Account_tx bug fix (#694)

Fixes #693

.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

.dockerignore

@@ -0,0 +1 @@
+build/

.git-blame-ignore-revs

@@ -6,3 +6,5 @@
 
 # clang-format
 e41150248a97e4bdc1cf21b54650c4bb7c63928e
+2e542e7b0d94451a933c88778461cc8d3d7e6417
+d816ef54abd8e8e979b9c795bdb657a8d18f5e95

.githooks/ensure_release_tag

@@ -0,0 +1,20 @@
+#!/bin/bash
+
+# Pushing a release branch requires an annotated tag at the released commit
+branch=$(git rev-parse --abbrev-ref HEAD)
+
+if [[ $branch =~ master ]]; then
+  # check if HEAD commit is tagged
+  if ! git describe --exact-match HEAD; then
+    echo "Commits to master must be tagged"
+    exit 1
+  fi
+elif [[ $branch =~ release/* ]]; then
+  IFS=/ read -r branch rel_ver <<< ${branch}
+  tag=$(git describe --tags --abbrev=0)
+  if [[ "${rel_ver}" != "${tag}" ]]; then
+    echo "release/${rel_ver} branches must have annotated tag ${rel_ver}"
+    echo "git tag -am\"${rel_ver}\" ${rel_ver}"
+    exit 1
+  fi
+fi

.githooks/pre-commit

@@ -0,0 +1,27 @@
+#!/bin/bash
+
+exec 1>&2
+
+# paths to check and re-format
+sources="src unittests"
+formatter="clang-format-11 -i"
+
+first=$(git diff $sources)
+find $sources -type f \( -name '*.cpp' -o -name '*.h' -o -name '*.ipp' \) -print0 | xargs -0 $formatter
+second=$(git diff $sources)
+changes=$(diff <(echo "$first") <(echo "$second") | wc -l | sed -e 's/^[[:space:]]*//')
+
+if [ "$changes" != "0" ]; then
+    cat <<\EOF
+
+                                   WARNING
+-----------------------------------------------------------------------------
+  Automatically re-formatted code with `clang-format` - commit was aborted.
+  Please manually add any updated files and commit again.
+-----------------------------------------------------------------------------
+
+EOF
+    exit 1
+fi
+
+.githooks/ensure_release_tag

.github/actions/lint/action.yml

@@ -0,0 +1,13 @@
+runs:
+  using: composite
+  steps:
+    # Github's ubuntu-20.04 image already has clang-format-11 installed
+    - run: |
+        find src unittests -type f \( -name '*.cpp' -o -name '*.h' -o -name '*.ipp' \) -print0 | xargs -0 clang-format-11 -i
+      shell: bash
+
+    - name: Check for differences
+      id: assert
+      shell: bash
+      run: |
+        git diff --color --exit-code | tee "clang-format.patch"

.github/actions/test/Dockerfile

@@ -0,0 +1,6 @@
+FROM cassandra:4.0.4
+
+RUN apt-get update && apt-get install -y postgresql
+COPY entrypoint.sh /entrypoint.sh
+
+ENTRYPOINT ["/entrypoint.sh"]

.github/actions/test/entrypoint.sh

@@ -0,0 +1,8 @@
+#!/bin/bash
+
+pg_ctlcluster 12 main start
+su postgres -c"psql -c\"alter user postgres with password 'postgres'\""
+su cassandra -c "/opt/cassandra/bin/cassandra -R"
+sleep 90
+chmod +x ./clio_tests
+./clio_tests

.github/workflows/build.yml

@@ -1,43 +1,212 @@
 name: Build Clio
 on:
   push:
-    branches:  [master, develop]
+    branches: [master, release/*, develop, develop-next]
   pull_request:
-    branches:  [master, develop]
-  
+    branches: [master, release/*, develop, develop-next]
+  workflow_dispatch:
+
 jobs:
   lint:
     name: Lint
     runs-on: ubuntu-20.04
     steps:
-      - name: Get source
-        uses: actions/checkout@v3
-      
+      - uses: actions/checkout@v3
       - name: Run clang-format
-        uses: XRPLF/clio-build/lint@main
-  
-  build_clio: 
-    name: Build
-    runs-on: ubuntu-20.04
+        uses: ./.github/actions/lint
+
+  build_clio:
+    name: Build Clio
+    runs-on: [self-hosted, heavy]
     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: ${{ matrix.type.image }}
     steps:
-      - name: Get clio repo
+      - uses: actions/checkout@v3
+        with:
+          path: clio
+          fetch-depth: 0
+
+      - name: Clone Clio packaging repo
         uses: actions/checkout@v3
         with:
-          path: clio_src
-      
-      - name: Get gha repo
-        uses: actions/checkout@v3
-        with:
-          repository: XRPLF/clio-build
+          path: clio-packages
+          repository: XRPLF/clio-packages
           ref: main
-          path: gha # must be the same as defined in XRPLF/clio-build
-      
+
       - name: Build
-        uses: XRPLF/clio-build/build@main
-      
-      # - name: Artifact clio_tests
-      #   uses: actions/upload-artifact@v2
-      #   with:
-      #     name: clio_output
-      #     path: clio_src/build/clio_tests
+        shell: bash
+        run: |
+          export CLIO_ROOT=$(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
+        uses: actions/upload-artifact@v3
+        with:
+          name: clio_${{ matrix.type.suffix }}_packages
+          path: ${{ github.workspace }}/*.${{ matrix.type.suffix }}
+
+      - name: Artifact clio_tests
+        uses: actions/upload-artifact@v3
+        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: [self-hosted, macOS]
+
+    steps:
+      - uses: actions/checkout@v3
+        with:
+          path: clio
+
+      - name: Check Boost cache
+        id: boost
+        uses: actions/cache@v3
+        with:
+          path: boost_1_77_0
+          key: ${{ runner.os }}-boost
+
+      - name: Build Boost
+        if: ${{ steps.boost.outputs.cache-hit != 'true' }}
+        run: |
+          rm -rf boost_1_77_0.tar.gz boost_1_77_0 # cleanup if needed first
+          curl -s -fOJL "https://boostorg.jfrog.io/artifactory/main/release/1.77.0/source/boost_1_77_0.tar.gz"
+          tar zxf boost_1_77_0.tar.gz
+          cd boost_1_77_0
+          ./bootstrap.sh
+          ./b2 define=BOOST_ASIO_HAS_STD_INVOKE_RESULT cxxflags="-std=c++20"
+
+      - name: Install dependencies
+        run: |
+          brew install llvm@14 pkg-config protobuf openssl ninja cassandra-cpp-driver bison cmake
+
+      - name: Setup environment for llvm-14
+        run: |
+          export PATH="/usr/local/opt/llvm@14/bin:$PATH"
+          export LDFLAGS="-L/usr/local/opt/llvm@14/lib -L/usr/local/opt/llvm@14/lib/c++ -Wl,-rpath,/usr/local/opt/llvm@14/lib/c++"
+          export CPPFLAGS="-I/usr/local/opt/llvm@14/include"
+
+      - name: Build clio
+        run: |
+          export BOOST_ROOT=$(pwd)/boost_1_77_0
+          cd clio
+          cmake -B build -DCMAKE_C_COMPILER='/usr/local/opt/llvm@14/bin/clang' -DCMAKE_CXX_COMPILER='/usr/local/opt/llvm@14/bin/clang++'
+          if ! cmake --build build -j; then
+            echo '# 🔥🔥 MacOS AppleClang build failed!💥' >> $GITHUB_STEP_SUMMARY
+            exit 1
+          fi
+      - name: Run Test
+        run: |
+          cd clio/build
+          ./clio_tests --gtest_filter="-BackendCassandraBaseTest*:BackendCassandraTest*:BackendCassandraFactoryTestWithDB*"
+
+  test_clio:
+    name: Test Clio
+    runs-on: [self-hosted, Linux]
+    needs: build_clio
+    strategy:
+      fail-fast: false
+      matrix:
+        suffix: [rpm, deb]
+    steps:
+      - uses: actions/checkout@v3
+
+      - name: Get clio_tests artifact
+        uses: actions/download-artifact@v3
+        with:
+          name: clio_tests-${{ 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: ${{ runner.os }}-boost
+
+      - name: Build boost
+        if: steps.boost.outputs.cache-hit != 'true'
+        run: |
+          curl -s -OJL "https://boostorg.jfrog.io/artifactory/main/release/1.77.0/source/boost_1_77_0.tar.gz"
+          tar zxf boost_1_77_0.tar.gz
+          mv boost_1_77_0 boost
+          cd boost
+          ./bootstrap.sh
+          ./b2
+      - name: install deps
+        run: |
+          sudo apt-get -y install git pkg-config protobuf-compiler libprotobuf-dev libssl-dev wget build-essential doxygen bison flex autoconf clang-format gcovr
+      - name: Build clio
+        run: |
+          export BOOST_ROOT=$(pwd)/boost
+          cd clio
+          cmake -B build -DCODE_COVERAGE=on -DTEST_PARAMETER='--gtest_filter="-BackendCassandraBaseTest*:BackendCassandraTest*:BackendCassandraFactoryTestWithDB*"'
+          if ! cmake --build build -j$(nproc); then
+            echo '# 🔥Ubuntu build🔥 failed!💥' >> $GITHUB_STEP_SUMMARY
+            exit 1
+          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
+          cat code-coverage-results.md > $GITHUB_STEP_SUMMARY
+      - name: Upload coverage reports to Codecov
+        uses: codecov/codecov-action@v3
+        with:
+          files: clio/build/clio_tests-gcc-cov/out.xml
+
+      - uses: actions/upload-artifact@v3
+        with:
+          name: UnitTestCoverage
+          path: UnitTestCoverage/
+
+      - uses: actions/upload-artifact@v3
+        with:
+          name: code_coverage_report
+          path: clio/build/clio_tests-gcc-cov/out.xml

.gitignore

@@ -1,2 +1,6 @@
-build/
+*clio*.log
+build*/
+.vscode
 .python-version
+config.json
+src/main/impl/Build.cpp

CMake/Build.cpp.in

@@ -0,0 +1,39 @@
+//------------------------------------------------------------------------------
+/*
+    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 <main/Build.h>
+
+namespace Build {
+static constexpr char versionString[] = "@VERSION@";
+
+std::string const&
+getClioVersionString()
+{
+    static std::string const value = versionString;
+    return value;
+}
+
+std::string const&
+getClioFullVersionString()
+{
+    static std::string const value = "clio-" + getClioVersionString();
+    return value;
+}
+
+}  // namespace Build

CMake/ClioVersion.cmake

@@ -0,0 +1,33 @@
+#[===================================================================[
+   write version to source
+#]===================================================================]
+
+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 branch --show-current)
+execute_process(COMMAND ${GIT_EXECUTABLE} ${GIT_COMMAND} OUTPUT_VARIABLE BRANCH OUTPUT_STRIP_TRAILING_WHITESPACE)
+
+if(BRANCH STREQUAL "")
+  set(BRANCH "dev")
+endif()
+
+if(NOT (BRANCH MATCHES master OR BRANCH MATCHES release/*)) # for develop and any other branch name YYYYMMDDHMS-<branch>-<git-ref>
+  execute_process(COMMAND date +%Y%m%d%H%M%S OUTPUT_VARIABLE DATE OUTPUT_STRIP_TRAILING_WHITESPACE)
+  set(VERSION "${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}-${REV}")
+endif()
+
+if(CMAKE_BUILD_TYPE MATCHES Debug)
+  set(VERSION "${VERSION}+DEBUG")
+endif()
+
+message(STATUS "Build version: ${VERSION}")
+set(clio_version "${VERSION}")
+
+configure_file(CMake/Build.cpp.in ${CMAKE_SOURCE_DIR}/src/main/impl/Build.cpp)

CMake/coverage.cmake

@@ -0,0 +1,126 @@
+# call add_converage(module_name) to add coverage targets for the given module
+function(add_converage module)
+  if("${CMAKE_C_COMPILER_ID}" MATCHES "(Apple)?[Cc]lang"
+     OR "${CMAKE_CXX_COMPILER_ID}" MATCHES "(Apple)?[Cc]lang")
+    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 ${LLVM_COV_PATH}/llvm-cov)
+      message(FATAL_ERROR "llvm-cov not found! Aborting.")
+    endif()
+
+    # set Flags
+    target_compile_options(${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)
+
+    # llvm-cov
+    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
+      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
+      COMMAND
+        ${LLVM_COV_PATH}/llvm-cov report $<TARGET_FILE:${module}>
+        -instr-profile=${module}.profdata
+        -ignore-filename-regex=".*_makefiles|.*unittests|.*_deps"
+        -show-region-summary=false
+      DEPENDS ${module}-ccov-preprocessing)
+
+    # exclude libs and unittests self
+    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|.*_deps" > /dev/null 2>&1
+      DEPENDS ${module}-ccov-preprocessing)
+
+    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")
+
+    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
+                                             -fPIC)
+    target_link_libraries(${module} PRIVATE gcov)
+
+    target_compile_options(clio PRIVATE -fprofile-arcs -ftest-coverage
+                                             -fPIC)
+    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
+      COMMAND ${module} ${TEST_PARAMETER}
+      COMMAND rm -rf ${COV_OUTPUT_PATH}
+      COMMAND mkdir ${COV_OUTPUT_PATH}
+      COMMAND
+        gcovr -r ${CMAKE_SOURCE_DIR} --object-directory=${PROJECT_BINARY_DIR} -x
+        ${COV_OUTPUT_PATH}/out.xml --exclude='${CMAKE_SOURCE_DIR}/unittests/'
+        --exclude='${PROJECT_BINARY_DIR}/'
+      COMMAND
+        gcovr -r ${CMAKE_SOURCE_DIR} --object-directory=${PROJECT_BINARY_DIR}
+        --html ${COV_OUTPUT_PATH}/report.html
+        --exclude='${CMAKE_SOURCE_DIR}/unittests/'
+        --exclude='${PROJECT_BINARY_DIR}/'
+      WORKING_DIRECTORY ${PROJECT_BINARY_DIR}
+      COMMENT "Running gcovr to produce Cobertura code coverage report.")
+
+    # generate the detail report
+    add_custom_target(
+      ${module}-ccov-report
+      COMMAND ${module} ${TEST_PARAMETER}
+      COMMAND rm -rf ${COV_OUTPUT_PATH}
+      COMMAND mkdir ${COV_OUTPUT_PATH}
+      COMMAND
+        gcovr -r ${CMAKE_SOURCE_DIR} --object-directory=${PROJECT_BINARY_DIR}
+        --html-details ${COV_OUTPUT_PATH}/index.html
+        --exclude='${CMAKE_SOURCE_DIR}/unittests/'
+        --exclude='${PROJECT_BINARY_DIR}/'
+      WORKING_DIRECTORY ${PROJECT_BINARY_DIR}
+      COMMENT "Running gcovr to produce Cobertura code coverage report.")
+    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()

CMake/deps/Postgres.cmake

@@ -1,30 +0,0 @@
-set(POSTGRES_INSTALL_DIR ${CMAKE_BINARY_DIR}/postgres)
-set(POSTGRES_LIBS pq pgcommon pgport)
-ExternalProject_Add(postgres
-    GIT_REPOSITORY https://github.com/postgres/postgres.git
-    GIT_TAG REL_14_1
-    GIT_SHALLOW 1
-    LOG_CONFIGURE 1
-    LOG_BUILD 1
-    CONFIGURE_COMMAND ./configure --prefix ${POSTGRES_INSTALL_DIR} --without-readline --verbose
-    BUILD_COMMAND ${CMAKE_COMMAND} -E env --unset=MAKELEVEL make VERBOSE=${CMAKE_VERBOSE_MAKEFILE} -j32
-    BUILD_IN_SOURCE 1
-    INSTALL_COMMAND  ${CMAKE_COMMAND} -E env make -s --no-print-directory install
-    UPDATE_COMMAND ""
-    BUILD_BYPRODUCTS
-        ${POSTGRES_INSTALL_DIR}/lib/${CMAKE_STATIC_LIBRARY_PREFIX}pq${CMAKE_STATIC_LIBRARY_SUFFIX}}
-        ${POSTGRES_INSTALL_DIR}/lib/${CMAKE_STATIC_LIBRARY_PREFIX}pgcommon${CMAKE_STATIC_LIBRARY_SUFFIX}}
-        ${POSTGRES_INSTALL_DIR}/lib/${CMAKE_STATIC_LIBRARY_PREFIX}pgport${CMAKE_STATIC_LIBRARY_SUFFIX}}
-        )
-
-foreach(_lib ${POSTGRES_LIBS})
-    add_library(${_lib} STATIC IMPORTED GLOBAL)
-    add_dependencies(${_lib} postgres)
-    set_target_properties(${_lib} PROPERTIES
-        IMPORTED_LOCATION ${POSTGRES_INSTALL_DIR}/lib/${CMAKE_STATIC_LIBRARY_PREFIX}${_lib}.a)
-    set_target_properties(${_lib} PROPERTIES
-        INTERFACE_INCLUDE_DIRECTORIES ${POSTGRES_INSTALL_DIR}/include)
-    target_link_libraries(clio PUBLIC ${POSTGRES_INSTALL_DIR}/lib/${CMAKE_STATIC_LIBRARY_PREFIX}${_lib}${CMAKE_STATIC_LIBRARY_SUFFIX})
-endforeach()
-
-target_include_directories(clio PUBLIC ${POSTGRES_INSTALL_DIR}/include)

CMake/deps/Remove-bitset-operator.patch

@@ -0,0 +1,24 @@
+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

@@ -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/cassandra.cmake

@@ -10,7 +10,7 @@ if(NOT cassandra)
         ExternalProject_Add(zlib_src
             PREFIX ${nih_cache_path}
             GIT_REPOSITORY https://github.com/madler/zlib.git
-            GIT_TAG master
+            GIT_TAG v1.2.12
             INSTALL_COMMAND ""
             BUILD_BYPRODUCTS <BINARY_DIR>/${CMAKE_STATIC_LIBRARY_PREFIX}z.a
             )
@@ -33,7 +33,7 @@ if(NOT cassandra)
         ExternalProject_Add(krb5_src
             PREFIX ${nih_cache_path}
             GIT_REPOSITORY https://github.com/krb5/krb5.git
-            GIT_TAG master
+            GIT_TAG krb5-1.20
             UPDATE_COMMAND ""
             CONFIGURE_COMMAND autoreconf src && CFLAGS=-fcommon ./src/configure --enable-static --disable-shared
             BUILD_IN_SOURCE 1
@@ -66,7 +66,7 @@ if(NOT cassandra)
         ExternalProject_Add(libuv_src
             PREFIX ${nih_cache_path}
             GIT_REPOSITORY https://github.com/libuv/libuv.git
-            GIT_TAG v1.x
+            GIT_TAG v1.44.1
             INSTALL_COMMAND ""
             BUILD_BYPRODUCTS <BINARY_DIR>/${CMAKE_STATIC_LIBRARY_PREFIX}uv_a.a
             )
@@ -89,7 +89,7 @@ if(NOT cassandra)
     ExternalProject_Add(cassandra_src
         PREFIX ${nih_cache_path}
         GIT_REPOSITORY https://github.com/datastax/cpp-driver.git
-        GIT_TAG master
+        GIT_TAG 2.16.2
         CMAKE_ARGS
         -DLIBUV_ROOT_DIR=${BINARY_DIR}
         -DLIBUV_INCLUDE_DIR=${SOURCE_DIR}/include
@@ -144,8 +144,10 @@ if(NOT cassandra)
 else()
     message("Found system installed cassandra cpp driver")
     message(${cassandra})
-
     find_path(cassandra_includes NAMES cassandra.h REQUIRED)
+    message(${cassandra_includes})
+    get_filename_component(CASSANDRA_HEADER ${cassandra_includes}/cassandra.h REALPATH)
+    get_filename_component(CASSANDRA_HEADER_DIR ${CASSANDRA_HEADER} DIRECTORY)
     target_link_libraries (clio PUBLIC ${cassandra})
-    target_include_directories(clio INTERFACE ${cassandra_includes})
+    target_include_directories(clio PUBLIC ${CASSANDRA_HEADER_DIR})
 endif()

CMake/deps/gtest.cmake

@@ -10,10 +10,13 @@ if(NOT googletest_POPULATED)
   add_subdirectory(${googletest_SOURCE_DIR} ${googletest_BINARY_DIR} EXCLUDE_FROM_ALL)
 endif()
 
-target_link_libraries(clio_tests PUBLIC clio gtest_main)
+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)
+#increase timeout for tests discovery to 10 seconds, by default it is 5s. As more unittests added, we start to hit this issue
+#https://github.com/google/googletest/issues/3475
+gtest_discover_tests(clio_tests DISCOVERY_TIMEOUT 10)

CMake/deps/libfmt.cmake

@@ -0,0 +1,14 @@
+FetchContent_Declare(
+  libfmt
+  URL https://github.com/fmtlib/fmt/releases/download/9.1.0/fmt-9.1.0.zip
+)
+
+FetchContent_GetProperties(libfmt)
+
+if(NOT libfmt_POPULATED)
+  FetchContent_Populate(libfmt)
+  add_subdirectory(${libfmt_SOURCE_DIR} ${libfmt_BINARY_DIR} EXCLUDE_FROM_ALL)
+endif()
+
+target_link_libraries(clio PUBLIC fmt)
+

CMake/deps/rippled.cmake

@@ -1,11 +1,13 @@
 set(RIPPLED_REPO "https://github.com/ripple/rippled.git")
-set(RIPPLED_BRANCH "develop")
+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)

CMake/install/PostInstall.cmake

@@ -1,4 +0,0 @@
-execute_process(
-  COMMAND ln -s ${CMAKE_INSTALL_PREFIX}/${CLIO_INSTALL_DIR}/bin/clio_server /usr/local/bin/clio_server
-  COMMAND_ECHO STDOUT
-  )

CMake/install/clio.service.in

@@ -0,0 +1,17 @@
+[Unit]
+Description=Clio XRPL API server
+Documentation=https://github.com/XRPLF/clio.git
+
+After=network-online.target
+Wants=network-online.target
+
+[Service]
+Type=simple
+ExecStart=@CLIO_INSTALL_DIR@/bin/clio_server @CLIO_INSTALL_DIR@/etc/config.json
+Restart=on-failure
+User=clio
+Group=clio
+LimitNOFILE=65536
+
+[Install]
+WantedBy=multi-user.target

CMake/install/install.cmake

@@ -1,7 +1,16 @@
-set(CLIO_INSTALL_DIR /opt/clio)
+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(FILES example-config.json DESTINATION etc/clio)
-install(SCRIPT "${CMAKE_SOURCE_DIR}/CMake/install/PostInstall.cmake")
+
+#install(FILES example-config.json DESTINATION etc RENAME config.json)
+file(READ example-config.json config)
+string(REGEX REPLACE "./clio_log" "/var/log/clio/" config "${config}")
+file(WRITE ${CMAKE_BINARY_DIR}/install-config.json "${config}")
+install(FILES ${CMAKE_BINARY_DIR}/install-config.json DESTINATION etc RENAME config.json)
+
+configure_file("${CMAKE_SOURCE_DIR}/CMake/install/clio.service.in" "${CMAKE_BINARY_DIR}/clio.service")
+
+install(FILES "${CMAKE_BINARY_DIR}/clio.service" DESTINATION /lib/systemd/system)
+

CMake/settings.cmake

@@ -1 +1,6 @@
-set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -Wno-narrowing -Wall -Werror -Wno-dangling-else")
+target_compile_options(clio  
+  PUBLIC -Wall
+         -Werror 
+         -Wno-narrowing
+         -Wno-deprecated-declarations 
+         -Wno-dangling-else)

CMake/version-config.h

@@ -1 +0,0 @@
-#define VERSION "@PROJECT_VERSION@"

CMakeLists.txt

@@ -1,6 +1,10 @@
 cmake_minimum_required(VERSION 3.16.3)
 
-project(clio VERSION 0.1.1)
+project(clio)
+
+if(CMAKE_CXX_COMPILER_VERSION VERSION_LESS 11)
+  message(FATAL_ERROR "GCC 11+ required for building clio")
+endif()
 
 option(BUILD_TESTS "Build tests" TRUE)
 
@@ -10,6 +14,14 @@ if(VERBOSE)
   set(FETCHCONTENT_QUIET FALSE CACHE STRING "Verbose FetchContent()")
 endif()
 
+
+if(PACKAGING)
+  add_definitions(-DPKG=1)
+endif()
+
+#c++20 removed std::result_of but boost 1.75 is still using it.
+add_definitions(-DBOOST_ASIO_HAS_STD_INVOKE_RESULT=1)
+
 add_library(clio)
 target_compile_features(clio PUBLIC cxx_std_20)
 target_include_directories(clio PUBLIC src)
@@ -17,67 +29,155 @@ target_include_directories(clio PUBLIC src)
 include(FetchContent)
 include(ExternalProject)
 include(CMake/settings.cmake)
+include(CMake/ClioVersion.cmake)
 include(CMake/deps/rippled.cmake)
+include(CMake/deps/libfmt.cmake)
 include(CMake/deps/Boost.cmake)
 include(CMake/deps/cassandra.cmake)
-include(CMake/deps/Postgres.cmake)
-
-# configure_file(CMake/version-config.h include/version.h) # NOTE: Not used, but an idea how to handle versioning.
+include(CMake/deps/SourceLocation.cmake)
 
 target_sources(clio PRIVATE
+  ## Main
+  src/main/impl/Build.cpp
   ## Backend
   src/backend/BackendInterface.cpp
-  src/backend/CassandraBackend.cpp
-  src/backend/LayeredCache.cpp
-  src/backend/Pg.cpp
-  src/backend/PostgresBackend.cpp
-  src/backend/SimpleCache.cpp
+  src/backend/LedgerCache.cpp
+  ## NextGen Backend
+  src/backend/cassandra/impl/Future.cpp
+  src/backend/cassandra/impl/Cluster.cpp
+  src/backend/cassandra/impl/Batch.cpp
+  src/backend/cassandra/impl/Result.cpp
+  src/backend/cassandra/impl/Tuple.cpp
+  src/backend/cassandra/impl/SslContext.cpp
+  src/backend/cassandra/Handle.cpp
+  src/backend/cassandra/SettingsProvider.cpp
   ## ETL
-  src/etl/ETLSource.cpp
-  src/etl/ReportingETL.cpp
+  src/etl/Source.cpp
+  src/etl/ProbingSource.cpp
+  src/etl/NFTHelpers.cpp
+  src/etl/ETLService.cpp
+  src/etl/LoadBalancer.cpp
+  src/etl/impl/ForwardCache.cpp
   ## Subscriptions
   src/subscriptions/SubscriptionManager.cpp
   ## RPC
-  src/rpc/RPC.cpp
+  src/rpc/Errors.cpp
+  src/rpc/Factories.cpp
   src/rpc/RPCHelpers.cpp
   src/rpc/Counters.cpp
-  ## RPC Methods
-  # Account
+  src/rpc/WorkQueue.cpp
+  src/rpc/common/Specs.cpp
+  src/rpc/common/Validators.cpp
+  # RPC impl
+  src/rpc/common/impl/HandlerProvider.cpp
+  ## RPC handler
   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/GatewayBalances.cpp
-  src/rpc/handlers/NoRippleCheck.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/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/rpc/handlers/TransactionEntry.cpp
-  src/rpc/handlers/AccountTx.cpp
-  # Dex
-  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
-  # Utility
-  src/rpc/handlers/Random.cpp)
+  src/rpc/handlers/Tx.cpp
+  ## Util
+  src/config/Config.cpp
+  src/log/Logger.cpp
+  src/util/Taggable.cpp)
 
