Fix empty log lines (#825)

Fixes #302

.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

@@ -7,3 +7,4 @@
 # 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/linux_build/build.sh

@@ -0,0 +1,20 @@
+#!/usr/bin/env bash
+
+conan profile new default --detect
+conan profile update settings.compiler.cppstd=20 default
+conan profile update settings.compiler.libcxx=libstdc++11 default
+conan remote add --insert 0 conan-non-prod http://18.143.149.228:8081/artifactory/api/conan/conan-non-prod
+
+cd rippled
+conan export external/snappy snappy/1.1.10@
+conan export external/soci soci/4.0.3@
+conan export .
+conan install --output-folder build_rippled -install-folder build_rippled --build missing --settings build_type=Release
+cmake -B build_rippled -DCMAKE_TOOLCHAIN_FILE:FILEPATH=build/generators/conan_toolchain.cmake -DCMAKE_BUILD_TYPE=Release
+cmake --build build_rippled --target xrpl_core --parallel $(($(nproc) - 2))
+cd ..
+
+conan export external/cassandra
+conan install . -if build_clio -of build_clio --build missing --settings build_type=Release -o tests=True
+cmake -DCMAKE_TOOLCHAIN_FILE:FILEPATH=build/generators/conan_toolchain.cmake -DCMAKE_BUILD_TYPE=Release -B build_clio
+cmake --build build_clio --parallel $(($(nproc) - 2))

.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

@@ -1,9 +1,9 @@
 name: Build Clio
 on:
   push:
-    branches:  [master, develop, develop-next]
+    branches: [master, release/*, develop]
   pull_request:
-    branches:  [master, develop, develop-next]
+    branches: [master, release/*, develop]
   workflow_dispatch:
 
 jobs:
@@ -11,59 +11,105 @@ jobs:
     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-gha/lint@main
+        uses: ./.github/actions/lint
 
-  build_clio:
-    name: Build
-    runs-on: [self-hosted, Linux]
+  build_mac:
+    name: Build macOS
     needs: lint
+    continue-on-error: true
+    runs-on: [self-hosted, macOS]
     steps:
-
-      - name: Clone Clio repo
-        uses: actions/checkout@v3
+      - uses: actions/checkout@v3
         with:
-          path: clio_src
+          path: clio
 
-      - name: Clone Clio CI repo
-        uses: actions/checkout@v3
+      - name: List conan artifactory
+        run: |
+          conan search
+          conan remote list
+          if [[ $(conan remote list |grep conan-non-prod| wc -c) -ne 0 ]]; then
+              echo "conan-non-prod is available"
+          else
+              echo "adding conan-non-prod"
+              conan remote add conan-non-prod http://18.143.149.228:8081/artifactory/api/conan/conan-non-prod
+          fi
+
+      - name: Install dependencies
+        run: |
+          brew install llvm@14 pkg-config ninja bison cmake
+
+      - name: Setup environment for llvm-14
+        run: |
+          export PATH="/usr/local/opt/llvm@14/bin:$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: |
+          cd clio
+          mkdir -p build
+          cd build
+          conan install .. -of . -b missing -s build_type=Release -o clio:tests=True
+          cmake -DCMAKE_TOOLCHAIN_FILE:FILEPATH=build/generators/conan_toolchain.cmake -DCMAKE_BUILD_TYPE=Release ..
+          cmake --build . --parallel $(($(sysctl -n hw.logicalcpu) - 2))
+
+      - name: Upload clio_tests
+        uses: actions/upload-artifact@v3
         with:
-          path: clio_ci
-          repository: 'XRPLF/clio-ci'
+          name: clio_tests_mac
+          path: ./clio/build/clio_tests
 
-      - name: Clone GitHub actions repo
-        uses: actions/checkout@v3
-        with:
-          repository: XRPLF/clio-gha
-          path: gha # must be the same as defined in XRPLF/clio-gha
-
-      - name: Build
-        uses: XRPLF/clio-gha/build@main
-
-      - name: Artifact clio_tests
-        uses: actions/upload-artifact@v2
-        with:
-          name: clio_tests
-          path: clio_tests
-
-      - name: Artifact Debian package
-        uses: actions/upload-artifact@v2
-        with:
-          name: deb_package-${{ github.sha }}
-          path: clio_ci/build/*.deb
-
-  test_clio:
-    name: Test Clio
+  build_linux:
+    name: Build linux
+    needs: lint
+    continue-on-error: true
     runs-on: [self-hosted, Linux]
-    needs: build_clio
+    container:
+      image: conanio/gcc11:1.60.2
+      options: --user root
     steps:
-      - name: Get clio_tests artifact
-        uses: actions/download-artifact@v3
-        with:
-          name: clio_tests
+      - name: Get Clio
+        uses: actions/checkout@v3
 
-      - name: Run tests
-        uses: XRPLF/clio-gha/test@main
+      - name: Get rippled
+        uses: actions/checkout@v3
+        with:
+          repository: thejohnfreeman/rippled
+          ref: clio
+          path: rippled
+
+      - name: Build Clio
+        run: |
+          ./.github/actions/linux_build/build.sh
+
+      - name: Upload clio_tests
+        uses: actions/upload-artifact@v3
+        with:
+          name: clio_tests_linux
+          path: ./build_clio/clio_tests
+
+  test_mac:
+    needs: build_mac
+    runs-on: [self-hosted, macOS]
+    steps:
+    - uses: actions/download-artifact@v3
+      with:
+        name: clio_tests_mac
+    - name: Run clio_tests
+      run: |
+        chmod +x ./clio_tests
+        ./clio_tests --gtest_filter="-BackendCassandraBaseTest*:BackendCassandraTest*:BackendCassandraFactoryTestWithDB*"
+
+  test_linux:
+    needs: build_linux
+    runs-on: [self-hosted, x-heavy]
+    steps:
+    - uses: actions/download-artifact@v3
+      with:
+        name: clio_tests_linux
+    - name: Run clio_tests
+      run: |
+        chmod +x ./clio_tests
+        ./clio_tests --gtest_filter="-BackendCassandraBaseTest*:BackendCassandraTest*:BackendCassandraFactoryTestWithDB*"

.gitignore

@@ -1,3 +1,8 @@
 *clio*.log
-build/
+build*/
+.build
+.vscode
 .python-version
+CMakeUserPresets.json
+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/CheckCompiler.cmake

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

CMake/ClioVersion.cmake

@@ -1,15 +1,33 @@
 #[===================================================================[
-   read version from source
+   write version to source
 #]===================================================================]
 
-file (STRINGS src/main/impl/Build.cpp BUILD_INFO)
-foreach (line_ ${BUILD_INFO})
-  if (line_ MATCHES "versionString[ ]*=[ ]*\"(.+)\"")
-    set (clio_version ${CMAKE_MATCH_1})
-  endif ()
-endforeach ()
-if (clio_version)
-  message (STATUS "clio version: ${clio_version}")
-else ()
-  message (FATAL_ERROR "unable to determine clio version")
+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 +1,125 @@
+# call add_coverage(module_name) to add coverage targets for the given module
+function (add_coverage module)
+  if ("${CMAKE_C_COMPILER_ID}" MATCHES "(Apple)?[Cc]lang"
+     OR "${CMAKE_CXX_COMPILER_ID}" MATCHES "(Apple)?[Cc]lang")
+    message ("[Coverage] Building with llvm Code Coverage Tools")
+    # 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/Docs.cmake

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

CMake/Settings.cmake

@@ -0,0 +1,7 @@
+target_compile_options (clio PUBLIC
+  -Wall
+  -Werror
+  -Wno-narrowing
+  -Wno-deprecated-declarations 
+  -Wno-dangling-else
+  -Wno-unused-but-set-variable)

CMake/SourceLocation.cmake

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

CMake/deps/Boost.cmake

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

CMake/deps/OpenSSL.cmake

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

CMake/deps/Postgres.cmake

@@ -1,31 +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}}
-        )
-ExternalProject_Get_Property (postgres BINARY_DIR)
-
-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()
-add_dependencies(clio postgres)
-target_include_directories(clio PUBLIC ${POSTGRES_INSTALL_DIR}/include)

CMake/deps/Threads.cmake

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

CMake/deps/cassandra.cmake

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

CMake/deps/gtest.cmake

@@ -1,19 +1,4 @@
-FetchContent_Declare(
-  googletest
-  URL https://github.com/google/googletest/archive/609281088cfefc76f9d0ce82e1ff6c30cc3591e5.zip
-)
+find_package (GTest REQUIRED)
 
-FetchContent_GetProperties(googletest)
-
-if(NOT googletest_POPULATED)
-  FetchContent_Populate(googletest)
-  add_subdirectory(${googletest_SOURCE_DIR} ${googletest_BINARY_DIR} EXCLUDE_FROM_ALL)
-endif()
-
-target_link_libraries(clio_tests PUBLIC clio gtest_main)
-
-enable_testing()
-
-include(GoogleTest)
-
-gtest_discover_tests(clio_tests)
+enable_testing ()
+include (GoogleTest)

CMake/deps/libfmt.cmake

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

CMake/deps/libxrpl.cmake

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

CMake/deps/rippled.cmake

@@ -1,18 +0,0 @@
-set(RIPPLED_REPO "https://github.com/ripple/rippled.git")
-set(RIPPLED_BRANCH "1.9.0")
-set(NIH_CACHE_ROOT "${CMAKE_CURRENT_BINARY_DIR}" CACHE INTERNAL "")
-message(STATUS "Cloning ${RIPPLED_REPO} branch ${RIPPLED_BRANCH}")
-FetchContent_Declare(rippled
-  GIT_REPOSITORY "${RIPPLED_REPO}"
-  GIT_TAG "${RIPPLED_BRANCH}"
-  GIT_SHALLOW ON
-)
-
-FetchContent_GetProperties(rippled)
-if(NOT rippled_POPULATED)
-  FetchContent_Populate(rippled)
-  add_subdirectory(${rippled_SOURCE_DIR} ${rippled_BINARY_DIR} EXCLUDE_FROM_ALL)
-endif()
-
-target_link_libraries(clio PUBLIC xrpl_core grpc_pbufs)
-target_include_directories(clio PUBLIC ${rippled_SOURCE_DIR}/src ) # TODO: Seems like this shouldn't be needed?

CMake/install/clio.service.in

@@ -11,6 +11,7 @@ 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
+WantedBy=multi-user.target

CMake/install/install.cmake

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

CMake/settings.cmake

@@ -1 +0,0 @@
-set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -Wno-narrowing -Wall -Werror -Wno-dangling-else")

CMakeLists.txt

@@ -1,105 +1,263 @@
 cmake_minimum_required(VERSION 3.16.3)
-
 project(clio)
 
-option(BUILD_TESTS "Build tests" TRUE)
+# ========================================================================== #
+#                                   Options                                  #
+# ========================================================================== #
+option (verbose   "Verbose build"                                     FALSE) 
+option (tests     "Build tests"                                       FALSE)
+option (docs      "Generate doxygen docs"                             FALSE)
+option (coverage  "Build test coverage report"                        FALSE)
+option (packaging "Create distribution packages"                      FALSE)
+# ========================================================================== #
+set (san "" CACHE STRING "Add sanitizer instrumentation")
+set_property (CACHE san PROPERTY STRINGS ";undefined;memory;address;thread")
+# ========================================================================== #
 
-option(VERBOSE "Verbose build" TRUE)
-if(VERBOSE)
-  set(CMAKE_VERBOSE_MAKEFILE TRUE)
-  set(FETCHCONTENT_QUIET FALSE CACHE STRING "Verbose FetchContent()")
-endif()
+# Include required modules
+include (CheckCXXCompilerFlag)
 
-if(NOT GIT_COMMIT_HASH)
-  if(VERBOSE)
-    message(WARNING "GIT_COMMIT_HASH not provided...looking for git")
-  endif()
-  find_package(Git)
-  if(Git_FOUND)
-      execute_process(COMMAND ${GIT_EXECUTABLE} describe --always --abbrev=8
-          OUTPUT_STRIP_TRAILING_WHITESPACE OUTPUT_VARIABLE gch)
-      if(gch)
-          set(GIT_COMMIT_HASH "${gch}")
-          message(STATUS "Git commit: ${GIT_COMMIT_HASH}")
-          add_definitions(-DCLIO_GIT_COMMIT_HASH="${GIT_COMMIT_HASH}")
-      endif()
-  endif()
-endif() #git
+if (verbose)
+  set (CMAKE_VERBOSE_MAKEFILE TRUE)
+endif ()
 
