMtx_;
+
+    std::shared_ptr<BackendInterface> backend_;
+    std::shared_ptr<feed::SubscriptionManager> subscriptions_;
+    LoadBalancer& balancer_;
+
+    etl::detail::ForwardCache forwardCache_;
+    boost::uuids::uuid uuid_;
+
+protected:
+    std::string ip_;
+    size_t numFailures_ = 0;
+
+    boost::asio::strand<boost::asio::io_context::executor_type> strand_;
+    boost::asio::steady_timer timer_;
+    boost::asio::ip::tcp::resolver resolver_;
+    boost::beast::flat_buffer readBuffer_;
+
+    std::unique_ptr<org::xrpl::rpc::v1::XRPLedgerAPIService::Stub> stub_;
+
+    std::atomic_bool closing_{false};
+    std::atomic_bool paused_{false};
+    std::atomic_bool connected_{false};
+
+    SourceHooks hooks_;
+
+public:
+    /**
+     * @brief Create the base portion of ETL source.
+     *
+     * @param config The configuration to use
+     * @param ioc The io_context to run on
+     * @param backend BackendInterface implementation
+     * @param subscriptions Subscription manager
+     * @param validatedLedgers The network validated ledgers datastructure
+     * @param balancer Load balancer to use
+     * @param hooks Hooks to use for connect/disconnect events
+     */
+    SourceImpl(
+        util::Config const& config,
+        boost::asio::io_context& ioc,
+        std::shared_ptr<BackendInterface> backend,
+        std::shared_ptr<feed::SubscriptionManager> subscriptions,
+        std::shared_ptr<NetworkValidatedLedgers> validatedLedgers,
+        LoadBalancer& balancer,
+        SourceHooks hooks)
+        : networkValidatedLedgers_(validatedLedgers)
+        , backend_(backend)
+        , subscriptions_(subscriptions)
+        , balancer_(balancer)
+        , forwardCache_(config, ioc, *this)
+        , strand_(boost::asio::make_strand(ioc))
+        , timer_(strand_)
+        , resolver_(strand_)
+        , hooks_(hooks)
+    {
+        static boost::uuids::random_generator uuidGenerator;
+        uuid_ = uuidGenerator();
+
+        ip_ = config.valueOr<std::string>("ip", {});
+        wsPort_ = config.valueOr<std::string>("ws_port", {});
+
+        if (auto value = config.maybeValue<std::string>("grpc_port"); value)
+        {
+            grpcPort_ = *value;
+            try
+            {
+                boost::asio::ip::tcp::endpoint endpoint{boost::asio::ip::make_address(ip_), std::stoi(grpcPort_)};
+                std::stringstream ss;
+                ss << endpoint;
+                grpc::ChannelArguments chArgs;
+                chArgs.SetMaxReceiveMessageSize(-1);
+                stub_ = org::xrpl::rpc::v1::XRPLedgerAPIService::NewStub(
+                    grpc::CreateCustomChannel(ss.str(), grpc::InsecureChannelCredentials(), chArgs));
+                LOG(log_.debug()) << "Made stub for remote = " << toString();
+            }
+            catch (std::exception const& e)
+            {
+                LOG(log_.debug()) << "Exception while creating stub = " << e.what() << " . Remote = " << toString();
+            }
+        }
+    }
+
+    ~SourceImpl()
+    {
+        derived().close(false);
+    }
+
+    bool
+    isConnected() const override
+    {
+        return connected_;
+    }
+
+    boost::uuids::uuid
+    token() const override
+    {
+        return uuid_;
+    }
+
+    std::optional<boost::json::object>
+    requestFromRippled(
+        boost::json::object const& request,
+        std::string const& clientIp,
+        boost::asio::yield_context yield) const override
+    {
+        LOG(log_.trace()) << "Attempting to forward request to tx. "
+                          << "request = " << boost::json::serialize(request);
+
+        boost::json::object response;
+        if (!isConnected())
+        {
+            LOG(log_.error()) << "Attempted to proxy but failed to connect to tx";
+            return {};
+        }
+
+        namespace beast = boost::beast;
+        namespace http = beast::http;
+        namespace websocket = beast::websocket;
+        namespace net = boost::asio;
+        using tcp = boost::asio::ip::tcp;
+
+        try
+        {
+            auto executor = boost::asio::get_associated_executor(yield);
+            boost::beast::error_code ec;
+            tcp::resolver resolver{executor};
+
+            auto ws = std::make_unique<websocket::stream<beast::tcp_stream>>(executor);
+
+            auto const results = resolver.async_resolve(ip_, wsPort_, yield[ec]);
+            if (ec)
+                return {};
+
+            ws->next_layer().expires_after(std::chrono::seconds(3));
+            ws->next_layer().async_connect(results, yield[ec]);
+            if (ec)
+                return {};
+
+            // Set a decorator to change the User-Agent of the handshake and to tell rippled to charge the client IP for
+            // RPC resources. See "secure_gateway" in
+            // https://github.com/ripple/rippled/blob/develop/cfg/rippled-example.cfg
+            ws->set_option(websocket::stream_base::decorator([&clientIp](websocket::request_type& req) {
+                req.set(http::field::user_agent, std::string(BOOST_BEAST_VERSION_STRING) + " websocket-client-coro");
+                req.set(http::field::forwarded, "for=" + clientIp);
+            }));
+
+            ws->async_handshake(ip_, "/", yield[ec]);
+            if (ec)
+                return {};
+
+            ws->async_write(net::buffer(boost::json::serialize(request)), yield[ec]);
+            if (ec)
+                return {};
+
+            beast::flat_buffer buffer;
+            ws->async_read(buffer, yield[ec]);
+            if (ec)
+                return {};
+
+            auto begin = static_cast<char const*>(buffer.data().data());
+            auto end = begin + buffer.data().size();
+            auto parsed = boost::json::parse(std::string(begin, end));
+
+            if (!parsed.is_object())
+            {
+                LOG(log_.error()) << "Error parsing response: " << std::string{begin, end};
+                return {};
+            }
+
+            response = parsed.as_object();
+            response["forwarded"] = true;
+
+            return response;
+        }
+        catch (std::exception const& e)
+        {
+            LOG(log_.error()) << "Encountered exception : " << e.what();
+            return {};
+        }
+    }
+
+    bool
+    hasLedger(uint32_t sequence) const override
+    {
+        std::lock_guard lck(mtx_);
+        for (auto& pair : validatedLedgers_)
+        {
+            if (sequence >= pair.first && sequence <= pair.second)
+            {
+                return true;
+            }
+            else if (sequence < pair.first)
+            {
+                // validatedLedgers_ is a sorted list of disjoint ranges
+                // if the sequence comes before this range, the sequence will
+                // come before all subsequent ranges
+                return false;
+            }
+        }
+        return false;
+    }
+
+    std::pair<grpc::Status, org::xrpl::rpc::v1::GetLedgerResponse>
+    fetchLedger(uint32_t sequence, bool getObjects = true, bool getObjectNeighbors = false) override
+    {
+        org::xrpl::rpc::v1::GetLedgerResponse response;
+        if (!stub_)
+            return {{grpc::StatusCode::INTERNAL, "No Stub"}, response};
+
+        // Ledger header with txns and metadata
+        org::xrpl::rpc::v1::GetLedgerRequest request;