-add_executable(clio_server src/main.cpp)
+add_executable(clio_server src/main/main.cpp)
 target_link_libraries(clio_server PUBLIC clio)
 
 if(BUILD_TESTS)
-  add_executable(clio_tests unittests/main.cpp)
+  set(TEST_TARGET clio_tests)
+  add_executable(${TEST_TARGET}
+    unittests/Playground.cpp
+    unittests/Logger.cpp
+    unittests/Config.cpp
+    unittests/ProfilerTest.cpp
+    unittests/DOSGuard.cpp
+    unittests/SubscriptionTest.cpp
+    unittests/SubscriptionManagerTest.cpp
+    unittests/util/TestObject.cpp
+    unittests/util/StringUtils.cpp
+    # ETL
+    unittests/etl/ExtractionDataPipeTest.cpp
+    unittests/etl/ExtractorTest.cpp
+    unittests/etl/TransformerTest.cpp
+    # RPC
+    unittests/rpc/ErrorTests.cpp
+    unittests/rpc/BaseTests.cpp
+    unittests/rpc/RPCHelpersTest.cpp
+    unittests/rpc/CountersTest.cpp
+    unittests/rpc/AdminVerificationTest.cpp
+    ## RPC handlers
+    unittests/rpc/handlers/DefaultProcessorTests.cpp
+    unittests/rpc/handlers/TestHandlerTests.cpp
+    unittests/rpc/handlers/AccountCurrenciesTest.cpp
+    unittests/rpc/handlers/AccountLinesTest.cpp
+    unittests/rpc/handlers/AccountTxTest.cpp
+    unittests/rpc/handlers/AccountOffersTest.cpp
+    unittests/rpc/handlers/AccountInfoTest.cpp
+    unittests/rpc/handlers/AccountChannelsTest.cpp
+    unittests/rpc/handlers/AccountNFTsTest.cpp
+    unittests/rpc/handlers/BookOffersTest.cpp
+    unittests/rpc/handlers/GatewayBalancesTest.cpp
+    unittests/rpc/handlers/TxTest.cpp
+    unittests/rpc/handlers/TransactionEntryTest.cpp
+    unittests/rpc/handlers/LedgerEntryTest.cpp
+    unittests/rpc/handlers/LedgerRangeTest.cpp
+    unittests/rpc/handlers/NoRippleCheckTest.cpp
+    unittests/rpc/handlers/ServerInfoTest.cpp
+    unittests/rpc/handlers/PingTest.cpp
+    unittests/rpc/handlers/RandomTest.cpp
+    unittests/rpc/handlers/NFTInfoTest.cpp
+    unittests/rpc/handlers/NFTBuyOffersTest.cpp
+    unittests/rpc/handlers/NFTSellOffersTest.cpp
+    unittests/rpc/handlers/NFTHistoryTest.cpp
+    unittests/rpc/handlers/SubscribeTest.cpp
+    unittests/rpc/handlers/UnsubscribeTest.cpp
+    unittests/rpc/handlers/LedgerDataTest.cpp
+    unittests/rpc/handlers/AccountObjectsTest.cpp
+    unittests/rpc/handlers/BookChangesTest.cpp
+    unittests/rpc/handlers/LedgerTest.cpp
+    # Backend
+    unittests/backend/BackendFactoryTest.cpp
+    unittests/backend/cassandra/BaseTests.cpp
+    unittests/backend/cassandra/BackendTests.cpp
+    unittests/backend/cassandra/RetryPolicyTests.cpp
+    unittests/backend/cassandra/SettingsProviderTests.cpp
+    unittests/backend/cassandra/ExecutionStrategyTests.cpp
+    unittests/backend/cassandra/AsyncExecutorTests.cpp
+    unittests/webserver/ServerTest.cpp
+    unittests/webserver/RPCExecutorTest.cpp)
   include(CMake/deps/gtest.cmake)
+
+  # test for dwarf5 bug on ci 
+  target_compile_options(clio PUBLIC -gdwarf-4)
+
+  # if CODE_COVERAGE enable, add clio_test-ccov
+  if(CODE_COVERAGE)
+    include(CMake/coverage.cmake)
+    add_converage(${TEST_TARGET})
+  endif()
 endif()
 
 include(CMake/install/install.cmake)
+if(PACKAGING)
+    include(CMake/packaging.cmake)
+endif()

CONTRIBUTING.md

@@ -0,0 +1,134 @@
+# Contributing
+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 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.
+8. Open a PR to the main repository onto the _develop_ branch, and follow the provided template.
+
+> **Note:** Please read the [Style guide](#style-guide).
+
+## Install git hooks
+Please run the following command in order to use git hooks that are helpful for `clio` development.
+
+``` bash
+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 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.
+
+``` bash
+# Create a backup of your branch
+git branch <your feature branch>_bk
+
+# Rebase and squash commits into one
+git checkout develop
+git pull origin develop
+git checkout <your feature branch>
+git rebase -i develop
+```
+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** 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
+```
+> **Note:** If there are merge conflicts, please resolve them now.
+
+``` bash
+# Use the same commit message as you did above
+git commit -m 'Your message'
+git rebase --continue
+```
+
+> **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 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 reviewer(s).
+This section describes the process of adding your fixes.
+
+We assume that you already made the required changes on your feature branch.
+
+``` bash
+# Add the changed code
+git add <paths to add>
+
+# 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
+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.
+
+## Formatting
+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
+* Proliferation of nearly identical code.
+* Proliferation of new files and classes unless it improves readability or/and compilation time.
+* Unmanaged memory allocation and raw pointers.
+* Macros (unless they add significant value.)
+* Lambda patterns (unless these add significant value.)
+* CPU or architecture-specific code unless there is a good reason to include it, and where it is used guard it with macros and provide explanatory comments.
+* Importing new libraries unless there is a very good reason to do so.
+
+## Seek to
+* Extend functionality of existing code rather than creating new code.
+* Prefer readability over terseness where important logic is concerned.
+* Inline functions that are not used or are not likely to be used elsewhere in the codebase.
+* Use clear and self-explanatory names for functions, variables, structs and classes.
+* Use TitleCase for classes, structs and filenames, camelCase for function and variable names, lower case for namespaces and folders.
+* Provide as many comments as you feel that a competent programmer would need to understand what your code does.
+
+# Maintainers
+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
+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.
+
+Existing maintainers can resign, or be subject to a vote for removal at the behest of two existing maintainers. A minimum of 60% agreement and 50% participation are required. The XRP Ledger Foundation will have the ability, for cause, to remove an existing maintainer without a vote.
+
+## Existing Maintainers
+
+* [cindyyan317](https://github.com/cindyyan317) (Ripple)
+* [godexsoft](https://github.com/godexsoft) (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

@@ -0,0 +1,3 @@
+PROJECT_NAME           = "Clio"
+INPUT                  = src
+RECURSIVE              = YES

README.md

@@ -1,9 +1,6 @@
-**Status:** This software is in beta mode. We encourage anyone to try it out and
-report any issues they discover. Version 1.0 coming soon.
-
 # 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 is 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,
@@ -12,9 +9,9 @@ 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 specified rippled node. Running Clio requires access to a rippled node
+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.
 
 
@@ -25,13 +22,15 @@ from which data can be extracted. The rippled node does not need to be running o
 
 ## Building
 
-Clio is built with cmake. Clio requires c++20, and boost 1.75.0 or later.
+Clio is built with CMake. Clio requires at least GCC-11/clang-14.0.0 (C++20), and Boost 1.75.0.
 
-Use these instructions to build a Clio executable from source. These instructions were tested on Ubuntu 20.04 LTS.
+Use these instructions to build a Clio executable from the source. These instructions were tested on Ubuntu 20.04 LTS.
 
-```
+```sh
 # Install dependencies
-  sudo apt-get -y install git pkg-config protobuf-compiler libprotobuf-dev libssl-dev wget build-essential bison flex autoconf cmake
+  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
 