-add_library(clio)
-target_compile_features(clio PUBLIC cxx_std_20)
-target_include_directories(clio PUBLIC src)
+if (packaging)
+  add_definitions (-DPKG=1)
+endif ()
 
-include(FetchContent)
-include(ExternalProject)
-include(CMake/settings.cmake)
-include(CMake/ClioVersion.cmake)
-include(CMake/deps/rippled.cmake)
-include(CMake/deps/Boost.cmake)
-include(CMake/deps/cassandra.cmake)
-include(CMake/deps/Postgres.cmake)
+add_library (clio)
 
-target_sources(clio PRIVATE
+# Clio tweaks and checks
+include (CMake/CheckCompiler.cmake)
+include (CMake/Settings.cmake)
+include (CMake/ClioVersion.cmake)
+include (CMake/SourceLocation.cmake)
+
+# Clio deps
+include (CMake/deps/libxrpl.cmake)
+include (CMake/deps/Boost.cmake)
+include (CMake/deps/OpenSSL.cmake)
+include (CMake/deps/Threads.cmake)
+include (CMake/deps/libfmt.cmake)
+include (CMake/deps/cassandra.cmake)
+
+# TODO: Include directory will be wrong when installed.
+target_include_directories (clio PUBLIC src)
+target_compile_features (clio PUBLIC cxx_std_20)
+
+target_link_libraries (clio
+  PUBLIC Boost::boost
+  PUBLIC Boost::coroutine
+  PUBLIC Boost::program_options
+  PUBLIC Boost::system
+  PUBLIC Boost::log
+  PUBLIC Boost::log_setup
+  PUBLIC cassandra-cpp-driver::cassandra-cpp-driver
+  PUBLIC fmt::fmt
+  PUBLIC OpenSSL::Crypto
+  PUBLIC OpenSSL::SSL
+  PUBLIC xrpl::libxrpl
+
+  INTERFACE Threads::Threads
+)
+
+if (is_gcc)
+  # FIXME: needed on gcc for now
+  target_compile_definitions (clio PUBLIC BOOST_ASIO_DISABLE_CONCEPTS)
+endif ()
+
+target_sources (clio PRIVATE
   ## Main
   src/main/impl/Build.cpp
   ## Backend
-  src/backend/BackendInterface.cpp
-  src/backend/CassandraBackend.cpp
-  src/backend/Pg.cpp
-  src/backend/PostgresBackend.cpp
-  src/backend/SimpleCache.cpp
+  src/data/BackendInterface.cpp
+  src/data/LedgerCache.cpp
+  src/data/cassandra/impl/Future.cpp
+  src/data/cassandra/impl/Cluster.cpp
+  src/data/cassandra/impl/Batch.cpp
+  src/data/cassandra/impl/Result.cpp
+  src/data/cassandra/impl/Tuple.cpp
+  src/data/cassandra/impl/SslContext.cpp
+  src/data/cassandra/Handle.cpp
+  src/data/cassandra/SettingsProvider.cpp
   ## ETL
-  src/etl/ETLSource.cpp
-  src/etl/ReportingETL.cpp
-  ## Subscriptions
-  src/subscriptions/SubscriptionManager.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
+  ## Feed
+  src/feed/SubscriptionManager.cpp
+  ## Web
+  src/web/IntervalSweepHandler.cpp
   ## RPC
-  src/rpc/RPC.cpp
+  src/rpc/Errors.cpp
+  src/rpc/Factories.cpp
   src/rpc/RPCHelpers.cpp
   src/rpc/Counters.cpp
   src/rpc/WorkQueue.cpp
-  ## RPC Methods
-  # Account
+  src/rpc/common/Specs.cpp
+  src/rpc/common/Validators.cpp
+  src/rpc/common/MetaProcessors.cpp
+  src/rpc/common/impl/APIVersionParser.cpp
+  src/rpc/common/impl/HandlerProvider.cpp
+  ## RPC handlers
   src/rpc/handlers/AccountChannels.cpp
   src/rpc/handlers/AccountCurrencies.cpp
   src/rpc/handlers/AccountInfo.cpp
   src/rpc/handlers/AccountLines.cpp
-  src/rpc/handlers/AccountOffers.cpp
+  src/rpc/handlers/AccountNFTs.cpp
   src/rpc/handlers/AccountObjects.cpp
+  src/rpc/handlers/AccountOffers.cpp
+  src/rpc/handlers/AccountTx.cpp
+  src/rpc/handlers/BookChanges.cpp
+  src/rpc/handlers/BookOffers.cpp
+  src/rpc/handlers/DepositAuthorized.cpp
   src/rpc/handlers/GatewayBalances.cpp
-  src/rpc/handlers/NoRippleCheck.cpp
-  # 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
-  # NFT
-  src/rpc/handlers/NFTOffers.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/util/config/Config.cpp
+  src/util/log/Logger.cpp
+  src/util/Taggable.cpp)
 
-add_executable(clio_server src/main/main.cpp)
-target_link_libraries(clio_server PUBLIC clio)
+# Clio server
+add_executable (clio_server src/main/Main.cpp)
+target_link_libraries (clio_server PUBLIC clio)
 
-if(BUILD_TESTS)
-  add_executable(clio_tests unittests/main.cpp)
-  include(CMake/deps/gtest.cmake)
-endif()
+# Unittesting
+if (tests)
+  set (TEST_TARGET clio_tests)
+  add_executable (${TEST_TARGET}
+    # Common
+    unittests/Main.cpp
+    unittests/Playground.cpp
+    unittests/LoggerTests.cpp
+    unittests/ConfigTests.cpp
+    unittests/ProfilerTests.cpp
+    unittests/JsonUtilTests.cpp
+    unittests/DOSGuardTests.cpp
+    unittests/SubscriptionTests.cpp
+    unittests/SubscriptionManagerTests.cpp
+    unittests/util/TestObject.cpp
+    unittests/util/StringUtils.cpp
+    # ETL
+    unittests/etl/ExtractionDataPipeTests.cpp
+    unittests/etl/ExtractorTests.cpp
+    unittests/etl/TransformerTests.cpp
+    unittests/etl/CacheLoaderTests.cpp
+    # RPC
+    unittests/rpc/ErrorTests.cpp
+    unittests/rpc/BaseTests.cpp
+    unittests/rpc/RPCHelpersTests.cpp
+    unittests/rpc/CountersTests.cpp
+    unittests/rpc/AdminVerificationTests.cpp
+    unittests/rpc/APIVersionTests.cpp
+    unittests/rpc/ForwardingProxyTests.cpp
+    unittests/rpc/WorkQueueTests.cpp
+    unittests/rpc/AmendmentsTests.cpp
+    ## RPC handlers
+    unittests/rpc/handlers/DefaultProcessorTests.cpp
+    unittests/rpc/handlers/TestHandlerTests.cpp
+    unittests/rpc/handlers/AccountCurrenciesTests.cpp
+    unittests/rpc/handlers/AccountLinesTests.cpp
+    unittests/rpc/handlers/AccountTxTests.cpp
+    unittests/rpc/handlers/AccountOffersTests.cpp
+    unittests/rpc/handlers/AccountInfoTests.cpp
+    unittests/rpc/handlers/AccountChannelsTests.cpp
+    unittests/rpc/handlers/AccountNFTsTests.cpp
+    unittests/rpc/handlers/BookOffersTests.cpp
+    unittests/rpc/handlers/DepositAuthorizedTests.cpp
+    unittests/rpc/handlers/GatewayBalancesTests.cpp
+    unittests/rpc/handlers/TxTests.cpp
+    unittests/rpc/handlers/TransactionEntryTests.cpp
+    unittests/rpc/handlers/LedgerEntryTests.cpp
+    unittests/rpc/handlers/LedgerRangeTests.cpp
+    unittests/rpc/handlers/NoRippleCheckTests.cpp
+    unittests/rpc/handlers/ServerInfoTests.cpp
+    unittests/rpc/handlers/PingTests.cpp
+    unittests/rpc/handlers/RandomTests.cpp
+    unittests/rpc/handlers/NFTInfoTests.cpp
+    unittests/rpc/handlers/NFTBuyOffersTests.cpp
+    unittests/rpc/handlers/NFTSellOffersTests.cpp
+    unittests/rpc/handlers/NFTHistoryTests.cpp
+    unittests/rpc/handlers/SubscribeTests.cpp
+    unittests/rpc/handlers/UnsubscribeTests.cpp
+    unittests/rpc/handlers/LedgerDataTests.cpp
+    unittests/rpc/handlers/AccountObjectsTests.cpp
+    unittests/rpc/handlers/BookChangesTests.cpp
+    unittests/rpc/handlers/LedgerTests.cpp
+    unittests/rpc/handlers/VersionHandlerTests.cpp
+    # Backend
+    unittests/data/BackendFactoryTests.cpp
+    unittests/data/cassandra/BaseTests.cpp
+    unittests/data/cassandra/BackendTests.cpp
+    unittests/data/cassandra/RetryPolicyTests.cpp
+    unittests/data/cassandra/SettingsProviderTests.cpp
+    unittests/data/cassandra/ExecutionStrategyTests.cpp
+    unittests/data/cassandra/AsyncExecutorTests.cpp
+    # Webserver
+    unittests/web/ServerTests.cpp
+    unittests/web/RPCServerHandlerTests.cpp
+    unittests/web/WhitelistHandlerTests.cpp
+    unittests/web/SweepHandlerTests.cpp)
 
-include(CMake/install/install.cmake)
-if(PACKAGING)
-    include(CMake/packaging.cmake)
-endif()
+  include (CMake/deps/gtest.cmake)
+
+  # See https://github.com/google/googletest/issues/3475
+  gtest_discover_tests (clio_tests DISCOVERY_TIMEOUT 10)
+
+  # Fix for dwarf5 bug on ci 
+  target_compile_options (clio PUBLIC -gdwarf-4)
+
+  target_compile_definitions (${TEST_TARGET} PUBLIC UNITTEST_BUILD)
+  target_include_directories (${TEST_TARGET} PRIVATE unittests)
+  target_link_libraries (${TEST_TARGET} PUBLIC clio gtest::gtest)
+
+  # Generate `clio_tests-ccov` if coverage is enabled
+  # Note: use `make clio_tests-ccov` to generate report
+  if (coverage)
+    include (CMake/Coverage.cmake)
+    add_coverage (${TEST_TARGET})
+  endif ()
+endif ()
+
+# Enable selected sanitizer if enabled via `san`
+if (san)
+  target_compile_options (clio
+    PUBLIC
+      # Sanitizers recommend minimum of -O1 for reasonable performance
+      $<$<CONFIG:Debug>:-O1>
+      ${SAN_FLAG}
+      -fno-omit-frame-pointer)
+  target_compile_definitions (clio
+    PUBLIC
+      $<$<STREQUAL:${san},address>:SANITIZER=ASAN>
+      $<$<STREQUAL:${san},thread>:SANITIZER=TSAN>
+      $<$<STREQUAL:${san},memory>:SANITIZER=MSAN>
+      $<$<STREQUAL:${san},undefined>:SANITIZER=UBSAN>)
+  target_link_libraries (clio INTERFACE ${SAN_FLAG} ${SAN_LIB})
+endif ()
+
+# Generate `docs` target for doxygen documentation if enabled
+# Note: use `make docs` to generate the documentation
+if (docs)
+  include (CMake/Docs.cmake)
+endif ()
+
+include (CMake/install/install.cmake)
+if (packaging)
+  include (CMake/packaging.cmake) # This file exists only in build runner
+endif ()

CONTRIBUTING.md

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

README.md

@@ -1,49 +1,102 @@
 # Clio
-Clio is an XRP Ledger API server. Clio is optimized for RPC calls, over WebSocket or JSON-RPC. Validated
-historical ledger and transaction data are stored in a more space-efficient format,
+
+Clio is an XRP Ledger API server. Clio is optimized for RPC calls, over WebSocket or JSON-RPC. 
+Validated historical ledger and transaction data are stored in a more space-efficient format,
 using up to 4 times less space than rippled. Clio can be configured to store data in Apache Cassandra or ScyllaDB,
-allowing for scalable read throughput. Multiple Clio nodes can share
-access to the same dataset, allowing for a highly available cluster of Clio nodes,
-without the need for redundant data storage or computation.
+allowing for scalable read throughput. Multiple Clio nodes can share access to the same dataset, 
+allowing for a highly available cluster of Clio nodes, without the need for redundant data storage or computation.
 