+        grpc::ClientContext context;
+
+        request.mutable_ledger()->set_sequence(sequence);
+        request.set_transactions(true);
+        request.set_expand(true);
+        request.set_get_objects(getObjects);
+        request.set_get_object_neighbors(getObjectNeighbors);
+        request.set_user("ETL");
+
+        grpc::Status status = stub_->GetLedger(&context, request, &response);
+
+        if (status.ok() && !response.is_unlimited())
+        {
+            log_.warn()
+                << "is_unlimited is false. Make sure secure_gateway is set correctly on the ETL source. source = "
+                << toString() << "; status = " << status.error_message();
+        }
+
+        return {status, std::move(response)};
+    }
+
+    std::string
+    toString() const override
+    {
+        return "{validated_ledger: " + getValidatedRange() + ", ip: " + ip_ + ", web socket port: " + wsPort_ +
+            ", grpc port: " + grpcPort_ + "}";
+    }
+
+    boost::json::object
+    toJson() const override
+    {
+        boost::json::object res;
+
+        res["validated_range"] = getValidatedRange();
+        res["is_connected"] = std::to_string(isConnected());
+        res["ip"] = ip_;
+        res["ws_port"] = wsPort_;
+        res["grpc_port"] = grpcPort_;
+
+        auto last = getLastMsgTime();
+        if (last.time_since_epoch().count() != 0)
+            res["last_msg_age_seconds"] = std::to_string(
+                std::chrono::duration_cast<std::chrono::seconds>(std::chrono::system_clock::now() - getLastMsgTime())
+                    .count());
+
+        return res;
+    }
+
+    std::pair<std::vector<std::string>, bool>
+    loadInitialLedger(std::uint32_t sequence, std::uint32_t numMarkers, bool cacheOnly = false) override
+    {
+        if (!stub_)
+            return {{}, false};
+
+        grpc::CompletionQueue cq;
+        void* tag;
+        bool ok = false;
+        std::vector<etl::detail::AsyncCallData> calls;
+        auto markers = getMarkers(numMarkers);
+
+        for (size_t i = 0; i < markers.size(); ++i)
+        {
+            std::optional<ripple::uint256> nextMarker;
+
+            if (i + 1 < markers.size())
+                nextMarker = markers[i + 1];
+
+            calls.emplace_back(sequence, markers[i], nextMarker);
+        }
+
+        LOG(log_.debug()) << "Starting data download for ledger " << sequence << ". Using source = " << toString();
+
+        for (auto& c : calls)
+            c.call(stub_, cq);
+
+        size_t numFinished = 0;
+        bool abort = false;
+        size_t incr = 500000;
+        size_t progress = incr;
+        std::vector<std::string> edgeKeys;
+
+        while (numFinished < calls.size() && cq.Next(&tag, &ok))
+        {
+            assert(tag);
+            auto ptr = static_cast<etl::detail::AsyncCallData*>(tag);
+
+            if (!ok)
+            {
+                LOG(log_.error()) << "loadInitialLedger - ok is false";
+                return {{}, false};  // handle cancelled
+            }
+            else
+            {
+                LOG(log_.trace()) << "Marker prefix = " << ptr->getMarkerPrefix();
+
+                auto result = ptr->process(stub_, cq, *backend_, abort, cacheOnly);
+                if (result != etl::detail::AsyncCallData::CallStatus::MORE)
+                {
+                    ++numFinished;
+                    LOG(log_.debug()) << "Finished a marker. "
+                                      << "Current number of finished = " << numFinished;
+
+                    std::string lastKey = ptr->getLastKey();
+
+                    if (lastKey.size())
+                        edgeKeys.push_back(ptr->getLastKey());
+                }
+
+                if (result == etl::detail::AsyncCallData::CallStatus::ERRORED)
+                    abort = true;
+
+                if (backend_->cache().size() > progress)
+                {
+                    LOG(log_.info()) << "Downloaded " << backend_->cache().size() << " records from rippled";
+                    progress += incr;
+                }
+            }
+        }
+
+        LOG(log_.info()) << "Finished loadInitialLedger. cache size = " << backend_->cache().size();
+        return {std::move(edgeKeys), !abort};
+    }
+
+    std::optional<boost::json::object>
+    forwardToRippled(boost::json::object const& request, std::string const& clientIp, boost::asio::yield_context yield)
+        const override
+    {
+        if (auto resp = forwardCache_.get(request); resp)
+        {
+            LOG(log_.debug()) << "request hit forwardCache";
+            return resp;
+        }
+
+        return requestFromRippled(request, clientIp, yield);
+    }
+
+    void
+    pause() override
+    {
+        paused_ = true;
+        derived().close(false);
+    }
+
+    void
+    resume() override
+    {
+        paused_ = false;
+        derived().close(true);
+    }
+
+    /**
+     * @brief Callback for resolving the server host.
+     *
+     * @param ec The error code
+     * @param results Result of the resolve operation
+     */
+    void
+    onResolve(boost::beast::error_code ec, boost::asio::ip::tcp::resolver::results_type results)
+    {
+        if (ec)
+        {
+            // try again
+            reconnect(ec);
+        }
+        else
+        {
+            boost::beast::get_lowest_layer(derived().ws()).expires_after(std::chrono::seconds(30));
+            boost::beast::get_lowest_layer(derived().ws()).async_connect(results, [this](auto ec, auto ep) {
+                derived().onConnect(ec, ep);
+            });
+        }
+    }
+
+    /**
+     * @brief Callback for handshake with the server.
+     *
+     * @param ec The error code
+     */
+    void
+    onHandshake(boost::beast::error_code ec)
+    {
+        if (auto action = hooks_.onConnected(ec); action == SourceHooks::Action::STOP)
+            return;
+
+        if (ec)
+        {
+            // start over
+            reconnect(ec);
+        }
+        else
+        {
+            boost::json::object jv{
+                {"command", "subscribe"},
+                {"streams", {"ledger", "manifests", "validations", "transactions_proposed"}},
+            };
+            std::string s = boost::json::serialize(jv);
+            LOG(log_.trace()) << "Sending subscribe stream message";
+
+            derived().ws().set_option(
+                boost::beast::websocket::stream_base::decorator([](boost::beast::websocket::request_type& req) {
+                    req.set(
+                        boost::beast::http::field::user_agent,
+                        std::string(BOOST_BEAST_VERSION_STRING) + " clio-client");
+                    req.set("X-User", "coro-client");
+                }));
+
+            // Send subscription message
+            derived().ws().async_write(boost::asio::buffer(s), [this](auto ec, size_t size) { onWrite(ec, size); });
+        }
+    }
+
+    /**
+     * @brief Callback for writing data.
+     *
+     * @param ec The error code
+     * @param size Amount of bytes written
+     */
+    void
+    onWrite(boost::beast::error_code ec, [[maybe_unused]] size_t size)
+    {
+        if (ec)
+            reconnect(ec);
+        else
+            derived().ws().async_read(readBuffer_, [this](auto ec, size_t size) { onRead(ec, size); });
+    }
+
+    /**
+     * @brief Callback for data available to read.