 # 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
@@ -49,30 +48,62 @@ Use these instructions to build a Clio executable from source. These instruction
 ```
 
 ## Running
-`./clio_server config.json`
+```sh
+./clio_server config.json
+```
 
 Clio needs access to a rippled server. The config files of rippled and Clio need
 to match in a certain sense.
 Clio needs to know:
-- the ip of rippled
-- the port on which rippled is accepting unencrypted websocket connections
+- the IP of rippled
+- the port on which rippled is accepting unencrypted WebSocket connections
 - the port on which rippled is handling gRPC requests
 
 rippled needs to open:
 - a port to accept unencrypted websocket connections
-- a port to handle gRPC requests, with the ip(s) of Clio specified in the `secure_gateway` entry
+- a port to handle gRPC requests, with the IP(s) of Clio specified in the `secure_gateway` entry
 
-The example configs of rippled and Clio are setup such that minimal changes are
+The example configs of rippled and Clio are setups such that minimal changes are
 required. When running locally, the only change needed is to uncomment the `port_grpc`
 section of the rippled config. When running Clio and rippled on separate machines,
 in addition to uncommenting the `port_grpc` section, a few other steps must be taken:
-1. change the `ip` of the first entry of `etl_sources` to the ip where your rippled
+1. change the `ip` of the first entry of `etl_sources` to the IP where your rippled
 server is running
-2. open a public, unencrypted websocket port on your rippled server
-3. change the ip specified in `secure_gateway` of `port_grpc` section of the rippled config
-to the ip of your Clio server. This entry can take the form of a comma separated list if
+2. open a public, unencrypted WebSocket port on your rippled server
+3. change the IP specified in `secure_gateway` of `port_grpc` section of the rippled config
+to the IP of your Clio server. This entry can take the form of a comma-separated list if
 you are running multiple Clio nodes.
 
+
+In addition, the parameter `start_sequence` can be included and configured within the top level of the config file. This parameter specifies the sequence of first ledger to extract if the database is empty. Note that ETL extracts ledgers in order and that no backfilling functionality currently exists, meaning Clio will not retroactively learn ledgers older than the one you specify. Choosing to specify this or not will yield the following behavior:
+- If this setting is absent and the database is empty, ETL will start with the next ledger validated by the network. 
+- If this setting is present and the database is not empty, an exception is thrown.
+
+In addition, the optional parameter `finish_sequence` can be added to the json file as well, specifying where the ledger can stop.
+
+To add `start_sequence` and/or `finish_sequence` to the config.json file appropriately, they will be on the same top level of precedence as other parameters (such as `database`, `etl_sources`, `read_only`, etc.) and be specified with an integer. Here is an example snippet from the config file:
+
+```json
+"start_sequence": 12345,
+"finish_sequence": 54321
+```
+
+The parameters `ssl_cert_file` and `ssl_key_file` can also be added to the top level of precedence of our Clio config. `ssl_cert_file` specifies the filepath for your SSL cert while `ssl_key_file` specifies the filepath for your SSL key. It is up to you how to change ownership of these folders for your designated Clio user. Your options include:
+- Copying the two files as root somewhere that's accessible by the Clio user, then running `sudo chown` to your user
+- Changing the permissions directly so it's readable by your Clio user
+- Running Clio as root (strongly discouraged)
+
+An example of how to specify `ssl_cert_file` and `ssl_key_file` in the config:
+
+```json
+"server":{
+    "ip": "0.0.0.0",
+    "port": 51233
+},
+"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
 start first, and it's fine to stop one or the other and restart at any given time.
 
@@ -84,7 +115,7 @@ the most recent ledger on the network, and then backfill. If Clio is extracting
 from rippled, and then rippled is stopped for a significant amount of time and then restarted, rippled
 will take time to backfill to the next ledger that Clio wants. The time it takes is proportional
 to the amount of time rippled was offline for. Also be aware that the amount rippled backfills
-is dependent on the online_delete and ledger_history config values; if these values
+are dependent on the online_delete and ledger_history config values; if these values
 are small, and rippled is stopped for a significant amount of time, rippled may never backfill
 to the ledger that Clio wants. To avoid this situation, it is advised to keep history
 proportional to the amount of time that you expect rippled to be offline. For example, if you
@@ -106,7 +137,7 @@ This can take some time, and depends on database throughput. With a moderately f
 database, this should take less than 10 minutes. If you did not properly set `secure_gateway`
 in the `port_grpc` section of rippled, this step will fail. Once the first ledger
 is fully downloaded, Clio only needs to extract the changed data for each ledger,
-so extraction is much faster and Clio can keep up with rippled in real time. Even under
+so extraction is much faster and Clio can keep up with rippled in real-time. Even under
 intense load, Clio should not lag behind the network, as Clio is not processing the data,
 and is simply writing to a database. The throughput of Clio is dependent on the throughput
 of your database, but a standard Cassandra or Scylla deployment can handle
@@ -140,3 +171,68 @@ are doing this, be aware that database traffic will be flowing across regions,
 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.
+
+## Developing against `rippled` in standalone mode
+
+If you wish you develop against a `rippled` instance running in standalone
+mode there are a few quirks of both clio and rippled you need to keep in mind.
+You must:
+
+1. Advance the `rippled` ledger to at least ledger 256
+2. Wait 10 minutes before first starting clio against this standalone node.
+
+## Logging
+Clio provides several logging options, all are configurable via the config file and are detailed below.
+
+`log_level`: The minimum level of severity at which the log message will be outputted by default.
+Severity options are `trace`, `debug`, `info`, `warning`, `error`, `fatal`. Defaults to `info`.
+
+`log_format`: The format of log lines produced by clio. Defaults to `"%TimeStamp% (%SourceLocation%) [%ThreadID%] %Channel%:%Severity% %Message%"`.
+Each of the variables expands like so
+- `TimeStamp`: The full date and time of the log entry
+- `SourceLocation`: A partial path to the c++ file and the line number in said file (`source/file/path:linenumber`)  
+- `ThreadID`: The ID of the thread the log entry is written from
+- `Channel`: The channel that this log entry was sent to
+- `Severity`: The severity (aka log level) the entry was sent at
+- `Message`: The actual log message
+
+`log_channels`: An array of json objects, each overriding properties for a logging `channel`. 
+At the moment of writing, only `log_level` can be overriden using this mechanism.
+
+Each object is of this format:
+```json
+{
+    "channel": "Backend",
+    "log_level": "fatal"
+}
+``` 
+If no override is present for a given channel, that channel will log at the severity specified by the global `log_level`.
+Overridable log channels: `Backend`, `WebServer`, `Subscriptions`, `RPC`, `ETL` and `Performance`.
+
+> **Note:** See `example-config.json` for more details. 
+
+`log_to_console`: Enable/disable log output to console. Options are `true`/`false`. Defaults to true.
+
+`log_directory`: Path to the directory where log files are stored. If such directory doesn't exist, Clio will create it. If not specified, logs are not written to a file.
+
+`log_rotation_size`: The max size of the log file in **megabytes** before it will rotate into a smaller file. Defaults to 2GB.
+
+`log_directory_max_size`: The max size of the log directory in **megabytes** before old log files will be
+deleted to free up space. Defaults to 50GB.
+
+`log_rotation_hour_interval`: The time interval in **hours** after the last log rotation to automatically
+rotate the current log file. Defaults to 12 hours.
+
+Note, time-based log rotation occurs dependently on size-based log rotation, where if a
+size-based log rotation occurs, the timer for the time-based rotation will reset.
+
+`log_tag_style`: Tag implementation to use. Must be one of:
+- `uint`: Lock free and threadsafe but outputs just a simple unsigned integer 
+- `uuid`: Threadsafe and outputs a UUID tag
+- `none`: Don't use tagging at all
+
+## Cassandra / Scylla Administration
+
+Since Clio relies on either Cassandra or Scylla for its database backend, here are some important considerations:
+
+- Scylla, by default, will reserve all free RAM on a machine for itself. If you are running `rippled` or other services on the same machine, restrict its memory usage using the `--memory` argument: https://docs.scylladb.com/getting-started/scylla-in-a-shared-environment/

REVIEW.md

@@ -1,121 +0,0 @@
-# How to review clio
-Clio is a massive project, and thus I don't expect the code to be reviewed the
-way a normal PR would. So I put this guide together to help reviewers look at
-the relevant pieces of code without getting lost in the weeds.
-
-One thing reviewers should keep in mind is that most of clio is designed to be
-lightweight and simple. We try not to introduce any uneccessary complexity and
-keep the code as simple and straightforward as possible. Sometimes complexity is
-unavoidable, but simplicity is the goal.
-
-## Order of review
-The code is organized into 4 main components, each with their own folder. The
-code in each folder is as self contained as possible. A good way to approach
-the review would be to review one folder at a time.
-
-### backend
-The code in the backend folder is the heart of the project, and reviewers should
-start here. This is the most complex part of the code, as well as the most
-performance sensitive. clio does not keep any data in memory, so performance
-generally depends on the data model and the way we talk to the database.
-
-Reviewers should start with the README in this folder to get a high level idea
-of the data model and to review the data model itself. Then, reviewers should
-dive into the implementation. The table schemas and queries for Cassandra are 
-defined in `CassandraBackend::open()`. The table schemas for Postgres are defined
-in Pg.cpp. The queries for Postgres are defined in each of the functions of `PostgresBackend`.
-A good way to approach the implementation would be to look at the table schemas,
-and then go through the functions declared in `BackendInterface`. Reviewers could
-also branch out to the rest of the code by looking at where these functions are
-called from.
-
-### webserver
-The code in the webserver folder implements the web server for handling RPC requests.
-This code was mostly copied and pasted from boost beast example code, so I would
-really appreciate review here.
-
-### rpc
-The rpc folder contains all of the handlers and any helper functions they need.
-This code is not too complicated, so reviewers don't need to dwell long here.
-
-### etl
-The etl folder contains all of the code for extracting data from rippled. This
-code is complex and important, but most of this code was just copied from rippled
-reporting mode, and thus has already been reviewed and is being used in prod.
-
-## Design decisions that should be reviewed
-
-### Data model
-Reviewers should review the general data model. The data model itself is described
-at a high level in the README in the backend folder. The table schemas and queries
-for Cassandra are defined in the `open()` function of `CassandraBackend`. The table
-schemas for Postgres are defined in Pg.cpp. 
-
-Particular attention should be paid to the keys table, and the problem that solves
-(successor/upper bound). I originally was going to have a special table for book_offers,
-but then I decided that we could use the keys table itself for that and save space.
-This makes book_offers somewhat slow compared to rippled, though still very usable.
-
-### Large rows
-I did some tricks with Cassandra to deal with very large rows in the keys and account_tx
-tables. For each of these, the partition key (the first component of the primary
-key) is a compound key. This is meant to break large rows into smaller rows. This
-is done to avoid hotspots. Data is sharded in Cassandra, and if some rows get very
-large, some nodes can have a lot more data than others.
-
-For account_tx, this has performance implications when iterating very far back
-in time. Refer to the `fetchAccountTransactions()` function in `CassandraBackend`.
-
-It is unclear if this needs to be done for other tables.
-
-### Postgres table partitioning
-Originally, Postgres exhibited performance problems when the dataset approach 1
-TB. This was solved by table partitioning.
-
-### Threading
-I used asio for multithreading. There are a lot of different io_contexts lying
-around the code. This needs to be cleaned up a bit. Most of these are really
-just ways to submit an async job to a single thread. I don't think it makes
-sense to have one io_context for the whole application, but some of the threading
-is a bit opaque and could be cleaned up.
-
-### Boost Json
-I used boost json for serializing data to json.
-
-### No cache
-As of now, there is no cache. I am not sure if a cache is even worth it. A
-transaction cache would not be hard, but a cache for ledger data will be hard.
-While a cache would improve performance, it would increase memory usage. clio
-is designed to be lightweight. Also, I've reached thousands of requests per
-second with a single clio node, so I'm not sure performance is even an issue.
-
-## Things I'm less than happy about
-
-#### BackendIndexer
-This is a particularly hairy piece of code that handles writing to the keys table.
-I am not too happy with this code. Parts of it need to execute in real time as
-part of ETL, and other parts are allowed to run in the background. There is also
-code that detects if a previous background job failed to complete before the
-server shutdown, and thus tries to rerun that job. The code feels tacked on, and
-I would like it to be more cleanly integrated with the rest of the code.
-
-#### Shifting
-There is some bit shifting going on with the keys table and the account_tx table.
-The keys table is written to every 2^20 ledgers. Maybe it would be better to just
-write every 1 million ledgers. 
-
-#### performance of book_offers
-book_offers is a bit slow. It could be sped up in a variety of ways. One is to
-keep a separate book_offers table. However, this is not straightforward and will
-use more space. Another is to keep a cache of book_offers for the most recent ledger
-(or few ledgers). I am not sure if this is worth it
-
-#### account_tx in Cassandra
-After the fix to deal with large rows, account_tx can be slow at times when using
-Cassandra. Specifically, if there are large gaps in time where the account was
-not affected by any transactions, the code will be reading empty records. I would
-like to sidestep this issue if possible.
-
-#### Implementation of fetchLedgerPage
-`fetchLedgerPage()` is rather complex. Part of this seems unavoidable, since this
-code is dealing with the keys table.

cloud-example-config.json

@@ -1,38 +1,35 @@
 {
-    "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": {
+            "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
         }
     },
-    "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
 }

docker/centos/Dockerfile

@@ -0,0 +1,49 @@
+# FROM centos:7 as deps
+FROM centos:7 as build
+
+ENV CLIO_DIR=/opt/clio/
+# ENV OPENSSL_DIR=/opt/openssl
+
+RUN yum -y install git epel-release centos-release-scl perl-IPC-Cmd openssl
+RUN yum install -y devtoolset-11
+ENV version=3.16
+ENV build=3
+# RUN curl -OJL https://cmake.org/files/v$version/cmake-$version.$build.tar.gz
+COPY docker/shared/install_cmake.sh /install_cmake.sh
+RUN /install_cmake.sh 3.16.3 /usr/local
+RUN source /opt/rh/devtoolset-11/enable
+WORKDIR /tmp
+# RUN mkdir $OPENSSL_DIR && cd $OPENSSL_DIR
+COPY docker/centos/build_git_centos7.sh build_git_centos7.sh
+
+RUN ./build_git_centos7.sh
+RUN git clone https://github.com/openssl/openssl
+WORKDIR /tmp/openssl
+RUN git checkout OpenSSL_1_1_1q
+#--prefix=/usr --openssldir=/etc/ssl --libdir=lib no-shared zlib-dynamic
+RUN SSLDIR=$(openssl version -d | cut -d: -f2 | tr -d [:space:]\") && ./config -fPIC --prefix=/usr --openssldir=${SSLDIR} zlib shared && \
+    make -j $(nproc) && \
+    make install_sw
+WORKDIR /tmp
+# FROM centos:7 as build
+
+RUN git clone https://github.com/xrplf/clio.git
+COPY docker/shared/build_boost.sh build_boost.sh
+ENV OPENSSL_ROOT=/opt/local/openssl
+ENV BOOST_ROOT=/boost
+RUN source scl_source enable devtoolset-11 && /tmp/build_boost.sh 1.75.0
+RUN yum install -y bison flex
+RUN yum install -y rpmdevtools rpmlint
+RUN source /opt/rh/devtoolset-11/enable && cd /tmp/clio && \
+    cmake -B build -DBUILD_TESTS=1 && \
+    cmake --build build --parallel $(nproc)
+RUN mkdir output
+RUN strip clio/build/clio_server && strip clio/build/clio_tests
+RUN cp clio/build/clio_tests output/ && cp clio/build/clio_server output/
+RUN cp clio/example-config.json output/example-config.json
+
+FROM centos:7
+COPY --from=build /tmp/output /clio
+RUN mkdir -p /opt/clio/etc && mv /clio/example-config.json /opt/clio/etc/config.json
+
+CMD ["/clio/clio_server", "/opt/clio/etc/config.json"]

docker/centos/build_git_centos7.sh

@@ -0,0 +1,18 @@
+#!/usr/bin/env bash
+
+set -ex
+GIT_VERSION="2.37.1"
+curl -OJL https://github.com/git/git/archive/refs/tags/v${GIT_VERSION}.tar.gz
+tar zxvf git-${GIT_VERSION}.tar.gz
+cd git-${GIT_VERSION}
+
+yum install -y centos-release-scl epel-release
+yum update -y
+yum install -y devtoolset-11 autoconf gnu-getopt gettext zlib-devel  libcurl-devel
+
+source /opt/rh/devtoolset-11/enable
+make configure
+./configure
+make git -j$(nproc)
+make install git
+git  --version | cut -d ' ' -f3

docker/centos/install_cmake.sh

@@ -0,0 +1,11 @@
+#!/bin/bash
+
+set -eo pipefail
+
+CMAKE_VERSION=${1:-"3.16.3"}
+cd /tmp
+URL="https://github.com/Kitware/CMake/releases/download/v${CMAKE_VERSION}/cmake-${CMAKE_VERSION}-linux-x86_64.tar.gz"
+curl -OJLs $URL
+tar xzvf cmake-${CMAKE_VERSION}-Linux-x86_64.tar.gz
+mv cmake-${CMAKE_VERSION}-Linux-x86_64 /opt/
+ln -s /opt/cmake-${CMAKE_VERSION}-Linux-x86_64/bin/cmake /usr/local/bin/cmake

docker/clio_docker/centos/build_boost.sh

@@ -0,0 +1,13 @@
+#!/usr/bin/env bash
+set -exu
+
+#yum install wget lz4 lz4-devel git llvm13-static.x86_64 llvm13-devel.x86_64 devtoolset-11-binutils zlib-static
+# it's either those or link=static that halves the failures. probably link=static
+BOOST_VERSION=$1
+BOOST_VERSION_=$(echo ${BOOST_VERSION} | tr . _)
+echo "BOOST_VERSION: ${BOOST_VERSION}"
+echo "BOOST_VERSION_: ${BOOST_VERSION_}"
+curl -OJLs "https://boostorg.jfrog.io/artifactory/main/release/${BOOST_VERSION}/source/boost_${BOOST_VERSION_}.tar.gz"
+tar zxf "boost_${BOOST_VERSION_}.tar.gz"
+cd boost_${BOOST_VERSION_} && ./bootstrap.sh && ./b2 --without-python link=static -j$(nproc)
+mkdir -p /boost && mv boost /boost && mv stage /boost

docker/clio_docker/centos/build_git_centos7.sh

@@ -0,0 +1,18 @@
+#!/usr/bin/env bash
+
+set -ex
+GIT_VERSION="2.37.1"
+curl -OJL https://github.com/git/git/archive/refs/tags/v${GIT_VERSION}.tar.gz
+tar zxvf git-${GIT_VERSION}.tar.gz
+cd git-${GIT_VERSION}
+
+yum install -y centos-release-scl epel-release
+yum update -y
+yum install -y devtoolset-11 autoconf gnu-getopt gettext zlib-devel  libcurl-devel
+
+source /opt/rh/devtoolset-11/enable
+make configure
+./configure
+make git -j$(nproc)
+make install git
+git  --version | cut -d ' ' -f3

docker/clio_docker/centos/dockerfile

@@ -0,0 +1,34 @@
+FROM centos:7
+
+ENV CLIO_DIR=/opt/clio/
+# ENV OPENSSL_DIR=/opt/openssl
+
+RUN yum -y install git epel-release centos-release-scl perl-IPC-Cmd openssl
+RUN yum install -y devtoolset-11
+ENV version=3.16
+ENV build=3
+# RUN curl -OJL https://cmake.org/files/v$version/cmake-$version.$build.tar.gz
+COPY install_cmake.sh /install_cmake.sh
+RUN /install_cmake.sh 3.16.3 /usr/local
+RUN source /opt/rh/devtoolset-11/enable
+WORKDIR /tmp
+# RUN mkdir $OPENSSL_DIR && cd $OPENSSL_DIR
+COPY build_git_centos7.sh build_git_centos7.sh
+
+RUN ./build_git_centos7.sh
+RUN git clone https://github.com/openssl/openssl
+WORKDIR /tmp/openssl
+RUN git checkout OpenSSL_1_1_1q
+#--prefix=/usr --openssldir=/etc/ssl --libdir=lib no-shared zlib-dynamic
+RUN SSLDIR=$(openssl version -d | cut -d: -f2 | tr -d [:space:]\") && ./config -fPIC --prefix=/usr --openssldir=${SSLDIR} zlib shared && \
+    make -j $(nproc) && \
+    make install_sw
+WORKDIR /tmp
+RUN git clone https://github.com/xrplf/clio.git
+COPY build_boost.sh build_boost.sh
+ENV OPENSSL_ROOT=/opt/local/openssl
+ENV BOOST_ROOT=/boost
+RUN source scl_source enable devtoolset-11 && /tmp/build_boost.sh 1.75.0
+RUN yum install -y bison flex
+RUN source /opt/rh/devtoolset-11/enable && \
+    cd /tmp/clio && cmake -B build -Dtests=0 -Dlocal_libarchive=1 -Dunity=0 -DBUILD_TESTS=0 && cmake --build build --parallel $(nproc)

docker/clio_docker/centos/install_cmake.sh

@@ -0,0 +1,11 @@
+#!/bin/bash
+
+set -eo pipefail
+
+CMAKE_VERSION=${1:-"3.16.3"}
+cd /tmp
+URL="https://github.com/Kitware/CMake/releases/download/v${CMAKE_VERSION}/cmake-${CMAKE_VERSION}-linux-x86_64.tar.gz"
+curl -OJLs $URL
+tar xzvf cmake-${CMAKE_VERSION}-Linux-x86_64.tar.gz
+mv cmake-${CMAKE_VERSION}-Linux-x86_64 /opt/
+ln -s /opt/cmake-${CMAKE_VERSION}-Linux-x86_64/bin/cmake /usr/local/bin/cmake

docker/shared/build_boost.sh

@@ -0,0 +1,13 @@
+#!/usr/bin/env bash
+set -exu
+
+#yum install wget lz4 lz4-devel git llvm13-static.x86_64 llvm13-devel.x86_64 devtoolset-11-binutils zlib-static
+# it's either those or link=static that halves the failures. probably link=static
+BOOST_VERSION=$1
+BOOST_VERSION_=$(echo ${BOOST_VERSION} | tr . _)
+echo "BOOST_VERSION: ${BOOST_VERSION}"
+echo "BOOST_VERSION_: ${BOOST_VERSION_}"
+curl -OJLs "https://boostorg.jfrog.io/artifactory/main/release/${BOOST_VERSION}/source/boost_${BOOST_VERSION_}.tar.gz"
+tar zxf "boost_${BOOST_VERSION_}.tar.gz"
+cd boost_${BOOST_VERSION_} && ./bootstrap.sh && ./b2 --without-python link=static -j$(nproc)
+mkdir -p /boost && mv boost /boost && mv stage /boost

docker/shared/install_cmake.sh

@@ -0,0 +1,11 @@
+#!/bin/bash
+
+set -eo pipefail
+
+CMAKE_VERSION=${1:-"3.16.3"}
+cd /tmp
+URL="https://github.com/Kitware/CMake/releases/download/v${CMAKE_VERSION}/cmake-${CMAKE_VERSION}-linux-x86_64.tar.gz"
+curl -OJLs $URL
+tar xzvf cmake-${CMAKE_VERSION}-Linux-x86_64.tar.gz
+mv cmake-${CMAKE_VERSION}-Linux-x86_64 /opt/
+ln -s /opt/cmake-${CMAKE_VERSION}-Linux-x86_64/bin/cmake /usr/local/bin/cmake

docker/shared/install_openssl.sh

@@ -0,0 +1,3 @@
+#!/bin/bash
+
+set -e

docker/ubuntu/Dockerfile

@@ -0,0 +1,24 @@
+FROM ubuntu:20.04 AS boost
+
+RUN apt-get update && apt-get install -y build-essential
+ARG BOOST_VERSION_=1_75_0
+ARG BOOST_VERSION=1.75.0
+COPY docker/shared/build_boost.sh .
+RUN apt install -y curl
+RUN ./build_boost.sh ${BOOST_VERSION}
+ENV BOOST_ROOT=/boost
+
+FROM ubuntu:20.04 AS build
+ENV BOOST_ROOT=/boost
+COPY --from=boost /boost /boost
+ENV DEBIAN_FRONTEND=noninteractive
+RUN apt-get update && apt-get install --no-install-recommends -y build-essential software-properties-common pkg-config libssl-dev wget curl gpg git zlib1g-dev bison flex autoconf lsb-release
+RUN apt install -y gpg-agent
+RUN wget https://apt.llvm.org/llvm.sh
+RUN chmod +x llvm.sh && ./llvm.sh 14 && ./llvm.sh 15
+# COPY . /clio
+## Install cmake
+ARG CMAKE_VERSION=3.16.3
+COPY docker/shared/install_cmake.sh .
+RUN ./install_cmake.sh ${CMAKE_VERSION}
+ENV PATH="/opt/local/cmake/bin:$PATH"

example-config.json

@@ -1,37 +1,93 @@
 {
-    "database":
-    {
-        "type":"cassandra",
-        "cassandra":
-        {
-            "contact_points":"127.0.0.1",
-            "port":9042,
-            "keyspace":"clio",
-            "replication_factor":1,
-            "table_prefix":"",
-            "max_requests_outstanding":25000,
-            "threads":8
+    "database": {
+        "type": "cassandra",
+        "cassandra": {
+            "contact_points": "127.0.0.1",
+            "port": 9042,
+            "keyspace": "clio",
+            "replication_factor": 1,
+            "table_prefix": "",
+            "max_write_requests_outstanding": 25000,
+            "max_read_requests_outstanding": 30000,
+            "threads": 8
         }
     },
-    "etl_sources":
-    [
+    "etl_sources": [
         {
-            "ip":"127.0.0.1",
-            "ws_port":"6006",
-            "grpc_port":"50051"
+            "ip": "127.0.0.1",
+            "ws_port": "6006",
+            "grpc_port": "50051"
         }
     ],
-    "dos_guard":
-    {
-        "whitelist":["127.0.0.1"]
+    "dos_guard": {
+        "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
     },
-    "server":{
-        "ip":"0.0.0.0",
-        "port":51233
+    "cache": {
+        "peers": [
+            {
+                "ip": "127.0.0.1",
+                "port": 51234
+            }
+        ]
     },
-    "log_level":"debug",
-    "log_file":"./clio.log",
-    "online_delete":0,
-    "extractor_threads":8,
-    "read_only":false
+    "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_queue_size": 500
+    },
+    "log_channels": [
+        {
+            "channel": "Backend",
+            "log_level": "fatal"
+        },
+        {
+            "channel": "WebServer",
+            "log_level": "info"
+        },
+        {
+            "channel": "Subscriptions",
+            "log_level": "info"
+        },
+        {
+            "channel": "RPC",
+            "log_level": "error"
+        },
+        {
+            "channel": "ETL",
+            "log_level": "debug"
+        },
+        {
+            "channel": "Performance",
+            "log_level": "trace"
+        }
+    ],
+    "log_level": "info",
+    "log_format": "%TimeStamp% (%SourceLocation%) [%ThreadID%] %Channel%:%Severity% %Message%", // This is the default format
+    "log_to_console": true,
+    "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"
 }

src/backend/BackendFactory.h

@@ -1,65 +1,62 @@
-#ifndef RIPPLE_APP_REPORTING_BACKENDFACTORY_H_INCLUDED
-#define RIPPLE_APP_REPORTING_BACKENDFACTORY_H_INCLUDED
+//------------------------------------------------------------------------------
+/*
+    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 <boost/algorithm/string.hpp>
 #include <backend/BackendInterface.h>
 #include <backend/CassandraBackend.h>
-#include <backend/PostgresBackend.h>
+#include <config/Config.h>
+#include <log/Logger.h>
+
+#include <boost/algorithm/string.hpp>
 
 namespace Backend {
 std::shared_ptr<BackendInterface>
-make_Backend(boost::asio::io_context& ioc, boost::json::object const& config)
+make_Backend(boost::asio::io_context& ioc, clio::Config const& config)
 {
-    BOOST_LOG_TRIVIAL(info) << __func__ << ": Constructing BackendInterface";
+    static clio::Logger log{"Backend"};
+    log.info() << "Constructing BackendInterface";
 
-    boost::json::object dbConfig = config.at("database").as_object();
-
-    bool readOnly = false;
-    if (config.contains("read_only"))
-        readOnly = config.at("read_only").as_bool();
-
-    auto type = dbConfig.at("type").as_string();
+    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"))
     {
-        if (config.contains("online_delete"))
-            dbConfig.at(type).as_object()["ttl"] =
-                config.at("online_delete").as_int64() * 4;
-        backend = std::make_shared<CassandraBackend>(
-            ioc, dbConfig.at(type).as_object());
-    }
-    else if (boost::iequals(type, "postgres"))
-    {
-        if (dbConfig.contains("experimental") &&
-            dbConfig.at("experimental").is_bool() &&
-            dbConfig.at("experimental").as_bool())
-            backend = std::make_shared<PostgresBackend>(
-                ioc, dbConfig.at(type).as_object());
-        else
-            BOOST_LOG_TRIVIAL(fatal)
-                << "Postgres support is experimental at this time. "
-                << "If you would really like to use Postgres, add "
-                   "\"experimental\":true to your database config";
+        auto cfg = config.section("database." + type);
+        backend =
+            std::make_shared<Backend::Cassandra::CassandraBackend>(Backend::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);
     }
 
-    BOOST_LOG_TRIVIAL(info)
-        << __func__ << ": Constructed BackendInterface Successfully";
-
+    log.info() << "Constructed BackendInterface Successfully";
     return backend;
 }
 }  // namespace Backend
-
-#endif  // RIPPLE_REPORTING_BACKEND_FACTORY

src/backend/BackendInterface.cpp

@@ -1,33 +1,59 @@
+//------------------------------------------------------------------------------
+/*
+    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 <backend/BackendInterface.h>
+#include <log/Logger.h>
+
 #include <ripple/protocol/Indexes.h>
 #include <ripple/protocol/STLedgerEntry.h>
-#include <backend/BackendInterface.h>
+
+using namespace clio;
+
+// local to compilation unit loggers
+namespace {
+clio::Logger gLog{"Backend"};
+}  // namespace
+
 namespace Backend {
 bool
 BackendInterface::finishWrites(std::uint32_t const ledgerSequence)
 {
+    gLog.debug() << "Want finish writes for " << ledgerSequence;
     auto commitRes = doFinishWrites();
     if (commitRes)
     {
+        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));
-    ripple::uint256 key256 = ripple::uint256::fromVoid(key.data());
     doWriteLedgerObject(std::move(key), seq, std::move(blob));
 }
 
 std::optional<LedgerRange>
-BackendInterface::hardFetchLedgerRangeNoThrow(
-    boost::asio::yield_context& yield) const
+BackendInterface::hardFetchLedgerRangeNoThrow(boost::asio::yield_context& yield) const
 {
-    BOOST_LOG_TRIVIAL(debug) << __func__;
+    gLog.trace() << "called";
     while (true)
     {
         try
@@ -44,7 +70,7 @@ BackendInterface::hardFetchLedgerRangeNoThrow(
 std::optional<LedgerRange>
 BackendInterface::hardFetchLedgerRangeNoThrow() const
 {
-    BOOST_LOG_TRIVIAL(debug) << __func__;
+    gLog.trace() << "called";
     return retryOnTimeout([&]() { return hardFetchLedgerRange(); });
 }
 
@@ -58,21 +84,17 @@ BackendInterface::fetchLedgerObject(
     auto obj = cache_.get(key, sequence);
     if (obj)
     {
-        BOOST_LOG_TRIVIAL(trace)
-            << __func__ << " - cache hit - " << ripple::strHex(key);
+        gLog.trace() << "Cache hit - " << ripple::strHex(key);
         return *obj;
     }
     else
     {
-        BOOST_LOG_TRIVIAL(trace)
-            << __func__ << " - cache miss - " << ripple::strHex(key);
+        gLog.trace() << "Cache miss - " << ripple::strHex(key);
         auto dbObj = doFetchLedgerObject(key, sequence, yield);
         if (!dbObj)
-            BOOST_LOG_TRIVIAL(trace)
-                << __func__ << " - missed cache and missed in db";
+            gLog.trace() << "Missed cache and missed in db";
         else
-            BOOST_LOG_TRIVIAL(trace)
-                << __func__ << " - missed cache but found in db";
+            gLog.trace() << "Missed cache but found in db";
         return dbObj;
     }
 }
@@ -94,9 +116,7 @@ BackendInterface::fetchLedgerObjects(
         else
             misses.push_back(keys[i]);
     }
-    BOOST_LOG_TRIVIAL(trace)
-        << __func__ << " - cache hits = " << keys.size() - misses.size()
-        << " - cache misses = " << misses.size();
+    gLog.trace() << "Cache hits = " << keys.size() - misses.size() << " - cache misses = " << misses.size();
 
     if (misses.size())
     {
@@ -122,11 +142,9 @@ BackendInterface::fetchSuccessorKey(
 {
     auto succ = cache_.getSuccessor(key, ledgerSequence);
     if (succ)
-        BOOST_LOG_TRIVIAL(trace)
-            << __func__ << " - cache hit - " << ripple::strHex(key);
+        gLog.trace() << "Cache hit - " << ripple::strHex(key);
     else
-        BOOST_LOG_TRIVIAL(trace)
-            << __func__ << " - cache miss - " << ripple::strHex(key);
+        gLog.trace() << "Cache miss - " << ripple::strHex(key);
     return succ ? succ->key : doFetchSuccessorKey(key, ledgerSequence, yield);
 }
 
@@ -153,7 +171,6 @@ 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
 {
     // TODO try to speed this up. This can take a few seconds. The goal is
@@ -162,10 +179,7 @@ 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;
@@ -180,30 +194,24 @@ BackendInterface::fetchBookOffers(
         succMillis += getMillis(mid2 - mid1);
         if (!offerDir || offerDir->key >= bookEnd)
         {
-            BOOST_LOG_TRIVIAL(debug) << __func__ << " - offerDir.has_value() "
-                                     << offerDir.has_value() << " breaking";
+            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)
             {
-                BOOST_LOG_TRIVIAL(debug)
-                    << __func__ << " next is empty. breaking";
+                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;
@@ -215,28 +223,21 @@ BackendInterface::fetchBookOffers(
     auto objs = fetchLedgerObjects(keys, ledgerSequence, yield);
     for (size_t i = 0; i < keys.size() && i < limit; ++i)
     {
-        BOOST_LOG_TRIVIAL(debug)
-            << __func__ << " key = " << ripple::strHex(keys[i])