 Clio offers the full rippled API, with the caveat that Clio by default only returns validated data.
 This means that `ledger_index` defaults to `validated` instead of `current` for all requests.
 Other non-validated data is also not returned, such as information about queued transactions.
-For requests that require access to the p2p network, such as `fee` or `submit`, Clio automatically forwards the request to a rippled node and propagates the response back to the client. To access non-validated data for *any* request, simply add `ledger_index: "current"` to the request, and Clio will forward the request to rippled.
+For requests that require access to the p2p network, such as `fee` or `submit`, Clio automatically forwards the request to a rippled node and propagates the response back to the client. 
+To access non-validated data for *any* request, simply add `ledger_index: "current"` to the request, and Clio will forward the request to rippled.
 
 Clio does not connect to the peer-to-peer network. Instead, Clio extracts data from a group of specified rippled nodes. Running Clio requires access to at least one rippled node
 from which data can be extracted. The rippled node does not need to be running on the same machine as Clio.
 
-
 ## Requirements
 1. Access to a Cassandra cluster or ScyllaDB cluster. Can be local or remote.
-
 2. Access to one or more rippled nodes. Can be local or remote.
 
 ## Building
 
-Clio is built with CMake. Clio requires c++20, and boost 1.75.0 or later.
+Clio is built with CMake and uses Conan for managing dependencies. 
+It is written in C++20 and therefore requires a modern compiler.
 
-Use these instructions to build a Clio executable from the source. These instructions were tested on Ubuntu 20.04 LTS.
+## Prerequisites 
 
-```sh
-# Install dependencies
-  sudo apt-get -y install git pkg-config protobuf-compiler libprotobuf-dev libssl-dev wget build-essential bison flex autoconf cmake
+### Minimum Requirements
 
-# Compile Boost
-  wget -O $HOME/boost_1_75_0.tar.gz https://boostorg.jfrog.io/artifactory/main/release/1.75.0/source/boost_1_75_0.tar.gz
-  tar xvzf $HOME/boost_1_75_0.tar.gz
-  cd $HOME/boost_1_75_0
-  ./bootstrap.sh
-  ./b2 -j$(nproc)
-  echo "export BOOST_ROOT=$HOME/boost_1_75_0" >> $HOME/.profile && source $HOME/.profile
+- [Python 3.7](https://www.python.org/downloads/)
+- [Conan 1.55](https://conan.io/downloads.html)
+- [CMake 3.16](https://cmake.org/download/)
+- [**Optional**] [GCovr](https://gcc.gnu.org/onlinedocs/gcc/Gcov.html) (needed for code coverage generation)
 
-# Clone the Clio Git repository & build Clio
-  cd $HOME
-  git clone https://github.com/XRPLF/clio.git
-  cd $HOME/clio
-  cmake -B build && cmake --build build --parallel $(nproc)
+| Compiler    | Version |
+|-------------|---------|
+| GCC         | 11      |
+| Clang       | 14      |
+| Apple Clang | 14.0.3  |
+
+### Conan configuration
+
+Clio does not require anything but default settings in your (`~/.conan/profiles/default`) Conan profile. It's best to have no extra flags specified. 
+> Mac example:
 ```
+[settings]
+os=Macos
+os_build=Macos
+arch=armv8
+arch_build=armv8
+compiler=apple-clang
+compiler.version=14
+compiler.libcxx=libc++
+build_type=Release
+compiler.cppstd=20
+```
+> Linux example:
+```
+[settings]
+os=Linux
+os_build=Linux
+arch=x86_64
+arch_build=x86_64
+compiler=gcc
+compiler.version=11
+compiler.libcxx=libstdc++11
+build_type=Release
+compiler.cppstd=20
+```
+
+### Artifactory
+
+1. Make sure artifactory is setup with Conan
+```sh
+conan remote add --insert 0 conan-non-prod http://18.143.149.228:8081/artifactory/api/conan/conan-non-prod
+```
+Now you should be able to download prebuilt `xrpl` package on some platforms.
+
+2. Remove old packages you may have cached: 
+```sh 
+conan remove -f xrpl
+conan remove -f cassandra-cpp-driver
+```
+
+## Building Clio
+
+Navigate to Clio's root directory and perform
+```sh
+mkdir build && cd build
+conan install .. --output-folder . --build missing --settings build_type=Release -o tests=True
+cmake -DCMAKE_TOOLCHAIN_FILE:FILEPATH=build/generators/conan_toolchain.cmake -DCMAKE_BUILD_TYPE=Release ..
+cmake --build . --parallel 8 # or without the number if you feel extra adventurous
+```
+If all goes well, `conan install` will find required packages and `cmake` will do the rest. you should end up with `clio_server` and `clio_tests` in the `build` directory (the current directory).
+
+> **Tip:** You can omit the `-o tests=True` in `conan install` command above if you don't want to build `clio_tests`.
+
+> **Tip:** To generate a Code Coverage report, include `-o coverage=True` in the `conan install` command above, along with `-o tests=True` to enable tests. After running the `cmake` commands, execute `make clio_tests-ccov`. The coverage report will be found at `clio_tests-llvm-cov/index.html`.
 
 ## Running
 ```sh
@@ -72,6 +125,36 @@ server is running
 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.
 
@@ -140,25 +223,81 @@ which can cause high latencies. A possible alternative to this is to just deploy
 a database in each region, and the Clio nodes in each region use their region's database.
 This is effectively two systems.
 
+Clio supports API versioning as [described here](https://xrpl.org/request-formatting.html#api-versioning).
+It's possible to configure `minimum`, `maximum` and `default` version like so: 
+```json
+"api_version": {
+    "min": 1,
+    "max": 2,
+    "default": 2 
+}
+```
+All of the above are optional. 
+Clio will fallback to hardcoded defaults when not specified in the config file or configured values are outside 
+of the minimum and maximum supported versions hardcoded in `src/rpc/common/APIVersion.h`.
+> **Note:** See `example-config.json` for more details. 
+
+## Developing against `rippled` in standalone mode
+
+If you wish you develop against a `rippled` instance running in standalone
+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. 
-Severity options are `trace`, `debug`, `info`, `warning`, `error`, `fatal`. 
+`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_to_console`: Enable/disable log output to console. Options are `true`/`false`.
+`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_to_file`: Enable/disable log saving to files in persistent local storage.  Options are `true`/`false`.
+`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.
 
-`log_directory`: Path to the directory where log files are stored. If such directory doesn't exist, Clio will create it.
+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`.
 
-`log_rotation_size`: The max size of the log file in **megabytes** before it will rotate into a smaller file.
+> **Note:** See `example-config.json` for more details. 
 
-`log_directory_max_size`: The max size of the log directory in **megabytes** before old log files will be 
-deleted to free up space.
+`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.
+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 +1,43 @@
+/*
+ * This is an example configuration file. Please do not use without modifying to suit your needs.
+ */
 {
-    "database":
-    {
-        "type":"cassandra",
-        "cassandra":
-        {
-            "secure_connect_bundle":"[path/to/zip. ignore if using contact_points]",
-            "contact_points":"[ip. ignore if using secure_connect_bundle]",
-            "port":"[port. ignore if using_secure_connect_bundle]",
-            "keyspace":"clio",
-            "username":"[username, if any]",
-            "password":"[password, if any]",
-            "max_requests_outstanding":25000,
-            "threads":8
+    "database": {
+        "type": "cassandra",
+        "cassandra": {
+            // This option can be used to setup a secure connect bundle connection
+            "secure_connect_bundle": "[path/to/zip. ignore if using contact_points]",
+            // The following options are used only if using contact_points
+            "contact_points": "[ip. ignore if using secure_connect_bundle]",
+            "port": "[port. ignore if using_secure_connect_bundle]",
+            // Authentication settings
+            "username": "[username, if any]",
+            "password": "[password, if any]",
+            // Other common settings
+            "keyspace": "clio",
+            "max_write_requests_outstanding": 25000,
+            "max_read_requests_outstanding": 30000,
+            "threads": 8
         }
     },
-    "etl_sources":
-    [
+    "etl_sources": [
         {
-            "ip":"[rippled ip]",
-            "ws_port":"6006",
-            "grpc_port":"50051"
+            "ip": "[rippled ip]",
+            "ws_port": "6006",
+            "grpc_port": "50051"
         }
     ],
-    "dos_guard":
-    {
-        "whitelist":["127.0.0.1"]
+    "dos_guard": {
+        "whitelist": [
+            "127.0.0.1"
+        ]
     },
-    "server":{
-        "ip":"0.0.0.0",
-        "port":8080
+    "server": {
+        "ip": "0.0.0.0",
+        "port": 8080
     },
-    "log_level":"debug",
-    "log_file":"./clio.log",
-    "online_delete":0,
-    "extractor_threads":8,
-    "read_only":false
+    "log_level": "debug",
+    "log_file": "./clio.log",
+    "extractor_threads": 8,
+    "read_only": false
 }

conanfile.py

@@ -0,0 +1,87 @@
+from conan import ConanFile
+from conan.tools.cmake import CMake, CMakeToolchain, cmake_layout
+import re
+
+class Clio(ConanFile):
+    name = 'clio'
+    license = 'ISC'
+    author = 'Alex Kremer <akremer@ripple.com>, John Freeman <jfreeman@ripple.com>'
+    url = 'https://github.com/xrplf/clio'
+    description = 'Clio RPC server'
+    settings = 'os', 'compiler', 'build_type', 'arch'
+    options = {
+        'fPIC': [True, False],
+        'verbose': [True, False],   
+        'tests': [True, False],     # build unit tests; create `clio_tests` binary
+        'docs': [True, False],      # doxygen API docs; create custom target 'docs'
+        'packaging': [True, False], # create distribution packages
+        'coverage': [True, False],  # build for test coverage report; create custom target `clio_tests-ccov`
+    }
+
+    requires = [
+        'boost/1.82.0',
+        'cassandra-cpp-driver/2.16.2',
+        'fmt/10.0.0',
+        'grpc/1.50.1',
+        'openssl/1.1.1u',
+        'xrpl/1.12.0-b2',
+    ]
+
+    default_options = {
+        'fPIC': True,
+        'verbose': False,
+        'tests': False,
+        'packaging': False,
+        'coverage': False,
+        'docs': False,
+        
+        'xrpl/*:tests': False,
+        'cassandra-cpp-driver/*:shared': False,
+        'date/*:header_only': True,
+        'grpc/*:shared': False,
+        'grpc/*:secure': True,
+        'libpq/*:shared': False,
+        'lz4/*:shared': False,
+        'openssl/*:shared': False,
+        'protobuf/*:shared': False,
+        'protobuf/*:with_zlib': True,
+        'snappy/*:shared': False,
+        'gtest/*:no_main': True,
+    }
+
+    exports_sources = (
+        'CMakeLists.txt', 'CMake/*', 'src/*'
+    )
+
+    def requirements(self):
+        if self.options.tests:
+            self.requires('gtest/1.13.0')
+
+    def configure(self):
+        if self.settings.compiler == 'apple-clang':
+            self.options['boost'].visibility = 'global'
+
+    def layout(self):
+        cmake_layout(self)
+        # Fix this setting to follow the default introduced in Conan 1.48 
+        # to align with our build instructions.
+        self.folders.generators = 'build/generators'
+
+    generators = 'CMakeDeps'
+    def generate(self):
+        tc = CMakeToolchain(self)
+        tc.variables['verbose'] = self.options.verbose
+        tc.variables['tests'] = self.options.tests
+        tc.variables['coverage'] = self.options.coverage
+        tc.variables['docs'] = self.options.docs
+        tc.variables['packaging'] = self.options.packaging
+        tc.generate()
+
+    def build(self):
+        cmake = CMake(self)
+        cmake.configure()
+        cmake.build()
+
+    def package(self):
+        cmake = CMake(self)
+        cmake.install()