+     *
+     * @param ec The error code
+     * @param size Amount of bytes read
+     */
+    void
+    onRead(boost::beast::error_code ec, size_t size)
+    {
+        if (ec)
+        {
+            reconnect(ec);
+        }
+        else
+        {
+            handleMessage(size);
+            derived().ws().async_read(readBuffer_, [this](auto ec, size_t size) { onRead(ec, size); });
+        }
+    }
+
+    /**
+     * @brief Handle the most recently received message.
+     *
+     * @param size Amount of bytes available in the read buffer
+     * @return true if the message was handled successfully; false otherwise
+     */
+    bool
+    handleMessage(size_t size)
+    {
+        setLastMsgTime();
+
+        try
+        {
+            auto const msg = boost::beast::buffers_to_string(readBuffer_.data());
+            readBuffer_.consume(size);
+
+            auto const raw = boost::json::parse(msg);
+            auto const response = raw.as_object();
+            uint32_t ledgerIndex = 0;
+
+            if (response.contains("result"))
+            {
+                auto const& result = response.at("result").as_object();
+                if (result.contains("ledger_index"))
+                    ledgerIndex = result.at("ledger_index").as_int64();
+
+                if (result.contains("validated_ledgers"))
+                {
+                    auto const& validatedLedgers = result.at("validated_ledgers").as_string();
+                    setValidatedRange({validatedLedgers.data(), validatedLedgers.size()});
+                }
+
+                LOG(log_.info()) << "Received a message on ledger "
+                                 << " subscription stream. Message : " << response << " - " << toString();
+            }
+            else if (response.contains("type") && response.at("type") == "ledgerClosed")
+            {
+                LOG(log_.info()) << "Received a message on ledger "
+                                 << " subscription stream. Message : " << response << " - " << toString();
+                if (response.contains("ledger_index"))
+                {
+                    ledgerIndex = response.at("ledger_index").as_int64();
+                }
+                if (response.contains("validated_ledgers"))
+                {
+                    auto const& validatedLedgers = response.at("validated_ledgers").as_string();
+                    setValidatedRange({validatedLedgers.data(), validatedLedgers.size()});
+                }
+            }
+            else
+            {
+                if (balancer_.shouldPropagateTxnStream(this))
+                {
+                    if (response.contains("transaction"))
+                    {
+                        forwardCache_.freshen();
+                        subscriptions_->forwardProposedTransaction(response);
+                    }
+                    else if (response.contains("type") && response.at("type") == "validationReceived")
+                    {
+                        subscriptions_->forwardValidation(response);
+                    }
+                    else if (response.contains("type") && response.at("type") == "manifestReceived")
+                    {
+                        subscriptions_->forwardManifest(response);
+                    }
+                }
+            }
+
+            if (ledgerIndex != 0)
+            {
+                LOG(log_.trace()) << "Pushing ledger sequence = " << ledgerIndex << " - " << toString();
+                networkValidatedLedgers_->push(ledgerIndex);
+            }
+
+            return true;
+        }
+        catch (std::exception const& e)
+        {
+            LOG(log_.error()) << "Exception in handleMessage : " << e.what();
+            return false;
+        }
+    }
+
+protected:
+    Derived&
+    derived()
+    {
+        return static_cast<Derived&>(*this);
+    }
+
+    void
+    run() override
+    {
+        resolver_.async_resolve(ip_, wsPort_, [this](auto ec, auto results) { onResolve(ec, results); });
+    }
+
+    void
+    reconnect(boost::beast::error_code ec)
+    {
+        if (paused_)
+            return;
+
+        if (isConnected())
+            hooks_.onDisconnected(ec);
+
+        connected_ = false;
+        readBuffer_ = {};
+
+        // These are somewhat normal errors. operation_aborted occurs on shutdown,
+        // when the timer is cancelled. connection_refused will occur repeatedly
+        std::string err = ec.message();
+        // if we cannot connect to the transaction processing process
+        if (ec.category() == boost::asio::error::get_ssl_category())
+        {
+            err = std::string(" (") + boost::lexical_cast<std::string>(ERR_GET_LIB(ec.value())) + "," +
+                boost::lexical_cast<std::string>(ERR_GET_REASON(ec.value())) + ") ";
+
+            // ERR_PACK /* crypto/err/err.h */
+            char buf[128];
+            ::ERR_error_string_n(ec.value(), buf, sizeof(buf));
+            err += buf;
+
+            LOG(log_.error()) << err;
+        }
+
+        if (ec != boost::asio::error::operation_aborted && ec != boost::asio::error::connection_refused)
+        {
+            LOG(log_.error()) << "error code = " << ec << " - " << toString();
+        }
+        else
+        {
+            LOG(log_.warn()) << "error code = " << ec << " - " << toString();
+        }
+
+        // exponentially increasing timeouts, with a max of 30 seconds
+        size_t waitTime = std::min(pow(2, numFailures_), 30.0);
+        numFailures_++;
+        timer_.expires_after(boost::asio::chrono::seconds(waitTime));
+        timer_.async_wait([this](auto ec) {
+            bool startAgain = (ec != boost::asio::error::operation_aborted);
+            derived().close(startAgain);
+        });
+    }
+
+private:
+    void
+    setLastMsgTime()
+    {
+        std::lock_guard lck(lastMsgTimeMtx_);
+        lastMsgTime_ = std::chrono::system_clock::now();
+    }
+
+    std::chrono::system_clock::time_point
+    getLastMsgTime() const
+    {
+        std::lock_guard lck(lastMsgTimeMtx_);
+        return lastMsgTime_;
+    }
+
+    void
+    setValidatedRange(std::string const& range)
+    {
+        std::vector<std::pair<uint32_t, uint32_t>> pairs;
+        std::vector<std::string> ranges;
+        boost::split(ranges, range, boost::is_any_of(","));
+        for (auto& pair : ranges)
+        {
+            std::vector<std::string> minAndMax;
+
+            boost::split(minAndMax, pair, boost::is_any_of("-"));
+
+            if (minAndMax.size() == 1)
+            {
+                uint32_t sequence = std::stoll(minAndMax[0]);
+                pairs.push_back(std::make_pair(sequence, sequence));
+            }
+            else
+            {
+                assert(minAndMax.size() == 2);
+                uint32_t min = std::stoll(minAndMax[0]);
+                uint32_t max = std::stoll(minAndMax[1]);
+                pairs.push_back(std::make_pair(min, max));
+            }
+        }
+        std::sort(pairs.begin(), pairs.end(), [](auto left, auto right) { return left.first < right.first; });
+
+        // we only hold the lock here, to avoid blocking while string processing
+        std::lock_guard lck(mtx_);
+        validatedLedgers_ = std::move(pairs);
+        validatedLedgersRaw_ = range;
+    }
+
+    std::string
+    getValidatedRange() const
+    {
+        std::lock_guard lck(mtx_);
+        return validatedLedgersRaw_;
+    }
+};
+
+/**
+ * @brief Implementation of a source that uses a regular, non-secure websocket connection.
+ */
+class PlainSource : public SourceImpl<PlainSource>
+{
+    using StreamType = boost::beast::websocket::stream<boost::beast::tcp_stream>;
+    std::unique_ptr<StreamType> ws_;
+
+public:
+    /**
+     * @brief Create a non-secure ETL source.