-            << " blob = " << ripple::strHex(objs[i])
-            << " ledgerSequence = " << ledgerSequence;
+        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();
-    BOOST_LOG_TRIVIAL(debug)
-        << __func__ << " "
-        << "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";
+    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;
 }
@@ -255,10 +256,8 @@ BackendInterface::fetchLedgerPage(
     bool reachedEnd = false;
     while (keys.size() < limit && !reachedEnd)
     {
-        ripple::uint256 const& curCursor = keys.size() ? keys.back()
-            : cursor                                   ? *cursor
-                                                       : firstKey;
-        uint32_t 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;
@@ -273,16 +272,14 @@ BackendInterface::fetchLedgerPage(
             page.objects.push_back({std::move(keys[i]), std::move(objects[i])});
         else if (!outOfOrder)
         {
-            BOOST_LOG_TRIVIAL(error)
-                << __func__ << " incorrect successor table. key = "
-                << ripple::strHex(keys[i]) << " - seq = " << ledgerSequence;
+            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]);
             }
-            BOOST_LOG_TRIVIAL(error) << __func__ << msg.str();
-            assert(false);
+            gLog.error() << msg.str();
         }
     }
     if (keys.size() && !reachedEnd)
@@ -292,9 +289,7 @@ 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;
 
@@ -303,7 +298,7 @@ BackendInterface::fetchFees(
 
     if (!bytes)
     {
-        BOOST_LOG_TRIVIAL(error) << __func__ << " - could not find fees";
+        gLog.error() << "Could not find fees";
         return {};
     }
 

src/backend/BackendInterface.h

@@ -1,16 +1,48 @@
-#ifndef RIPPLE_APP_REPORTING_BACKENDINTERFACE_H_INCLUDED
-#define RIPPLE_APP_REPORTING_BACKENDINTERFACE_H_INCLUDED
+//------------------------------------------------------------------------------
+/*
+    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/asio.hpp>
 #include <backend/DBHelpers.h>
-#include <backend/SimpleCache.h>
+#include <backend/LedgerCache.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
     {
@@ -18,10 +50,20 @@ class DatabaseTimeout : public std::exception
     }
 };
 
+/**
+ * @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
@@ -30,48 +72,82 @@ retryOnTimeout(F func, size_t waitMs = 500)
         }
         catch (DatabaseTimeout& t)
         {
+            log.error() << "Database request timed out. Sleeping and retrying ... ";
             std::this_thread::sleep_for(std::chrono::milliseconds(waitMs));
-            BOOST_LOG_TRIVIAL(error)
-                << __func__ << " function timed out. Retrying ... ";
         }
     }
 }
 
+/**
+ * @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);
 
-    using R = typename std::result_of<F(boost::asio::yield_context&)>::type;
+    /**
+     * @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 boost::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();
-            });
+        boost::asio::spawn(strand, [&f, &work, &res](boost::asio::yield_context yield) {
+            res = f(yield);
+            work.reset();
+        });
 
         ctx.run();
         return res;
     }
     else
     {
-        boost::asio::spawn(
-            strand, [&f, &work](boost::asio::yield_context yield) {
-                f(yield);
-                work.reset();
-            });
+        /*! @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)
@@ -79,57 +155,66 @@ 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_;
+    LedgerCache cache_;
 
-    // mutex used for open() and close()
-    mutable std::mutex mutex_;
+    /**
+     * @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(boost::json::object const& config)
-    {
-    }
-    virtual ~BackendInterface()
-    {
-    }
+    BackendInterface() = default;
+    virtual ~BackendInterface() = default;
 
-    // *** 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:
-    // *** ledger methods
-    //
-
-    SimpleCache const&
+    /**
+     * @brief Cache that holds states of the ledger
+     * @return Immutable cache
+     */
+    LedgerCache const&
     cache() const
     {
         return cache_;
     }
 
-    SimpleCache&
+    /**
+     * @brief Cache that holds states of the ledger
+     * @return Mutable cache
+     */
+    LedgerCache&
     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;
+    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;
+    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
     {
@@ -137,10 +222,18 @@ public:
         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::unique_lock lck(rngMtx_);
+        std::scoped_lock lck(rngMtx_);
         assert(!range || newMax >= range->maxSequence);
         if (!range)
             range = {newMax, newMax};
@@ -148,70 +241,171 @@ public:
             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;
 
-    // *** transaction methods
+    /*! @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;
+    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;
+    fetchTransactions(std::vector<ripple::uint256> const& hashes, boost::asio::yield_context& yield) const = 0;
 
-    virtual AccountTransactions
+    /**
+     * @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<AccountTransactionsCursor> const& cursor,
+        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;
+    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,
+    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;
 
-    // *** state data methods
+    /*! @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;
+    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;
+    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;
+    fetchLedgerDiff(std::uint32_t const ledgerSequence, boost::asio::yield_context& yield) const = 0;
 
-    // Fetches a page of ledger objects, ordered by key/index.
-    // Used by ledger_data
+    /**
+     * @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,
@@ -220,63 +414,94 @@ public:
         bool outOfOrder,
         boost::asio::yield_context& yield) const;
 
-    // Fetches the successor to key/index
+    /*! @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;
+    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;
-    // Fetches the successor to key/index
+    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;
+    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);
-        });
+        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;
 
-    // Doesn't throw DatabaseTimeout. Should be used with care.
+    /*! @brief Fetches ledger range but doesn't throw timeout. Use with care. */
     std::optional<LedgerRange>
     hardFetchLedgerRangeNoThrow() const;
-    // Doesn't throw DatabaseTimeout. Should be used with care.
+    /*! @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;
+    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);
+    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,
@@ -285,49 +510,70 @@ public:
         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
-    writeSuccessor(
-        std::string&& key,
-        std::uint32_t const seq,
-        std::string&& successor) = 0;
+    writeNFTTransactions(std::vector<NFTTransactionsData>&& data) = 0;
 
-    // Tell the database we are about to begin writing data for a particular
-    // ledger.
+    /**
+     * @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;
 
-    // Tell the database we have finished writing all data for a particular
-    // ledger
-    // TODO change the return value to represent different results. committed,
-    // write conflict, errored, successful but not committed
+    /**
+     * @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);
 
     virtual bool
-    doOnlineDelete(
-        std::uint32_t numLedgersToKeep,
-        boost::asio::yield_context& yield) const = 0;
+    isTooBusy() const = 0;
 
-    // Open the database. Set up all of the necessary objects and
-    // datastructures. After this call completes, the database is ready for
-    // use.
-    virtual void
-    open(bool readOnly) = 0;
-
-    // Close the database, releasing any resources
-    virtual void
-    close(){};
-
-    // *** private helper methods
 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;
+    doWriteLedgerObject(std::string&& key, std::uint32_t const seq, std::string&& blob) = 0;
 
     virtual bool
     doFinishWrites() = 0;
@@ -335,4 +581,3 @@ private:
 
 }  // namespace Backend
 using BackendInterface = Backend::BackendInterface;
-#endif

src/backend/DBHelpers.h

@@ -1,16 +1,38 @@
-#ifndef CLIO_BACKEND_DBHELPERS_H_INCLUDED
-#define CLIO_BACKEND_DBHELPERS_H_INCLUDED
+//------------------------------------------------------------------------------
+/*
+    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/Pg.h>
+
 #include <backend/Types.h>
 
-/// Struct used to keep track of what to write to transactions and
-/// account_transactions tables in Postgres
+/**
+ * @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;
@@ -18,10 +40,7 @@ struct AccountTransactionsData
     std::uint32_t transactionIndex;
     ripple::uint256 txHash;
 
-    AccountTransactionsData(
-        ripple::TxMeta& meta,
-        ripple::uint256 const& txHash,
-        beast::Journal& j)
+    AccountTransactionsData(ripple::TxMeta& meta, ripple::uint256 const& txHash, beast::Journal& j)
         : accounts(meta.getAffectedAccounts())
         , ledgerSequence(meta.getLgrSeq())
         , transactionIndex(meta.getIndex())
@@ -32,6 +51,94 @@ struct AccountTransactionsData
     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)
+    {
+    }
+};
+
 template <class T>
 inline bool
 isOffer(T const& object)
@@ -68,8 +175,7 @@ isBookDir(T const& key, R const& object)
     if (!isDirNode(object))
         return false;
 
-    ripple::STLedgerEntry const sle{
-        ripple::SerialIter{object.data(), object.size()}, key};
+    ripple::STLedgerEntry const sle{ripple::SerialIter{object.data(), object.size()}, key};
     return !sle[~ripple::sfOwner].has_value();
 }
 
@@ -96,30 +202,6 @@ getBookBase(T const& key)
     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)
 {
@@ -127,4 +209,3 @@ uint256ToString(ripple::uint256 const& uint)
 }
 
 static constexpr std::uint32_t rippleEpochStart = 946684800;
-#endif

src/backend/LayeredCache.cpp

@@ -1,110 +0,0 @@
-#include <backend/LayeredCache.h>
-namespace Backend {
-
-void
-LayeredCache::insert(
-    ripple::uint256 const& key,
-    Blob const& value,
-    uint32_t seq)
-{
-    auto entry = map_[key];
-    // stale insert, do nothing
-    if (seq <= entry.recent.seq)
-        return;
-    entry.old = entry.recent;
-    entry.recent = {seq, value};
-    if (value.empty())
-        pendingDeletes_.push_back(key);
-    if (!entry.old.blob.empty())
-        pendingSweeps_.push_back(key);
-}
-
-std::optional<Blob>
-LayeredCache::select(CacheEntry const& entry, uint32_t seq) const
-{
-    if (seq < entry.old.seq)
-        return {};
-    if (seq < entry.recent.seq && !entry.old.blob.empty())
-        return entry.old.blob;
-    if (!entry.recent.blob.empty())
-        return entry.recent.blob;
-    return {};
-}
-void
-LayeredCache::update(std::vector<LedgerObject> const& blobs, uint32_t seq)
-{
-    std::unique_lock lck{mtx_};
-    if (seq > mostRecentSequence_)
-        mostRecentSequence_ = seq;
-    for (auto const& k : pendingSweeps_)
-    {
-        auto e = map_[k];
-        e.old = {};
-    }
-    for (auto const& k : pendingDeletes_)
-    {
-        map_.erase(k);
-    }
-    for (auto const& b : blobs)
-    {
-        insert(b.key, b.blob, seq);
-    }
-}
-std::optional<LedgerObject>
-LayeredCache::getSuccessor(ripple::uint256 const& key, uint32_t seq) const
-{
-    ripple::uint256 curKey = key;
-    while (true)
-    {
-        std::shared_lock lck{mtx_};
-        if (seq < mostRecentSequence_ - 1)
-            return {};
-        auto e = map_.upper_bound(curKey);
-        if (e == map_.end())
-            return {};
-        auto const& entry = e->second;
-        auto blob = select(entry, seq);
-        if (!blob)
-        {
-            curKey = e->first;
-            continue;
-        }
-        else
-            return {{e->first, *blob}};
-    }
-}
-std::optional<LedgerObject>
-LayeredCache::getPredecessor(ripple::uint256 const& key, uint32_t seq) const
-{
-    ripple::uint256 curKey = key;
-    std::shared_lock lck{mtx_};
-    while (true)
-    {
-        if (seq < mostRecentSequence_ - 1)
-            return {};
-        auto e = map_.lower_bound(curKey);
-        --e;
-        if (e == map_.begin())
-            return {};
-        auto const& entry = e->second;
-        auto blob = select(entry, seq);
-        if (!blob)
-        {
-            curKey = e->first;
-            continue;
-        }
-        else
-            return {{e->first, *blob}};
-    }
-}
-std::optional<Blob>
-LayeredCache::get(ripple::uint256 const& key, uint32_t seq) const
-{
-    std::shared_lock lck{mtx_};
-    auto e = map_.find(key);
-    if (e == map_.end())
-        return {};
-    auto const& entry = e->second;
-    return select(entry, seq);
-}
-}  // namespace Backend

src/backend/LayeredCache.h

@@ -1,73 +0,0 @@
-#ifndef CLIO_LAYEREDCACHE_H_INCLUDED
-#define CLIO_LAYEREDCACHE_H_INCLUDED
-
-#include <ripple/basics/base_uint.h>
-#include <backend/Types.h>
-#include <map>
-#include <mutex>
-#include <shared_mutex>
-#include <utility>
-#include <vector>
-namespace Backend {
-class LayeredCache
-{
-    struct SeqBlobPair
-    {
-        uint32_t seq;
-        Blob blob;
-    };
-    struct CacheEntry
-    {
-        SeqBlobPair recent;
-        SeqBlobPair old;
-    };
-
-    std::map<ripple::uint256, CacheEntry> map_;
-    std::vector<ripple::uint256> pendingDeletes_;
-    std::vector<ripple::uint256> pendingSweeps_;
-    mutable std::shared_mutex mtx_;
-    uint32_t mostRecentSequence_;
-
-    void
-    insert(ripple::uint256 const& key, Blob const& value, uint32_t seq);
-
-    /*
-    void
-    insert(ripple::uint256 const& key, Blob const& value, uint32_t seq)
-    {
-        map_.emplace(key,{{seq,value,{}});
-    }
-    void
-    update(ripple::uint256 const& key, Blob const& value, uint32_t seq)
-    {
-            auto& entry = map_.find(key);
-            entry.old = entry.recent;
-            entry.recent = {seq, value};
-            pendingSweeps_.push_back(key);
-    }
-    void
-    erase(ripple::uint256 const& key, uint32_t seq)
-    {
-            update(key, {}, seq);
-            pendingDeletes_.push_back(key);
-    }
-    */
-    std::optional<Blob>
-    select(CacheEntry const& entry, uint32_t seq) const;
-
-public:
-    void
-    update(std::vector<LedgerObject> const& blobs, uint32_t seq);
-
-    std::optional<Blob>
-    get(ripple::uint256 const& key, uint32_t seq) const;
-
-    std::optional<LedgerObject>
-    getSuccessor(ripple::uint256 const& key, uint32_t seq) const;
-
-    std::optional<LedgerObject>
-    getPredecessor(ripple::uint256 const& key, uint32_t seq) const;
-};
-
-}  // namespace Backend
-#endif

src/backend/LedgerCache.cpp

@@ -0,0 +1,160 @@
+//------------------------------------------------------------------------------
+/*
+    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 <backend/LedgerCache.h>
+
+namespace Backend {
+
+uint32_t
+LedgerCache::latestLedgerSequence() const
+{
+    std::shared_lock lck{mtx_};
+    return latestSeq_;
+}
+
+void
+LedgerCache::update(std::vector<LedgerObject> const& objs, uint32_t seq, bool isBackground)
+{
+    if (disabled_)
+        return;
+
+    {
+        std::scoped_lock lck{mtx_};
+        if (seq > latestSeq_)
+        {
+            assert(seq == latestSeq_ + 1 || latestSeq_ == 0);
+            latestSeq_ = seq;
+        }
+        for (auto const& obj : objs)
+        {
+            if (obj.blob.size())
+            {
+                if (isBackground && deletes_.count(obj.key))
+                    continue;
+
+                auto& e = map_[obj.key];
+                if (seq > e.seq)
+                {
+                    e = {seq, obj.blob};
+                }
+            }
+            else
+            {
+                map_.erase(obj.key);
+                if (!full_ && !isBackground)
+                    deletes_.insert(obj.key);
+            }
+        }
+    }
+}
+
+std::optional<LedgerObject>
+LedgerCache::getSuccessor(ripple::uint256 const& key, uint32_t seq) const
+{
+    if (!full_)
+        return {};
+    std::shared_lock{mtx_};
+    successorReqCounter_++;
+    if (seq != latestSeq_)
+        return {};
+    auto e = map_.upper_bound(key);
+    if (e == map_.end())
+        return {};
+    successorHitCounter_++;
+    return {{e->first, e->second.blob}};
+}
+
+std::optional<LedgerObject>
+LedgerCache::getPredecessor(ripple::uint256 const& key, uint32_t seq) const
+{
+    if (!full_)
+        return {};
+    std::shared_lock lck{mtx_};
+    if (seq != latestSeq_)
+        return {};
+    auto e = map_.lower_bound(key);
+    if (e == map_.begin())
+        return {};
+    --e;
+    return {{e->first, e->second.blob}};
+}
+
+std::optional<Blob>
+LedgerCache::get(ripple::uint256 const& key, uint32_t seq) const
+{
+    std::shared_lock lck{mtx_};
+    if (seq > latestSeq_)
+        return {};
+    objectReqCounter_++;
+    auto e = map_.find(key);
+    if (e == map_.end())
+        return {};
+    if (seq < e->second.seq)
+        return {};
+    objectHitCounter_++;
+    return {e->second.blob};
+}
+
+void
+LedgerCache::setDisabled()
+{
+    disabled_ = true;
+}
+
+void
+LedgerCache::setFull()
+{
+    if (disabled_)
+        return;
+
+    full_ = true;
+    std::scoped_lock lck{mtx_};
+    deletes_.clear();
+}
+
+bool
+LedgerCache::isFull() const
+{
+    return full_;
+}
+
+size_t
+LedgerCache::size() const
+{
+    std::shared_lock lck{mtx_};
+    return map_.size();
+}
+
+float
+LedgerCache::getObjectHitRate() const
+{
+    if (!objectReqCounter_)
+        return 1;
+    return ((float)objectHitCounter_) / objectReqCounter_;
+}
+
+float
+LedgerCache::getSuccessorHitRate() const
+{
+    if (!successorReqCounter_)
+        return 1;
+    return ((float)successorHitCounter_) / successorReqCounter_;
+}
+
+}  // namespace Backend

src/backend/LedgerCache.h

@@ -0,0 +1,98 @@
+//------------------------------------------------------------------------------
+/*
+    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 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:
+    // 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/Pg.h

@@ -1,571 +0,0 @@
-#ifndef RIPPLE_CORE_PG_H_INCLUDED
-#define RIPPLE_CORE_PG_H_INCLUDED
-
-#include <ripple/basics/StringUtilities.h>
-#include <ripple/basics/chrono.h>
-#include <ripple/ledger/ReadView.h>
-#include <boost/asio/io_context.hpp>
-#include <boost/asio/ip/tcp.hpp>
-#include <boost/asio/spawn.hpp>
-#include <boost/icl/closed_interval.hpp>
-#include <boost/json.hpp>
-#include <boost/lexical_cast.hpp>
-#include <boost/log/trivial.hpp>
-#include <atomic>
-#include <chrono>
-#include <condition_variable>
-#include <functional>
-#include <libpq-fe.h>
-#include <memory>
-#include <mutex>
-#include <optional>
-#include <string>
-#include <string_view>
-#include <utility>
-#include <vector>
-
-// These postgres structs must be freed only by the postgres API.
-using pg_result_type = std::unique_ptr<PGresult, void (*)(PGresult*)>;
-using pg_connection_type = std::unique_ptr<PGconn, void (*)(PGconn*)>;
-using asio_socket_type = std::unique_ptr<
-    boost::asio::ip::tcp::socket,
-    void (*)(boost::asio::ip::tcp::socket*)>;
-
-/** first: command
- * second: parameter values
- *
- * The 2nd member takes an optional string to
- * distinguish between NULL parameters and empty strings. An empty
- * item corresponds to a NULL parameter.
- *
- * Postgres reads each parameter as a c-string, regardless of actual type.
- * Binary types (bytea) need to be converted to hex and prepended with
- * \x ("\\x").
- */
-using pg_params =
-    std::pair<char const*, std::vector<std::optional<std::string>>>;
-
-/** Parameter values for pg API. */
-using pg_formatted_params = std::vector<char const*>;
-
-/** Parameters for managing postgres connections. */
-struct PgConfig
-{
-    /** Maximum connections allowed to db. */
-    std::size_t max_connections{1000};
-    /** Close idle connections past this duration. */
-    std::chrono::seconds timeout{600};
-
-    /** Index of DB connection parameter names. */
-    std::vector<char const*> keywordsIdx;
-    /** DB connection parameter names. */
-    std::vector<std::string> keywords;
-    /** Index of DB connection parameter values. */
-    std::vector<char const*> valuesIdx;
-    /** DB connection parameter values. */
-    std::vector<std::string> values;
-};
-
-//-----------------------------------------------------------------------------
-
-/** Class that operates on postgres query results.
- *
- * The functions that return results do not check first whether the
- * expected results are actually there. Therefore, the caller first needs
- * to check whether or not a valid response was returned using the operator
- * bool() overload. If number of tuples or fields are unknown, then check
- * those. Each result field should be checked for null before attempting
- * to return results. Finally, the caller must know the type of the field
- * before calling the corresponding function to return a field. Postgres
- * internally stores each result field as null-terminated strings.
- */
-class PgResult
-{
-    // The result object must be freed using the libpq API PQclear() call.
-    pg_result_type result_{nullptr, [](PGresult* result) { PQclear(result); }};
-    std::optional<std::pair<ExecStatusType, std::string>> error_;
-
-public:
-    /** Constructor for when the process is stopping.
-     *
-     */
-    PgResult()
-    {
-    }
-
-    /** Constructor for successful query results.
-     *
-     * @param result Query result.
-     */
-    explicit PgResult(pg_result_type&& result) : result_(std::move(result))
-    {
-    }
-
-    /** Constructor for failed query results.
-     *
-     * @param result Query result that contains error information.
-     * @param conn Postgres connection that contains error information.
-     */
-    PgResult(PGresult* result, PGconn* conn)
-        : error_({PQresultStatus(result), PQerrorMessage(conn)})
-    {
-    }
-
-    /** Return field as a null-terminated string pointer.
-     *
-     * Note that this function does not guarantee that the result struct
-     * exists, or that the row and fields exist, or that the field is
-     * not null.
-     *
-     * @param ntuple Row number.
-     * @param nfield Field number.
-     * @return Field contents.
-     */
-    char const*
-    c_str(int ntuple = 0, int nfield = 0) const
-    {
-        return PQgetvalue(result_.get(), ntuple, nfield);
-    }
-
-    std::vector<unsigned char>
-    asUnHexedBlob(int ntuple = 0, int nfield = 0) const
-    {
-        std::string_view view{c_str(ntuple, nfield) + 2};
-        auto res = ripple::strUnHex(view.size(), view.cbegin(), view.cend());
-        if (res)
-            return *res;
-        return {};
-    }
-
-    ripple::uint256
-    asUInt256(int ntuple = 0, int nfield = 0) const
-    {
-        ripple::uint256 val;
-        if (!val.parseHex(c_str(ntuple, nfield) + 2))
-            throw std::runtime_error("Pg - failed to parse hex into uint256");
-        return val;
-    }
-
-    /** Return field as equivalent to Postgres' INT type (32 bit signed).
-     *
-     * Note that this function does not guarantee that the result struct
-     * exists, or that the row and fields exist, or that the field is
-     * not null, or that the type is that requested.
-
-     * @param ntuple Row number.
-     * @param nfield Field number.
-     * @return Field contents.
-     */
-    std::int32_t
-    asInt(int ntuple = 0, int nfield = 0) const
-    {
-        return boost::lexical_cast<std::int32_t>(
-            PQgetvalue(result_.get(), ntuple, nfield));
-    }
-
-    /** Return field as equivalent to Postgres' BIGINT type (64 bit signed).
-     *
-     * Note that this function does not guarantee that the result struct
-     * exists, or that the row and fields exist, or that the field is
-     * not null, or that the type is that requested.
-
-     * @param ntuple Row number.
-     * @param nfield Field number.
-     * @return Field contents.
-     */
-    std::int64_t
-    asBigInt(int ntuple = 0, int nfield = 0) const
-    {
-        return boost::lexical_cast<std::int64_t>(
-            PQgetvalue(result_.get(), ntuple, nfield));
-    }
-
-    /** Returns whether the field is NULL or not.
-     *
-     * Note that this function does not guarantee that the result struct
-     * exists, or that the row and fields exist.
-     *
-     * @param ntuple Row number.
-     * @param nfield Field number.
-     * @return Whether field is NULL.
-     */
-    bool
-    isNull(int ntuple = 0, int nfield = 0) const
-    {
-        return PQgetisnull(result_.get(), ntuple, nfield);
-    }
-
-    /** Check whether a valid response occurred.
-     *
-     * @return Whether or not the query returned a valid response.
-     */
-    operator bool() const
-    {
-        return result_ != nullptr;
-    }
-
-    /** Message describing the query results suitable for diagnostics.
-     *
-     * If error, then the postgres error type and message are returned.
-     * Otherwise, "ok"
-     *
-     * @return Query result message.
-     */
-    std::string
-    msg() const;
-
-    /** Get number of rows in result.
-     *
-     * Note that this function does not guarantee that the result struct
-     * exists.
-     *
-     * @return Number of result rows.
-     */
-    int
-    ntuples() const
-    {
-        return PQntuples(result_.get());
-    }
-
-    /** Get number of fields in result.
-     *
-     * Note that this function does not guarantee that the result struct
-     * exists.
-     *
-     * @return Number of result fields.
-     */
-    int
-    nfields() const
-    {
-        return PQnfields(result_.get());
-    }
-
-    /** Return result status of the command.
-     *
-     * Note that this function does not guarantee that the result struct
-     * exists.
-     *
-     * @return
-     */
-    ExecStatusType
-    status() const
-    {
-        return PQresultStatus(result_.get());
-    }
-};
-
-/* Class that contains and operates upon a postgres connection. */
-class Pg
-{
-    friend class PgPool;
-    friend class PgQuery;
-
-    PgConfig const& config_;
-    boost::asio::io_context::strand strand_;
-    bool& stop_;
-    std::mutex& mutex_;
-
-    asio_socket_type socket_{nullptr, [](boost::asio::ip::tcp::socket*) {}};
-
-    // The connection object must be freed using the libpq API PQfinish() call.
-    pg_connection_type conn_{nullptr, [](PGconn* conn) { PQfinish(conn); }};
-
-    inline asio_socket_type
-    getSocket(boost::asio::yield_context& strand);
-
-    inline PgResult
-    waitForStatus(boost::asio::yield_context& yield, ExecStatusType expected);
-
-    inline void
-    flush(boost::asio::yield_context& yield);
-
-    /** Clear results from the connection.
-     *
-     * Results from previous commands must be cleared before new commands
-     * can be processed. This function should be called on connections
-     * that weren't processed completely before being reused, such as
-     * when being checked-in.
-     *
-     * @return whether or not connection still exists.
-     */
-    bool
-    clear();
-
-    /** Connect to postgres.
-     *
-     * Idempotently connects to postgres by first checking whether an
-     * existing connection is already present. If connection is not present
-     * or in an errored state, reconnects to the database.
-     */
-    void
-    connect(boost::asio::yield_context& yield);
-
-    /** Disconnect from postgres. */
-    void
-    disconnect()
-    {
-        conn_.reset();
-        socket_.reset();
-    }
-
-    /** Execute postgres query.
-     *
-     * If parameters are included, then the command should contain only a
-     * single SQL statement. If no parameters, then multiple SQL statements
-     * delimited by semi-colons can be processed. The response is from
-     * the last command executed.
-     *
-     * @param command postgres API command string.
-     * @param nParams postgres API number of parameters.
-     * @param values postgres API array of parameter.
-     * @return Query result object.
-     */
-    PgResult
-    query(
-        char const* command,
-        std::size_t const nParams,
-        char const* const* values,
-        boost::asio::yield_context& yield);
-
-    /** Execute postgres query with no parameters.
-     *
-     * @param command Query string.
-     * @return Query result object;
-     */
-    PgResult
-    query(char const* command, boost::asio::yield_context& yield)
-    {
-        return query(command, 0, nullptr, yield);
-    }
-
-    /** Execute postgres query with parameters.
-     *
-     * @param dbParams Database command and parameter values.
-     * @return Query result object.
-     */
-    PgResult
-    query(pg_params const& dbParams, boost::asio::yield_context& yield);
-
-    /** Insert multiple records into a table using Postgres' bulk COPY.
-     *
-     * Throws upon error.
-     *
-     * @param table Name of table for import.
-     * @param records Records in the COPY IN format.
-     */
-    void
-    bulkInsert(
-        char const* table,
-        std::string const& records,
-        boost::asio::yield_context& yield);
-
-public:
-    /** Constructor for Pg class.
-     *
-     * @param config Config parameters.
-     * @param j Logger object.
-     * @param stop Reference to connection pool's stop flag.
-     * @param mutex Reference to connection pool's mutex.
-     */
-    Pg(PgConfig const& config,
-       boost::asio::io_context& ctx,
-       bool& stop,
-       std::mutex& mutex)
-        : config_(config), strand_(ctx), stop_(stop), mutex_(mutex)
-    {
-    }
-};
-
-//-----------------------------------------------------------------------------
-
-/** Database connection pool.
- *
- * Allow re-use of postgres connections. Postgres connections are created
- * as needed until configurable limit is reached. After use, each connection
- * is placed in a container ordered by time of use. Each request for
- * a connection grabs the most recently used connection from the container.
- * If none are available, a new connection is used (up to configured limit).
- * Idle connections are destroyed periodically after configurable
- * timeout duration.
- *
- * This should be stored as a shared pointer so PgQuery objects can safely
- * outlive it.
- */
-class PgPool
-{
-    friend class PgQuery;
-
-    using clock_type = std::chrono::steady_clock;
-
-    boost::asio::io_context& ioc_;
-    PgConfig config_;
-    std::mutex mutex_;
-    std::condition_variable cond_;
-    std::size_t connections_{};
-    bool stop_{false};
-
-    /** Idle database connections ordered by timestamp to allow timing out. */
-    std::multimap<std::chrono::time_point<clock_type>, std::unique_ptr<Pg>>
-        idle_;
-
-    /** Get a postgres connection object.
-     *
-     * Return the most recent idle connection in the pool, if available.
-     * Otherwise, return a new connection unless we're at the threshold.
-     * If so, then wait until a connection becomes available.
-     *
-     * @return Postgres object.
-     */
-    std::unique_ptr<Pg>
-    checkout();
-
-    /** Return a postgres object to the pool for reuse.
-     *
-     * If connection is healthy, place in pool for reuse. After calling this,
-     * the container no longer have a connection unless checkout() is called.
-     *
-     * @param pg Pg object.
-     */
-    void
-    checkin(std::unique_ptr<Pg>& pg);
-
-public:
-    /** Connection pool constructor.
-     *
-     * @param pgConfig Postgres config.
-     * @param j Logger object.
-     * @param parent Stoppable parent.
-     */
-    PgPool(boost::asio::io_context& ioc, boost::json::object const& config);
-
-    ~PgPool()
-    {
-        onStop();
-    }
-
-    PgConfig&
-    config()
-    {
-        return config_;
-    }
-
-    /** Initiate idle connection timer.
-     *
-     * The PgPool object needs to be fully constructed to support asynchronous
-     * operations.
-     */
-    void
-    setup();
-
-    /** Prepare for process shutdown. (Stoppable) */
-    void
-    onStop();
-};
-
-//-----------------------------------------------------------------------------
-
-/** Class to query postgres.
- *
- * This class should be used by functions outside of this
- * compilation unit for querying postgres. It automatically acquires and
- * relinquishes a database connection to handle each query.
- */
-class PgQuery
-{
-private:
-    std::shared_ptr<PgPool> pool_;
-    std::unique_ptr<Pg> pg_;
-
-public:
-    PgQuery() = delete;
-
-    PgQuery(std::shared_ptr<PgPool> const& pool)
-        : pool_(pool), pg_(pool->checkout())
-    {
-    }
-
-    ~PgQuery()
-    {
-        pool_->checkin(pg_);
-    }
-
-    // TODO. add sendQuery and getResult, for sending the query and getting the
-    // result asynchronously. This could be useful for sending a bunch of
-    // requests concurrently
-
-    /** Execute postgres query with parameters.
-     *
-     * @param dbParams Database command with parameters.
-     * @return Result of query, including errors.
-     */
-    PgResult
-    operator()(pg_params const& dbParams, boost::asio::yield_context& yield)
-    {
-        if (!pg_)  // It means we're stopping. Return empty result.