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

@@ -1,42 +1,127 @@
+/*
+ * This is an example configuration file. Please do not use without modifying to suit your needs.
+ */
 {
-    "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,
+            //
+            // Advanced options. USE AT OWN RISK:
+            // ---
+            "max_connections_per_host": 1, // Defaults to 2
+            "core_connections_per_host": 1, // Defaults to 2
+            "max_concurrent_requests_threshold": 55000 // Defaults to ((max_read + max_write) / core_connections_per_host)
+            //
+            // Below options will use defaults from cassandra driver if left unspecified.
+            // See https://docs.datastax.com/en/developer/cpp-driver/2.0/api/struct.CassCluster/ for details.
+            // 
+            // "queue_size_event": 1,
+            // "queue_size_io": 2,
+            // "write_bytes_high_water_mark": 3,
+            // "write_bytes_low_water_mark": 4,
+            // "pending_requests_high_water_mark": 5,
+            // "pending_requests_low_water_mark": 6,
+            // "max_requests_per_flush": 7,
+            // "max_concurrent_creation": 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": {
+        // Comma-separated list of IPs to exclude from rate limiting
+        "whitelist": [
+            "127.0.0.1"
+        ],
+        //
+        // The below values are the default values and are only specified here
+        // for documentation purposes. The rate limiter currently limits
+        // connections and bandwidth per IP. The rate limiter looks at the raw
+        // IP of a client connection, and so requests routed through a load
+        // balancer will all have the same IP and be treated as a single client.
+        //
+        "max_fetches": 1000000, // Max bytes per IP per sweep interval
+        "max_connections": 20, // Max connections per IP
+        "max_requests": 20, // Max connections per IP per sweep interval
+        "sweep_interval": 1 // Time in seconds before resetting max_fetches and max_requests
     },
-    "server":{
-        "ip":"0.0.0.0",
-        "port":51233
+    "cache": {
+        // Comma-separated list of peer nodes that Clio can use to download cache from at startup
+        "peers": [
+            {
+                "ip": "127.0.0.1",
+                "port": 51234
+            }
+        ]
     },
-    "log_level":"debug",
+    "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
+    },
+    // Overrides log level on a per logging channel.
+    // Defaults to global "log_level" for each unspecified channel.
+    "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 (this is the default format)
+    "log_format": "%TimeStamp% (%SourceLocation%) [%ThreadID%] %Channel%:%Severity% %Message%",
     "log_to_console": true,
-    "log_to_file": true,
-    "log_directory":"./clio_log",
+    // Clio logs to file in the specified directory only if "log_directory" is set
+    // "log_directory": "./clio_log",
     "log_rotation_size": 2048,
     "log_directory_max_size": 51200,
     "log_rotation_hour_interval": 12,
-    "online_delete":0,
-    "extractor_threads":8,
-    "read_only":false
+    "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"
+    "api_version": {
+        "min": 2,
+        "max": 2,
+        "default": 2 // Clio only supports API v2 and newer
+    }
 }

external/cassandra/conanfile.py

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

src/backend/BackendFactory.h

@@ -1,65 +0,0 @@
-#ifndef RIPPLE_APP_REPORTING_BACKENDFACTORY_H_INCLUDED
-#define RIPPLE_APP_REPORTING_BACKENDFACTORY_H_INCLUDED
-
-#include <boost/algorithm/string.hpp>
-#include <backend/BackendInterface.h>
-#include <backend/CassandraBackend.h>
-#include <backend/PostgresBackend.h>
-
-namespace Backend {
-std::shared_ptr<BackendInterface>
-make_Backend(boost::asio::io_context& ioc, boost::json::object const& config)
-{
-    BOOST_LOG_TRIVIAL(info) << __func__ << ": 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();
-
-    std::shared_ptr<BackendInterface> backend = nullptr;
-
-    if (boost::iequals(type, "cassandra"))
-    {
-        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";
-    }
-
-    if (!backend)
-        throw std::runtime_error("Invalid database type");
-
-    backend->open(readOnly);
-    auto rng = backend->hardFetchLedgerRangeNoThrow();
-    if (rng)
-    {
-        backend->updateRange(rng->minSequence);
-        backend->updateRange(rng->maxSequence);
-    }
-
-    BOOST_LOG_TRIVIAL(info)
-        << __func__ << ": Constructed BackendInterface Successfully";
-
-    return backend;
-}
-}  // namespace Backend
-
-#endif  // RIPPLE_REPORTING_BACKEND_FACTORY

src/backend/BackendInterface.h

@@ -1,338 +0,0 @@
-#ifndef RIPPLE_APP_REPORTING_BACKENDINTERFACE_H_INCLUDED
-#define RIPPLE_APP_REPORTING_BACKENDINTERFACE_H_INCLUDED
-#include <ripple/ledger/ReadView.h>
-#include <boost/asio.hpp>
-#include <backend/DBHelpers.h>
-#include <backend/SimpleCache.h>
-#include <backend/Types.h>
-#include <thread>
-#include <type_traits>
-namespace Backend {
-
-class DatabaseTimeout : public std::exception
-{
-    const char*
-    what() const throw() override
-    {
-        return "Database read timed out. Please retry the request";
-    }
-};
-
-template <class F>
-auto
-retryOnTimeout(F func, size_t waitMs = 500)
-{
-    while (true)
-    {
-        try
-        {
-            return func();
-        }
-        catch (DatabaseTimeout& t)
-        {
-            std::this_thread::sleep_for(std::chrono::milliseconds(waitMs));
-            BOOST_LOG_TRIVIAL(error)
-                << __func__ << " function timed out. Retrying ... ";
-        }
-    }
-}
-
-template <class F>
-auto
-synchronous(F&& f)
-{
-    boost::asio::io_context ctx;
-    boost::asio::io_context::strand strand(ctx);
-    std::optional<boost::asio::io_context::work> work;
-
-    work.emplace(ctx);
-
-    using R = typename std::result_of<F(boost::asio::yield_context&)>::type;
-    if constexpr (!std::is_same<R, void>::value)
-    {
-        R res;
-        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();
-            });
-
-        ctx.run();
-    }
-}
-
-template <class F>
-auto
-synchronousAndRetryOnTimeout(F&& f)
-{
-    return retryOnTimeout([&]() { return synchronous(f); });
-}
-
-class BackendInterface
-{
-protected:
-    mutable std::shared_mutex rngMtx_;
-    std::optional<LedgerRange> range;
-    SimpleCache cache_;
-
-    // mutex used for open() and close()
-    mutable std::mutex mutex_;
-
-public:
-    BackendInterface(boost::json::object const& config)
-    {
-    }
-    virtual ~BackendInterface()
-    {
-    }
-
-    // *** 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&
-    cache() const
-    {
-        return cache_;
-    }
-
-    SimpleCache&
-    cache()
-    {
-        return cache_;
-    }
-
-    virtual std::optional<ripple::LedgerInfo>
-    fetchLedgerBySequence(
-        std::uint32_t const sequence,
-        boost::asio::yield_context& yield) const = 0;
-
-    virtual std::optional<ripple::LedgerInfo>
-    fetchLedgerByHash(
-        ripple::uint256 const& hash,
-        boost::asio::yield_context& yield) const = 0;
-
-    virtual std::optional<std::uint32_t>
-    fetchLatestLedgerSequence(boost::asio::yield_context& yield) const = 0;
-
-    std::optional<LedgerRange>
-    fetchLedgerRange() const
-    {
-        std::shared_lock lck(rngMtx_);
-        return range;
-    }
-
-    void
-    updateRange(uint32_t newMax)
-    {
-        std::unique_lock lck(rngMtx_);
-        assert(!range || newMax >= range->maxSequence);
-        if (!range)
-            range = {newMax, newMax};
-        else
-            range->maxSequence = newMax;
-    }
-
-    std::optional<ripple::Fees>
-    fetchFees(std::uint32_t const seq, boost::asio::yield_context& yield) const;
-
-    // *** transaction methods
-    virtual std::optional<TransactionAndMetadata>
-    fetchTransaction(
-        ripple::uint256 const& hash,
-        boost::asio::yield_context& yield) const = 0;
-
-    virtual std::vector<TransactionAndMetadata>
-    fetchTransactions(
-        std::vector<ripple::uint256> const& hashes,
-        boost::asio::yield_context& yield) const = 0;
-
-    virtual AccountTransactions
-    fetchAccountTransactions(
-        ripple::AccountID const& account,
-        std::uint32_t const limit,
-        bool forward,
-        std::optional<AccountTransactionsCursor> const& cursor,
-        boost::asio::yield_context& yield) const = 0;
-
-    virtual std::vector<TransactionAndMetadata>
-    fetchAllTransactionsInLedger(
-        std::uint32_t const ledgerSequence,
-        boost::asio::yield_context& yield) const = 0;
-
-    virtual std::vector<ripple::uint256>
-    fetchAllTransactionHashesInLedger(
-        std::uint32_t const ledgerSequence,
-        boost::asio::yield_context& yield) const = 0;
-
-    // *** state data methods
-    std::optional<Blob>
-    fetchLedgerObject(
-        ripple::uint256 const& key,
-        std::uint32_t const sequence,
-        boost::asio::yield_context& yield) const;
-
-    std::vector<Blob>
-    fetchLedgerObjects(
-        std::vector<ripple::uint256> const& keys,
-        std::uint32_t const sequence,
-        boost::asio::yield_context& yield) const;
-
-    virtual std::optional<Blob>
-    doFetchLedgerObject(
-        ripple::uint256 const& key,
-        std::uint32_t const sequence,
-        boost::asio::yield_context& yield) const = 0;
-
-    virtual std::vector<Blob>
-    doFetchLedgerObjects(
-        std::vector<ripple::uint256> const& keys,
-        std::uint32_t const sequence,
-        boost::asio::yield_context& yield) const = 0;
-
-    virtual std::vector<LedgerObject>
-    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
-    LedgerPage
-    fetchLedgerPage(
-        std::optional<ripple::uint256> const& cursor,
-        std::uint32_t const ledgerSequence,
-        std::uint32_t const limit,
-        bool outOfOrder,
-        boost::asio::yield_context& yield) const;
-
-    // Fetches the successor to key/index
-    std::optional<LedgerObject>
-    fetchSuccessorObject(
-        ripple::uint256 key,
-        std::uint32_t const ledgerSequence,
-        boost::asio::yield_context& yield) const;
-
-    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
-
-    virtual std::optional<ripple::uint256>
-    doFetchSuccessorKey(
-        ripple::uint256 key,
-        std::uint32_t const ledgerSequence,
-        boost::asio::yield_context& yield) const = 0;
-
-    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;
-
-    std::optional<LedgerRange>
-    hardFetchLedgerRange() const
-    {
-        return synchronous([&](boost::asio::yield_context yield) {
-            return hardFetchLedgerRange(yield);
-        });
-    }
-
-    virtual std::optional<LedgerRange>
-    hardFetchLedgerRange(boost::asio::yield_context& yield) const = 0;
-
-    // Doesn't throw DatabaseTimeout. Should be used with care.
-    std::optional<LedgerRange>
-    hardFetchLedgerRangeNoThrow() const;
-    // Doesn't throw DatabaseTimeout. Should be used with care.
-    std::optional<LedgerRange>
-    hardFetchLedgerRangeNoThrow(boost::asio::yield_context& yield) const;
-
-    virtual void
-    writeLedger(
-        ripple::LedgerInfo const& ledgerInfo,
-        std::string&& ledgerHeader) = 0;
-
-    virtual void
-    writeLedgerObject(
-        std::string&& key,
-        std::uint32_t const seq,
-        std::string&& blob);
-
-    virtual void
-    writeTransaction(
-        std::string&& hash,
-        std::uint32_t const seq,
-        std::uint32_t const date,
-        std::string&& transaction,
-        std::string&& metadata) = 0;
-
-    virtual void
-    writeAccountTransactions(std::vector<AccountTransactionsData>&& data) = 0;
-
-    virtual void
-    writeSuccessor(
-        std::string&& key,
-        std::uint32_t const seq,
-        std::string&& successor) = 0;
-
-    // Tell the database we are about to begin writing data for a particular
-    // 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
-    bool
-    finishWrites(std::uint32_t const ledgerSequence);
-
-    virtual bool
-    doOnlineDelete(
-        std::uint32_t numLedgersToKeep,
-        boost::asio::yield_context& yield) 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:
-    virtual void
-    doWriteLedgerObject(
-        std::string&& key,
-        std::uint32_t const seq,
-        std::string&& blob) = 0;
-
-    virtual bool
-    doFinishWrites() = 0;
-};
-
-}  // namespace Backend
-using BackendInterface = Backend::BackendInterface;
-#endif