+     *
+     * @param config The configuration to use
+     * @param ioc The io_context to run on
+     * @param backend BackendInterface implementation
+     * @param subscriptions Subscription manager
+     * @param validatedLedgers The network validated ledgers datastructure
+     * @param balancer Load balancer to use
+     * @param hooks Hooks to use for connect/disconnect events
+     */
+    PlainSource(
+        util::Config const& config,
+        boost::asio::io_context& ioc,
+        std::shared_ptr<BackendInterface> backend,
+        std::shared_ptr<feed::SubscriptionManager> subscriptions,
+        std::shared_ptr<NetworkValidatedLedgers> validatedLedgers,
+        LoadBalancer& balancer,
+        SourceHooks hooks)
+        : SourceImpl(config, ioc, backend, subscriptions, validatedLedgers, balancer, std::move(hooks))
+        , ws_(std::make_unique<StreamType>(strand_))
+    {
+    }
+
+    /**
+     * @brief Callback for connection to the server.
+     *
+     * @param ec The error code
+     * @param endpoint The resolved endpoint
+     */
+    void
+    onConnect(boost::beast::error_code ec, boost::asio::ip::tcp::resolver::results_type::endpoint_type endpoint);
+
+    /**
+     * @brief Close the websocket.
+     *
+     * @param startAgain Whether to automatically reconnect
+     */
+    void
+    close(bool startAgain);
+
+    /** @return The underlying TCP stream */
+    StreamType&
+    ws()
+    {
+        return *ws_;
+    }
+};
+
+/**
+ * @brief Implementation of a source that uses a secure websocket connection.
+ */
+class SslSource : public SourceImpl<SslSource>
+{
+    using StreamType = boost::beast::websocket::stream<boost::beast::ssl_stream<boost::beast::tcp_stream>>;
+    std::optional<std::reference_wrapper<boost::asio::ssl::context>> sslCtx_;
+    std::unique_ptr<StreamType> ws_;
+
+public:
+    /**
+     * @brief Create a secure ETL source.
+     *
+     * @param config The configuration to use
+     * @param ioc The io_context to run on
+     * @param sslCtx The SSL context if any
+     * @param backend BackendInterface implementation
+     * @param subscriptions Subscription manager
+     * @param validatedLedgers The network validated ledgers datastructure
+     * @param balancer Load balancer to use
+     * @param hooks Hooks to use for connect/disconnect events
+     */
+    SslSource(
+        util::Config const& config,
+        boost::asio::io_context& ioc,
+        std::optional<std::reference_wrapper<boost::asio::ssl::context>> sslCtx,
+        std::shared_ptr<BackendInterface> backend,
+        std::shared_ptr<feed::SubscriptionManager> subscriptions,
+        std::shared_ptr<NetworkValidatedLedgers> validatedLedgers,
+        LoadBalancer& balancer,
+        SourceHooks hooks)
+        : SourceImpl(config, ioc, backend, subscriptions, validatedLedgers, balancer, std::move(hooks))
+        , sslCtx_(sslCtx)
+        , ws_(std::make_unique<StreamType>(strand_, *sslCtx_))
+    {
+    }
+
+    /**
+     * @brief Callback for connection to the server.
+     *
+     * @param ec The error code
+     * @param endpoint The resolved endpoint
+     */
+    void
+    onConnect(boost::beast::error_code ec, boost::asio::ip::tcp::resolver::results_type::endpoint_type endpoint);
+
+    /**
+     * @brief Callback for SSL handshake completion.
+     *
+     * @param ec The error code
+     * @param endpoint The resolved endpoint
+     */
+    void
+    onSslHandshake(boost::beast::error_code ec, boost::asio::ip::tcp::resolver::results_type::endpoint_type endpoint);
+
+    /**
+     * @brief Close the websocket.
+     *
+     * @param startAgain Whether to automatically reconnect
+     */
+    void
+    close(bool startAgain);
+
+    /** @return The underlying SSL stream */
+    StreamType&
+    ws()
+    {
+        return *ws_;
+    }
+};
+}  // namespace etl

src/etl/SystemState.h

@@ -0,0 +1,53 @@
+//------------------------------------------------------------------------------
+/*
+    This file is part of clio: https://github.com/XRPLF/clio
+    Copyright (c) 2023, the clio developers.
+
+    Permission to use, copy, modify, and distribute this software for any
+    purpose with or without fee is hereby granted, provided that the above
+    copyright notice and this permission notice appear in all copies.
+
+    THE  SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL WARRANTIES
+    WITH  REGARD  TO  THIS  SOFTWARE  INCLUDING  ALL  IMPLIED  WARRANTIES  OF
+    MERCHANTABILITY  AND  FITNESS. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR
+    ANY  SPECIAL,  DIRECT,  INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES
+    WHATSOEVER  RESULTING  FROM  LOSS  OF USE, DATA OR PROFITS, WHETHER IN AN
+    ACTION  OF  CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF
+    OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE.
+*/
+//==============================================================================
+
+#pragma once
+
+#include <atomic>
+
+namespace etl {
+
+/**
+ * @brief Represents the state of the ETL subsystem.
+ */
+struct SystemState
+{
+    /**
+     * @brief Whether the process is in strict read-only mode.
+     *
+     * In strict read-only mode, the process will never attempt to become the ETL writer, and will only publish ledgers
+     * as they are written to the database.
+     */
+    bool isReadOnly = false;
+
+    std::atomic_bool isWriting = false;     /**< @brief Whether the process is writing to the database. */
+    std::atomic_bool isStopping = false;    /**< @brief Whether the software is stopping. */
+    std::atomic_bool writeConflict = false; /**< @brief Whether a write conflict was detected. */
+
+    /**
+     * @brief Whether clio detected an amendment block.
+     *
+     * Being amendment blocked means that Clio was compiled with libxrpl that does not yet support some field that
+     * arrived from rippled and therefore can't extract the ledger diff. When this happens, Clio can't proceed with ETL
+     * and should log this error and only handle RPC requests.
+     */
+    std::atomic_bool isAmendmentBlocked = false;
+};
+
+}  // namespace etl

src/etl/impl/AmendmentBlock.h

@@ -0,0 +1,96 @@
+//------------------------------------------------------------------------------
+/*
+    This file is part of clio: https://github.com/XRPLF/clio
+    Copyright (c) 2023, the clio developers.
+
+    Permission to use, copy, modify, and distribute this software for any
+    purpose with or without fee is hereby granted, provided that the above
+    copyright notice and this permission notice appear in all copies.
+
+    THE  SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL WARRANTIES
+    WITH  REGARD  TO  THIS  SOFTWARE  INCLUDING  ALL  IMPLIED  WARRANTIES  OF
+    MERCHANTABILITY  AND  FITNESS. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR
+    ANY  SPECIAL,  DIRECT,  INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES
+    WHATSOEVER  RESULTING  FROM  LOSS  OF USE, DATA OR PROFITS, WHETHER IN AN
+    ACTION  OF  CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF
+    OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE.