-            return PgResult();
-        return pg_->query(dbParams, yield);
-    }
-
-    /** Execute postgres query with only command statement.
-     *
-     * @param command Command statement.
-     * @return Result of query, including errors.
-     */
-    PgResult
-    operator()(char const* command, boost::asio::yield_context& yield)
-    {
-        return operator()(pg_params{command, {}}, yield);
-    }
-
-    /** Insert multiple records into a table using Postgres' bulk COPY.
-     *
-     * Throws upon error.
-     *
-     * @param table Name of table for import.
-     * @param records Records in the COPY IN format.
-     */
-    void
-    bulkInsert(
-        char const* table,
-        std::string const& records,
-        boost::asio::yield_context& yield)
-    {
-        pg_->bulkInsert(table, records, yield);
-    }
-};
-
-//-----------------------------------------------------------------------------
-
-/** Create Postgres connection pool manager.
- *
- * @param pgConfig Configuration for Postgres.
- * @param j Logger object.
- * @param parent Stoppable parent object.
- * @return Postgres connection pool manager
- */
-std::shared_ptr<PgPool>
-make_PgPool(boost::asio::io_context& ioc, boost::json::object const& pgConfig);
-
-/** Initialize the Postgres schema.
- *
- * This function ensures that the database is running the latest version
- * of the schema.
- *
- * @param pool Postgres connection pool manager.
- */
-void
-initSchema(std::shared_ptr<PgPool> const& pool);
-void
-initAccountTx(std::shared_ptr<PgPool> const& pool);
-
-// Load the ledger info for the specified ledger/s from the database
-// @param whichLedger specifies the ledger to load via ledger sequence, ledger
-// hash or std::monostate (which loads the most recent)
-// @return vector of LedgerInfos
-std::optional<ripple::LedgerInfo>
-getLedger(
-    std::variant<std::monostate, ripple::uint256, std::uint32_t> const&
-        whichLedger,
-    std::shared_ptr<PgPool>& pgPool);
-
-#endif  // RIPPLE_CORE_PG_H_INCLUDED

src/backend/PostgresBackend.cpp

@@ -1,860 +0,0 @@
-#include <boost/asio.hpp>
-#include <boost/format.hpp>
-#include <backend/PostgresBackend.h>
-#include <thread>
-namespace Backend {
-
-// Type alias for async completion handlers
-using completion_token = boost::asio::yield_context;
-using function_type = void(boost::system::error_code);
-using result_type = boost::asio::async_result<completion_token, function_type>;
-using handler_type = typename result_type::completion_handler_type;
-
-struct HandlerWrapper
-{
-    handler_type handler;
-
-    HandlerWrapper(handler_type&& handler_) : handler(std::move(handler_))
-    {
-    }
-};
-
-PostgresBackend::PostgresBackend(
-    boost::asio::io_context& ioc,
-    boost::json::object const& config)
-    : BackendInterface(config)
-    , pgPool_(make_PgPool(ioc, config))
-    , writeConnection_(pgPool_)
-{
-    if (config.contains("write_interval"))
-    {
-        writeInterval_ = config.at("write_interval").as_int64();
-    }
-}
-void
-PostgresBackend::writeLedger(
-    ripple::LedgerInfo const& ledgerInfo,
-    std::string&& ledgerHeader)
-{
-    synchronous([&](boost::asio::yield_context yield) {
-        auto cmd = boost::format(
-            R"(INSERT INTO ledgers
-            VALUES (%u,'\x%s', '\x%s',%u,%u,%u,%u,%u,'\x%s','\x%s'))");
-
-        auto ledgerInsert = boost::str(
-            cmd % ledgerInfo.seq % ripple::strHex(ledgerInfo.hash) %
-            ripple::strHex(ledgerInfo.parentHash) % ledgerInfo.drops.drops() %
-            ledgerInfo.closeTime.time_since_epoch().count() %
-            ledgerInfo.parentCloseTime.time_since_epoch().count() %
-            ledgerInfo.closeTimeResolution.count() % ledgerInfo.closeFlags %
-            ripple::strHex(ledgerInfo.accountHash) %
-            ripple::strHex(ledgerInfo.txHash));
-
-        auto res = writeConnection_(ledgerInsert.data(), yield);
-        abortWrite_ = !res;
-        inProcessLedger = ledgerInfo.seq;
-    });
-}
-
-void
-PostgresBackend::writeAccountTransactions(
-    std::vector<AccountTransactionsData>&& data)
-{
-    if (abortWrite_)
-        return;
-    PgQuery pg(pgPool_);
-    for (auto const& record : data)
-    {
-        for (auto const& a : record.accounts)
-        {
-            std::string acct = ripple::strHex(a);
-            accountTxBuffer_ << "\\\\x" << acct << '\t'
-                             << std::to_string(record.ledgerSequence) << '\t'
-                             << std::to_string(record.transactionIndex) << '\t'
-                             << "\\\\x" << ripple::strHex(record.txHash)
-                             << '\n';
-        }
-    }
-}
-
-void
-PostgresBackend::doWriteLedgerObject(
-    std::string&& key,
-    std::uint32_t const seq,
-    std::string&& blob)
-{
-    synchronous([&](boost::asio::yield_context yield) {
-        if (abortWrite_)
-            return;
-        objectsBuffer_ << "\\\\x" << ripple::strHex(key) << '\t'
-                       << std::to_string(seq) << '\t' << "\\\\x"
-                       << ripple::strHex(blob) << '\n';
-        numRowsInObjectsBuffer_++;
-        // If the buffer gets too large, the insert fails. Not sure why. So we
-        // insert after 1 million records
-        if (numRowsInObjectsBuffer_ % writeInterval_ == 0)
-        {
-            BOOST_LOG_TRIVIAL(info)
-                << __func__ << " Flushing large buffer. num objects = "
-                << numRowsInObjectsBuffer_;
-            writeConnection_.bulkInsert("objects", objectsBuffer_.str(), yield);
-            BOOST_LOG_TRIVIAL(info) << __func__ << " Flushed large buffer";
-            objectsBuffer_.str("");
-        }
-    });
-}
-
-void
-PostgresBackend::writeSuccessor(
-    std::string&& key,
-    std::uint32_t const seq,
-    std::string&& successor)
-{
-    synchronous([&](boost::asio::yield_context yield) {
-        if (range)
-        {
-            if (successors_.count(key) > 0)
-                return;
-            successors_.insert(key);
-        }
-        successorBuffer_ << "\\\\x" << ripple::strHex(key) << '\t'
-                         << std::to_string(seq) << '\t' << "\\\\x"
-                         << ripple::strHex(successor) << '\n';
-        BOOST_LOG_TRIVIAL(trace)
-            << __func__ << ripple::strHex(key) << " - " << std::to_string(seq);
-        numRowsInSuccessorBuffer_++;
-        if (numRowsInSuccessorBuffer_ % writeInterval_ == 0)
-        {
-            BOOST_LOG_TRIVIAL(info)
-                << __func__ << " Flushing large buffer. num successors = "
-                << numRowsInSuccessorBuffer_;
-            writeConnection_.bulkInsert(
-                "successor", successorBuffer_.str(), yield);
-            BOOST_LOG_TRIVIAL(info) << __func__ << " Flushed large buffer";
-            successorBuffer_.str("");
-        }
-    });
-}
-
-void
-PostgresBackend::writeTransaction(
-    std::string&& hash,
-    std::uint32_t const seq,
-    std::uint32_t const date,
-    std::string&& transaction,
-    std::string&& metadata)
-{
-    if (abortWrite_)
-        return;
-    transactionsBuffer_ << "\\\\x" << ripple::strHex(hash) << '\t'
-                        << std::to_string(seq) << '\t' << std::to_string(date)
-                        << '\t' << "\\\\x" << ripple::strHex(transaction)
-                        << '\t' << "\\\\x" << ripple::strHex(metadata) << '\n';
-}
-
-std::uint32_t
-checkResult(PgResult const& res, std::uint32_t const numFieldsExpected)
-{
-    if (!res)
-    {
-        auto msg = res.msg();
-        BOOST_LOG_TRIVIAL(error) << __func__ << " - " << msg;
-        if (msg.find("statement timeout"))
-            throw DatabaseTimeout();
-        assert(false);
-        throw DatabaseTimeout();
-    }
-    if (res.status() != PGRES_TUPLES_OK)
-    {
-        std::stringstream msg;
-        msg << " : Postgres response should have been "
-               "PGRES_TUPLES_OK but instead was "
-            << res.status() << " - msg  = " << res.msg();
-        BOOST_LOG_TRIVIAL(error) << __func__ << " - " << msg.str();
-        assert(false);
-        throw DatabaseTimeout();
-    }
-
-    BOOST_LOG_TRIVIAL(trace)
-        << __func__ << " Postgres result msg  : " << res.msg();
-    if (res.isNull() || res.ntuples() == 0)
-    {
-        return 0;
-    }
-    else if (res.ntuples() > 0)
-    {
-        if (res.nfields() != numFieldsExpected)
-        {
-            std::stringstream msg;
-            msg << "Wrong number of fields in Postgres "
-                   "response. Expected "
-                << numFieldsExpected << ", but got " << res.nfields();
-            throw std::runtime_error(msg.str());
-            assert(false);
-        }
-    }
-    return res.ntuples();
-}
-
-ripple::LedgerInfo
-parseLedgerInfo(PgResult const& res)
-{
-    std::int64_t ledgerSeq = res.asBigInt(0, 0);
-    ripple::uint256 hash = res.asUInt256(0, 1);
-    ripple::uint256 prevHash = res.asUInt256(0, 2);
-    std::int64_t totalCoins = res.asBigInt(0, 3);
-    std::int64_t closeTime = res.asBigInt(0, 4);
-    std::int64_t parentCloseTime = res.asBigInt(0, 5);
-    std::int64_t closeTimeRes = res.asBigInt(0, 6);
-    std::int64_t closeFlags = res.asBigInt(0, 7);
-    ripple::uint256 accountHash = res.asUInt256(0, 8);
-    ripple::uint256 txHash = res.asUInt256(0, 9);
-
-    using time_point = ripple::NetClock::time_point;
-    using duration = ripple::NetClock::duration;
-
-    ripple::LedgerInfo info;
-    info.seq = ledgerSeq;
-    info.hash = hash;
-    info.parentHash = prevHash;
-    info.drops = totalCoins;
-    info.closeTime = time_point{duration{closeTime}};
-    info.parentCloseTime = time_point{duration{parentCloseTime}};
-    info.closeFlags = closeFlags;
-    info.closeTimeResolution = duration{closeTimeRes};
-    info.accountHash = accountHash;
-    info.txHash = txHash;
-    info.validated = true;
-    return info;
-}
-std::optional<std::uint32_t>
-PostgresBackend::fetchLatestLedgerSequence(
-    boost::asio::yield_context& yield) const
-{
-    PgQuery pgQuery(pgPool_);
-    pgQuery(set_timeout, yield);
-
-    auto const query =
-        "SELECT ledger_seq FROM ledgers ORDER BY ledger_seq DESC LIMIT 1";
-
-    if (auto res = pgQuery(query, yield); checkResult(res, 1))
-        return res.asBigInt(0, 0);
-
-    return {};
-}
-
-std::optional<ripple::LedgerInfo>
-PostgresBackend::fetchLedgerBySequence(
-    std::uint32_t const sequence,
-    boost::asio::yield_context& yield) const
-{
-    PgQuery pgQuery(pgPool_);
-    pgQuery(set_timeout, yield);
-
-    std::stringstream sql;
-    sql << "SELECT * FROM ledgers WHERE ledger_seq = "
-        << std::to_string(sequence);
-
-    if (auto res = pgQuery(sql.str().data(), yield); checkResult(res, 10))
-        return parseLedgerInfo(res);
-
-    return {};
-}
-
-std::optional<ripple::LedgerInfo>
-PostgresBackend::fetchLedgerByHash(
-    ripple::uint256 const& hash,
-    boost::asio::yield_context& yield) const
-{
-    PgQuery pgQuery(pgPool_);
-    pgQuery(set_timeout, yield);
-
-    std::stringstream sql;
-    sql << "SELECT * FROM ledgers WHERE ledger_hash = \'\\x"
-        << ripple::to_string(hash) << "\'";
-
-    if (auto res = pgQuery(sql.str().data(), yield); checkResult(res, 10))
-        return parseLedgerInfo(res);
-
-    return {};
-}
-
-std::optional<LedgerRange>
-PostgresBackend::hardFetchLedgerRange(boost::asio::yield_context& yield) const
-{
-    auto range = PgQuery(pgPool_)("SELECT complete_ledgers()", yield);
-    if (!range)
-        return {};
-
-    std::string res{range.c_str()};
-    BOOST_LOG_TRIVIAL(debug) << "range is = " << res;
-    try
-    {
-        size_t minVal = 0;
-        size_t maxVal = 0;
-        if (res == "empty" || res == "error" || res.empty())
-            return {};
-        else if (size_t delim = res.find('-'); delim != std::string::npos)
-        {
-            minVal = std::stol(res.substr(0, delim));
-            maxVal = std::stol(res.substr(delim + 1));
-        }
-        else
-        {
-            minVal = maxVal = std::stol(res);
-        }
-        return LedgerRange{minVal, maxVal};
-    }
-    catch (std::exception&)
-    {
-        BOOST_LOG_TRIVIAL(error)
-            << __func__ << " : "
-            << "Error parsing result of getCompleteLedgers()";
-    }
-    return {};
-}
-
-std::optional<Blob>
-PostgresBackend::doFetchLedgerObject(
-    ripple::uint256 const& key,
-    std::uint32_t const sequence,
-    boost::asio::yield_context& yield) const
-{
-    PgQuery pgQuery(pgPool_);
-    pgQuery(set_timeout, yield);
-
-    std::stringstream sql;
-    sql << "SELECT object FROM objects WHERE key = "
-        << "\'\\x" << ripple::strHex(key) << "\'"
-        << " AND ledger_seq <= " << std::to_string(sequence)
-        << " ORDER BY ledger_seq DESC LIMIT 1";
-
-    if (auto res = pgQuery(sql.str().data(), yield); checkResult(res, 1))
-    {
-        auto blob = res.asUnHexedBlob(0, 0);
-        if (blob.size())
-            return blob;
-    }
-
-    return {};
-}
-
-// returns a transaction, metadata pair
-std::optional<TransactionAndMetadata>
-PostgresBackend::fetchTransaction(
-    ripple::uint256 const& hash,
-    boost::asio::yield_context& yield) const
-{
-    PgQuery pgQuery(pgPool_);
-    pgQuery(set_timeout, yield);
-
-    std::stringstream sql;
-    sql << "SELECT transaction,metadata,ledger_seq,date FROM transactions "
-           "WHERE hash = "
-        << "\'\\x" << ripple::strHex(hash) << "\'";
-
-    if (auto res = pgQuery(sql.str().data(), yield); checkResult(res, 4))
-    {
-        return {
-            {res.asUnHexedBlob(0, 0),
-             res.asUnHexedBlob(0, 1),
-             res.asBigInt(0, 2),
-             res.asBigInt(0, 3)}};
-    }
-
-    return {};
-}
-std::vector<TransactionAndMetadata>
-PostgresBackend::fetchAllTransactionsInLedger(
-    std::uint32_t const ledgerSequence,
-    boost::asio::yield_context& yield) const
-{
-    PgQuery pgQuery(pgPool_);
-    pgQuery(set_timeout, yield);
-
-    std::stringstream sql;
-    sql << "SELECT transaction, metadata, ledger_seq,date FROM transactions "
-           "WHERE "
-        << "ledger_seq = " << std::to_string(ledgerSequence);
-
-    auto res = pgQuery(sql.str().data(), yield);
-    if (size_t numRows = checkResult(res, 4))
-    {
-        std::vector<TransactionAndMetadata> txns;
-        for (size_t i = 0; i < numRows; ++i)
-        {
-            txns.push_back(
-                {res.asUnHexedBlob(i, 0),
-                 res.asUnHexedBlob(i, 1),
-                 res.asBigInt(i, 2),
-                 res.asBigInt(i, 3)});
-        }
-        return txns;
-    }
-    return {};
-}
-std::vector<ripple::uint256>
-PostgresBackend::fetchAllTransactionHashesInLedger(
-    std::uint32_t const ledgerSequence,
-    boost::asio::yield_context& yield) const
-{
-    PgQuery pgQuery(pgPool_);
-    pgQuery(set_timeout, yield);
-
-    std::stringstream sql;
-    sql << "SELECT hash FROM transactions WHERE "
-        << "ledger_seq = " << std::to_string(ledgerSequence);
-
-    auto res = pgQuery(sql.str().data(), yield);
-    if (size_t numRows = checkResult(res, 1))
-    {
-        std::vector<ripple::uint256> hashes;
-        for (size_t i = 0; i < numRows; ++i)
-        {
-            hashes.push_back(res.asUInt256(i, 0));
-        }
-        return hashes;
-    }
-
-    return {};
-}
-
-std::optional<ripple::uint256>
-PostgresBackend::doFetchSuccessorKey(
-    ripple::uint256 key,
-    std::uint32_t const ledgerSequence,
-    boost::asio::yield_context& yield) const
-{
-    PgQuery pgQuery(pgPool_);
-    pgQuery(set_timeout, yield);
-
-    std::stringstream sql;
-    sql << "SELECT next FROM successor WHERE key = "
-        << "\'\\x" << ripple::strHex(key) << "\'"
-        << " AND ledger_seq <= " << std::to_string(ledgerSequence)
-        << " ORDER BY ledger_seq DESC LIMIT 1";
-
-    if (auto res = pgQuery(sql.str().data(), yield); checkResult(res, 1))
-    {
-        auto next = res.asUInt256(0, 0);
-        if (next == lastKey)
-            return {};
-        return next;
-    }
-
-    return {};
-}
-
-std::vector<TransactionAndMetadata>
-PostgresBackend::fetchTransactions(
-    std::vector<ripple::uint256> const& hashes,
-    boost::asio::yield_context& yield) const
-{
-    if (!hashes.size())
-        return {};
-
-    std::vector<TransactionAndMetadata> results;
-    results.resize(hashes.size());
-
-    handler_type handler(std::forward<decltype(yield)>(yield));
-    result_type result(handler);
-
-    auto hw = new HandlerWrapper(std::move(handler));
-
-    auto start = std::chrono::system_clock::now();
-
-    std::atomic_uint numRemaining = hashes.size();
-    std::atomic_bool errored = false;
-
-    for (size_t i = 0; i < hashes.size(); ++i)
-    {
-        auto const& hash = hashes[i];
-        boost::asio::spawn(
-            get_associated_executor(yield),
-            [this, &hash, &results, hw, &numRemaining, &errored, i](
-                boost::asio::yield_context yield) {
-                BOOST_LOG_TRIVIAL(trace) << __func__ << " getting txn = " << i;
-
-                PgQuery pgQuery(pgPool_);
-
-                std::stringstream sql;
-                sql << "SELECT transaction,metadata,ledger_seq,date FROM "
-                       "transactions "
-                       "WHERE HASH = \'\\x"
-                    << ripple::strHex(hash) << "\'";
-
-                try
-                {
-                    if (auto const res = pgQuery(sql.str().data(), yield);
-                        checkResult(res, 4))
-                    {
-                        results[i] = {
-                            res.asUnHexedBlob(0, 0),
-                            res.asUnHexedBlob(0, 1),
-                            res.asBigInt(0, 2),
-                            res.asBigInt(0, 3)};
-                    }
-                }
-                catch (DatabaseTimeout const&)
-                {
-                    errored = true;
-                }
-
-                if (--numRemaining == 0)
-                {
-                    handler_type h(std::move(hw->handler));
-                    h(boost::system::error_code{});
-                }
-            });
-    }
-
-    // Yields the worker to the io_context until handler is called.
-    result.get();
-
-    delete hw;
-
-    auto end = std::chrono::system_clock::now();
-    auto duration =
-        std::chrono::duration_cast<std::chrono::milliseconds>(end - start);
-
-    BOOST_LOG_TRIVIAL(info)
-        << __func__ << " fetched " << std::to_string(hashes.size())
-        << " transactions asynchronously. took "
-        << std::to_string(duration.count());
-    if (errored)
-    {
-        BOOST_LOG_TRIVIAL(error) << __func__ << " Database fetch timed out";
-        throw DatabaseTimeout();
-    }
-
-    return results;
-}
-
-std::vector<Blob>
-PostgresBackend::doFetchLedgerObjects(
-    std::vector<ripple::uint256> const& keys,
-    std::uint32_t const sequence,
-    boost::asio::yield_context& yield) const
-{
-    if (!keys.size())
-        return {};
-
-    PgQuery pgQuery(pgPool_);
-    pgQuery(set_timeout, yield);
-
-    std::vector<Blob> results;
-    results.resize(keys.size());
-
-    handler_type handler(std::forward<decltype(yield)>(yield));
-    result_type result(handler);
-
-    auto hw = new HandlerWrapper(std::move(handler));
-
-    std::atomic_uint numRemaining = keys.size();
-    std::atomic_bool errored = false;
-    auto start = std::chrono::system_clock::now();
-    for (size_t i = 0; i < keys.size(); ++i)
-    {
-        auto const& key = keys[i];
-        boost::asio::spawn(
-            boost::asio::get_associated_executor(yield),
-            [this, &key, &results, &numRemaining, &errored, hw, i, sequence](
-                boost::asio::yield_context yield) {
-                PgQuery pgQuery(pgPool_);
-
-                std::stringstream sql;
-                sql << "SELECT object FROM "
-                       "objects "
-                       "WHERE key = \'\\x"
-                    << ripple::strHex(key) << "\'"
-                    << " AND ledger_seq <= " << std::to_string(sequence)
-                    << " ORDER BY ledger_seq DESC LIMIT 1";
-
-                try
-                {
-                    if (auto const res = pgQuery(sql.str().data(), yield);
-                        checkResult(res, 1))
-                        results[i] = res.asUnHexedBlob();
-                }
-                catch (DatabaseTimeout const& ex)
-                {
-                    errored = true;
-                }
-
-                if (--numRemaining == 0)
-                {
-                    handler_type h(std::move(hw->handler));
-                    h(boost::system::error_code{});
-                }
-            });
-    }
-
-    // Yields the worker to the io_context until handler is called.
-    result.get();
-
-    delete hw;
-
-    auto end = std::chrono::system_clock::now();
-    auto duration =
-        std::chrono::duration_cast<std::chrono::milliseconds>(end - start);
-
-    BOOST_LOG_TRIVIAL(info)
-        << __func__ << " fetched " << std::to_string(keys.size())
-        << " objects asynchronously. ms = " << std::to_string(duration.count());
-    if (errored)
-    {
-        BOOST_LOG_TRIVIAL(error) << __func__ << " Database fetch timed out";
-        throw DatabaseTimeout();
-    }
-
-    return results;
-}
-
-std::vector<LedgerObject>
-PostgresBackend::fetchLedgerDiff(
-    std::uint32_t const ledgerSequence,
-    boost::asio::yield_context& yield) const
-{
-    PgQuery pgQuery(pgPool_);
-    pgQuery(set_timeout, yield);
-
-    std::stringstream sql;
-    sql << "SELECT key,object FROM objects "
-           "WHERE "
-        << "ledger_seq = " << std::to_string(ledgerSequence);
-
-    auto res = pgQuery(sql.str().data(), yield);
-    if (size_t numRows = checkResult(res, 2))
-    {
-        std::vector<LedgerObject> objects;
-        for (size_t i = 0; i < numRows; ++i)
-        {
-            objects.push_back({res.asUInt256(i, 0), res.asUnHexedBlob(i, 1)});
-        }
-        return objects;
-    }
-
-    return {};
-}
-
-AccountTransactions
-PostgresBackend::fetchAccountTransactions(
-    ripple::AccountID const& account,
-    std::uint32_t const limit,
-    bool forward,
-    std::optional<AccountTransactionsCursor> const& cursor,
-    boost::asio::yield_context& yield) const
-{
-    PgQuery pgQuery(pgPool_);
-    pgQuery(set_timeout, yield);
-    pg_params dbParams;
-
-    char const*& command = dbParams.first;
-    std::vector<std::optional<std::string>>& values = dbParams.second;
-    command =
-        "SELECT account_tx($1::bytea, $2::bigint, $3::bool, "
-        "$4::bigint, $5::bigint)";
-    values.resize(5);
-    values[0] = "\\x" + strHex(account);
-
-    values[1] = std::to_string(limit);
-
-    values[2] = std::to_string(forward);
-
-    if (cursor)
-    {
-        values[3] = std::to_string(cursor->ledgerSequence);
-        values[4] = std::to_string(cursor->transactionIndex);
-    }
-    for (size_t i = 0; i < values.size(); ++i)
-    {
-        BOOST_LOG_TRIVIAL(debug) << "value " << std::to_string(i) << " = "
-                                 << (values[i] ? values[i].value() : "null");
-    }
-
-    auto start = std::chrono::system_clock::now();
-    auto res = pgQuery(dbParams, yield);
-    auto end = std::chrono::system_clock::now();
-
-    auto duration = ((end - start).count()) / 1000000000.0;
-    BOOST_LOG_TRIVIAL(info)
-        << __func__ << " : executed stored_procedure in "
-        << std::to_string(duration)
-        << " num records = " << std::to_string(checkResult(res, 1));
-
-    checkResult(res, 1);
-
-    char const* resultStr = res.c_str();
-    BOOST_LOG_TRIVIAL(debug) << __func__ << " : "
-                             << "postgres result = " << resultStr
-                             << " : account = " << strHex(account);
-
-    boost::json::value raw = boost::json::parse(resultStr);
-    boost::json::object responseObj = raw.as_object();
-    BOOST_LOG_TRIVIAL(debug) << " parsed = " << responseObj;
-    if (responseObj.contains("transactions"))
-    {
-        auto txns = responseObj.at("transactions").as_array();
-        std::vector<ripple::uint256> hashes;
-        for (auto& hashHex : txns)
-        {
-            ripple::uint256 hash;
-            if (hash.parseHex(hashHex.at("hash").as_string().c_str() + 2))
-                hashes.push_back(hash);
-        }
-        if (responseObj.contains("cursor"))
-        {
-            return {
-                fetchTransactions(hashes, yield),
-                {{responseObj.at("cursor").at("ledger_sequence").as_int64(),
-                  responseObj.at("cursor")
-                      .at("transaction_index")
-                      .as_int64()}}};
-        }
-        return {fetchTransactions(hashes, yield), {}};
-    }
-    return {{}, {}};
-}  // namespace Backend
-
-void
-PostgresBackend::open(bool readOnly)
-{
-    initSchema(pgPool_);
-    initAccountTx(pgPool_);
-}
-
-void
-PostgresBackend::close()
-{
-}
-
-void
-PostgresBackend::startWrites() const
-{
-    synchronous([&](boost::asio::yield_context yield) {
-        numRowsInObjectsBuffer_ = 0;
-        abortWrite_ = false;
-        auto res = writeConnection_("BEGIN", yield);
-        if (!res || res.status() != PGRES_COMMAND_OK)
-        {
-            std::stringstream msg;
-            msg << "Postgres error creating transaction: " << res.msg();
-            throw std::runtime_error(msg.str());
-        }
-    });
-}
-
-bool
-PostgresBackend::doFinishWrites()
-{
-    synchronous([&](boost::asio::yield_context yield) {
-        if (!abortWrite_)
-        {
-            std::string txStr = transactionsBuffer_.str();
-            writeConnection_.bulkInsert("transactions", txStr, yield);
-            writeConnection_.bulkInsert(
-                "account_transactions", accountTxBuffer_.str(), yield);
-            std::string objectsStr = objectsBuffer_.str();
-            if (objectsStr.size())
-                writeConnection_.bulkInsert("objects", objectsStr, yield);
-            BOOST_LOG_TRIVIAL(debug)
-                << __func__ << " objects size = " << objectsStr.size()
-                << " txns size = " << txStr.size();
-            std::string successorStr = successorBuffer_.str();
-            if (successorStr.size())
-                writeConnection_.bulkInsert("successor", successorStr, yield);
-            if (!range)
-            {
-                std::stringstream indexCreate;
-                indexCreate
-                    << "CREATE INDEX diff ON objects USING hash(ledger_seq) "
-                       "WHERE NOT "
-                       "ledger_seq = "
-                    << std::to_string(inProcessLedger);
-                writeConnection_(indexCreate.str().data(), yield);
-            }
-        }
-        auto res = writeConnection_("COMMIT", yield);
-        if (!res || res.status() != PGRES_COMMAND_OK)
-        {
-            std::stringstream msg;
-            msg << "Postgres error committing transaction: " << res.msg();
-            throw std::runtime_error(msg.str());
-        }
-        transactionsBuffer_.str("");
-        transactionsBuffer_.clear();
-        objectsBuffer_.str("");
-        objectsBuffer_.clear();
-        successorBuffer_.str("");
-        successorBuffer_.clear();
-        successors_.clear();
-        accountTxBuffer_.str("");
-        accountTxBuffer_.clear();
-        numRowsInObjectsBuffer_ = 0;
-    });
-
-    return !abortWrite_;
-}
-
-bool
-PostgresBackend::doOnlineDelete(
-    std::uint32_t const numLedgersToKeep,
-    boost::asio::yield_context& yield) const
-{
-    auto rng = fetchLedgerRange();
-    if (!rng)
-        return false;
-    std::uint32_t minLedger = rng->maxSequence - numLedgersToKeep;
-    if (minLedger <= rng->minSequence)
-        return false;
-    PgQuery pgQuery(pgPool_);
-    pgQuery("SET statement_timeout TO 0", yield);
-    std::optional<ripple::uint256> cursor;
-    while (true)
-    {
-        auto [objects, curCursor] = retryOnTimeout([&]() {
-            return fetchLedgerPage(cursor, minLedger, 256, false, yield);
-        });
-        BOOST_LOG_TRIVIAL(debug) << __func__ << " fetched a page";
-        std::stringstream objectsBuffer;
-
-        for (auto& obj : objects)
-        {
-            objectsBuffer << "\\\\x" << ripple::strHex(obj.key) << '\t'
-                          << std::to_string(minLedger) << '\t' << "\\\\x"
-                          << ripple::strHex(obj.blob) << '\n';
-        }
-        pgQuery.bulkInsert("objects", objectsBuffer.str(), yield);
-        cursor = curCursor;
-        if (!cursor)
-            break;
-    }
-    BOOST_LOG_TRIVIAL(info) << __func__ << " finished inserting into objects";
-    {
-        std::stringstream sql;
-        sql << "DELETE FROM ledgers WHERE ledger_seq < "
-            << std::to_string(minLedger);
-        auto res = pgQuery(sql.str().data(), yield);
-        if (res.msg() != "ok")
-            throw std::runtime_error("Error deleting from ledgers table");
-    }
-    {
-        std::stringstream sql;
-        sql << "DELETE FROM keys WHERE ledger_seq < "
-            << std::to_string(minLedger);
-        auto res = pgQuery(sql.str().data(), yield);
-        if (res.msg() != "ok")
-            throw std::runtime_error("Error deleting from keys table");
-    }
-    {
-        std::stringstream sql;
-        sql << "DELETE FROM books WHERE ledger_seq < "
-            << std::to_string(minLedger);
-        auto res = pgQuery(sql.str().data(), yield);
-        if (res.msg() != "ok")
-            throw std::runtime_error("Error deleting from books table");
-    }
-    return true;
-}
-
-}  // namespace Backend

src/backend/PostgresBackend.h

@@ -1,145 +0,0 @@
-#ifndef RIPPLE_APP_REPORTING_POSTGRESBACKEND_H_INCLUDED
-#define RIPPLE_APP_REPORTING_POSTGRESBACKEND_H_INCLUDED
-#include <boost/json.hpp>
-#include <backend/BackendInterface.h>
-
-namespace Backend {
-class PostgresBackend : public BackendInterface
-{
-private:
-    mutable size_t numRowsInObjectsBuffer_ = 0;
-    mutable std::stringstream objectsBuffer_;
-    mutable size_t numRowsInSuccessorBuffer_ = 0;
-    mutable std::stringstream successorBuffer_;
-    mutable std::stringstream transactionsBuffer_;
-    mutable std::stringstream accountTxBuffer_;
-    std::shared_ptr<PgPool> pgPool_;
-    mutable PgQuery writeConnection_;
-    mutable bool abortWrite_ = false;
-    std::uint32_t writeInterval_ = 1000000;
-    std::uint32_t inProcessLedger = 0;
-    mutable std::unordered_set<std::string> successors_;
-
-    const char* const set_timeout = "SET statement_timeout TO 10000";
-
-public:
-    PostgresBackend(
-        boost::asio::io_context& ioc,
-        boost::json::object const& config);
-
-    std::optional<std::uint32_t>
-    fetchLatestLedgerSequence(boost::asio::yield_context& yield) const override;
-
-    std::optional<ripple::LedgerInfo>
-    fetchLedgerBySequence(
-        std::uint32_t const sequence,
-        boost::asio::yield_context& yield) const override;
-
-    std::optional<ripple::LedgerInfo>
-    fetchLedgerByHash(
-        ripple::uint256 const& hash,
-        boost::asio::yield_context& yield) const override;
-
-    std::optional<Blob>
-    doFetchLedgerObject(
-        ripple::uint256 const& key,
-        std::uint32_t const sequence,
-        boost::asio::yield_context& yield) const override;
-
-    // returns a transaction, metadata pair
-    std::optional<TransactionAndMetadata>
-    fetchTransaction(
-        ripple::uint256 const& hash,
-        boost::asio::yield_context& yield) const override;
-
-    std::vector<TransactionAndMetadata>
-    fetchAllTransactionsInLedger(
-        std::uint32_t const ledgerSequence,
-        boost::asio::yield_context& yield) const override;
-
-    std::vector<ripple::uint256>
-    fetchAllTransactionHashesInLedger(
-        std::uint32_t const ledgerSequence,
-        boost::asio::yield_context& yield) const override;
-
-    std::vector<LedgerObject>
-    fetchLedgerDiff(
-        std::uint32_t const ledgerSequence,
-        boost::asio::yield_context& yield) const override;
-
-    std::optional<LedgerRange>
-    hardFetchLedgerRange(boost::asio::yield_context& yield) const override;
-
-    std::optional<ripple::uint256>
-    doFetchSuccessorKey(
-        ripple::uint256 key,
-        std::uint32_t const ledgerSequence,
-        boost::asio::yield_context& yield) const override;
-
-    std::vector<TransactionAndMetadata>
-    fetchTransactions(
-        std::vector<ripple::uint256> const& hashes,
-        boost::asio::yield_context& yield) const override;
-
-    std::vector<Blob>
-    doFetchLedgerObjects(
-        std::vector<ripple::uint256> const& keys,
-        std::uint32_t const sequence,
-        boost::asio::yield_context& yield) const override;
-
-    AccountTransactions
-    fetchAccountTransactions(
-        ripple::AccountID const& account,
-        std::uint32_t const limit,
-        bool forward,
-        std::optional<AccountTransactionsCursor> const& cursor,
-        boost::asio::yield_context& yield) const override;
-
-    void
-    writeLedger(
-        ripple::LedgerInfo const& ledgerInfo,
-        std::string&& ledgerHeader) override;
-
-    void
-    doWriteLedgerObject(
-        std::string&& key,
-        std::uint32_t const seq,
-        std::string&& blob) override;
-
-    void
-    writeSuccessor(
-        std::string&& key,
-        std::uint32_t const seq,
-        std::string&& successor) override;
-
-    void
-    writeTransaction(
-        std::string&& hash,
-        std::uint32_t const seq,
-        std::uint32_t const date,
-        std::string&& transaction,
-        std::string&& metadata) override;
-
-    void
-    writeAccountTransactions(