src/backend/DBHelpers.h

@@ -1,130 +0,0 @@
-#ifndef CLIO_BACKEND_DBHELPERS_H_INCLUDED
-#define CLIO_BACKEND_DBHELPERS_H_INCLUDED
-
-#include <ripple/basics/Log.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
-struct AccountTransactionsData
-{
-    boost::container::flat_set<ripple::AccountID> accounts;
-    std::uint32_t ledgerSequence;
-    std::uint32_t transactionIndex;
-    ripple::uint256 txHash;
-
-    AccountTransactionsData(
-        ripple::TxMeta& meta,
-        ripple::uint256 const& txHash,
-        beast::Journal& j)
-        : accounts(meta.getAffectedAccounts())
-        , ledgerSequence(meta.getLgrSeq())
-        , transactionIndex(meta.getIndex())
-        , txHash(txHash)
-    {
-    }
-
-    AccountTransactionsData() = default;
-};
-
-template <class T>
-inline bool
-isOffer(T const& object)
-{
-    short offer_bytes = (object[1] << 8) | object[2];
-    return offer_bytes == 0x006f;
-}
-
-template <class T>
-inline bool
-isOfferHex(T const& object)
-{
-    auto blob = ripple::strUnHex(4, object.begin(), object.begin() + 4);
-    if (blob)
-    {
-        short offer_bytes = ((*blob)[1] << 8) | (*blob)[2];
-        return offer_bytes == 0x006f;
-    }
-    return false;
-}
-
-template <class T>
-inline bool
-isDirNode(T const& object)
-{
-    short spaceKey = (object.data()[1] << 8) | object.data()[2];
-    return spaceKey == 0x0064;
-}
-
-template <class T, class R>
-inline bool
-isBookDir(T const& key, R const& object)
-{
-    if (!isDirNode(object))
-        return false;
-
-    ripple::STLedgerEntry const sle{
-        ripple::SerialIter{object.data(), object.size()}, key};
-    return !sle[~ripple::sfOwner].has_value();
-}
-
-template <class T>
-inline ripple::uint256
-getBook(T const& offer)
-{
-    ripple::SerialIter it{offer.data(), offer.size()};
-    ripple::SLE sle{it, {}};
-    ripple::uint256 book = sle.getFieldH256(ripple::sfBookDirectory);
-    return book;
-}
-
-template <class T>
-inline ripple::uint256
-getBookBase(T const& key)
-{
-    assert(key.size() == ripple::uint256::size());
-    ripple::uint256 ret;
-    for (size_t i = 0; i < 24; ++i)
-    {
-        ret.data()[i] = key.data()[i];
-    }
-    return ret;
-}
-
-inline ripple::LedgerInfo
-deserializeHeader(ripple::Slice data)
-{
-    ripple::SerialIter sit(data.data(), data.size());
-
-    ripple::LedgerInfo info;
-
-    info.seq = sit.get32();
-    info.drops = sit.get64();
-    info.parentHash = sit.get256();
-    info.txHash = sit.get256();
-    info.accountHash = sit.get256();
-    info.parentCloseTime =
-        ripple::NetClock::time_point{ripple::NetClock::duration{sit.get32()}};
-    info.closeTime =
-        ripple::NetClock::time_point{ripple::NetClock::duration{sit.get32()}};
-    info.closeTimeResolution = ripple::NetClock::duration{sit.get8()};
-    info.closeFlags = sit.get8();
-
-    info.hash = sit.get256();
-
-    return info;
-}
-
-inline std::string
-uint256ToString(ripple::uint256 const& uint)
-{
-    return {reinterpret_cast<const char*>(uint.data()), uint.size()};
-}
-
-static constexpr std::uint32_t rippleEpochStart = 946684800;
-#endif

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 +0,0 @@
-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 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.
-
-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).
-
-The below examples are a sort of pseudo query language
-
-## Ledgers
-
-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.
-
-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.
-
-## Transactions
-Transactions are stored in a very basic table, with a schema like so:
-
-```
-CREATE TABLE transactions (
-hash blob,
-ledger_sequence int,
-transaction blob,
-PRIMARY KEY(hash))
-```
-The primary key is the hash.
-
-A common query pattern is fetching all transactions in a ledger. In PostgreSQL,
-nothing special is needed for this. We just query:
-```
-SELECT * FROM transactions WHERE ledger_sequence = s;
-```
-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 transaction_hashes (
-ledger_sequence int,
-hash blob,
-PRIMARY KEY(ledger_sequence, blob))
-```
-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 objects (
-id blob,
-ledger_sequence int,
-object blob,
-PRIMARY KEY(key,ledger_sequence))
-```
-
-The `objects` table has a compound primary key. This is essential. Looking up
-a ledger object as of a given ledger then is just:
-```
-SELECT object FROM objects WHERE id = ? and ledger_sequence <= ?
-    ORDER BY ledger_sequence DESC LIMIT 1;
-```
-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:
-```
-SELECT * FROM objects WHERE id > ? AND ledger_sequence <= s ORDER BY id ASC, ledger_sequence DESC LIMIT 1;
-```
-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:
-```
-CREATE TABLE keys (
-ledger_sequence int,
-id blob,
-PRIMARY KEY(ledger_sequence, id)
-)
-```
-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 account_tx (
-account blob,
-ledger_sequence int,
-transaction_index int,
-hash blob,
-PRIMARY KEY(account,ledger_sequence,transaction_index))
-```
-
-An example of looking up from this table going backwards in time is:
-```
-SELECT hash FROM account_tx WHERE account = ? 
-    AND ledger_sequence <= ? and transaction_index <= ? 
-    ORDER BY ledger_sequence DESC, transaction_index DESC;
-```
-
-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

@@ -1,120 +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)
-{
-    if (disabled_)
-        return;
-
-    {
-        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::setDisabled()
-{
-    disabled_ = true;
-}
-
-void
-SimpleCache::setFull()
-{
-    if (disabled_)
-        return;
-
-    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

@@ -1,68 +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;
-    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;
-};
-
-}  // namespace Backend
-#endif

src/backend/Types.h

@@ -1,73 +0,0 @@
-#ifndef CLIO_TYPES_H_INCLUDED
-#define CLIO_TYPES_H_INCLUDED
-#include <ripple/basics/base_uint.h>
-#include <optional>
-#include <string>
-#include <vector>
-
-namespace Backend {
-
-// *** return types
-
-using Blob = std::vector<unsigned char>;
-
-struct LedgerObject
-{
-    ripple::uint256 key;
-    Blob blob;
-    bool
-    operator==(const LedgerObject& other) const
-    {
-        return key == other.key && blob == other.blob;
-    }
-};
-
-struct LedgerPage
-{
-    std::vector<LedgerObject> objects;
-    std::optional<ripple::uint256> cursor;
-};
-struct BookOffersPage
-{
-    std::vector<LedgerObject> offers;
-    std::optional<ripple::uint256> cursor;
-};
-struct TransactionAndMetadata
-{
-    Blob transaction;
-    Blob metadata;
-    std::uint32_t ledgerSequence;
-    std::uint32_t date;
-    bool
-    operator==(const TransactionAndMetadata& other) const
-    {
-        return transaction == other.transaction && metadata == other.metadata &&
-            ledgerSequence == other.ledgerSequence && date == other.date;
-    }
-};
-
-struct AccountTransactionsCursor
-{
-    std::uint32_t ledgerSequence;
-    std::uint32_t transactionIndex;
-};
-
-struct AccountTransactions
-{
-    std::vector<TransactionAndMetadata> txns;
-    std::optional<AccountTransactionsCursor> cursor;
-};
-
-struct LedgerRange
-{
-    std::uint32_t minSequence;
-    std::uint32_t maxSequence;
-};
-constexpr ripple::uint256 firstKey{
-    "0000000000000000000000000000000000000000000000000000000000000000"};
-constexpr ripple::uint256 lastKey{
-    "FFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFF"};
-constexpr ripple::uint256 hi192{
-    "0000000000000000000000000000000000000000000000001111111111111111"};
-}  // namespace Backend
-#endif

src/data/BackendFactory.h

@@ -0,0 +1,68 @@
+//------------------------------------------------------------------------------
+/*
+    This file is part of clio: https://github.com/XRPLF/clio
+    Copyright (c) 2022, the clio developers.
+
+    Permission to use, copy, modify, and distribute this software for any
+    purpose with or without fee is hereby granted, provided that the above
+    copyright notice and this permission notice appear in all copies.
+
+    THE  SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL WARRANTIES
+    WITH  REGARD  TO  THIS  SOFTWARE  INCLUDING  ALL  IMPLIED  WARRANTIES  OF
+    MERCHANTABILITY  AND  FITNESS. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR
+    ANY  SPECIAL,  DIRECT,  INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES
+    WHATSOEVER  RESULTING  FROM  LOSS  OF USE, DATA OR PROFITS, WHETHER IN AN
+    ACTION  OF  CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF
+    OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE.
+*/
+//==============================================================================
+
+#pragma once
+
+#include <data/BackendInterface.h>
+#include <data/CassandraBackend.h>
+#include <util/config/Config.h>
+#include <util/log/Logger.h>
+
+#include <boost/algorithm/string.hpp>
+
+namespace data {
+
+/**
+ * @brief A factory function that creates the backend based on a config.
+ *
+ * @param config The clio config to use
+ * @return A shared_ptr<BackendInterface> with the selected implementation
+ */
+std::shared_ptr<BackendInterface>
+make_Backend(util::Config const& config)
+{
+    static util::Logger log{"Backend"};
+    LOG(log.info()) << "Constructing BackendInterface";
+
+    auto const readOnly = config.valueOr("read_only", false);
+
+    auto const type = config.value<std::string>("database.type");
+    std::shared_ptr<BackendInterface> backend = nullptr;
+
+    // TODO: retire `cassandra-new` by next release after 2.0
+    if (boost::iequals(type, "cassandra") or boost::iequals(type, "cassandra-new"))
+    {
+        auto cfg = config.section("database." + type);
+        backend = std::make_shared<data::cassandra::CassandraBackend>(data::cassandra::SettingsProvider{cfg}, readOnly);
+    }
+
+    if (!backend)
+        throw std::runtime_error("Invalid database type");
+
+    auto const rng = backend->hardFetchLedgerRangeNoThrow();
+    if (rng)
+    {
+        backend->updateRange(rng->minSequence);
+        backend->updateRange(rng->maxSequence);
+    }
+
+    LOG(log.info()) << "Constructed BackendInterface Successfully";
+    return backend;
+}
+}  // namespace data

src/data/BackendInterface.cpp

@@ -1,50 +1,56 @@
+//------------------------------------------------------------------------------
+/*
+    This file is part of clio: https://github.com/XRPLF/clio
+    Copyright (c) 2022, the clio developers.
+
+    Permission to use, copy, modify, and distribute this software for any
+    purpose with or without fee is hereby granted, provided that the above
+    copyright notice and this permission notice appear in all copies.
+
+    THE  SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL WARRANTIES
+    WITH  REGARD  TO  THIS  SOFTWARE  INCLUDING  ALL  IMPLIED  WARRANTIES  OF
+    MERCHANTABILITY  AND  FITNESS. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR
+    ANY  SPECIAL,  DIRECT,  INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES
+    WHATSOEVER  RESULTING  FROM  LOSS  OF USE, DATA OR PROFITS, WHETHER IN AN
+    ACTION  OF  CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF
+    OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE.
+*/
+//==============================================================================
+
+#include <data/BackendInterface.h>
+#include <util/log/Logger.h>
+
 #include <ripple/protocol/Indexes.h>
 #include <ripple/protocol/STLedgerEntry.h>
-#include <backend/BackendInterface.h>
-namespace Backend {
+
+// local to compilation unit loggers
+namespace {
+util::Logger gLog{"Backend"};
+}  // namespace
+
+namespace data {
 bool
 BackendInterface::finishWrites(std::uint32_t const ledgerSequence)
 {
+    LOG(gLog.debug()) << "Want finish writes for " << ledgerSequence;
     auto commitRes = doFinishWrites();
     if (commitRes)
     {
+        LOG(gLog.debug()) << "Successfully commited. Updating range now to " << ledgerSequence;
         updateRange(ledgerSequence);
     }
     return commitRes;
 }
 void
-BackendInterface::writeLedgerObject(
-    std::string&& key,
-    std::uint32_t const seq,
-    std::string&& blob)
+BackendInterface::writeLedgerObject(std::string&& key, std::uint32_t const seq, std::string&& blob)
 {
     assert(key.size() == sizeof(ripple::uint256));
-    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
-{
-    BOOST_LOG_TRIVIAL(debug) << __func__;
-    while (true)
-    {
-        try
-        {
-            return hardFetchLedgerRange(yield);
-        }
-        catch (DatabaseTimeout& t)
-        {
-            ;
-        }
-    }
-}
-
 std::optional<LedgerRange>
 BackendInterface::hardFetchLedgerRangeNoThrow() const
 {
-    BOOST_LOG_TRIVIAL(debug) << __func__;
     return retryOnTimeout([&]() { return hardFetchLedgerRange(); });
 }
 