+*/
+//==============================================================================
+
+#pragma once
+
+#include <etl/SystemState.h>
+#include <util/log/Logger.h>
+
+#include <boost/asio/io_context.hpp>
+#include <boost/asio/steady_timer.hpp>
+
+#include <chrono>
+#include <functional>
+
+namespace etl::detail {
+
+struct AmendmentBlockAction
+{
+    void
+    operator()()
+    {
+        static util::Logger log{"ETL"};
+        LOG(log.fatal())
+            << "Can't process new ledgers: The current ETL source is not compatible with the version of the "
+               "libxrpl Clio is currently using. Please upgrade Clio to a newer version.";
+    }
+};
+
+template <typename ActionCallableType = AmendmentBlockAction>
+class AmendmentBlockHandler
+{
+    std::reference_wrapper<boost::asio::io_context> ctx_;
+    std::reference_wrapper<SystemState> state_;
+    boost::asio::steady_timer timer_;
+    std::chrono::milliseconds interval_;
+
+    ActionCallableType action_;
+
+public:
+    template <typename DurationType = std::chrono::seconds>
+    AmendmentBlockHandler(
+        boost::asio::io_context& ioc,
+        SystemState& state,
+        DurationType interval = DurationType{1},
+        ActionCallableType&& action = ActionCallableType())
+        : ctx_{std::ref(ioc)}
+        , state_{std::ref(state)}
+        , timer_{ioc}
+        , interval_{std::chrono::duration_cast<std::chrono::milliseconds>(interval)}
+        , action_{std::move(action)}
+    {
+    }
+
+    ~AmendmentBlockHandler()
+    {
+        boost::asio::post(ctx_.get(), [this]() { timer_.cancel(); });
+    }
+
+    void
+    onAmendmentBlock()
+    {
+        state_.get().isAmendmentBlocked = true;
+        startReportingTimer();
+    }
+
+private:
+    void
+    startReportingTimer()
+    {
+        action_();
+
+        timer_.expires_after(interval_);
+        timer_.async_wait([this](auto ec) {
+            if (!ec)
+                boost::asio::post(ctx_.get(), [this] { startReportingTimer(); });
+        });
+    }
+};
+
+}  // namespace etl::detail

src/etl/impl/AsyncData.h

@@ -0,0 +1,182 @@
+//------------------------------------------------------------------------------
+/*
+    This file is part of clio: https://github.com/XRPLF/clio
+    Copyright (c) 2023, the clio developers.
+
+    Permission to use, copy, modify, and distribute this software for any
+    purpose with or without fee is hereby granted, provided that the above
+    copyright notice and this permission notice appear in all copies.
+
+    THE  SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL WARRANTIES
+    WITH  REGARD  TO  THIS  SOFTWARE  INCLUDING  ALL  IMPLIED  WARRANTIES  OF
+    MERCHANTABILITY  AND  FITNESS. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR
+    ANY  SPECIAL,  DIRECT,  INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES
+    WHATSOEVER  RESULTING  FROM  LOSS  OF USE, DATA OR PROFITS, WHETHER IN AN
+    ACTION  OF  CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF
+    OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE.
+*/
+//==============================================================================
+
+#pragma once
+
+#include <etl/NFTHelpers.h>
+#include <util/log/Logger.h>
+
+#include <ripple/proto/org/xrpl/rpc/v1/xrp_ledger.grpc.pb.h>
+#include <grpcpp/grpcpp.h>
+
+namespace etl::detail {
+
+class AsyncCallData
+{
+    util::Logger log_{"ETL"};
+
+    std::unique_ptr<org::xrpl::rpc::v1::GetLedgerDataResponse> cur_;
+    std::unique_ptr<org::xrpl::rpc::v1::GetLedgerDataResponse> next_;
+
+    org::xrpl::rpc::v1::GetLedgerDataRequest request_;
+    std::unique_ptr<grpc::ClientContext> context_;
+
+    grpc::Status status_;
+    unsigned char nextPrefix_;
+
+    std::string lastKey_;
+
+public:
+    AsyncCallData(uint32_t seq, ripple::uint256 const& marker, std::optional<ripple::uint256> const& nextMarker)
+    {
+        request_.mutable_ledger()->set_sequence(seq);
+        if (marker.isNonZero())
+        {
+            request_.set_marker(marker.data(), marker.size());
+        }
+        request_.set_user("ETL");
+        nextPrefix_ = 0x00;
+        if (nextMarker)
+            nextPrefix_ = nextMarker->data()[0];
+
+        unsigned char prefix = marker.data()[0];
+
+        LOG(log_.debug()) << "Setting up AsyncCallData. marker = " << ripple::strHex(marker)
+                          << " . prefix = " << ripple::strHex(std::string(1, prefix))
+                          << " . nextPrefix_ = " << ripple::strHex(std::string(1, nextPrefix_));
+
+        assert(nextPrefix_ > prefix || nextPrefix_ == 0x00);
+
+        cur_ = std::make_unique<org::xrpl::rpc::v1::GetLedgerDataResponse>();
+        next_ = std::make_unique<org::xrpl::rpc::v1::GetLedgerDataResponse>();
+        context_ = std::make_unique<grpc::ClientContext>();
+    }
+
+    enum class CallStatus { MORE, DONE, ERRORED };
+
+    CallStatus
+    process(
+        std::unique_ptr<org::xrpl::rpc::v1::XRPLedgerAPIService::Stub>& stub,
+        grpc::CompletionQueue& cq,
+        BackendInterface& backend,
+        bool abort,
+        bool cacheOnly = false)
+    {
+        LOG(log_.trace()) << "Processing response. "
+                          << "Marker prefix = " << getMarkerPrefix();
+        if (abort)
+        {
+            LOG(log_.error()) << "AsyncCallData aborted";
+            return CallStatus::ERRORED;
+        }
+        if (!status_.ok())
+        {
+            LOG(log_.error()) << "AsyncCallData status_ not ok: "
+                              << " code = " << status_.error_code() << " message = " << status_.error_message();
+            return CallStatus::ERRORED;
+        }
+        if (!next_->is_unlimited())
+        {
+            LOG(log_.warn()) << "AsyncCallData is_unlimited is false. Make sure "
+                                "secure_gateway is set correctly at the ETL source";
+        }
+
+        std::swap(cur_, next_);
+
+        bool more = true;
+
+        // if no marker returned, we are done
+        if (cur_->marker().size() == 0)
+            more = false;
+
+        // if returned marker is greater than our end, we are done
+        unsigned char prefix = cur_->marker()[0];
+        if (nextPrefix_ != 0x00 && prefix >= nextPrefix_)
+            more = false;
+
+        // if we are not done, make the next async call
+        if (more)
+        {
+            request_.set_marker(std::move(cur_->marker()));
+            call(stub, cq);
+        }
+
+        auto const numObjects = cur_->ledger_objects().objects_size();
+        LOG(log_.debug()) << "Writing " << numObjects << " objects";
+
+        std::vector<data::LedgerObject> cacheUpdates;
+        cacheUpdates.reserve(numObjects);
+
+        for (int i = 0; i < numObjects; ++i)
+        {
+            auto& obj = *(cur_->mutable_ledger_objects()->mutable_objects(i));
+            if (!more && nextPrefix_ != 0x00)
+            {
+                if (static_cast<unsigned char>(obj.key()[0]) >= nextPrefix_)
+                    continue;
+            }
+            cacheUpdates.push_back(
+                {*ripple::uint256::fromVoidChecked(obj.key()),
+                 {obj.mutable_data()->begin(), obj.mutable_data()->end()}});
+            if (!cacheOnly)
+            {
+                if (lastKey_.size())
+                    backend.writeSuccessor(std::move(lastKey_), request_.ledger().sequence(), std::string{obj.key()});
+                lastKey_ = obj.key();
+                backend.writeNFTs(getNFTDataFromObj(request_.ledger().sequence(), obj.key(), obj.data()));
+                backend.writeLedgerObject(
+                    std::move(*obj.mutable_key()), request_.ledger().sequence(), std::move(*obj.mutable_data()));
+            }
+        }
+        backend.cache().update(cacheUpdates, request_.ledger().sequence(), cacheOnly);
+        LOG(log_.debug()) << "Wrote " << numObjects << " objects. Got more: " << (more ? "YES" : "NO");
+
+        return more ? CallStatus::MORE : CallStatus::DONE;
+    }
+
+    void
+    call(std::unique_ptr<org::xrpl::rpc::v1::XRPLedgerAPIService::Stub>& stub, grpc::CompletionQueue& cq)
+    {
+        context_ = std::make_unique<grpc::ClientContext>();
+
+        std::unique_ptr<grpc::ClientAsyncResponseReader<org::xrpl::rpc::v1::GetLedgerDataResponse>> rpc(
+            stub->PrepareAsyncGetLedgerData(context_.get(), request_, &cq));
+
+        rpc->StartCall();
+
+        rpc->Finish(next_.get(), &status_, this);
+    }
+
+    std::string
+    getMarkerPrefix()
+    {
+        if (next_->marker().size() == 0)
+            return "";
+        else
+            return ripple::strHex(std::string{next_->marker().data()[0]});
+    }
+
+    std::string
+    getLastKey()
+    {
+        return lastKey_;
+    }
+};
+
+}  // namespace etl::detail

src/etl/impl/CacheLoader.h

@@ -0,0 +1,435 @@
+//------------------------------------------------------------------------------
+/*
+    This file is part of clio: https://github.com/XRPLF/clio
+    Copyright (c) 2023, the clio developers.