-        std::vector<AccountTransactionsData>&& data) override;
-
-    void
-    open(bool readOnly) override;
-
-    void
-    close() override;
-
-    void
-    startWrites() const override;
-
-    bool
-    doFinishWrites() override;
-
-    bool
-    doOnlineDelete(
-        std::uint32_t const numLedgersToKeep,
-        boost::asio::yield_context& yield) const override;
-};
-}  // namespace Backend
-#endif

src/backend/README.md

@@ -1,174 +1,220 @@
-The data model used by clio is different than that used by rippled.
-rippled uses what is known as a SHAMap, which is a tree structure, with
-actual ledger and transaction data at the leaves of the tree. Looking up a record
-is a tree traversal, where the key is used to determine the path to the proper
-leaf node. The path from root to leaf is used as a proof-tree on the p2p network,
-where nodes can prove that a piece of data is present in a ledger by sending
-the path from root to leaf. Other nodes can verify this path and be certain
-that the data does actually exist in the ledger in question.
+# Clio 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. 
 
-clio instead flattens the data model, so lookups are O(1). This results in time
-and space savings. This is possible because clio does not participate in the peer
-to peer protocol, and thus does not need to verify any data. clio fully trusts the
-rippled nodes that are being used as a data source.
+## Data Model
+The data model used by Clio to read and write ledger data is different from what Rippled uses. Rippled uses a novel data structure named [*SHAMap*](https://github.com/ripple/rippled/blob/master/src/ripple/shamap/README.md), which is a combination of a Merkle Tree and a Radix Trie. In a SHAMap, ledger objects are stored in the root vertices of the tree. Thus, looking up a record located at the leaf node of the SHAMap executes a tree search, where the path from the root node to the leaf node is the key of the record. Rippled nodes can also generate a proof-tree by forming a subtree with all the path nodes and their neighbors, which can then be used to prove the existnce of the leaf node data to other Rippled nodes. In short, the main purpose of the SHAMap data structure is to facilitate the fast validation of data integrity between different decentralized Rippled nodes.
 
-clio uses certain features of database query languages to make this happen. Many
-databases provide the necessary features to implement the clio data model. At the
-time of writing, the data model is implemented in PostgreSQL and CQL (the query
-language used by Apache Cassandra and ScyllaDB).
+Since Clio only extracts past validated ledger data from a group of trusted Rippled nodes, it can be safely assumed that these ledger data are correct without the need to validate with other  nodes in the XRP peer-to-peer network. Because of this, Clio is able to use a flattened data model to store the past validated ledger data, which allows for direct record lookup with much faster constant time operations. 
 
-The below examples are a sort of pseudo query language
+There are three main types of data in each XRP ledger version, they are [Ledger Header](https://xrpl.org/ledger-header.html), [Transaction Set](https://xrpl.org/transaction-formats.html) and [State Data](https://xrpl.org/ledger-object-types.html). Due to the structural differences of the different types of databases, Clio may choose to represent these data using a different schema for each unique database type. 
 
-## Ledgers
+**Keywords**  
+*Sequence*: A unique incrementing identification number used to label the different ledger versions.  
+*Hash*: The SHA512-half (calculate SHA512 and take the first 256 bits) hash of various ledger data like the entire ledger or specific ledger objects.
+*Ledger Object*: The [binary-encoded](https://xrpl.org/serialization.html) STObject containing specific data (i.e. metadata, transaction data).  
+*Metadata*: The data containing [detailed information](https://xrpl.org/transaction-metadata.html#transaction-metadata) of the outcome of a specific transaction, regardless of whether the transaction was successful.  
+*Transaction data*: The data containing the [full details](https://xrpl.org/transaction-common-fields.html) of a specific transaction.  
+*Object Index*: The pseudo-random unique identifier of a ledger object, created by hashing the data of the object.  
 
-We store ledger headers in a ledgers table. In PostgreSQL, we store
-the headers in their deserialized form, so we can look up by sequence or hash.
+## Cassandra Implementation
+Cassandra is a distributed wide-column NoSQL database designed to handle large data throughput with high availability and no single point of failure. By leveraging Cassandra, Clio will be able to quickly and reliably scale up when needed simply by adding more Cassandra nodes to the Cassandra cluster configuration.
 
-In Cassandra, we store the headers as blobs. The primary table maps a ledger sequence
-to the blob, and a secondary table maps a ledger hash to a ledger sequence.
+In Cassandra, Clio will be creating 9 tables to store the ledger data, they are `ledger_transactions`, `transactions`, `ledger_hashes`, `ledger_range`, `objects`, `ledgers`, `diff`, `account_tx`, and `successor`.  Their schemas and how they work are detailed below.
 
-## Transactions
-Transactions are stored in a very basic table, with a schema like so:
+*Note, if you would like visually explore the data structure of the Cassandra database, you can first run Clio server with database `type` configured as `cassandra` to fill ledger data from Rippled nodes into Cassandra, then use a GUI database management tool like [Datastax's Opcenter](https://docs.datastax.com/en/install/6.0/install/opscInstallOpsc.html) to interactively view it.* 
 
+
+### `ledger_transactions`
 ```
-CREATE TABLE transactions (
-hash blob,
-ledger_sequence int,
-transaction blob,
-PRIMARY KEY(hash))
+CREATE TABLE clio.ledger_transactions (  
+	ledger_sequence bigint,  # The sequence number of the ledger version
+	hash blob,               # Hash of all the transactions on this ledger version
+	PRIMARY KEY (ledger_sequence, hash)  
+) WITH CLUSTERING ORDER BY (hash ASC) ...
+ ```
+This table stores the hashes of all transactions in a given ledger sequence ordered by the hash value in ascending order. 
+
+### `transactions`
 ```
-The primary key is the hash.
+CREATE TABLE clio.transactions (  
+	hash blob PRIMARY KEY,   # The transaction hash
+	date bigint,             # Date of the transaction
+	ledger_sequence bigint,  # The sequence that the transaction was validated
+	metadata blob,           # Metadata of the transaction
+	transaction blob         # Data of the transaction
+) ...
+ ```
+This table stores the full transaction and metadata of each ledger version with the transaction hash as the primary key.
 
-A common query pattern is fetching all transactions in a ledger. In PostgreSQL,
-nothing special is needed for this. We just query:
+To look up all the transactions that were validated in a ledger version with sequence `n`, one can first get the all the transaction hashes in that ledger version by querying `SELECT * FROM ledger_transactions WHERE ledger_sequence = n;`. Then, iterate through the list of hashes and query `SELECT * FROM transactions WHERE hash = one_of_the_hash_from_the_list;` to get the detailed transaction data.  
+
+### `ledger_hashes`
 ```
-SELECT * FROM transactions WHERE ledger_sequence = s;
+CREATE TABLE clio.ledger_hashes (
+	hash blob PRIMARY KEY,  # Hash of entire ledger version's data
+	sequence bigint         # The sequence of the ledger version
+) ...
+ ```
+This table stores the hash of all ledger versions by their sequences. 
+### `ledger_range`
 ```
-Cassandra doesn't handle queries like this well, since `ledger_sequence` is not
-the primary key, so we use a second table that maps a ledger sequence number
-to all of the hashes in that ledger:
+CREATE TABLE clio.ledger_range (
+	is_latest boolean PRIMARY KEY,  # Whether this sequence is the stopping range
+	sequence bigint                 # The sequence number of the starting/stopping range
+) ...
+ ```
+This table marks the range of ledger versions that is stored on this specific Cassandra node. Because of its nature, there are only two records in this table with `false` and `true` values for `is_latest`, marking the starting and ending sequence of the ledger range. 
 
+### `objects`
 ```
-CREATE TABLE transaction_hashes (
-ledger_sequence int,
-hash blob,
-PRIMARY KEY(ledger_sequence, blob))
+CREATE TABLE clio.objects (
+	key blob,         # Object index of the object
+	sequence bigint,  # The sequence this object was last updated
+	object blob,      # Data of the object
+	PRIMARY KEY (key, sequence)
+) WITH CLUSTERING ORDER BY (sequence DESC) ...
+ ```
+This table stores the specific data of all objects that ever existed on the XRP network, even if they are deleted (which is represented with a special `0x` value). The records are ordered by descending sequence, where the newest validated ledger objects are at the top. 
+
+This table is updated when all data for a given ledger sequence has been written to the various tables in the database. For each ledger, many associated records are written to different tables. This table is used as a synchronization mechanism, to prevent the application from reading data from a ledger for which all data has not yet been fully written.
+
+### `ledgers`
 ```
-This table uses a compound primary key, so we can have multiple records with
-the same ledger sequence but different hash. Looking up all of the transactions
-in a given ledger then requires querying the transaction_hashes table to get the hashes of
-all of the transactions in the ledger, and then using those hashes to query the
-transactions table. Sometimes we only want the hashes though.
-
-## Ledger data
-
-Ledger data is more complicated than transaction data. Objects have different versions,
-where applying transactions in a particular ledger changes an object with a given
-key. A basic example is an account root object: the balance changes with every
-transaction sent or received, though the key (object ID) for this object remains the same.
-
-Ledger data then is modeled like so:
+CREATE TABLE clio.ledgers (
+	sequence bigint PRIMARY KEY,  # Sequence of the ledger version
+	header blob                   # Data of the header
+) ...
+ ```
+This table stores the ledger header data of specific ledger versions by their sequence.
 
+### `diff`
 ```
-CREATE TABLE objects (
-id blob,
-ledger_sequence int,
-object blob,
-PRIMARY KEY(key,ledger_sequence))
+CREATE TABLE clio.diff (
+	seq bigint,  # Sequence of the ledger version
+	key blob,    # Hash of changes in the ledger version
+	PRIMARY KEY (seq, key)
+) WITH CLUSTERING ORDER BY (key ASC) ...
+ ```
+This table stores the object index of all the changes in each ledger version.
+
+### `account_tx`
 ```
+CREATE TABLE clio.account_tx (
+	account blob,
+	seq_idx frozen<tuple<bigint, bigint>>,  # Tuple of (ledger_index, transaction_index)
+	hash blob,                              # Hash of the transaction
+	PRIMARY KEY (account, seq_idx)
+) WITH CLUSTERING ORDER BY (seq_idx DESC) ...
+ ```
+This table stores the list of transactions affecting a given account. This includes transactions made by the account, as well as transactions received.
 
-The `objects` table has a compound primary key. This is essential. Looking up
-a ledger object as of a given ledger then is just:
+
+### `successor`
 ```
-SELECT object FROM objects WHERE id = ? and ledger_sequence <= ?
-    ORDER BY ledger_sequence DESC LIMIT 1;
+CREATE TABLE clio.successor (
+	key blob,    # Object index
+	seq bigint,  # The sequnce that this ledger object's predecessor and successor was updated
+	next blob,   # Index of the next object that existed in this sequence
+	PRIMARY KEY (key, seq)
+) WITH CLUSTERING ORDER BY (seq ASC) ...
+  ```
+This table is the important backbone of how histories of ledger objects are stored in Cassandra. The successor table stores the object index of all ledger objects that were validated on the XRP network along with the ledger sequence that the object was upated on. Due to the unique nature of the table with each key being ordered by the sequence, by tracing through the table with a specific sequence number, Clio can recreate a Linked List data structure that represents all the existing ledger object at that ledger sequence. The special value of `0x00...00` and `0xFF...FF` are used to label the head and tail of the Linked List in the successor table. The diagram below showcases how tracing through the same table but with different sequence parameter filtering can result in different Linked List data representing the corresponding past state of the ledger objects. A query like `SELECT * FROM successor WHERE key = ? AND seq <= n ORDER BY seq DESC LIMIT 1;` can effectively trace through the successor table and get the Linked List of a specific sequence `n`.
+
+![Successor Table Trace Diagram](https://raw.githubusercontent.com/Shoukozumi/clio/9b2ea3efb6b164b02e9a5f0ef6717065a70f078c/src/backend/README.png)
+*P.S.: The `diff` is `(DELETE 0x00...02, CREATE 0x00...03)` for `seq=1001` and `(CREATE 0x00...04)` for `seq=1002`, which is both accurately reflected with the Linked List trace*
+
+In each new ledger version with sequence `n`, a ledger object `v` can either be **created**, **modified**, or **deleted**. For all three of these operations, the procedure to update the successor table can be broken down in to two steps: 
+ 1. Trace through the Linked List of the previous sequence to to find the ledger object `e` with the greatest object index smaller or equal than the `v`'s index. Save `e`'s `next` value (the index of the next ledger object) as `w`.
+ 2. If `v` is...
+	 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`
 ```
-This gives us the most recent ledger object written at or before a specified ledger.
-
-When a ledger object is deleted, we write a record where `object` is just an empty blob.
-
-### Next
-Generally RPCs that read ledger data will just use the above query pattern. However,
-a few RPCs (`book_offers` and `ledger_data`) make use of a certain tree operation
-called `successor`, which takes in an object id and ledger sequence, and returns
-the id of the successor object in the ledger. This is the object in the ledger with the smallest id
-greater than the input id.
-
-This problem is quite difficult for clio's data model, since computing this
-generally requires the inner nodes of the tree, which clio doesn't store. A naive
-way to do this with PostgreSQL is like so:
+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) ...
 ```
-SELECT * FROM objects WHERE id > ? AND ledger_sequence <= s ORDER BY id ASC, ledger_sequence DESC LIMIT 1;
+This table indexes NFT IDs with their owner at a given ledger. So
 ```
-This query is not really possible with Cassandra, unless you use ALLOW FILTERING, which
-is an anti pattern (for good reason!). It would require contacting basically every node
-in the entire cluster.
-
-But even with Postgres, this query is not scalable. Why? Consider what the query
-is doing at the database level. The database starts at the input id, and begins scanning
-the table in ascending order of id. It needs to skip over any records that don't actually
-exist in the desired ledger, which are objects that have been deleted, or objects that
-were created later. As ledger history grows, this query skips over more and more records,
-which results in the query taking longer and longer. The time this query takes grows
-unbounded then, as ledger history just keeps growing. With under a million ledgers, this
-query is usable, but as we approach 10 million ledgers are more, the query starts to become very slow.
-
-To alleviate this issue, the data model uses a checkpointing method. We create a second
-table called keys, like so:
+SELECT * FROM nf_tokens
+WHERE token_id = N AND seq <= Y
+ORDER BY seq DESC LIMIT 1;
 ```
-CREATE TABLE keys (
-ledger_sequence int,
-id blob,
-PRIMARY KEY(ledger_sequence, id)
-)
+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`
 ```
-However, this table does not have an entry for every ledger sequence. Instead,
-this table has an entry for rougly every 1 million ledgers. We call these ledgers
-flag ledgers. For each flag ledger, the keys table contains every object id in that
-ledger, as well as every object id that existed in any ledger between the last flag
-ledger and this one. This is a lot of keys, but not every key that ever existed (which
-is what the naive attempt at implementing successor was iterating over). In this manner,
-the performance is bounded. If we wanted to increase the performance of the successor operation,
-we can increase the frequency of flag ledgers. However, this will use more space. 1 million
-was chosen as a reasonable tradeoff to bound the performance, but not use too much space,
-especially since this is only needed for two RPC calls.
-
-We write to this table every ledger, for each new key. However, we also need to handle
-keys that existed in the previous flag ledger. To do that, at each flag ledger, we
-iterate through the previous flag ledger, and write any keys that are still present
-in the new flag ledger. This is done asynchronously.
-
-## Account Transactions
-rippled offers a RPC called `account_tx`. This RPC returns all transactions that
-affect a given account, and allows users to page backwards or forwards in time.
-Generally, this is a modeled with a table like so:
+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) ...
 ```
-CREATE TABLE account_tx (
-account blob,
-ledger_sequence int,
-transaction_index int,
-hash blob,
-PRIMARY KEY(account,ledger_sequence,transaction_index))
+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`
 ```
-
-An example of looking up from this table going backwards in time is:
+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) ...
 ```
-SELECT hash FROM account_tx WHERE account = ? 
-    AND ledger_sequence <= ? and transaction_index <= ? 
-    ORDER BY ledger_sequence DESC, transaction_index 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.
 
-This query returns the hashes, and then we use those hashes to read from the 
-transactions table.
-
-## Comments
-There are various nuances around how these data models are tuned and optimized
-for each database implementation. Cassandra and PostgreSQL are very different,
-so some slight modifications are needed. However, the general model outlined here
-is implemented by both databases, and when adding a new database, this general model
-should be followed, unless there is a good reason not to. Generally, a database will be
-decently similar to either PostgreSQL or Cassandra, so using those as a basis should
-be sufficient.
-
-Whatever database is used, clio requires strong consistency, and durability. For this
-reason, any replication strategy needs to maintain strong consistency.

src/backend/SimpleCache.cpp

@@ -1,104 +0,0 @@
-#include <backend/SimpleCache.h>
-namespace Backend {
-uint32_t
-SimpleCache::latestLedgerSequence() const
-{
-    std::shared_lock lck{mtx_};
-    return latestSeq_;
-}
-
-void
-SimpleCache::update(
-    std::vector<LedgerObject> const& objs,
-    uint32_t seq,
-    bool isBackground)
-{
-    {
-        std::unique_lock lck{mtx_};
-        if (seq > latestSeq_)
-        {
-            assert(seq == latestSeq_ + 1 || latestSeq_ == 0);
-            latestSeq_ = seq;
-        }
-        for (auto const& obj : objs)
-        {
-            if (obj.blob.size())
-            {
-                if (isBackground && deletes_.count(obj.key))
-                    continue;
-                auto& e = map_[obj.key];
-                if (seq > e.seq)
-                {
-                    e = {seq, obj.blob};
-                }
-            }
-            else
-            {
-                map_.erase(obj.key);
-                if (!full_ && !isBackground)
-                    deletes_.insert(obj.key);
-            }
-        }
-    }
-}
-std::optional<LedgerObject>
-SimpleCache::getSuccessor(ripple::uint256 const& key, uint32_t seq) const
-{
-    if (!full_)
-        return {};
-    std::shared_lock{mtx_};
-    if (seq != latestSeq_)
-        return {};
-    auto e = map_.upper_bound(key);
-    if (e == map_.end())
-        return {};
-    return {{e->first, e->second.blob}};
-}
-std::optional<LedgerObject>
-SimpleCache::getPredecessor(ripple::uint256 const& key, uint32_t seq) const
-{
-    if (!full_)
-        return {};
-    std::shared_lock lck{mtx_};
-    if (seq != latestSeq_)
-        return {};
-    auto e = map_.lower_bound(key);
-    if (e == map_.begin())
-        return {};
-    --e;
-    return {{e->first, e->second.blob}};
-}
-std::optional<Blob>
-SimpleCache::get(ripple::uint256 const& key, uint32_t seq) const
-{
-    if (seq > latestSeq_)
-        return {};
-    std::shared_lock lck{mtx_};
-    auto e = map_.find(key);
-    if (e == map_.end())
-        return {};
-    if (seq < e->second.seq)
-        return {};
-    return {e->second.blob};
-}
-
-void
-SimpleCache::setFull()
-{
-    full_ = true;
-    std::unique_lock lck{mtx_};
-    deletes_.clear();
-}
-
-bool
-SimpleCache::isFull() const
-{
-    return full_;
-}
-size_t
-SimpleCache::size() const
-{
-    std::shared_lock lck{mtx_};
-    return map_.size();
-}
-}  // namespace Backend

src/backend/SimpleCache.h

@@ -1,63 +0,0 @@
-#ifndef CLIO_SIMPLECACHE_H_INCLUDED
-#define CLIO_SIMPLECACHE_H_INCLUDED
-
-#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;
-    };
-    std::map<ripple::uint256, CacheEntry> map_;
-    mutable std::shared_mutex mtx_;
-    uint32_t latestSeq_ = 0;
-    std::atomic_bool full_ = 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
-    setFull();
-
-    uint32_t
-    latestLedgerSequence() const;
-
-    // whether the cache has all data for the most recent ledger
-    bool
-    isFull() const;
-
-    size_t
-    size() const;
-};
-
-}  // namespace Backend
-#endif

src/backend/Types.h

@@ -1,6 +1,26 @@
-#ifndef CLIO_TYPES_H_INCLUDED
-#define CLIO_TYPES_H_INCLUDED
+//------------------------------------------------------------------------------
+/*
+    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>
@@ -36,8 +56,27 @@ struct TransactionAndMetadata
 {
     Blob transaction;
     Blob metadata;
-    std::uint32_t ledgerSequence;
-    std::uint32_t date;
+    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
     {
@@ -46,16 +85,72 @@ struct TransactionAndMetadata
     }
 };
 
-struct AccountTransactionsCursor
+struct TransactionsCursor
 {
     std::uint32_t ledgerSequence;
     std::uint32_t transactionIndex;
+
+    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&
+    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);
+    }
 };
 
-struct AccountTransactions
+struct TransactionsAndCursor
 {
     std::vector<TransactionAndMetadata> txns;
-    std::optional<AccountTransactionsCursor> cursor;
+    std::optional<TransactionsCursor> cursor;
+};
+
+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;
+    }
 };
 
 struct LedgerRange
@@ -63,11 +158,7 @@ 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"};
+constexpr ripple::uint256 firstKey{"0000000000000000000000000000000000000000000000000000000000000000"};
+constexpr ripple::uint256 lastKey{"FFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFF"};
+constexpr ripple::uint256 hi192{"0000000000000000000000000000000000000000000000001111111111111111"};
 }  // namespace Backend
-#endif

src/backend/cassandra/Concepts.h

@@ -0,0 +1,79 @@
+//------------------------------------------------------------------------------
+/*
+    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 <backend/cassandra/Types.h>
+
+#include <boost/asio/spawn.hpp>
+
+#include <chrono>
+#include <concepts>
+#include <optional>
+#include <string>
+
+namespace Backend::Cassandra {
+
+// 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
+
+// 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
+
+// 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 Backend::Cassandra

src/backend/cassandra/Error.h

@@ -0,0 +1,99 @@
+//------------------------------------------------------------------------------
+/*
+    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 Backend::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;
+    }
+
+    std::string
+    message() const
+    {
+        return message_;
+    }
+
+    uint32_t
+    code() const
+    {
+        return code_;
+    }
+
+    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;
+    }
+
+    bool
+    isInvalidQuery() const
+    {
+        return code_ == CASS_ERROR_SERVER_INVALID_QUERY;
+    }
+};
+
+}  // namespace Backend::Cassandra

src/backend/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 <backend/cassandra/Handle.h>
+
+namespace Backend::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 Backend::Cassandra

src/backend/cassandra/Handle.h

@@ -0,0 +1,295 @@
+//------------------------------------------------------------------------------
+/*
+    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 <backend/cassandra/Error.h>
+#include <backend/cassandra/Types.h>
+#include <backend/cassandra/impl/Batch.h>
+#include <backend/cassandra/impl/Cluster.h>
+#include <backend/cassandra/impl/Future.h>
+#include <backend/cassandra/impl/ManagedObject.h>
+#include <backend/cassandra/impl/Result.h>
+#include <backend/cassandra/impl/Session.h>
+#include <backend/cassandra/impl/Statement.h>
+#include <util/Expected.h>
+
+#include <cassandra.h>
+
+#include <chrono>
+#include <compare>
+#include <iterator>
+#include <vector>
+
+namespace Backend::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 Settings object
+     */
+    explicit Handle(Settings clusterSettings = Settings::defaultSettings());
+
+    /**
+     * @brief Construct a new handle with default settings and only by setting
+     * the contact points
+     */
+    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.
+     */
+    [[nodiscard]] MaybeErrorType
+    connect() const;
+
+    /**
+     * @brief Connect to the the specified keyspace asynchronously
+     *
+     * @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.
+     */
+    [[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.
+     */
+    [[maybe_unused]] MaybeErrorType
+    disconnect() const;
+
+    /**
+     * @brief Reconnect to the the specified keyspace asynchronously
+     *
+     * @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.
+     */
+    [[nodiscard]] MaybeErrorType
+    reconnect(std::string_view keyspace) const;
+
+    /**
+     * @brief Execute a simple query with optional args asynchronously
+     *
+     * @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 @ref asyncExecute(std::string_view, Args&&...) const for how this
+     * works.
+     */
+    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.
+     *
+     * @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.
+     */
+    [[maybe_unused]] MaybeErrorType
+    executeEach(std::vector<StatementType> const& statements) const;
+
+    /**
+     * @brief Execute a prepared statement with optional args asynchronously
+     *
+     * @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 @ref asyncExecute(std::vector<StatementType> const&, Args&&...) const
+     * for how this works.
+     */
+    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
+     *
+     * @return A future
+     */
+    [[nodiscard]] FutureType
+    asyncExecute(StatementType const& statement) const;
+
+    /**
+     * @brief Execute one (bound or simple) statements asynchronously with a
+     * callback
+     *
+     * @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.
+     */
+    [[maybe_unused]] ResultOrErrorType
+    execute(StatementType const& statement) const;
+
+    /**
+     * @brief Execute a batch of (bound or simple) statements asynchronously
+     *
+     * @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.
+     */
+    [[maybe_unused]] MaybeErrorType
+    execute(std::vector<StatementType> const& statements) const;
+
+    /**
+     * @brief Execute a batch of (bound or simple) statements asynchronously
+     * with a completion callback
+     *
+     * @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
+     *
+     * @return A @ref PreparedStatementType
+     * @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.
+ */
+template <typename... Types>
+[[nodiscard]] detail::ResultExtractor<Types...>
+extract(Handle::ResultType const& result)
+{
+    return {result};
+}
+
+}  // namespace Backend::Cassandra

src/backend/cassandra/Schema.h

@@ -0,0 +1,669 @@
+//------------------------------------------------------------------------------
+/*
+    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 <backend/cassandra/Concepts.h>
+#include <backend/cassandra/Handle.h>
+#include <backend/cassandra/SettingsProvider.h>
+#include <backend/cassandra/Types.h>
+#include <config/Config.h>
+#include <log/Logger.h>
+#include <util/Expected.h>
+
+#include <fmt/compile.h>
+
+namespace Backend::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
+{
+    // Current schema version.
+    // Update this everytime you update the schema.
+    // Migrations will be ran automatically based on this value.
+    static constexpr uint16_t version = 1u;
+
+    clio::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")));
+        }();
+
+        // the smallest transaction idx is 0, we use uint to store the transaction index, so we shall use ">=" to
+        // include it(the transaction with idx 0) in the result
+        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_.info() << "Preparing cassandra statements";
+        statements_ = std::make_unique<Statements>(settingsProvider_, handle);
+        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 Backend::Cassandra

src/backend/cassandra/SettingsProvider.cpp

@@ -0,0 +1,125 @@
+//------------------------------------------------------------------------------
+/*
+    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 <backend/cassandra/SettingsProvider.h>
+#include <backend/cassandra/impl/Cluster.h>
+#include <backend/cassandra/impl/Statement.h>
+#include <config/Config.h>
+
+#include <boost/json.hpp>
+
+#include <string>
+#include <thread>
+
+namespace Backend::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");
+
+    clio::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(clio::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.certificate = parseOptionalCertificate();
+    settings.username = config_.maybeValue<std::string>("username");
+    settings.password = config_.maybeValue<std::string>("password");
+
+    return settings;
+}
+
+}  // namespace Backend::Cassandra

src/backend/cassandra/SettingsProvider.h

@@ -0,0 +1,86 @@
+//------------------------------------------------------------------------------
+/*
+    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 <backend/cassandra/Handle.h>
+#include <backend/cassandra/Types.h>
+#include <config/Config.h>
+#include <log/Logger.h>
+#include <util/Expected.h>
+
+namespace Backend::Cassandra {
+
+/**
+ * @brief Provides settings for @ref CassandraBackend
+ */
+class SettingsProvider
+{
+    clio::Config config_;
+
+    std::string keyspace_;
+    std::optional<std::string> tablePrefix_;
+    uint16_t replicationFactor_;
+    uint16_t ttl_;
+    Settings settings_;
+
+public:
+    explicit SettingsProvider(clio::Config const& cfg, uint16_t ttl = 0);
+
+    /*! Get the cluster settings */
+    [[nodiscard]] Settings
+    getSettings() const;
+
+    /*! Get the specified keyspace */
+    [[nodiscard]] inline std::string
+    getKeyspace() const
+    {
+        return keyspace_;
+    }
+
+    /*! Get an optional table prefix to use in all queries */
+    [[nodiscard]] inline std::optional<std::string>
+    getTablePrefix() const
+    {
+        return tablePrefix_;
+    }
+
+    /*! Get the replication factor */
+    [[nodiscard]] inline uint16_t
+    getReplicationFactor() const
+    {
+        return replicationFactor_;
+    }
+
+    /*! Get 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 Backend::Cassandra

src/backend/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 Backend::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 Backend::Cassandra

src/backend/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 <backend/cassandra/Concepts.h>
+#include <backend/cassandra/Handle.h>
+#include <backend/cassandra/Types.h>
+#include <backend/cassandra/impl/RetryPolicy.h>
+#include <log/Logger.h>
+#include <util/Expected.h>
+
+#include <boost/asio.hpp>
+
+#include <functional>
+#include <memory>
+
+namespace Backend::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)>;
+
+    clio::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 Backend::Cassandra::detail

src/backend/cassandra/impl/Batch.cpp

@@ -0,0 +1,56 @@
+//------------------------------------------------------------------------------
+/*
+    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 <backend/cassandra/Error.h>
+#include <backend/cassandra/impl/Batch.h>
+#include <backend/cassandra/impl/Statement.h>
+#include <util/Expected.h>
+
+#include <exception>
+#include <vector>
+
+namespace {
+static constexpr auto batchDeleter = [](CassBatch* ptr) { cass_batch_free(ptr); };
+};
+
+namespace Backend::Cassandra::detail {
+
+// todo: use an appropritae 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 Backend::Cassandra::detail

src/backend/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 <backend/cassandra/Types.h>
+#include <backend/cassandra/impl/ManagedObject.h>
+