@@ -53,26 +59,22 @@ std::optional<Blob>
 BackendInterface::fetchLedgerObject(
     ripple::uint256 const& key,
     std::uint32_t const sequence,
-    boost::asio::yield_context& yield) const
+    boost::asio::yield_context yield) const
 {
     auto obj = cache_.get(key, sequence);
     if (obj)
     {
-        BOOST_LOG_TRIVIAL(trace)
-            << __func__ << " - cache hit - " << ripple::strHex(key);
+        LOG(gLog.trace()) << "Cache hit - " << ripple::strHex(key);
         return *obj;
     }
     else
     {
-        BOOST_LOG_TRIVIAL(trace)
-            << __func__ << " - cache miss - " << ripple::strHex(key);
+        LOG(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";
+            LOG(gLog.trace()) << "Missed cache and missed in db";
         else
-            BOOST_LOG_TRIVIAL(trace)
-                << __func__ << " - missed cache but found in db";
+            LOG(gLog.trace()) << "Missed cache but found in db";
         return dbObj;
     }
 }
@@ -81,7 +83,7 @@ std::vector<Blob>
 BackendInterface::fetchLedgerObjects(
     std::vector<ripple::uint256> const& keys,
     std::uint32_t const sequence,
-    boost::asio::yield_context& yield) const
+    boost::asio::yield_context yield) const
 {
     std::vector<Blob> results;
     results.resize(keys.size());
@@ -94,9 +96,7 @@ BackendInterface::fetchLedgerObjects(
         else
             misses.push_back(keys[i]);
     }
-    BOOST_LOG_TRIVIAL(trace)
-        << __func__ << " - cache hits = " << keys.size() - misses.size()
-        << " - cache misses = " << misses.size();
+    LOG(gLog.trace()) << "Cache hits = " << keys.size() - misses.size() << " - cache misses = " << misses.size();
 
     if (misses.size())
     {
@@ -118,15 +118,13 @@ std::optional<ripple::uint256>
 BackendInterface::fetchSuccessorKey(
     ripple::uint256 key,
     std::uint32_t const ledgerSequence,
-    boost::asio::yield_context& yield) const
+    boost::asio::yield_context yield) const
 {
     auto succ = cache_.getSuccessor(key, ledgerSequence);
     if (succ)
-        BOOST_LOG_TRIVIAL(trace)
-            << __func__ << " - cache hit - " << ripple::strHex(key);
+        LOG(gLog.trace()) << "Cache hit - " << ripple::strHex(key);
     else
-        BOOST_LOG_TRIVIAL(trace)
-            << __func__ << " - cache miss - " << ripple::strHex(key);
+        LOG(gLog.trace()) << "Cache miss - " << ripple::strHex(key);
     return succ ? succ->key : doFetchSuccessorKey(key, ledgerSequence, yield);
 }
 
@@ -134,7 +132,7 @@ std::optional<LedgerObject>
 BackendInterface::fetchSuccessorObject(
     ripple::uint256 key,
     std::uint32_t const ledgerSequence,
-    boost::asio::yield_context& yield) const
+    boost::asio::yield_context yield) const
 {
     auto succ = fetchSuccessorKey(key, ledgerSequence, yield);
     if (succ)
@@ -153,8 +151,7 @@ BackendInterface::fetchBookOffers(
     ripple::uint256 const& book,
     std::uint32_t const ledgerSequence,
     std::uint32_t const limit,
-    std::optional<ripple::uint256> const& cursor,
-    boost::asio::yield_context& yield) const
+    boost::asio::yield_context yield) const
 {
     // TODO try to speed this up. This can take a few seconds. The goal is
     // to get it down to a few hundred milliseconds.
@@ -162,10 +159,7 @@ BackendInterface::fetchBookOffers(
     const ripple::uint256 bookEnd = ripple::getQualityNext(book);
     ripple::uint256 uTipIndex = book;
     std::vector<ripple::uint256> keys;
-    auto getMillis = [](auto diff) {
-        return std::chrono::duration_cast<std::chrono::milliseconds>(diff)
-            .count();
-    };
+    auto getMillis = [](auto diff) { return std::chrono::duration_cast<std::chrono::milliseconds>(diff).count(); };
     auto begin = std::chrono::system_clock::now();
     std::uint32_t numSucc = 0;
     std::uint32_t numPages = 0;
@@ -180,30 +174,24 @@ BackendInterface::fetchBookOffers(
         succMillis += getMillis(mid2 - mid1);
         if (!offerDir || offerDir->key >= bookEnd)
         {
-            BOOST_LOG_TRIVIAL(trace) << __func__ << " - offerDir.has_value() "
-                                     << offerDir.has_value() << " breaking";
+            LOG(gLog.trace()) << "offerDir.has_value() " << offerDir.has_value() << " breaking";
             break;
         }
         uTipIndex = offerDir->key;
         while (keys.size() < limit)
         {
             ++numPages;
-            ripple::STLedgerEntry sle{
-                ripple::SerialIter{
-                    offerDir->blob.data(), offerDir->blob.size()},
-                offerDir->key};
+            ripple::STLedgerEntry sle{ripple::SerialIter{offerDir->blob.data(), offerDir->blob.size()}, offerDir->key};
             auto indexes = sle.getFieldV256(ripple::sfIndexes);
             keys.insert(keys.end(), indexes.begin(), indexes.end());
             auto next = sle.getFieldU64(ripple::sfIndexNext);
             if (!next)
             {
-                BOOST_LOG_TRIVIAL(trace)
-                    << __func__ << " next is empty. breaking";
+                LOG(gLog.trace()) << "Next is empty. breaking";
                 break;
             }
             auto nextKey = ripple::keylet::page(uTipIndex, next);
-            auto nextDir =
-                fetchLedgerObject(nextKey.key, ledgerSequence, yield);
+            auto nextDir = fetchLedgerObject(nextKey.key, ledgerSequence, yield);
             assert(nextDir);
             offerDir->blob = *nextDir;
             offerDir->key = nextKey.key;
@@ -215,40 +203,56 @@ BackendInterface::fetchBookOffers(
     auto objs = fetchLedgerObjects(keys, ledgerSequence, yield);
     for (size_t i = 0; i < keys.size() && i < limit; ++i)
     {
-        BOOST_LOG_TRIVIAL(trace)
-            << __func__ << " key = " << ripple::strHex(keys[i])
-            << " blob = " << ripple::strHex(objs[i])
-            << " ledgerSequence = " << ledgerSequence;
+        LOG(gLog.trace()) << "Key = " << ripple::strHex(keys[i]) << " blob = " << ripple::strHex(objs[i])
+                          << " ledgerSequence = " << ledgerSequence;
         assert(objs[i].size());
         page.offers.push_back({keys[i], objs[i]});
     }
     auto end = std::chrono::system_clock::now();
-    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"
-        << " book = " << ripple::strHex(book);
+    LOG(gLog.debug()) << "Fetching " << std::to_string(keys.size()) << " offers took "
+                      << std::to_string(getMillis(mid - begin)) << " milliseconds. Fetching next dir took "
+                      << std::to_string(succMillis) << " milliseonds. Fetched next dir " << std::to_string(numSucc)
+                      << " times"
+                      << " Fetching next page of dir took " << std::to_string(pageMillis) << " milliseconds"
+                      << ". num pages = " << std::to_string(numPages) << ". Fetching all objects took "
+                      << std::to_string(getMillis(end - mid))
+                      << " milliseconds. total time = " << std::to_string(getMillis(end - begin)) << " milliseconds"
+                      << " book = " << ripple::strHex(book);
 
     return page;
 }
 
+std::optional<LedgerRange>
+BackendInterface::hardFetchLedgerRange() const
+{
+    return synchronous([this](auto yield) { return hardFetchLedgerRange(yield); });
+}
+
+std::optional<LedgerRange>
+BackendInterface::fetchLedgerRange() const
+{
+    std::shared_lock lck(rngMtx_);
+    return range;
+}
+
+void
+BackendInterface::updateRange(uint32_t newMax)
+{
+    std::scoped_lock lck(rngMtx_);
+    assert(!range || newMax >= range->maxSequence);
+    if (!range)
+        range = {newMax, newMax};
+    else
+        range->maxSequence = newMax;
+}
+
 LedgerPage
 BackendInterface::fetchLedgerPage(
     std::optional<ripple::uint256> const& cursor,
     std::uint32_t const ledgerSequence,
     std::uint32_t const limit,
     bool outOfOrder,
-    boost::asio::yield_context& yield) const
+    boost::asio::yield_context yield) const
 {
     LedgerPage page;
 
@@ -256,11 +260,8 @@ BackendInterface::fetchLedgerPage(
     bool reachedEnd = false;
     while (keys.size() < limit && !reachedEnd)
     {
-        ripple::uint256 const& curCursor = keys.size() ? keys.back()
-            : cursor                                   ? *cursor
-                                                       : firstKey;
-        std::uint32_t const seq =
-            outOfOrder ? range->maxSequence : ledgerSequence;
+        ripple::uint256 const& curCursor = keys.size() ? keys.back() : cursor ? *cursor : firstKey;
+        std::uint32_t const seq = outOfOrder ? range->maxSequence : ledgerSequence;
         auto succ = fetchSuccessorKey(curCursor, seq, yield);
         if (!succ)
             reachedEnd = true;
@@ -275,16 +276,14 @@ 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;
+            LOG(gLog.error()) << "Deleted or non-existent object in successor table. key = " << ripple::strHex(keys[i])
+                              << " - seq = " << ledgerSequence;
             std::stringstream msg;
             for (size_t j = 0; j < objects.size(); ++j)
             {
                 msg << " - " << ripple::strHex(keys[j]);
             }
-            BOOST_LOG_TRIVIAL(error) << __func__ << msg.str();
-            assert(false);
+            LOG(gLog.error()) << msg.str();
         }
     }
     if (keys.size() && !reachedEnd)
@@ -294,9 +293,7 @@ BackendInterface::fetchLedgerPage(
 }
 
 std::optional<ripple::Fees>
-BackendInterface::fetchFees(
-    std::uint32_t const seq,
-    boost::asio::yield_context& yield) const
+BackendInterface::fetchFees(std::uint32_t const seq, boost::asio::yield_context yield) const
 {
     ripple::Fees fees;
 
@@ -305,7 +302,7 @@ BackendInterface::fetchFees(
 
     if (!bytes)
     {
-        BOOST_LOG_TRIVIAL(error) << __func__ << " - could not find fees";
+        LOG(gLog.error()) << "Could not find fees";
         return {};
     }
 
@@ -315,9 +312,6 @@ BackendInterface::fetchFees(
     if (sle.getFieldIndex(ripple::sfBaseFee) != -1)
         fees.base = sle.getFieldU64(ripple::sfBaseFee);
 
-    if (sle.getFieldIndex(ripple::sfReferenceFeeUnits) != -1)
-        fees.units = sle.getFieldU32(ripple::sfReferenceFeeUnits);
-
     if (sle.getFieldIndex(ripple::sfReserveBase) != -1)
         fees.reserve = sle.getFieldU32(ripple::sfReserveBase);
 