+
+    Permission to use, copy, modify, and distribute this software for any
+    purpose with or without fee is hereby granted, provided that the above
+    copyright notice and this permission notice appear in all copies.
+
+    THE  SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL WARRANTIES
+    WITH  REGARD  TO  THIS  SOFTWARE  INCLUDING  ALL  IMPLIED  WARRANTIES  OF
+    MERCHANTABILITY  AND  FITNESS. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR
+    ANY  SPECIAL,  DIRECT,  INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES
+    WHATSOEVER  RESULTING  FROM  LOSS  OF USE, DATA OR PROFITS, WHETHER IN AN
+    ACTION  OF  CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF
+    OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE.
+*/
+//==============================================================================
+
+#pragma once
+
+#include <data/BackendInterface.h>
+#include <util/log/Logger.h>
+
+#include <ripple/proto/org/xrpl/rpc/v1/xrp_ledger.grpc.pb.h>
+#include <boost/algorithm/string.hpp>
+#include <boost/asio/spawn.hpp>
+#include <boost/beast/core.hpp>
+#include <boost/beast/core/string.hpp>
+#include <boost/beast/websocket.hpp>
+#include <grpcpp/grpcpp.h>
+
+#include <chrono>
+#include <mutex>
+#include <thread>
+
+namespace etl::detail {
+
+/**
+ * @brief Cache loading interface
+ */
+template <typename CacheType>
+class CacheLoader
+{
+    enum class LoadStyle { ASYNC, SYNC, NOT_AT_ALL };
+
+    util::Logger log_{"ETL"};
+
+    std::reference_wrapper<boost::asio::io_context> ioContext_;
+    std::shared_ptr<BackendInterface> backend_;
+    std::reference_wrapper<CacheType> cache_;
+    LoadStyle cacheLoadStyle_ = LoadStyle::ASYNC;
+
+    // number of diffs to use to generate cursors to traverse the ledger in parallel during initial cache download
+    size_t numCacheDiffs_ = 32;
+
+    // number of markers to use at one time to traverse the ledger in parallel during initial cache download
+    size_t numCacheMarkers_ = 48;
+
+    // number of ledger objects to fetch concurrently per marker during cache download
+    size_t cachePageFetchSize_ = 512;
+
+    struct ClioPeer
+    {
+        std::string ip;
+        int port;
+    };
+
+    std::vector<ClioPeer> clioPeers_;
+
+    std::thread thread_;
+    std::atomic_bool stopping_ = false;
+
+public:
+    CacheLoader(
+        util::Config const& config,
+        boost::asio::io_context& ioc,
+        std::shared_ptr<BackendInterface> const& backend,
+        CacheType& ledgerCache)
+        : ioContext_{std::ref(ioc)}, backend_{backend}, cache_{ledgerCache}
+    {
+        if (config.contains("cache"))
+        {
+            auto const cache = config.section("cache");
+            if (auto entry = cache.maybeValue<std::string>("load"); entry)
+            {
+                if (boost::iequals(*entry, "sync"))
+                    cacheLoadStyle_ = LoadStyle::SYNC;
+                if (boost::iequals(*entry, "async"))
+                    cacheLoadStyle_ = LoadStyle::ASYNC;
+                if (boost::iequals(*entry, "none") or boost::iequals(*entry, "no"))
+                    cacheLoadStyle_ = LoadStyle::NOT_AT_ALL;
+            }
+
+            numCacheDiffs_ = cache.valueOr<size_t>("num_diffs", numCacheDiffs_);
+            numCacheMarkers_ = cache.valueOr<size_t>("num_markers", numCacheMarkers_);
+            cachePageFetchSize_ = cache.valueOr<size_t>("page_fetch_size", cachePageFetchSize_);
+
+            if (auto peers = cache.maybeArray("peers"); peers)
+            {
+                for (auto const& peer : *peers)
+                {
+                    auto ip = peer.value<std::string>("ip");
+                    auto port = peer.value<uint32_t>("port");
+
+                    // todo: use emplace_back when clang is ready
+                    clioPeers_.push_back({ip, port});
+                }
+
+                unsigned seed = std::chrono::system_clock::now().time_since_epoch().count();
+                std::shuffle(std::begin(clioPeers_), std::end(clioPeers_), std::default_random_engine(seed));
+            }
+        }
+    }
+
+    ~CacheLoader()
+    {
+        stop();
+        if (thread_.joinable())
+            thread_.join();
+    }
+
+    /**
+     * @brief Populates the cache by walking through the given ledger.
+     *
+     * Should only be called once. The default behavior is to return immediately and populate the cache in the
+     * background. This can be overridden via config parameter, to populate synchronously, or not at all.