+#include <cassandra.h>
+
+namespace Backend::Cassandra::detail {
+
+struct Batch : public ManagedObject<CassBatch>
+{
+    Batch(std::vector<Statement> const& statements);
+
+    MaybeError
+    add(Statement const& statement);
+};
+
+}  // namespace Backend::Cassandra::detail

src/backend/cassandra/impl/Cluster.cpp

@@ -0,0 +1,154 @@
+//------------------------------------------------------------------------------
+/*
+    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 <backend/cassandra/impl/Cluster.h>
+#include <backend/cassandra/impl/SslContext.h>
+#include <backend/cassandra/impl/Statement.h>
+#include <util/Expected.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 Backend::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(std::string{"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(
+            std::string{"Error setting cassandra io threads to "} + to_string(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());
+
+    // TODO: other options to experiment with and consider later:
+    // cass_cluster_set_max_concurrent_requests_threshold(*this, 10000);
+    // cass_cluster_set_queue_size_event(*this, 100000);
+    // cass_cluster_set_queue_size_io(*this, 100000);
+    // cass_cluster_set_write_bytes_high_water_mark(*this, 16 * 1024 * 1024);  // 16mb
+    // cass_cluster_set_write_bytes_low_water_mark(*this, 8 * 1024 * 1024);  // half of allowance
+    // cass_cluster_set_pending_requests_high_water_mark(*this, 5000);
+    // cass_cluster_set_pending_requests_low_water_mark(*this, 2500);  // half
+    // cass_cluster_set_max_requests_per_flush(*this, 1000);
+    // cass_cluster_set_max_concurrent_creation(*this, 8);
+    // cass_cluster_set_max_connections_per_host(*this, 6);
+    // cass_cluster_set_core_connections_per_host(*this, 4);
+    // cass_cluster_set_constant_speculative_execution_policy(*this, 1000, 1024);
+
+    if (auto const rc = cass_cluster_set_queue_size_io(
+            *this, settings.maxWriteRequestsOutstanding + settings.maxReadRequestsOutstanding);
+        rc != CASS_OK)
+    {
+        throw std::runtime_error(std::string{"Could not set queue size for IO per host: "} + cass_error_desc(rc));
+    }
+
+    setupConnection(settings);
+    setupCertificate(settings);
+    setupCredentials(settings);
+}
+
+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("Cassandra: Error setting " + label + " [" + value + "]: " + cass_error_desc(rc));
+    };
+
+    {
+        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_.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_.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_.debug() << "Set credentials; username: " << settings.username.value();
+    cass_cluster_set_credentials(*this, settings.username.value().c_str(), settings.password.value().c_str());
+}
+
+}  // namespace Backend::Cassandra::detail

src/backend/cassandra/impl/Cluster.h

@@ -0,0 +1,99 @@
+//------------------------------------------------------------------------------
+/*
+    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 <backend/cassandra/impl/ManagedObject.h>
+#include <log/Logger.h>
+
+#include <cassandra.h>
+
+#include <chrono>
+#include <optional>
+#include <string>
+#include <string_view>
+#include <thread>
+#include <variant>
+
+namespace Backend::Cassandra::detail {
+
+struct Settings
+{
+    struct ContactPoints
+    {
+        std::string contactPoints = "127.0.0.1";  // defaults to localhost
+        std::optional<uint16_t> port;
+    };
+
+    struct SecureConnectionBundle
+    {
+        std::string bundle;  // no meaningful default
+    };
+
+    bool enableLog = false;
+    std::chrono::milliseconds connectionTimeout = std::chrono::milliseconds{10000};
+    std::chrono::milliseconds requestTimeout = std::chrono::milliseconds{0};  // no timeout at all
+    std::variant<ContactPoints, SecureConnectionBundle> connectionInfo = ContactPoints{};
+    uint32_t threads = std::thread::hardware_concurrency();
+    uint32_t maxWriteRequestsOutstanding = 10'000;
+    uint32_t maxReadRequestsOutstanding = 100'000;
+    std::optional<std::string> certificate;  // ssl context
+    std::optional<std::string> username;
+    std::optional<std::string> password;
+
+    Settings
+    withContactPoints(std::string_view contactPoints)
+    {
+        auto tmp = *this;
+        tmp.connectionInfo = ContactPoints{std::string{contactPoints}};
+        return tmp;
+    }
+
+    static Settings
+    defaultSettings()
+    {
+        return Settings();
+    }
+};
+
+class Cluster : public ManagedObject<CassCluster>
+{
+    clio::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 Backend::Cassandra::detail

src/backend/cassandra/impl/ExecutionStrategy.h

@@ -0,0 +1,443 @@
+//------------------------------------------------------------------------------
+/*
+    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 <backend/cassandra/Handle.h>
+#include <backend/cassandra/Types.h>
+#include <backend/cassandra/impl/AsyncExecutor.h>
+#include <log/Logger.h>
+#include <util/Expected.h>
+
+#include <boost/asio/async_result.hpp>
+#include <boost/asio/spawn.hpp>
+
+#include <atomic>
+#include <condition_variable>
+#include <functional>
+#include <memory>
+#include <mutex>
+#include <optional>
+#include <thread>
+
+namespace Backend::Cassandra::detail {
+
+/**
+ * @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
+{
+    clio::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;
+    using FunctionType = void(boost::system::error_code);
+    using AsyncResultType = boost::asio::async_result<CompletionTokenType, FunctionType>;
+    using HandlerType = typename AsyncResultType::completion_handler_type;
+
+    DefaultExecutionStrategy(Settings settings, HandleType const& handle)
+        : maxWriteRequestsOutstanding_{settings.maxWriteRequestsOutstanding}
+        , maxReadRequestsOutstanding_{settings.maxReadRequestsOutstanding}
+        , work_{ioc_}
+        , handle_{std::cref(handle)}
+        , thread_{[this]() { ioc_.run(); }}
+    {
+        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_.debug() << "Waiting to sync all writes...";
+        std::unique_lock<std::mutex> lck(syncMutex_);
+        syncCv_.wait(lck, [this]() { return finishedAllWriteRequests(); });
+        log_.debug() << "Sync done.";
+    }
+
+    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_.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 prepradeStatement 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 prepradeStatement 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 handler = HandlerType{token};
+        auto result = AsyncResultType{handler};
+        auto const numStatements = statements.size();
+
+        // todo: perhaps use policy instead
+        while (true)
+        {
+            numReadRequestsOutstanding_ += numStatements;
+
+            auto const future = handle_.get().asyncExecute(statements, [handler](auto&&) mutable {
+                boost::asio::post(boost::asio::get_associated_executor(handler), [handler]() mutable {
+                    handler(boost::system::error_code{});
+                });
+            });
+
+            // suspend coroutine until completion handler is called
+            result.get();
+
+            numReadRequestsOutstanding_ -= numStatements;
+
+            // it's safe to call blocking get on future here as we already
+            // waited for the coroutine to resume above.
+            if (auto res = future.get(); res)
+            {
+                return res;
+            }
+            else
+            {
+                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)
+    {
+        auto handler = HandlerType{token};
+        auto result = AsyncResultType{handler};
+
+        // todo: perhaps use policy instead
+        while (true)
+        {
+            ++numReadRequestsOutstanding_;
+
+            auto const future = handle_.get().asyncExecute(statement, [handler](auto const&) mutable {
+                boost::asio::post(boost::asio::get_associated_executor(handler), [handler]() mutable {
+                    handler(boost::system::error_code{});
+                });
+            });
+
+            // suspend coroutine until completion handler is called
+            result.get();
+
+            --numReadRequestsOutstanding_;
+
+            // it's safe to call blocking get on future here as we already
+            // waited for the coroutine to resume above.
+            if (auto res = future.get(); res)
+            {
+                return res;
+            }
+            else
+            {
+                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)
+    {
+        auto handler = HandlerType{token};
+        auto result = AsyncResultType{handler};
+
+        std::atomic_bool hadError = false;
+        std::atomic_int numOutstanding = statements.size();
+        numReadRequestsOutstanding_ += statements.size();
+
+        auto futures = std::vector<FutureWithCallbackType>{};
+        futures.reserve(numOutstanding);
+
+        // used as the handler for each async statement individually
+        auto executionHandler = [handler, &hadError, &numOutstanding](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(handler), [handler]() mutable {
+                    handler(boost::system::error_code{});
+                });
+        };
+
+        std::transform(
+            std::cbegin(statements),
+            std::cend(statements),
+            std::back_inserter(futures),
+            [this, &executionHandler](auto const& statement) {
+                return handle_.get().asyncExecute(statement, executionHandler);
+            });
+
+        // suspend coroutine until completion handler is called
+        result.get();
+
+        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_.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 Backend::Cassandra::detail

src/backend/cassandra/impl/Future.cpp

@@ -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.
+*/
+//==============================================================================
+
+#include <backend/cassandra/Error.h>
+#include <backend/cassandra/impl/Future.h>
+#include <backend/cassandra/impl/Result.h>
+
+#include <exception>
+#include <vector>
+
+namespace {
+static constexpr auto futureDeleter = [](CassFuture* ptr) { cass_future_free(ptr); };
+}  // namespace
+
+namespace Backend::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 :/
+    auto* cb = static_cast<FutureWithCallback::fn_t*>(cbPtr);
+    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");
+        (*cb)(Error{CassandraError{errMsg, rc}});
+    }
+    else
+    {
+        (*cb)(Result{cass_future_get_result(ptr)});
+    }
+}
+
+/* implicit */ FutureWithCallback::FutureWithCallback(CassFuture* ptr, fn_t&& cb)
+    : Future{ptr}, cb_{std::make_unique<fn_t>(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 Backend::Cassandra::detail

src/backend/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 <backend/cassandra/Types.h>
+#include <backend/cassandra/impl/ManagedObject.h>
+
+#include <cassandra.h>
+
+namespace Backend::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 fn_t = std::function<void(ResultOrError)>;
+    using fn_ptr_t = std::unique_ptr<fn_t>;
+
+    /* implicit */ FutureWithCallback(CassFuture* ptr, fn_t&& cb);
+    FutureWithCallback(FutureWithCallback const&) = delete;
+    FutureWithCallback(FutureWithCallback&&) = default;
+
+private:
+    /*! Wrapped in a unique_ptr so it can survive std::move :/ */
+    fn_ptr_t cb_;
+};
+
+}  // namespace Backend::Cassandra::detail

src/backend/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 Backend::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() const
+    {
+        return ptr_.get();
+    }
+};
+
+}  // namespace Backend::Cassandra::detail

src/backend/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 <backend/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 Backend::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 Backend::Cassandra::detail

src/backend/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 <backend/cassandra/impl/ManagedObject.h>
+#include <backend/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 Backend::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 decayed_t = std::decay_t<Type>;
+    using uint_tuple_t = std::tuple<uint32_t, uint32_t>;
+    using uchar_vector_t = std::vector<unsigned char>;
+
+    if constexpr (std::is_same_v<decayed_t, 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<decayed_t, 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<decayed_t, uchar_vector_t>)
+    {
+        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 = uchar_vector_t{buf, buf + bufSize};
+    }
+    else if constexpr (std::is_same_v<decayed_t, uint_tuple_t>)
+    {
+        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<decayed_t, 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<decayed_t, 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<decayed_t, 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<decayed_t>(out);
+    }
+    else
+    {
+        // type not supported for extraction
+        static_assert(unsupported_v<decayed_t>);
+    }
+
+    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 Backend::Cassandra::detail

src/backend/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 <backend/cassandra/Handle.h>
+#include <backend/cassandra/Types.h>
+#include <log/Logger.h>
+#include <util/Expected.h>
+
+#include <boost/asio.hpp>
+
+#include <algorithm>
+#include <chrono>
+#include <cmath>
+
+namespace Backend::Cassandra::detail {
+
+/**
+ * @brief A retry policy that employs exponential backoff
+ */
+class ExponentialBackoffRetryPolicy
+{
+    clio::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_{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_.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 Backend::Cassandra::detail

src/backend/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 <backend/cassandra/impl/ManagedObject.h>
+
+#include <cassandra.h>
+
+namespace Backend::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 Backend::Cassandra::detail

src/backend/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 <backend/cassandra/impl/SslContext.h>
+
+namespace {
+static constexpr auto contextDeleter = [](CassSsl* ptr) { cass_ssl_free(ptr); };
+}  // namespace
+
+namespace Backend::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 Backend::Cassandra::detail

src/backend/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 <backend/cassandra/impl/ManagedObject.h>
+
+#include <cassandra.h>
+
+#include <string>
+
+namespace Backend::Cassandra::detail {
+
+struct SslContext : public ManagedObject<CassSsl>
+{
+    explicit SslContext(std::string const& certificate);
+};
+
+}  // namespace Backend::Cassandra::detail

src/backend/cassandra/impl/Statement.h

@@ -0,0 +1,164 @@
+//------------------------------------------------------------------------------
+/*
+    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 <backend/cassandra/Types.h>
+#include <backend/cassandra/impl/ManagedObject.h>
+#include <backend/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 Backend::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;
+
+    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)
+                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 decayed_t = std::decay_t<Type>;
+        using uchar_vec_t = std::vector<unsigned char>;
+        using uint_tuple_t = std::tuple<uint32_t, uint32_t>;
+
+        if constexpr (std::is_same_v<decayed_t, ripple::uint256>)
+        {
+            auto const rc = bindBytes(value.data(), value.size());
+            throwErrorIfNeeded(rc, "Bind ripple::uint256");
+        }
+        else if constexpr (std::is_same_v<decayed_t, ripple::AccountID>)
+        {
+            auto const rc = bindBytes(value.data(), value.size());
+            throwErrorIfNeeded(rc, "Bind ripple::AccountID");
+        }
+        else if constexpr (std::is_same_v<decayed_t, uchar_vec_t>)
+        {
+            auto const rc = bindBytes(value.data(), value.size());
+            throwErrorIfNeeded(rc, "Bind vector<unsigned char>");
+        }
+        else if constexpr (std::is_convertible_v<decayed_t, 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<decayed_t, uint_tuple_t>)
+        {
+            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<decayed_t, 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<decayed_t, 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<decayed_t, 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<decayed_t>);
+        }
+    }
+};
+
+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}
+    {
+    }
+
+    template <typename... Args>
+    Statement
+    bind(Args&&... args) const
+    {
+        Statement statement = cass_prepared_bind(*this);
+        statement.bind<Args...>(std::forward<Args>(args)...);
+        return statement;
+    }
+};
+
+}  // namespace Backend::Cassandra::detail

src/backend/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 <backend/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 Backend::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 Backend::Cassandra::detail

src/backend/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 <backend/cassandra/impl/ManagedObject.h>
+
+#include <cassandra.h>
+
+#include <functional>
+#include <string>
+#include <string_view>
+#include <tuple>
+
+namespace Backend::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 decayed_t = std::decay_t<Type>;
+
+        if constexpr (std::is_same_v<decayed_t, 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<decayed_t, 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<decayed_t>);
+        }
+    }
+};
+
+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 decayed_t = std::decay_t<Type>;
+
+        // clio only uses bigint (int64_t) so we convert any incoming type
+        if constexpr (std::is_convertible_v<decayed_t, 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<decayed_t>(out);
+        }
+        else
+        {
+            // type not supported for extraction
+            static_assert(unsupported_v<decayed_t>);
+        }
+
+        return output;
+    }
+};
+
+}  // namespace Backend::Cassandra::detail

src/config/Config.cpp

@@ -0,0 +1,183 @@
+//------------------------------------------------------------------------------
+/*
+    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 <config/Config.h>
+#include <log/Logger.h>
+
+#include <fstream>
+
+namespace clio {
+
+// Note: `store_(store)` MUST use `()` instead of `{}` otherwise gcc
+// picks `initializer_list` constructor and anything passed becomes an
+// array :-D
+Config::Config(boost::json::value store) : store_(std::move(store))
+{
+}
+
+Config::operator bool() const noexcept
+{
+    return not store_.is_null();
+}
+
+bool
+Config::contains(key_type key) const
+{
+    return lookup(key).has_value();
+}
+
+std::optional<boost::json::value>
+Config::lookup(key_type key) const
+{
+    if (store_.is_null())
+        return std::nullopt;
+
+    std::reference_wrapper<boost::json::value const> cur = std::cref(store_);
+    auto hasBrokenPath = false;
+    auto tokenized = detail::Tokenizer<key_type, Separator>{key};
+    std::string subkey{};
+
+    auto maybeSection = tokenized.next();
+    while (maybeSection.has_value())
+    {
+        auto section = maybeSection.value();
+        subkey += section;
+
+        if (not hasBrokenPath)
+        {
+            if (not cur.get().is_object())
+                throw detail::StoreException("Not an object at '" + subkey + "'");
+            if (not cur.get().as_object().contains(section))
+                hasBrokenPath = true;
+            else
+                cur = std::cref(cur.get().as_object().at(section));
+        }
+
+        subkey += Separator;
+        maybeSection = tokenized.next();
+    }
+
+    if (hasBrokenPath)
+        return std::nullopt;
+    return std::make_optional(cur);
+}
+
+std::optional<Config::array_type>
+Config::maybeArray(key_type key) const
+{
+    try
+    {
+        auto maybe_arr = lookup(key);
+        if (maybe_arr && maybe_arr->is_array())
+        {
+            auto& arr = maybe_arr->as_array();
+            array_type out;
+            out.reserve(arr.size());
+
+            std::transform(std::begin(arr), std::end(arr), std::back_inserter(out), [](auto&& element) {
+                return Config{std::move(element)};
+            });
+            return std::make_optional<array_type>(std::move(out));
+        }
+    }
+    catch (detail::StoreException const&)
+    {
+        // ignore store error, but rethrow key errors
+    }
+
+    return std::nullopt;
+}
+
+Config::array_type
+Config::array(key_type key) const
+{
+    if (auto maybe_arr = maybeArray(key); maybe_arr)
+        return maybe_arr.value();
+    throw std::logic_error("No array found at '" + key + "'");
+}
+
+Config::array_type
+Config::arrayOr(key_type key, array_type fallback) const
+{
+    if (auto maybe_arr = maybeArray(key); maybe_arr)
+        return maybe_arr.value();
+    return fallback;
+}
+
+Config::array_type
+Config::arrayOrThrow(key_type key, std::string_view err) const
+{
+    try
+    {
+        return maybeArray(key).value();
+    }
+    catch (std::exception const&)
+    {
+        throw std::runtime_error(err.data());
+    }
+}
+
+Config
+Config::section(key_type key) const
+{
+    auto maybe_element = lookup(key);
+    if (maybe_element && maybe_element->is_object())
+        return Config{std::move(*maybe_element)};
+    throw std::logic_error("No section found at '" + key + "'");
+}
+
+Config::array_type
+Config::array() const
+{
+    if (not store_.is_array())
+        throw std::logic_error("_self_ is not an array");
+
+    array_type out;
+    auto const& arr = store_.as_array();
+    out.reserve(arr.size());
+
+    std::transform(
+        std::cbegin(arr), std::cend(arr), std::back_inserter(out), [](auto const& element) { return Config{element}; });
+    return out;
+}
+
+Config
+ConfigReader::open(std::filesystem::path path)
+{
+    try
+    {
+        std::ifstream in(path, std::ios::in | std::ios::binary);
+        if (in)
+        {
+            std::stringstream contents;
+            contents << in.rdbuf();
+            auto opts = boost::json::parse_options{};
+            opts.allow_comments = true;
+            return Config{boost::json::parse(contents.str(), {}, opts)};
+        }
+    }
+    catch (std::exception const& e)
+    {
+        LogService::error() << "Could not read configuration file from '" << path.string() << "': " << e.what();
+    }
+
+    return Config{};
+}
+
+}  // namespace clio

src/config/Config.h

@@ -0,0 +1,401 @@
+//------------------------------------------------------------------------------
+/*
+    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 <config/detail/Helpers.h>
+
+#include <boost/json.hpp>
+#include <filesystem>
+#include <optional>
+#include <string>
+#include <string_view>
+
+namespace clio {
+
+/**
+ * @brief Convenience wrapper to query a JSON configuration file.
+ *
+ * Any custom data type can be supported by implementing the right `tag_invoke`
+ * for `boost::json::value_to`.
+ */
+class Config final
+{
+    boost::json::value store_;
+    static constexpr char Separator = '.';
+
+public:
+    using key_type = std::string;           /*! The type of key used */
+    using array_type = std::vector<Config>; /*! The type of array used */
+    using write_cursor_type = std::pair<std::optional<std::reference_wrapper<boost::json::value>>, key_type>;
+
+    /**
+     * @brief Construct a new Config object.
+     * @param store boost::json::value that backs this instance
+     */
+    explicit Config(boost::json::value store = {});
+
+    //
+    // Querying the store
+    //
+
+    /**
+     * @brief Checks whether underlying store is not null.
+     *
+     * @return true If the store is null
+     * @return false If the store is not null
+     */
+    operator bool() const noexcept;
+
+    /**
+     * @brief Checks whether something exists under given key.
+     *
+     * @param key The key to check
+     * @return true If something exists under key
+     * @return false If nothing exists under key
+     * @throws std::logic_error If the key is of invalid format
+     */
+    [[nodiscard]] bool
+    contains(key_type key) const;
+
+    //
+    // Key value access
+    //
+
+    /**
+     * @brief Interface for fetching values by key that returns std::optional.
+     *
+     * Will attempt to fetch the value under the desired key. If the value
+     * exists and can be represented by the desired type Result then it will be
+     * returned wrapped in an optional. If the value exists but the conversion
+     * to Result is not possible - a runtime_error will be thrown. If the value
+     * does not exist under the specified key - std::nullopt is returned.
+     *
+     * @tparam Result The desired return type
+     * @param key The key to check
+     * @return std::optional<Result> Optional value of desired type
+     * @throws std::logic_error Thrown if conversion to Result is not possible
+     * or key is of invalid format
+     */
+    template <typename Result>
+    [[nodiscard]] std::optional<Result>
+    maybeValue(key_type key) const
+    {
+        auto maybe_element = lookup(key);
+        if (maybe_element)
+            return std::make_optional<Result>(checkedAs<Result>(key, *maybe_element));
+        return std::nullopt;
+    }
+
+    /**
+     * @brief Interface for fetching values by key.
+     *
+     * Will attempt to fetch the value under the desired key. If the value
+     * exists and can be represented by the desired type Result then it will be
+     * returned. If the value exists but the conversion
+     * to Result is not possible OR the value does not exist - a logic_error
+     * will be thrown.
+     *
+     * @tparam Result The desired return type
+     * @param key The key to check
+     * @return Result Value of desired type
+     * @throws std::logic_error Thrown if conversion to Result is not
+     * possible, value does not exist under specified key path or the key is of
+     * invalid format
+     */
+    template <typename Result>
+    [[nodiscard]] Result
+    value(key_type key) const
+    {
+        return maybeValue<Result>(key).value();
+    }
+
+    /**
+     * @brief Interface for fetching values by key with fallback.
+     *
+     * Will attempt to fetch the value under the desired key. If the value
+     * exists and can be represented by the desired type Result then it will be
+     * returned. If the value exists but the conversion
+     * to Result is not possible - a logic_error will be thrown. If the value
+     * does not exist under the specified key - user specified fallback is
+     * returned.
+     *
+     * @tparam Result The desired return type
+     * @param key The key to check
+     * @param fallback The fallback value
+     * @return Result Value of desired type
+     * @throws std::logic_error Thrown if conversion to Result is not possible
+     * or the key is of invalid format
+     */
+    template <typename Result>
+    [[nodiscard]] Result
+    valueOr(key_type key, Result fallback) const
+    {
+        try
+        {
+            return maybeValue<Result>(key).value_or(fallback);
+        }
+        catch (detail::StoreException const&)
+        {
+            return fallback;
+        }
+    }
+
+    /**
+     * @brief Interface for fetching values by key with custom error handling.
+     *
+     * Will attempt to fetch the value under the desired key. If the value
+     * exists and can be represented by the desired type Result then it will be
+     * returned. If the value exists but the conversion
+     * to Result is not possible OR the value does not exist - a runtime_error
+     * will be thrown with the user specified message.
+     *
+     * @tparam Result The desired return type
+     * @param key The key to check
+     * @param err The custom error message
+     * @return Result Value of desired type
+     * @throws std::runtime_error Thrown if conversion to Result is not possible
+     * or value does not exist under key
+     */
+    template <typename Result>
+    [[nodiscard]] Result
+    valueOrThrow(key_type key, std::string_view err) const
+    {
+        try
+        {
+            return maybeValue<Result>(key).value();
+        }
+        catch (std::exception const&)
+        {
+            throw std::runtime_error(err.data());
+        }
+    }
+
+    /**
+     * @brief Interface for fetching an array by key that returns std::optional.
+     *
+     * Will attempt to fetch an array under the desired key. If the array
+     * exists then it will be
+     * returned wrapped in an optional. If the array does not exist under the
+     * specified key - std::nullopt is returned.
+     *
+     * @param key The key to check
+     * @return std::optional<array_type> Optional array
+     * @throws std::logic_error Thrown if the key is of invalid format
+     */
+    [[nodiscard]] std::optional<array_type>
+    maybeArray(key_type key) const;
+
+    /**
+     * @brief Interface for fetching an array by key.
+     *
+     * Will attempt to fetch an array under the desired key. If the array
+     * exists then it will be
+     * returned. If the array does not exist under the
+     * specified key an std::logic_error is thrown.
+     *
+     * @param key The key to check
+     * @return array_type The array
+     * @throws std::logic_error Thrown if there is no array under the desired
+     * key or the key is of invalid format
+     */
+    [[nodiscard]] array_type
+    array(key_type key) const;
+
+    /**
+     * @brief Interface for fetching an array by key with fallback.
+     *
+     * Will attempt to fetch an array under the desired key. If the array
+     * exists then it will be returned.
+     * If the array does not exist or another type is stored under the desired
+     * key - user specified fallback is returned.
+     *
+     * @param key The key to check
+     * @param fallback The fallback array
+     * @return array_type The array
+     * @throws std::logic_error Thrown if the key is of invalid format
+     */
+    [[nodiscard]] array_type
+    arrayOr(key_type key, array_type fallback) const;
+
+    /**
+     * @brief Interface for fetching an array by key with custom error handling.
+     *
+     * Will attempt to fetch an array under the desired key. If the array
+     * exists then it will be returned.
+     * If the array does not exist or another type is stored under the desired
+     * key - std::runtime_error is thrown with the user specified error message.
+     *
+     * @param key The key to check
+     * @param err The custom error message
+     * @return array_type The array
+     * @throws std::runtime_error Thrown if there is no array under the desired
+     * key
+     */
+    [[nodiscard]] array_type
+    arrayOrThrow(key_type key, std::string_view err) const;
+
+    /**
+     * @brief Interface for fetching a sub section by key.
+     *
+     * Will attempt to fetch an entire section under the desired key and return
+     * it as a Config instance. If the section does not exist or another type is
+     * stored under the desired key - std::logic_error is thrown.
+     *
+     * @param key The key to check
+     * @return Config Section represented as a separate instance of Config
+     * @throws std::logic_error Thrown if there is no section under the
+     * desired key or the key is of invalid format
+     */
+    [[nodiscard]] Config
+    section(key_type key) const;
+
+    //
+    // Direct self-value access
+    //
+
+    /**
+     * @brief Interface for reading the value directly referred to by the
+     * instance. Wraps as std::optional.
+     *
+     * See @ref maybeValue(key_type) const for how this works.
+     */
+    template <typename Result>
+    [[nodiscard]] std::optional<Result>
+    maybeValue() const
+    {
+        if (store_.is_null())
+            return std::nullopt;
+        return std::make_optional<Result>(checkedAs<Result>("_self_", store_));
+    }
+
+    /**
+     * @brief Interface for reading the value directly referred to by the
+     * instance.
+     *
+     * See @ref value(key_type) const for how this works.
+     */
+    template <typename Result>
+    [[nodiscard]] Result
+    value() const
+    {
+        return maybeValue<Result>().value();
+    }
+
+    /**
+     * @brief Interface for reading the value directly referred to by the
+     * instance with user-specified fallback.
+     *
+     * See @ref valueOr(key_type, Result) const for how this works.
+     */
+    template <typename Result>
+    [[nodiscard]] Result
+    valueOr(Result fallback) const
+    {
+        return maybeValue<Result>().valueOr(fallback);
+    }
+
+    /**
+     * @brief Interface for reading the value directly referred to by the
+     * instance with user-specified error message.
+     *
+     * See @ref valueOrThrow(key_type, std::string_view) const for how this
+     * works.
+     */
+    template <typename Result>
+    [[nodiscard]] Result
+    valueOrThrow(std::string_view err) const
+    {
+        try
+        {
+            return maybeValue<Result>().value();
+        }
+        catch (std::exception const&)
+        {
+            throw std::runtime_error(err.data());
+        }
+    }
+
+    /**
+     * @brief Interface for reading the array directly referred to by the
+     * instance.
+     *
+     * See @ref array(key_type) const for how this works.
+     */
+    [[nodiscard]] array_type
+    array() const;
+
+private:
+    template <typename Return>