@@ -327,4 +321,4 @@ BackendInterface::fetchFees(
     return fees;
 }
 
-}  // namespace Backend
+}  // namespace data

src/data/BackendInterface.h

@@ -0,0 +1,567 @@
+//------------------------------------------------------------------------------
+/*
+    This file is part of clio: https://github.com/XRPLF/clio
+    Copyright (c) 2022, the clio developers.
+
+    Permission to use, copy, modify, and distribute this software for any
+    purpose with or without fee is hereby granted, provided that the above
+    copyright notice and this permission notice appear in all copies.
+
+    THE  SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL WARRANTIES
+    WITH  REGARD  TO  THIS  SOFTWARE  INCLUDING  ALL  IMPLIED  WARRANTIES  OF
+    MERCHANTABILITY  AND  FITNESS. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR
+    ANY  SPECIAL,  DIRECT,  INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES
+    WHATSOEVER  RESULTING  FROM  LOSS  OF USE, DATA OR PROFITS, WHETHER IN AN
+    ACTION  OF  CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF
+    OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE.
+*/
+//==============================================================================
+
+#pragma once
+
+#include <data/DBHelpers.h>
+#include <data/LedgerCache.h>
+#include <data/Types.h>
+#include <util/config/Config.h>
+#include <util/log/Logger.h>
+
+#include <ripple/protocol/Fees.h>
+#include <ripple/protocol/LedgerHeader.h>
+#include <boost/asio/spawn.hpp>
+#include <boost/json.hpp>
+
+#include <thread>
+#include <type_traits>
+
+namespace data {
+
+/**
+ * @brief Represents a database timeout error.
+ */
+class DatabaseTimeout : public std::exception
+{
+public:
+    const char*
+    what() const throw() override
+    {
+        return "Database read timed out. Please retry the request";
+    }
+};
+
+/**
+ * @brief A helper function that catches DatabaseTimout exceptions and retries indefinitely.
+ *
+ * @tparam FnType The type of function object to execute
+ * @param func The function object to execute
+ * @param waitMs Delay between retry attempts
+ * @return auto The same as the return type of func
+ */
+template <class FnType>
+auto
+retryOnTimeout(FnType func, size_t waitMs = 500)
+{
+    static util::Logger log{"Backend"};
+
+    while (true)
+    {
+        try
+        {
+            return func();
+        }
+        catch (DatabaseTimeout const&)
+        {
+            LOG(log.error()) << "Database request timed out. Sleeping and retrying ... ";
+            std::this_thread::sleep_for(std::chrono::milliseconds(waitMs));
+        }
+    }
+}
+
+/**
+ * @brief Synchronously executes the given function object inside a coroutine.
+ *
+ * @tparam FnType The type of function object to execute
+ * @param func The function object to execute
+ * @return auto The same as the return type of func
+ */
+template <class FnType>
+auto
+synchronous(FnType&& func)
+{
+    boost::asio::io_context ctx;
+
+    using R = typename boost::result_of<FnType(boost::asio::yield_context)>::type;
+    if constexpr (!std::is_same<R, void>::value)
+    {
+        R res;
+        boost::asio::spawn(ctx, [&func, &res](auto yield) { res = func(yield); });
+
+        ctx.run();
+        return res;
+    }
+    else
+    {
+        boost::asio::spawn(ctx, [&func](auto yield) { func(yield); });
+        ctx.run();
+    }
+}
+
+/**
+ * @brief Synchronously execute the given function object and retry until no DatabaseTimeout is thrown.
+ *
+ * @tparam FnType The type of function object to execute
+ * @param func The function object to execute
+ * @return auto The same as the return type of func
+ */
+template <class FnType>
+auto
+synchronousAndRetryOnTimeout(FnType&& func)
+{
+    return retryOnTimeout([&]() { return synchronous(func); });
+}
+
+/**
+ * @brief The interface to the database used by Clio.
+ */
+class BackendInterface
+{
+protected:
+    mutable std::shared_mutex rngMtx_;
+    std::optional<LedgerRange> range;
+    LedgerCache cache_;
+
+public:
+    BackendInterface() = default;
+    virtual ~BackendInterface() = default;
+
+    // TODO: Remove this hack. Cache should not be exposed thru BackendInterface
+    /**
+     * @return Immutable cache
+     */
+    LedgerCache const&
+    cache() const
+    {
+        return cache_;
+    }
+
+    /**
+     * @return Mutable cache
+     */
+    LedgerCache&
+    cache()
+    {
+        return cache_;
+    }
+
+    /**
+     * @brief Fetches a specific ledger by sequence number.
+     *
+     * @param sequence The sequence number to fetch for
+     * @param yield The coroutine context
+     * @return The ripple::LedgerHeader if found; nullopt otherwise
+     */
+    virtual std::optional<ripple::LedgerHeader>
+    fetchLedgerBySequence(std::uint32_t const sequence, boost::asio::yield_context yield) const = 0;
+
+    /**
+     * @brief Fetches a specific ledger by hash.
+     *
+     * @param hash The hash to fetch for
+     * @param yield The coroutine context
+     * @return The ripple::LedgerHeader if found; nullopt otherwise
+     */
+    virtual std::optional<ripple::LedgerHeader>
+    fetchLedgerByHash(ripple::uint256 const& hash, boost::asio::yield_context yield) const = 0;
+
+    /**
+     * @brief Fetches the latest ledger sequence.
+     *
+     * @param yield The coroutine context
+     * @return Latest sequence wrapped in an optional if found; nullopt otherwise
+     */
+    virtual std::optional<std::uint32_t>
+    fetchLatestLedgerSequence(boost::asio::yield_context yield) const = 0;
+
+    /**
+     * @brief Fetch the current ledger range.
+     *
+     * @return The current ledger range if populated; nullopt otherwise
+     */
+    std::optional<LedgerRange>
+    fetchLedgerRange() const;
+
+    /**
+     * @brief Updates the range of sequences that are stored in the DB.
+     *
+     * @param newMax The new maximum sequence available
+     */
+    void
+    updateRange(uint32_t newMax);
+
+    /**
+     * @brief Fetch the fees from a specific ledger sequence.
+     *
+     * @param seq The sequence to fetch for
+     * @param yield The coroutine context
+     * @return ripple::Fees if fees are found; nullopt otherwise
+     */
+    std::optional<ripple::Fees>
+    fetchFees(std::uint32_t const seq, boost::asio::yield_context yield) const;
+
+    /**
+     * @brief Fetches a specific transaction.
+     *
+     * @param hash The hash of the transaction to fetch
+     * @param yield The coroutine context
+     * @return TransactionAndMetadata if transaction is found; nullopt otherwise
+     */
+    virtual std::optional<TransactionAndMetadata>
+    fetchTransaction(ripple::uint256 const& hash, boost::asio::yield_context yield) const = 0;
+
+    /**
+     * @brief Fetches multiple transactions.
+     *
+     * @param hashes A vector of hashes to fetch transactions for
+     * @param yield The coroutine context
+     * @return A vector of TransactionAndMetadata matching the given hashes
+     */
+    virtual std::vector<TransactionAndMetadata>
+    fetchTransactions(std::vector<ripple::uint256> const& hashes, boost::asio::yield_context yield) const = 0;
+
+    /**
+     * @brief Fetches all transactions for a specific account.
+     *
+     * @param account The account to fetch transactions for
+     * @param limit The maximum number of transactions per result page
+     * @param forward Whether to fetch the page forwards or backwards from the given cursor
+     * @param cursor The cursor to resume fetching from
+     * @param yield The coroutine context
+     * @return Results and a cursor to resume from
+     */
+    virtual TransactionsAndCursor
+    fetchAccountTransactions(
+        ripple::AccountID const& account,
+        std::uint32_t const limit,
+        bool forward,
+        std::optional<TransactionsCursor> const& cursor,
+        boost::asio::yield_context yield) const = 0;
+
+    /**
+     * @brief Fetches all transactions from a specific ledger.
+     *
+     * @param ledgerSequence The ledger sequence to fetch for
+     * @param yield The coroutine context
+     * @return Results as a vector of TransactionAndMetadata
+     */
+    virtual std::vector<TransactionAndMetadata>
+    fetchAllTransactionsInLedger(std::uint32_t const ledgerSequence, boost::asio::yield_context yield) const = 0;
+
+    /**
+     * @brief Fetches all transaction hashes from a specific ledger.
+     *
+     * @param ledgerSequence The ledger sequence to fetch for
+     * @param yield The coroutine context
+     * @return Hashes as ripple::uint256 in a vector
+     */
+    virtual std::vector<ripple::uint256>
+    fetchAllTransactionHashesInLedger(std::uint32_t const ledgerSequence, boost::asio::yield_context yield) const = 0;
+
+    /**
+     * @brief Fetches a specific NFT.
+     *
+     * @param tokenID The ID of the NFT
+     * @param ledgerSequence The ledger sequence to fetch for
+     * @param yield The coroutine context
+     * @return NFT object on success; nullopt otherwise
+     */
+    virtual std::optional<NFT>
+    fetchNFT(ripple::uint256 const& tokenID, std::uint32_t const ledgerSequence, boost::asio::yield_context yield)
+        const = 0;
+
+    /**
+     * @brief Fetches all transactions for a specific NFT.
+     *
+     * @param tokenID The ID of the NFT
+     * @param limit The maximum number of transactions per result page
+     * @param forward Whether to fetch the page forwards or backwards from the given cursor
+     * @param cursorIn The cursor to resume fetching from
+     * @param yield The coroutine context
+     * @return Results and a cursor to resume from
+     */
+    virtual TransactionsAndCursor
+    fetchNFTTransactions(
+        ripple::uint256 const& tokenID,
+        std::uint32_t const limit,
+        bool const forward,
+        std::optional<TransactionsCursor> const& cursorIn,
+        boost::asio::yield_context yield) const = 0;
+
+    /**
+     * @brief Fetches a specific ledger object.
+     *
+     * Currently the real fetch happens in doFetchLedgerObject and fetchLedgerObject attempts to fetch from Cache first
+     * and only calls out to the real DB if a cache miss ocurred.
+     *
+     * @param key The key of the object
+     * @param sequence The ledger sequence to fetch for
+     * @param yield The coroutine context
+     * @return The object as a Blob on success; nullopt otherwise
+     */
+    std::optional<Blob>
+    fetchLedgerObject(ripple::uint256 const& key, std::uint32_t const sequence, boost::asio::yield_context yield) const;
+
+    /**
+     * @brief Fetches all ledger objects by their keys.
+     *
+     * Currently the real fetch happens in doFetchLedgerObjects and fetchLedgerObjects attempts to fetch from Cache
+     * first and only calls out to the real DB for each of the keys that was not found in the cache.
+     *
+     * @param keys A vector with the keys of the objects to fetch
+     * @param sequence The ledger sequence to fetch for
+     * @param yield The coroutine context
+     * @return A vector of ledger objects as Blobs
+     */
+    std::vector<Blob>
+    fetchLedgerObjects(
+        std::vector<ripple::uint256> const& keys,
+        std::uint32_t const sequence,
+        boost::asio::yield_context yield) const;
+
+    /**
+     * @brief The database-specific implementation for fetching a ledger object.
+     *
+     * @param key The key to fetch for
+     * @param sequence The ledger sequence to fetch for
+     * @param yield The coroutine context
+     * @return The object as a Blob on success; nullopt otherwise
+     */
+    virtual std::optional<Blob>
+    doFetchLedgerObject(ripple::uint256 const& key, std::uint32_t const sequence, boost::asio::yield_context yield)
+        const = 0;
+
+    /**
+     * @brief The database-specific implementation for fetching ledger objects.
+     *
+     * @param keys The keys to fetch for
+     * @param sequence The ledger sequence to fetch for
+     * @param yield The coroutine context
+     * @return A vector of Blobs representing each fetched object
+     */
+    virtual std::vector<Blob>
+    doFetchLedgerObjects(
+        std::vector<ripple::uint256> const& keys,
+        std::uint32_t const sequence,
+        boost::asio::yield_context yield) const = 0;
+
+    /**
+     * @brief Returns the difference between ledgers.
+     *
+     * @param ledgerSequence The ledger sequence to fetch for
+     * @param yield The coroutine context
+     * @return A vector of LedgerObject representing the diff
+     */
+    virtual std::vector<LedgerObject>
+    fetchLedgerDiff(std::uint32_t const ledgerSequence, boost::asio::yield_context yield) const = 0;
+
+    /**
+     * @brief Fetches a page of ledger objects, ordered by key/index.
+     *
+     * @param cursor The cursor to resume fetching from
+     * @param ledgerSequence The ledger sequence to fetch for
+     * @param limit The maximum number of transactions per result page
+     * @param outOfOrder If set to true max available sequence is used instead of ledgerSequence
+     * @param yield The coroutine context
+     * @return The ledger page
+     */
+    LedgerPage
+    fetchLedgerPage(
+        std::optional<ripple::uint256> const& cursor,
+        std::uint32_t const ledgerSequence,
+        std::uint32_t const limit,
+        bool outOfOrder,
+        boost::asio::yield_context yield) const;
+
+    /**
+     * @brief Fetches the successor object.
+     *
+     * @param key The key to fetch for
+     * @param ledgerSequence The ledger sequence to fetch for
+     * @param yield The coroutine context
+     * @return The sucessor on success; nullopt otherwise
+     */
+    std::optional<LedgerObject>
+    fetchSuccessorObject(ripple::uint256 key, std::uint32_t const ledgerSequence, boost::asio::yield_context yield)
+        const;
+
+    /**
+     * @brief Fetches the successor key.
+     *
+     * Thea real fetch happens in doFetchSuccessorKey. This function will attempt to lookup the successor in the cache
+     * first and only if it's not found in the cache will it fetch from the actual DB.
+     *
+     * @param key The key to fetch for
+     * @param ledgerSequence The ledger sequence to fetch for
+     * @param yield The coroutine context
+     * @return The sucessor key on success; nullopt otherwise
+     */
+    std::optional<ripple::uint256>
+    fetchSuccessorKey(ripple::uint256 key, std::uint32_t const ledgerSequence, boost::asio::yield_context yield) const;
+
+    /**
+     * @brief Database-specific implementation of fetching the successor key
+     *
+     * @param key The key to fetch for
+     * @param ledgerSequence The ledger sequence to fetch for
+     * @param yield The coroutine context
+     * @return The sucessor on success; nullopt otherwise
+     */
+    virtual std::optional<ripple::uint256>
+    doFetchSuccessorKey(ripple::uint256 key, std::uint32_t const ledgerSequence, boost::asio::yield_context yield)
+        const = 0;
+
+    /**
+     * @brief Fetches book offers.
+     *
+     * @param book Unsigned 256-bit integer.
+     * @param ledgerSequence The ledger sequence to fetch for
+     * @param limit Pagaing limit as to how many transactions returned per page.
+     * @param yield The coroutine context
+     * @return The book offers page
+     */
+    BookOffersPage
+    fetchBookOffers(
+        ripple::uint256 const& book,
+        std::uint32_t const ledgerSequence,
+        std::uint32_t const limit,
+        boost::asio::yield_context yield) const;
+
+    /**
+     * @brief Synchronously fetches the ledger range from DB.
+     *
+     * This function just wraps hardFetchLedgerRange(boost::asio::yield_context) using synchronous(FnType&&).
+     *
+     * @return The ledger range if available; nullopt otherwise
+     */
+    std::optional<LedgerRange>
+    hardFetchLedgerRange() const;
+
+    /**
+     * @brief Fetches the ledger range from DB.
+     *
+     * @return The ledger range if available; nullopt otherwise
+     */
+    virtual std::optional<LedgerRange>
+    hardFetchLedgerRange(boost::asio::yield_context yield) const = 0;
+
+    /**
+     * @brief Fetches the ledger range from DB retrying until no DatabaseTimeout is thrown.
+     *
+     * @return The ledger range if available; nullopt otherwise
+     */
+    std::optional<LedgerRange>
+    hardFetchLedgerRangeNoThrow() const;
+
+    /**
+     * @brief Writes to a specific ledger.
+     *
+     * @param ledgerHeader Ledger header.
+     * @param blob r-value string serialization of ledger header.
+     */
+    virtual void
+    writeLedger(ripple::LedgerHeader const& ledgerHeader, std::string&& blob) = 0;
+
+    /**
+     * @brief Writes a new ledger object.
+     *
+     * @param key The key to write the ledger object under
+     * @param seq The ledger sequence to write for
+     * @param blob The data to write
+     */
+    virtual void
+    writeLedgerObject(std::string&& key, std::uint32_t const seq, std::string&& blob);
+
+    /**
+     * @brief Writes a new transaction.
+     *
+     * @param hash The hash of the transaction
+     * @param seq The ledger sequence to write for
+     * @param date The timestamp of the entry
+     * @param transaction The transaction data to write
+     * @param metadata The metadata to write
+     */
+    virtual void
+    writeTransaction(
+        std::string&& hash,
+        std::uint32_t const seq,
+        std::uint32_t const date,
+        std::string&& transaction,
+        std::string&& metadata) = 0;
+
+    /**
+     * @brief Writes NFTs to the database.
+     *
+     * @param data A vector of NFTsData objects representing the NFTs
+     */
+    virtual void
+    writeNFTs(std::vector<NFTsData>&& data) = 0;
+
+    /**
+     * @brief Write a new set of account transactions.
+     *
+     * @param data A vector of AccountTransactionsData objects representing the account transactions
+     */
+    virtual void
+    writeAccountTransactions(std::vector<AccountTransactionsData>&& data) = 0;
+
+    /**
+     * @brief Write NFTs transactions.
+     *
+     * @param data A vector of NFTTransactionsData objects
+     */
+    virtual void
+    writeNFTTransactions(std::vector<NFTTransactionsData>&& data) = 0;
+
+    /**
+     * @brief Write a new successor.
+     *
+     * @param key Key of the object that the passed successor will be the successor for
+     * @param seq The ledger sequence to write for
+     * @param successor The successor data to write
+     */
+    virtual void
+    writeSuccessor(std::string&& key, std::uint32_t const seq, std::string&& successor) = 0;
+
+    /**
+     * @brief Starts a write transaction with the DB. No-op for cassandra.
+     *
+     * Note: Can potentially be deprecated and removed.
+     */
+    virtual void
+    startWrites() const = 0;
+
+    /**
+     * @brief Tells database we finished writing all data for a specific ledger.
+     *
+     * Uses doFinishWrites to synchronize with the pending writes.
+     *
+     * @param ledgerSequence The ledger sequence to finish writing for
+     * @return true on success; false otherwise
+     */
+    bool
+    finishWrites(std::uint32_t const ledgerSequence);
+
+    /**
+     * @return true if database is overwhelmed; false otherwise
+     */
+    virtual bool
+    isTooBusy() const = 0;
+
+private:
+    virtual void
+    doWriteLedgerObject(std::string&& key, std::uint32_t const seq, std::string&& blob) = 0;
+
+    virtual bool
+    doFinishWrites() = 0;
+};
+
+}  // namespace data
+using BackendInterface = data::BackendInterface;