+     */
+    void
+    load(uint32_t seq)
+    {
+        if (cacheLoadStyle_ == LoadStyle::NOT_AT_ALL)
+        {
+            cache_.get().setDisabled();
+            LOG(log_.warn()) << "Cache is disabled. Not loading";
+            return;
+        }
+
+        if (cache_.get().isFull())
+        {
+            assert(false);
+            return;
+        }
+
+        if (clioPeers_.size() > 0)
+        {
+            boost::asio::spawn(ioContext_.get(), [this, seq](boost::asio::yield_context yield) {
+                for (auto const& peer : clioPeers_)
+                {
+                    // returns true on success
+                    if (loadCacheFromClioPeer(seq, peer.ip, std::to_string(peer.port), yield))
+                        return;
+                }
+
+                // if we couldn't successfully load from any peers, load from db
+                loadCacheFromDb(seq);
+            });
+            return;
+        }
+        else
+        {
+            loadCacheFromDb(seq);
+        }
+
+        // If loading synchronously, poll cache until full
+        while (cacheLoadStyle_ == LoadStyle::SYNC && not cache_.get().isFull())
+        {
+            LOG(log_.debug()) << "Cache not full. Cache size = " << cache_.get().size() << ". Sleeping ...";
+            std::this_thread::sleep_for(std::chrono::seconds(10));
+            if (cache_.get().isFull())
+                LOG(log_.info()) << "Cache is full. Cache size = " << cache_.get().size();
+        }
+    }
+
+    void
+    stop()
+    {
+        stopping_ = true;
+    }
+
+private:
+    bool
+    loadCacheFromClioPeer(
+        uint32_t ledgerIndex,
+        std::string const& ip,
+        std::string const& port,
+        boost::asio::yield_context yield)
+    {
+        LOG(log_.info()) << "Loading cache from peer. ip = " << ip << " . port = " << port;
+        namespace beast = boost::beast;          // from <boost/beast.hpp>
+        namespace http = beast::http;            // from <boost/beast/http.hpp>
+        namespace websocket = beast::websocket;  // from
+        namespace net = boost::asio;             // from
+        using tcp = boost::asio::ip::tcp;        // from
+        try
+        {
+            boost::beast::error_code ec;
+            // These objects perform our I/O
+            tcp::resolver resolver{ioContext_.get()};
+
+            LOG(log_.trace()) << "Creating websocket";
+            auto ws = std::make_unique<websocket::stream<beast::tcp_stream>>(ioContext_.get());
+
+            // Look up the domain name
+            auto const results = resolver.async_resolve(ip, port, yield[ec]);
+            if (ec)
+                return {};
+
+            LOG(log_.trace()) << "Connecting websocket";
+            // Make the connection on the IP address we get from a lookup
+            ws->next_layer().async_connect(results, yield[ec]);
+            if (ec)
+                return false;
+
+            LOG(log_.trace()) << "Performing websocket handshake";
+            // Perform the websocket handshake
+            ws->async_handshake(ip, "/", yield[ec]);
+            if (ec)
+                return false;
+
+            std::optional<boost::json::value> marker;
+
+            LOG(log_.trace()) << "Sending request";
+            auto getRequest = [&](auto marker) {
+                boost::json::object request = {
+                    {"command", "ledger_data"},
+                    {"ledger_index", ledgerIndex},
+                    {"binary", true},
+                    {"out_of_order", true},
+                    {"limit", 2048}};
+
+                if (marker)
+                    request["marker"] = *marker;
+                return request;
+            };
+
+            bool started = false;
+            size_t numAttempts = 0;
+            do
+            {
+                // Send the message
+                ws->async_write(net::buffer(boost::json::serialize(getRequest(marker))), yield[ec]);
+                if (ec)
+                {
+                    LOG(log_.error()) << "error writing = " << ec.message();
+                    return false;
+                }
+
+                beast::flat_buffer buffer;
+                ws->async_read(buffer, yield[ec]);
+                if (ec)
+                {
+                    LOG(log_.error()) << "error reading = " << ec.message();
+                    return false;
+                }
+
+                auto raw = beast::buffers_to_string(buffer.data());
+                auto parsed = boost::json::parse(raw);
+
+                if (!parsed.is_object())
+                {
+                    LOG(log_.error()) << "Error parsing response: " << raw;
+                    return false;
+                }
+                LOG(log_.trace()) << "Successfully parsed response " << parsed;
+
+                if (auto const& response = parsed.as_object(); response.contains("error"))
+                {
+                    LOG(log_.error()) << "Response contains error: " << response;
+                    auto const& err = response.at("error");
+                    if (err.is_string() && err.as_string() == "lgrNotFound")
+                    {
+                        ++numAttempts;
+                        if (numAttempts >= 5)
+                        {
+                            LOG(log_.error()) << " ledger not found at peer after 5 attempts. "
+                                                 "peer = "
+                                              << ip << " ledger = " << ledgerIndex
+                                              << ". Check your config and the health of the peer";
+                            return false;
+                        }
+                        LOG(log_.warn()) << "Ledger not found. ledger = " << ledgerIndex
+                                         << ". Sleeping and trying again";
+                        std::this_thread::sleep_for(std::chrono::seconds(1));
+                        continue;
+                    }
+                    return false;
+                }
+                started = true;
+                auto const& response = parsed.as_object()["result"].as_object();
+
+                if (!response.contains("cache_full") || !response.at("cache_full").as_bool())
+                {
+                    LOG(log_.error()) << "cache not full for clio node. ip = " << ip;
+                    return false;
+                }
+                if (response.contains("marker"))
+                    marker = response.at("marker");
+                else
+                    marker = {};
+
+                auto const& state = response.at("state").as_array();
+
+                std::vector<data::LedgerObject> objects;
+                objects.reserve(state.size());
+                for (auto const& ledgerObject : state)
+                {
+                    auto const& obj = ledgerObject.as_object();
+
+                    data::LedgerObject stateObject = {};
+
+                    if (!stateObject.key.parseHex(obj.at("index").as_string().c_str()))
+                    {
+                        LOG(log_.error()) << "failed to parse object id";
+                        return false;
+                    }
+                    boost::algorithm::unhex(obj.at("data").as_string().c_str(), std::back_inserter(stateObject.blob));
+                    objects.push_back(std::move(stateObject));
+                }
+                cache_.get().update(objects, ledgerIndex, true);
+
+                if (marker)
+                    LOG(log_.debug()) << "At marker " << *marker;
+            } while (marker || !started);
+
+            LOG(log_.info()) << "Finished downloading ledger from clio node. ip = " << ip;
+
+            cache_.get().setFull();
+            return true;
+        }
+        catch (std::exception const& e)
+        {
+            LOG(log_.error()) << "Encountered exception : " << e.what() << " - ip = " << ip;
+            return false;
+        }
+    }
+
+    void
+    loadCacheFromDb(uint32_t seq)
+    {
+        std::vector<data::LedgerObject> diff;
+        std::vector<std::optional<ripple::uint256>> cursors;
+
+        auto append = [](auto&& a, auto&& b) { a.insert(std::end(a), std::begin(b), std::end(b)); };
+
+        for (size_t i = 0; i < numCacheDiffs_; ++i)