+    [[nodiscard]] Return
+    checkedAs(key_type key, boost::json::value const& value) const
+    {
+        using boost::json::value_to;
+
+        auto has_error = false;
+        if constexpr (std::is_same_v<Return, bool>)
+        {
+            if (not value.is_bool())
+                has_error = true;
+        }
+        else if constexpr (std::is_same_v<Return, std::string>)
+        {
+            if (not value.is_string())
+                has_error = true;
+        }
+        else if constexpr (std::is_same_v<Return, double>)
+        {
+            if (not value.is_number())
+                has_error = true;
+        }
+        else if constexpr (std::is_convertible_v<Return, uint64_t> || std::is_convertible_v<Return, int64_t>)
+        {
+            if (not value.is_int64() && not value.is_uint64())
+                has_error = true;
+        }
+
+        if (has_error)
+            throw std::runtime_error(
+                "Type for key '" + key + "' is '" + std::string{to_string(value.kind())} + "' in JSON but requested '" +
+                detail::typeName<Return>() + "'");
+
+        return value_to<Return>(value);
+    }
+
+    std::optional<boost::json::value>
+    lookup(key_type key) const;
+
+    write_cursor_type
+    lookupForWrite(key_type key);
+};
+
+/**
+ * @brief Simple configuration file reader.
+ *
+ * Reads the JSON file under specified path and creates a @ref Config object
+ * from its contents.
+ */
+class ConfigReader final
+{
+public:
+    static Config
+    open(std::filesystem::path path);
+};
+
+}  // namespace clio

src/config/detail/Helpers.h

@@ -0,0 +1,164 @@
+//------------------------------------------------------------------------------
+/*
+    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 <optional>
+#include <queue>
+#include <stdexcept>
+#include <string>
+
+namespace clio::detail {
+
+/**
+ * @brief Thrown when a KeyPath related error occurs
+ */
+struct KeyException : public ::std::logic_error
+{
+    KeyException(::std::string msg) : ::std::logic_error{msg}
+    {
+    }
+};
+
+/**
+ * @brief Thrown when a Store (config's storage) related error occurs.
+ */
+struct StoreException : public ::std::logic_error
+{
+    StoreException(::std::string msg) : ::std::logic_error{msg}
+    {
+    }
+};
+
+/**
+ * @brief Simple string tokenizer. Used by @ref Config.
+ *
+ * @tparam KeyType The type of key to use
+ * @tparam Separator The separator character
+ */
+template <typename KeyType, char Separator>
+class Tokenizer final
+{
+    using opt_key_t = std::optional<KeyType>;
+    KeyType key_;
+    KeyType token_{};
+    std::queue<KeyType> tokens_{};
+
+public:
+    explicit Tokenizer(KeyType key) : key_{key}
+    {
+        if (key.empty())
+            throw KeyException("Empty key");
+
+        for (auto const& c : key)
+        {
+            if (c == Separator)
+                saveToken();
+            else
+                token_ += c;
+        }
+
+        saveToken();
+    }
+
+    [[nodiscard]] opt_key_t
+    next()
+    {
+        if (tokens_.empty())
+            return std::nullopt;
+        auto token = tokens_.front();
+        tokens_.pop();
+        return std::make_optional(std::move(token));
+    }
+
+private:
+    void
+    saveToken()
+    {
+        if (token_.empty())
+            throw KeyException("Empty token in key '" + key_ + "'.");
+        tokens_.push(std::move(token_));
+        token_ = {};
+    }
+};
+
+template <typename T>
+static constexpr const char*
+typeName()
+{
+    return typeid(T).name();
+}
+
+template <>
+constexpr const char*
+typeName<uint64_t>()
+{
+    return "uint64_t";
+}
+
+template <>
+constexpr const char*
+typeName<int64_t>()
+{
+    return "int64_t";
+}
+
+template <>
+constexpr const char*
+typeName<uint32_t>()
+{
+    return "uint32_t";
+}
+
+template <>
+constexpr const char*
+typeName<int32_t>()
+{
+    return "int32_t";
+}
+
+template <>
+constexpr const char*
+typeName<bool>()
+{
+    return "bool";
+}
+
+template <>
+constexpr const char*
+typeName<std::string>()
+{
+    return "std::string";
+}
+
+template <>
+constexpr const char*
+typeName<const char*>()
+{
+    return "const char*";
+}
+
+template <>
+constexpr const char*
+typeName<double>()
+{
+    return "double";
+}
+
+};  // namespace clio::detail

src/etl/ETLHelpers.h

@@ -1,5 +1,24 @@
-#ifndef RIPPLE_APP_REPORTING_ETLHELPERS_H_INCLUDED
-#define RIPPLE_APP_REPORTING_ETLHELPERS_H_INCLUDED
+//------------------------------------------------------------------------------
+/*
+    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 <condition_variable>
 #include <mutex>
@@ -7,24 +26,22 @@
 #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.
+/**
+ * @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_;
 
-    bool stopping_ = false;
-
 public:
     static std::shared_ptr<NetworkValidatedLedgers>
     make_ValidatedLedgers()
@@ -32,8 +49,11 @@ public:
         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)
     {
@@ -43,10 +63,13 @@ 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()
     {
@@ -55,19 +78,18 @@ 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 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
@@ -76,10 +98,13 @@ 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
 {
@@ -87,75 +112,96 @@ 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. elt is moved from
+     */
     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::unique_lock lck(m_);
+        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)
 {
@@ -173,5 +219,3 @@ getMarkers(size_t numMarkers)
     }
     return markers;
 }
-
-#endif  // RIPPLE_APP_REPORTING_ETLHELPERS_H_INCLUDED

src/etl/ETLService.cpp

@@ -0,0 +1,259 @@
+//------------------------------------------------------------------------------
+/*
+    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>
+
+using namespace clio;
+
+// Database must be populated when this starts
+std::optional<uint32_t>
+ETLService::runETLPipeline(uint32_t startSequence, int numExtractors)
+{
+    if (finishSequence_ && startSequence > *finishSequence_)
+        return {};
+
+    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_, 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_.debug() << "Extracted and wrote " << lastPublishedSeq.value_or(startSequence) - startSequence << " in "
+                 << ((end - begin).count()) / 1000000000.0;
+
+    state_.isWriting = false;
+
+    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_.info() << "Database is empty. Will download a ledger "
+                       "from the network.";
+        std::optional<ripple::LedgerInfo> ledger;
+
+        if (startSequence_)
+        {
+            log_.info() << "ledger sequence specified in config. "
+                        << "Will begin ETL process starting with ledger " << *startSequence_;
+            ledger = ledgerLoader_.loadInitialLedger(*startSequence_);
+        }
+        else
+        {
+            log_.info() << "Waiting for next ledger to be validated by network...";
+            std::optional<uint32_t> mostRecentValidated = networkValidatedLedgers_->getMostRecent();
+
+            if (mostRecentValidated)
+            {
+                log_.info() << "Ledger " << *mostRecentValidated << " has been validated. "
+                            << "Downloading...";
+                ledger = ledgerLoader_.loadInitialLedger(*mostRecentValidated);
+            }
+            else
+            {
+                log_.info() << "The wait for the next validated "
+                            << "ledger has been aborted. "
+                            << "Exiting monitor loop";
+                return;
+            }
+        }
+
+        if (ledger)
+        {
+            rng = backend_->hardFetchLedgerRangeNoThrow();
+        }
+        else
+        {
+            log_.error() << "Failed to load initial ledger. Exiting monitor loop";
+            return;
+        }
+    }
+    else
+    {
+        if (startSequence_)
+            log_.warn() << "start sequence specified but db is already populated";
+
+        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_.debug() << "Database is populated. "
+                 << "Starting monitor loop. sequence = " << nextSequence;
+
+    while (true)
+    {
+        if (auto rng = backend_->hardFetchLedgerRangeNoThrow(); rng && rng->maxSequence >= nextSequence)
+        {
+            ledgerPublisher_.publish(nextSequence, {});
+            ++nextSequence;
+        }
+        else if (networkValidatedLedgers_->waitUntilValidatedByNetwork(nextSequence, 1000))
+        {
+            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_.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_.info() << "Aborting ETL. Falling back to publishing";
+
+                // if no ledger was published, don't increment nextSequence
+                if (lastPublished)
+                    nextSequence = *lastPublished + 1;
+            }
+            else
+            {
+                ++nextSequence;
+            }
+        }
+    }
+}
+
+void
+ETLService::monitorReadOnly()
+{
+    log_.debug() << "Starting reporting in strict read only mode";
+
+    auto rng = backend_->hardFetchLedgerRangeNoThrow();
+    uint32_t latestSequence;
+
+    if (!rng)
+    {
+        if (auto net = networkValidatedLedgers_->getMostRecent())
+            latestSequence = *net;
+        else
+            return;
+    }
+    else
+    {
+        latestSequence = rng->maxSequence;
+    }
+
+    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_.info() << "Starting reporting etl";
+    state_.isStopping = false;
+
+    doWork();
+}
+
+void
+ETLService::doWork()
+{
+    worker_ = std::thread([this]() {
+        beast::setCurrentThreadName("rippled: ETLService worker");
+
+        if (state_.isReadOnly)
+            monitorReadOnly();
+        else
+            monitor();
+    });
+}
+
+ETLService::ETLService(
+    clio::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_)
+{
+    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_);
+}

src/etl/ETLService.h

@@ -0,0 +1,239 @@
+//------------------------------------------------------------------------------
+/*
+    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 <backend/BackendInterface.h>
+#include <backend/LedgerCache.h>
+#include <etl/LoadBalancer.h>
+#include <etl/Source.h>
+#include <etl/SystemState.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 <log/Logger.h>
+#include <subscriptions/SubscriptionManager.h>
+
+#include "org/xrpl/rpc/v1/xrp_ledger.grpc.pb.h"
+#include <grpcpp/grpcpp.h>
+
+#include <memory>
+
+struct AccountTransactionsData;
+struct NFTTransactionsData;
+struct NFTsData;
+class SubscriptionManager;
+
+/**
+ * @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 = SubscriptionManager;
+    using LoadBalancerType = LoadBalancer;
+    using NetworkValidatedLedgersType = NetworkValidatedLedgers;
+    using DataPipeType = clio::detail::ExtractionDataPipe<org::xrpl::rpc::v1::GetLedgerResponse>;
+    using CacheLoaderType = clio::detail::CacheLoader<Backend::LedgerCache>;
+    using LedgerFetcherType = clio::detail::LedgerFetcher<LoadBalancerType>;
+    using ExtractorType = clio::detail::Extractor<DataPipeType, NetworkValidatedLedgersType, LedgerFetcherType>;
+    using LedgerLoaderType = clio::detail::LedgerLoader<LoadBalancerType, LedgerFetcherType>;
+    using LedgerPublisherType = clio::detail::LedgerPublisher<SubscriptionManagerType>;
+    using TransformerType = clio::detail::Transformer<DataPipeType, LedgerLoaderType, LedgerPublisherType>;
+
+    clio::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_;
+
+    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(
+        clio::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);
+
+    static std::shared_ptr<ETLService>
+    make_ETLService(
+        clio::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_.info() << "onStop called";
+        log_.debug() << "Stopping Reporting ETL";
+
+        state_.isStopping = true;
+        cacheLoader_.stop();
+
+        if (worker_.joinable())
+            worker_.join();
+
+        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 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
+     * @return the last ledger written to the database, if any
+     */
+    std::optional<uint32_t>
+    runETLPipeline(uint32_t startSequence, int offset);
+
+    /**
+     * @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 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();
+};

src/etl/ETLSource.h

@@ -1,546 +0,0 @@
-#ifndef RIPPLE_APP_REPORTING_ETLSOURCE_H_INCLUDED
-#define RIPPLE_APP_REPORTING_ETLSOURCE_H_INCLUDED
-
-#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 <backend/BackendInterface.h>
-#include <subscriptions/SubscriptionManager.h>
-
-#include "org/xrpl/rpc/v1/xrp_ledger.grpc.pb.h"
-#include <etl/ETLHelpers.h>
-#include <grpcpp/grpcpp.h>
-
-class ETLLoadBalancer;
-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 ETLSource
-{
-public:
-    virtual bool
-    isConnected() const = 0;
-
-    virtual boost::json::object
-    toJson() const = 0;
-
-    virtual void
-    run() = 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()
-    {
-    }
-};
-
-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_;
-
-    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_;
-
-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};
-
-    void
-    run() override
-    {
-        BOOST_LOG_TRIVIAL(trace) << __func__ << " : " << toString();
-
-        auto const host = ip_;
-        auto const port = wsPort_;
-
-        resolver_.async_resolve(host, port, [this](auto ec, auto results) {
-            onResolve(ec, results);
-        });
-    }
-
-public:
-    ~ETLSourceImpl()
-    {
-        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(
-        boost::json::object const& config,
-        boost::asio::io_context& ioContext,
-        std::shared_ptr<BackendInterface> backend,
-        std::shared_ptr<SubscriptionManager> subscriptions,
-        std::shared_ptr<NetworkValidatedLedgers> networkValidatedLedgers,
-        ETLLoadBalancer& balancer);
-
-    /// @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);
-
-    /// 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(
-        boost::json::object const& config,
-        boost::asio::io_context& ioc,
-        std::shared_ptr<BackendInterface> backend,
-        std::shared_ptr<SubscriptionManager> subscriptions,
-        std::shared_ptr<NetworkValidatedLedgers> nwvl,
-        ETLLoadBalancer& balancer)
-        : ETLSourceImpl(config, ioc, backend, subscriptions, nwvl, balancer)
-        , 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(
-        boost::json::object 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)
-        : ETLSourceImpl(config, ioc, backend, subscriptions, nwvl, balancer)
-        , 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:
-    std::vector<std::unique_ptr<ETLSource>> sources_;
-
-    std::uint32_t downloadRanges_ = 16;
-
-public:
-    ETLLoadBalancer(
-        boost::json::object const& config,
-        boost::asio::io_context& ioContext,
-        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);
-
-    static std::shared_ptr<ETLLoadBalancer>
-    make_ETLLoadBalancer(
-        boost::json::object 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> validatedLedgers)
-    {
-        return std::make_shared<ETLLoadBalancer>(
-            config, ioc, sslCtx, 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);
-};
-
-#endif

src/etl/LoadBalancer.cpp

@@ -0,0 +1,235 @@
+//------------------------------------------------------------------------------
+/*
+    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 <backend/DBHelpers.h>
+#include <etl/ETLService.h>
+#include <etl/NFTHelpers.h>
+#include <etl/ProbingSource.h>
+#include <etl/Source.h>
+#include <log/Logger.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 <boost/json/src.hpp>
+
+#include <thread>
+
+using namespace clio;
+
+std::unique_ptr<Source>
+LoadBalancer::make_Source(
+    clio::Config const& config,
+    boost::asio::io_context& ioContext,
+    std::shared_ptr<BackendInterface> backend,
+    std::shared_ptr<SubscriptionManager> subscriptions,
+    std::shared_ptr<NetworkValidatedLedgers> networkValidatedLedgers,
+    LoadBalancer& balancer)
+{
+    auto src =
+        std::make_unique<ProbingSource>(config, ioContext, backend, subscriptions, networkValidatedLedgers, balancer);
+
+    src->run();
+
+    return src;
+}
+
+std::shared_ptr<LoadBalancer>
+LoadBalancer::make_LoadBalancer(
+    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<LoadBalancer>(config, ioc, backend, subscriptions, validatedLedgers);
+}
+
+LoadBalancer::LoadBalancer(
+    clio::Config const& config,
+    boost::asio::io_context& ioContext,
+    std::shared_ptr<BackendInterface> backend,
+    std::shared_ptr<SubscriptionManager> subscriptions,
+    std::shared_ptr<NetworkValidatedLedgers> nwvl)
+{
+    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, ioContext, backend, subscriptions, nwvl, *this);
+
+        sources_.push_back(std::move(source));
+        log_.info() << "Added etl source - " << sources_.back()->toString();
+    }
+}
+
+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_.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.info() << "Successfully fetched ledger = " << ledgerSequence
+                           << " from source = " << source->toString();
+                return true;
+            }
+            else
+            {
+                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((unsigned)time(0));
+    auto sourceIdx = rand() % sources_.size();
+    auto numAttempts = 0;
+
+    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((unsigned)time(0));
+    auto sourceIdx = rand() % sources_.size();
+    auto numAttempts = 0;
+
+    while (true)
+    {
+        auto& source = sources_[sourceIdx];
+
+        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_.debug() << "Successfully executed func at source = " << source->toString()
+                             << " - ledger sequence = " << ledgerSequence;
+                break;
+            }
+            else
+            {
+                log_.warn() << "Failed to execute func at source = " << source->toString()
+                            << " - ledger sequence = " << ledgerSequence;
+            }
+        }
+        else
+        {
+            log_.warn() << "Ledger not present at source = " << source->toString()
+                        << " - ledger sequence = " << ledgerSequence;
+        }
+        sourceIdx = (sourceIdx + 1) % sources_.size();
+        numAttempts++;
+        if (numAttempts % sources_.size() == 0)
+        {
+            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;
+}

src/etl/LoadBalancer.h

@@ -0,0 +1,161 @@
+//------------------------------------------------------------------------------
+/*
+    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 <boost/asio.hpp>
+#include "org/xrpl/rpc/v1/xrp_ledger.grpc.pb.h"
+#include <grpcpp/grpcpp.h>
+
+class Source;
+class ProbingSource;
+class SubscriptionManager;
+
+/**
+ * @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:
+    clio::Logger log_{"ETL"};
+    std::vector<std::unique_ptr<Source>> sources_;
+    std::uint32_t downloadRanges_ = 16;
+
+public:
+    /**
+     * @brief Create an instance of the load balancer
+     *
+     * @param config The configuration to use
+     * @param ioContext io context to run on
+     * @param backend BackendInterface implementation
+     * @param subscriptions Subscription manager
+     * @param nwvl The network validated ledgers datastructure
+     */
+    LoadBalancer(
+        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<LoadBalancer>
+    make_LoadBalancer(
+        clio::Config const& config,
+        boost::asio::io_context& ioc,
+        std::shared_ptr<BackendInterface> backend,
+        std::shared_ptr<SubscriptionManager> subscriptions,
+        std::shared_ptr<NetworkValidatedLedgers> validatedLedgers);
+
+    static std::unique_ptr<Source>
+    make_Source(
+        clio::Config const& config,
+        boost::asio::io_context& ioContext,
+        std::shared_ptr<BackendInterface> backend,
+        std::shared_ptr<SubscriptionManager> subscriptions,
+        std::shared_ptr<NetworkValidatedLedgers> networkValidatedLedgers,
+        LoadBalancer& balancer);
+
+    ~LoadBalancer()
+    {
+        sources_.clear();
+    }
+
+    /**
+     * @brief Load the initial ledger, writing data to the queue
+     *
+     * @param sequence sequence of ledger to download
+     */
+    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 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
+     */
+    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
+     * @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:
+    /**
+     * @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);
+};

src/etl/NFTHelpers.cpp

@@ -0,0 +1,340 @@
+//------------------------------------------------------------------------------
+/*
+    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 <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>
+
+std::pair<std::vector<NFTTransactionsData>, std::optional<NFTsData>>
+getNFTokenMintData(ripple::TxMeta const& txMeta, ripple::STTx const& sttx)
+{
+    // To find the minted token ID, we put all tokenIDs referenced in the
+    // metadata from prior to the tx application into one vector, then all
+    // tokenIDs referenced in the metadata from after the tx application into
+    // another, then find the one tokenID that was added by this tx
+    // application.
+    std::vector<ripple::uint256> prevIDs;
+    std::vector<ripple::uint256> finalIDs;
+
+    // The owner is not necessarily the issuer, if using authorized minter
+    // flow. Determine owner from the ledger object ID of the NFTokenPages
+    // that were changed.
+    std::optional<ripple::AccountID> owner;
+
+    for (ripple::STObject const& node : txMeta.getNodes())
+    {
+        if (node.getFieldU16(ripple::sfLedgerEntryType) != ripple::ltNFTOKEN_PAGE)
+            continue;
+
+        if (!owner)
+            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);
+            std::transform(
+                toAddNFTs.begin(), toAddNFTs.end(), std::back_inserter(finalIDs), [](ripple::STObject const& nft) {
+                    return nft.getFieldH256(ripple::sfNFTokenID);
+                });
+        }
+        // Else it's modified, as there should never be a deleted NFToken page
+        // as a result of a mint.
+        else
+        {
+            // When a mint results in splitting an existing page,
+            // it results in a created page and a modified node. Sometimes,
+            // the created node needs to be linked to a third page, resulting
+            // in modifying that third page's PreviousPageMin or NextPageMin
+            // field changing, but no NFTs within that page changing. In this
+            // case, there will be no previous NFTs and we need to skip.
+            // However, there will always be NFTs listed in the final fields,
+            // 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>();
+            if (!previousFields.isFieldPresent(ripple::sfNFTokens))
+                continue;
+
+            ripple::STArray const& toAddNFTs = previousFields.getFieldArray(ripple::sfNFTokens);
+            std::transform(
+                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);
+            std::transform(
+                toAddFinalNFTs.begin(),
+                toAddFinalNFTs.end(),
+                std::back_inserter(finalIDs),
+                [](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, sttx.getFieldVL(ripple::sfURI), txMeta)};
+
+    std::stringstream msg;
+    msg << " - unexpected NFTokenMint data in tx " << sttx.getTransactionID();
+    throw std::runtime_error(msg.str());
+}
+
+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())};
+
+    // 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 ||
+            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.
+        std::optional<ripple::STArray> prevNFTs;
+
+        if (node.isFieldPresent(ripple::sfPreviousFields))
+        {
+            ripple::STObject const& previousFields =
+                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);
+
+        if (!prevNFTs)
+            continue;
+
+        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())
+            return std::make_pair(
+                txs,
+                NFTsData(
+                    tokenID,
+                    ripple::AccountID::fromVoid(node.getFieldH256(ripple::sfLedgerIndex).data()),
+                    txMeta,
+                    true));
+    }
+
+    std::stringstream msg;
+    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)
+{
+    // 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);
+            });
+        if (affectedBuyOffer == txMeta.getNodes().end())
+        {
+            std::stringstream msg;
+            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::AccountID const owner = affectedBuyOffer->peekAtField(ripple::sfFinalFields)
+                                            .downcast<ripple::STObject>()
+                                            .getAccountID(ripple::sfOwner);
+        return {
+            {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);
+        });
+    if (affectedSellOffer == txMeta.getNodes().end())
+    {
+        std::stringstream msg;
+        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::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 ||
+            node.getFName() == ripple::sfDeletedNode)
+            continue;
+
+        ripple::AccountID const nodeOwner =
+            ripple::AccountID::fromVoid(node.getFieldH256(ripple::sfLedgerIndex).data());
+        if (nodeOwner == seller)
+            continue;
+
+        ripple::STArray const& nfts = [&node] {
+            if (node.getFName() == ripple::sfCreatedNode)
+                return node.peekAtField(ripple::sfNewFields)
+                    .downcast<ripple::STObject>()
+                    .getFieldArray(ripple::sfNFTokens);
+            return node.peekAtField(ripple::sfFinalFields)
+                .downcast<ripple::STObject>()
+                .getFieldArray(ripple::sfNFTokens);
+        }();
+
+        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())},
+                NFTsData(tokenID, nodeOwner, txMeta, false)};
+    }
+
+    std::stringstream msg;
+    msg << " - unexpected NFTokenAcceptOffer data in tx " << sttx.getTransactionID();
+    throw std::runtime_error(msg.str());
+}
+
+// This is the only transaction where there can be more than 1 element in
+// the returned vector, because you can cancel multiple offers in one
+// 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)
+{
+    std::vector<NFTTransactionsData> txs;
+    for (ripple::STObject const& node : txMeta.getNodes())
+    {
+        if (node.getFieldU16(ripple::sfLedgerEntryType) != ripple::ltNFTOKEN_OFFER)
+            continue;
+
+        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;
+    });
+    txs.erase(last, txs.end());
+    return {txs, {}};
+}
+
+// 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)
+{
+    return {{NFTTransactionsData(sttx.getFieldH256(ripple::sfNFTokenID), txMeta, sttx.getTransactionID())}, {}};
+}
+
+std::pair<std::vector<NFTTransactionsData>, std::optional<NFTsData>>
+getNFTDataFromTx(ripple::TxMeta const& txMeta, ripple::STTx const& sttx)
+{
+    if (txMeta.getResultTER() != ripple::tesSUCCESS)
+        return {{}, {}};
+
+    switch (sttx.getTxnType())
+    {
+        case ripple::TxType::ttNFTOKEN_MINT:
+            return getNFTokenMintData(txMeta, sttx);
+
+        case ripple::TxType::ttNFTOKEN_BURN:
+            return getNFTokenBurnData(txMeta, sttx);
+
+        case ripple::TxType::ttNFTOKEN_ACCEPT_OFFER:
+            return getNFTokenAcceptOfferData(txMeta, sttx);
+
+        case ripple::TxType::ttNFTOKEN_CANCEL_OFFER:
+            return getNFTokenCancelOfferData(txMeta, sttx);
+
+        case ripple::TxType::ttNFTOKEN_CREATE_OFFER:
+            return getNFTokenCreateOfferData(txMeta, sttx);
+
+        default:
+            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;
+}

src/etl/NFTHelpers.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 <backend/DBHelpers.h>
+
+#include <ripple/protocol/STTx.h>
+#include <ripple/protocol/TxMeta.h>
+
+/**
+ * @brief Pull NFT data from TX via ETLService
+ */
+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
+ */
+std::vector<NFTsData>
+getNFTDataFromObj(std::uint32_t const seq, std::string const& key, std::string const& blob);

src/etl/ProbingSource.cpp

@@ -0,0 +1,202 @@
+//------------------------------------------------------------------------------
+/*
+    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 <etl/ProbingSource.h>
+#include <log/Logger.h>
+
+using namespace clio;
+
+ProbingSource::ProbingSource(
+    clio::Config const& config,
+    boost::asio::io_context& ioc,
+    std::shared_ptr<BackendInterface> backend,
+    std::shared_ptr<SubscriptionManager> subscriptions,
+    std::shared_ptr<NetworkValidatedLedgers> nwvl,
+    LoadBalancer& balancer,
+    boost::asio::ssl::context sslCtx)
+    : sslCtx_{std::move(sslCtx)}
+    , 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
+ProbingSource::run()
+{
+    sslSrc_->run();
+    plainSrc_->run();
+}
+
+void
+ProbingSource::pause()
+{
+    sslSrc_->pause();
+    plainSrc_->pause();
+}
+
+void
+ProbingSource::resume()
+{
+    sslSrc_->resume();
+    plainSrc_->resume();
+}
+
+bool
+ProbingSource::isConnected() const
+{
+    return currentSrc_ && currentSrc_->isConnected();
+}
+
+bool
+ProbingSource::hasLedger(uint32_t sequence) const
+{
+    if (!currentSrc_)
+        return false;
+    return currentSrc_->hasLedger(sequence);
+}
+
+boost::json::object
+ProbingSource::toJson() const
+{
+    if (!currentSrc_)
+    {
+        boost::json::object sourcesJson = {
+            {"ws", plainSrc_->toJson()},
+            {"wss", sslSrc_->toJson()},
+        };
+
+        return {
+            {"probing", sourcesJson},
+        };
+    }
+    return currentSrc_->toJson();
+}
+
+std::string
+ProbingSource::toString() const
+{
+    if (!currentSrc_)
+        return "{probing... ws: " + plainSrc_->toString() + ", wss: " + sslSrc_->toString() + "}";
+    return currentSrc_->toString();
+}
+
+boost::uuids::uuid
+ProbingSource::token() const
+{
+    if (!currentSrc_)
+        return boost::uuids::nil_uuid();
+    return currentSrc_->token();
+}
+
+std::pair<std::vector<std::string>, bool>
+ProbingSource::loadInitialLedger(std::uint32_t ledgerSequence, std::uint32_t numMarkers, bool cacheOnly)
+{
+    if (!currentSrc_)
+        return {{}, false};
+    return currentSrc_->loadInitialLedger(ledgerSequence, numMarkers, cacheOnly);
+}
+
+std::pair<grpc::Status, ProbingSource::GetLedgerResponseType>
+ProbingSource::fetchLedger(uint32_t ledgerSequence, bool getObjects, bool getObjectNeighbors)
+{
+    if (!currentSrc_)
+        return {};
+    return currentSrc_->fetchLedger(ledgerSequence, getObjects, getObjectNeighbors);
+}
+
+std::optional<boost::json::object>
+ProbingSource::forwardToRippled(
+    boost::json::object const& request,
+    std::string const& clientIp,
+    boost::asio::yield_context& yield) const
+{
+    if (!currentSrc_)
+        return {};
+    return currentSrc_->forwardToRippled(request, clientIp, yield);
+}
+
+std::optional<boost::json::object>
+ProbingSource::requestFromRippled(
+    boost::json::object const& request,
+    std::string const& clientIp,
+    boost::asio::yield_context& yield) const
+{
+    if (!currentSrc_)
+        return {};
+    return currentSrc_->requestFromRippled(request, clientIp, yield);
+}
+
+SourceHooks
+ProbingSource::make_SSLHooks() noexcept
+{
+    return {// onConnected
+            [this](auto ec) {
+                std::lock_guard lck(mtx_);
+                if (currentSrc_)
+                    return SourceHooks::Action::STOP;
+
+                if (!ec)
+                {
+                    plainSrc_->pause();
+                    currentSrc_ = sslSrc_;
+                    log_.info() << "Selected WSS as the main source: " << currentSrc_->toString();
+                }
+                return SourceHooks::Action::PROCEED;
+            },
+            // onDisconnected
+            [this](auto ec) {
+                std::lock_guard lck(mtx_);
+                if (currentSrc_)
+                {
+                    currentSrc_ = nullptr;
+                    plainSrc_->resume();
+                }
+                return SourceHooks::Action::STOP;
+            }};
+}
+
+SourceHooks
+ProbingSource::make_PlainHooks() noexcept
+{
+    return {// onConnected
+            [this](auto ec) {
+                std::lock_guard lck(mtx_);
+                if (currentSrc_)
+                    return SourceHooks::Action::STOP;
+
+                if (!ec)
+                {
+                    sslSrc_->pause();
+                    currentSrc_ = plainSrc_;
+                    log_.info() << "Selected Plain WS as the main source: " << currentSrc_->toString();
+                }
+                return SourceHooks::Action::PROCEED;
+            },
+            // onDisconnected
+            [this](auto ec) {
+                std::lock_guard lck(mtx_);
+                if (currentSrc_)
+                {
+                    currentSrc_ = nullptr;
+                    sslSrc_->resume();
+                }
+                return SourceHooks::Action::STOP;
+            }};
+}