+        {
+            append(diff, data::synchronousAndRetryOnTimeout([&](auto yield) {
+                       return backend_->fetchLedgerDiff(seq - i, yield);
+                   }));
+        }
+
+        std::sort(diff.begin(), diff.end(), [](auto a, auto b) {
+            return a.key < b.key || (a.key == b.key && a.blob.size() < b.blob.size());
+        });
+
+        diff.erase(std::unique(diff.begin(), diff.end(), [](auto a, auto b) { return a.key == b.key; }), diff.end());
+
+        cursors.push_back({});
+        for (auto const& obj : diff)
+            if (obj.blob.size())
+                cursors.push_back({obj.key});
+        cursors.push_back({});
+
+        std::stringstream cursorStr;
+        for (auto const& c : cursors)
+            if (c)
+                cursorStr << ripple::strHex(*c) << ", ";
+
+        LOG(log_.info()) << "Loading cache. num cursors = " << cursors.size() - 1;
+        LOG(log_.trace()) << "cursors = " << cursorStr.str();
+
+        thread_ = std::thread{[this, seq, cursors = std::move(cursors)]() {
+            auto startTime = std::chrono::system_clock::now();
+            auto markers = std::make_shared<std::atomic_int>(0);
+            auto numRemaining = std::make_shared<std::atomic_int>(cursors.size() - 1);
+
+            for (size_t i = 0; i < cursors.size() - 1; ++i)
+            {
+                auto const start = cursors.at(i);
+                auto const end = cursors.at(i + 1);
+
+                markers->wait(numCacheMarkers_);
+                ++(*markers);
+
+                boost::asio::spawn(
+                    ioContext_.get(),
+                    [this, seq, start, end, numRemaining, startTime, markers](boost::asio::yield_context yield) {
+                        auto cursor = start;
+                        std::string cursorStr =
+                            cursor.has_value() ? ripple::strHex(cursor.value()) : ripple::strHex(data::firstKey);
+                        LOG(log_.debug()) << "Starting a cursor: " << cursorStr << " markers = " << *markers;
+
+                        while (not stopping_)
+                        {
+                            auto res = data::retryOnTimeout([this, seq, &cursor, yield]() {
+                                return backend_->fetchLedgerPage(cursor, seq, cachePageFetchSize_, false, yield);
+                            });
+
+                            cache_.get().update(res.objects, seq, true);
+
+                            if (!res.cursor || (end && *(res.cursor) > *end))
+                                break;
+
+                            LOG(log_.trace()) << "Loading cache. cache size = " << cache_.get().size()
+                                              << " - cursor = " << ripple::strHex(res.cursor.value())
+                                              << " start = " << cursorStr << " markers = " << *markers;
+
+                            cursor = std::move(res.cursor);
+                        }
+
+                        --(*markers);
+                        markers->notify_one();
+
+                        if (--(*numRemaining) == 0)
+                        {
+                            auto endTime = std::chrono::system_clock::now();
+                            auto duration = std::chrono::duration_cast<std::chrono::seconds>(endTime - startTime);
+
+                            LOG(log_.info()) << "Finished loading cache. cache size = " << cache_.get().size()
+                                             << ". Took " << duration.count() << " seconds";
+                            cache_.get().setFull();
+                        }
+                        else
+                        {
+                            LOG(log_.info()) << "Finished a cursor. num remaining = " << *numRemaining
+                                             << " start = " << cursorStr << " markers = " << *markers;
+                        }
+                    });
+            }
+        }};
+    }
+};
+
+}  // namespace etl::detail

src/etl/impl/ExtractionDataPipe.h

@@ -0,0 +1,134 @@
+//------------------------------------------------------------------------------
+/*
+    This file is part of clio: https://github.com/XRPLF/clio
+    Copyright (c) 2023, the clio developers.
+
+    Permission to use, copy, modify, and distribute this software for any
+    purpose with or without fee is hereby granted, provided that the above
+    copyright notice and this permission notice appear in all copies.
+
+    THE  SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL WARRANTIES
+    WITH  REGARD  TO  THIS  SOFTWARE  INCLUDING  ALL  IMPLIED  WARRANTIES  OF
+    MERCHANTABILITY  AND  FITNESS. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR
+    ANY  SPECIAL,  DIRECT,  INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES
+    WHATSOEVER  RESULTING  FROM  LOSS  OF USE, DATA OR PROFITS, WHETHER IN AN
+    ACTION  OF  CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF
+    OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE.
+*/
+//==============================================================================
+
+#pragma once
+
+#include <etl/ETLHelpers.h>
+#include <util/log/Logger.h>
+
+#include <memory>
+#include <vector>
+
+namespace etl::detail {
+
+/**
+ * @brief A collection of thread safe async queues used by Extractor and Transformer to communicate
+ */
+template <typename RawDataType>
+class ExtractionDataPipe
+{
+public:
+    using DataType = std::optional<RawDataType>;
+    using QueueType = ThreadSafeQueue<DataType>;  // TODO: probably should use boost::lockfree::queue instead?
+
+    constexpr static auto TOTAL_MAX_IN_QUEUE = 1000u;
+
+private:
+    util::Logger log_{"ETL"};
+
+    uint32_t stride_;
+    uint32_t startSequence_;
+
+    std::vector<std::shared_ptr<QueueType>> queues_;
+
+public:
+    /**
+     * @brief Create a new instance of the extraction data pipe
+     *
+     * @param stride
+     * @param startSequence
+     */
+    ExtractionDataPipe(uint32_t stride, uint32_t startSequence) : stride_{stride}, startSequence_{startSequence}
+    {
+        auto const maxQueueSize = TOTAL_MAX_IN_QUEUE / stride;
+        for (size_t i = 0; i < stride_; ++i)
+            queues_.push_back(std::make_unique<QueueType>(maxQueueSize));
+    }
+
+    /**
+     * @brief Push new data package for the specified sequence.
+     *
+     * Note: Potentially blocks until the underlying queue can accomodate another entry.
+     *
+     * @param sequence The sequence for which to enqueue the data package
+     * @param data The data to store
+     */
+    void
+    push(uint32_t sequence, DataType&& data)
+    {
+        getQueue(sequence)->push(std::move(data));
+    }
+
+    /**
+     * @brief Get data package for the given sequence
+     *
+     * Note: Potentially blocks until data is available.
+     *
+     * @param sequence The sequence for which data is required
+     * @return The data wrapped in an optional; nullopt means that there is no more data to expect
+     */
+    DataType
+    popNext(uint32_t sequence)
+    {
+        return getQueue(sequence)->pop();
+    }
+
+    /**
+     * @return Get the stride
+     */
+    uint32_t
+    getStride() const
+    {
+        return stride_;
+    }
+
+    /**
+     * @brief Hint the Transformer that the queue is done sending data
+     * @param sequence The sequence for which the extractor queue is to be hinted
+     */
+    void
+    finish(uint32_t sequence)
+    {
+        // empty optional hints the Transformer to shut down
+        push(sequence, std::nullopt);
+    }
+
+    /**
+     * @brief Unblocks internal queues
+     *
+     * Note: For now this must be called by the ETL when Transformer exits.
+     */
+    void
+    cleanup()
+    {
+        // TODO: this should not have to be called by hand. it should be done via RAII
+        for (auto i = 0u; i < stride_; ++i)
+            getQueue(i)->tryPop();  // pop from each queue that might be blocked on a push
+    }
+
+private:
+    std::shared_ptr<QueueType>
+    getQueue(uint32_t sequence)
+    {
+        LOG(log_.debug()) << "Grabbing extraction queue for " << sequence << "; start was " << startSequence_;
+        return queues_[(sequence - startSequence_) % stride_];
+    }
+};
+
+}  // namespace etl::detail