src/data/CassandraBackend.h

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

src/data/DBHelpers.h

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

src/data/LedgerCache.cpp

@@ -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 <data/LedgerCache.h>
+
+namespace data {
+
+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 data

src/data/LedgerCache.h

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

src/data/README.md

@@ -0,0 +1,221 @@
+# 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. 
+
+## 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.
+
+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. 
+
+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. 
+
+**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.  
+
+## 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, 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.
+
+*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 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`
+```
+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.
+
+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`
+```
+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`
+```
+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 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`
+```
+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 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.
+
+
+### `successor`
+```
+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`
+```
+CREATE TABLE clio.nf_tokens (
+	token_id blob,         # The NFT's ID
+	sequence bigint,       # Sequence of ledger version
+	owner blob,            # The account ID of the owner of this NFT at this ledger
+	is_burned boolean,     # True if token was burned in this ledger
+	PRIMARY KEY (token_id, sequence)
+) WITH CLUSTERING ORDER BY (sequence DESC) ...
+```
+This table indexes NFT IDs with their owner at a given ledger. So
+```
+SELECT * FROM nf_tokens
+WHERE token_id = N AND seq <= Y
+ORDER BY seq DESC LIMIT 1;
+```
+will give you the owner of token N at ledger Y and whether it was burned. If
+the token is burned, the owner field indicates the account that owned the
+token at the time it was burned; it does not indicate the person who burned
+the token, necessarily. If you need to determine who burned the token you can
+use the `nft_history` API, which will give you the NFTokenBurn transaction
+that burned this token, along with the account that submitted that
+transaction.
+
+#### `issuer_nf_tokens_v2`
+```
+CREATE TABLE clio.issuer_nf_tokens_v2 (
+	issuer blob,       # The NFT issuer's account ID
+	taxon bigint,      # The NFT's token taxon
+	token_id blob,     # The NFT's ID
+	PRIMARY KEY (issuer, taxon, token_id)
+) WITH CLUSTERING ORDER BY (taxon ASC, token_id ASC) ...
+```
+This table indexes token IDs against their issuer and issuer/taxon
+combination. This is useful for determining all the NFTs a specific account
+issued, or all the NFTs a specific account issued with a specific taxon. It is
+not useful to know all the NFTs with a given taxon while excluding issuer, since the
+meaning of a taxon is left to an issuer.
+
+#### `nf_token_uris`
+```
+CREATE TABLE clio.nf_token_uris (
+	token_id blob,    # The NFT's ID
+	sequence bigint,  # Sequence of ledger version
+	uri blob,         # The NFT's URI
+	PRIMARY KEY (token_id, sequence)
+) WITH CLUSTERING ORDER BY (sequence DESC) ...
+```
+This table is used to store an NFT's URI. Without storing this here, we would
+need to traverse the NFT owner's entire set of NFTs to find the URI, again due
+to the way that NFTs are stored in rippled. Furthermore, instead of storing
+this in the `nf_tokens` table, we store it here to save space. A given NFT
+will have only one entry in this table (see caveat below), written to this
+table as soon as clio sees the NFTokenMint transaction, or when clio loads an
+NFTokenPage from the initial ledger it downloaded. However, the `nf_tokens`
+table is written to every time an NFT changes ownership, or if it is burned.
+
+Given this, why do we have to store the sequence? Unfortunately there is an
+extreme edge case where a given NFT ID can be burned, and then re-minted with
+a different URI. This is extremely unlikely, and might be fixed in a future
+version to rippled, but just in case we can handle that edge case by allowing
+a given NFT ID to have a new URI assigned in this case, without removing the
+prior URI.
+
+#### `nf_token_transactions`
+```
+CREATE TABLE clio.nf_token_transactions (
+	token_id blob,                  # The NFT's ID
+	seq_idx tuple<bigint, bigint>,  # Tuple of (ledger_index, transaction_index)
+	hash blob,                      # Hash of the transaction
+	PRIMARY KEY (token_id, seq_idx)
+) WITH CLUSTERING ORDER BY (seq_idx DESC) ...
+```
+This table is the NFT equivalent of `account_tx`. It's motivated by the exact
+same reasons and serves the analogous purpose here. It drives the
+`nft_history` API.
+

src/data/Types.h

@@ -0,0 +1,191 @@
+//------------------------------------------------------------------------------
+/*
+    This file is part of clio: https://github.com/XRPLF/clio
+    Copyright (c) 2022, the clio developers.
+
+    Permission to use, copy, modify, and distribute this software for any
+    purpose with or without fee is hereby granted, provided that the above
+    copyright notice and this permission notice appear in all copies.
+
+    THE  SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL WARRANTIES
+    WITH  REGARD  TO  THIS  SOFTWARE  INCLUDING  ALL  IMPLIED  WARRANTIES  OF
+    MERCHANTABILITY  AND  FITNESS. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR
+    ANY  SPECIAL,  DIRECT,  INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES
+    WHATSOEVER  RESULTING  FROM  LOSS  OF USE, DATA OR PROFITS, WHETHER IN AN
+    ACTION  OF  CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF
+    OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE.
+*/
+//==============================================================================
+
+#pragma once
+
+#include <ripple/basics/base_uint.h>
+#include <ripple/protocol/AccountID.h>
+
+#include <optional>
+#include <string>
+#include <vector>
+
+namespace data {
+
+using Blob = std::vector<unsigned char>;
+
+/**
+ * @brief Represents an object in the ledger.
+ */
+struct LedgerObject
+{
+    ripple::uint256 key;
+    Blob blob;
+
+    bool
+    operator==(const LedgerObject& other) const
+    {
+        return key == other.key && blob == other.blob;
+    }
+};
+
+/**
+ * @brief Represents a page of LedgerObjects.
+ */
+struct LedgerPage
+{
+    std::vector<LedgerObject> objects;
+    std::optional<ripple::uint256> cursor;
+};
+
+/**
+ * @brief Represents a page of book offer objects.
+ */
+struct BookOffersPage
+{
+    std::vector<LedgerObject> offers;
+    std::optional<ripple::uint256> cursor;
+};
+
+/**
+ * @brief Represents a transaction and its metadata bundled together.
+ */
+struct TransactionAndMetadata
+{
+    Blob transaction;
+    Blob metadata;
+    std::uint32_t ledgerSequence = 0;
+    std::uint32_t date = 0;
+
+    TransactionAndMetadata() = default;
+    TransactionAndMetadata(
+        Blob const& transaction,
+        Blob const& metadata,
+        std::uint32_t ledgerSequence,
+        std::uint32_t date)
+        : transaction{transaction}, metadata{metadata}, ledgerSequence{ledgerSequence}, date{date}
+    {
+    }
+
+    TransactionAndMetadata(std::tuple<Blob, Blob, std::uint32_t, std::uint32_t> data)
+        : transaction{std::get<0>(data)}
+        , metadata{std::get<1>(data)}
+        , ledgerSequence{std::get<2>(data)}
+        , date{std::get<3>(data)}
+    {
+    }
+
+    bool
+    operator==(const TransactionAndMetadata& other) const
+    {
+        return transaction == other.transaction && metadata == other.metadata &&
+            ledgerSequence == other.ledgerSequence && date == other.date;
+    }
+};
+
+/**
+ * @brief Represents a cursor into the transactions table.
+ */
+struct TransactionsCursor
+{
+    std::uint32_t ledgerSequence = 0;
+    std::uint32_t transactionIndex = 0;
+
+    TransactionsCursor() = default;
+    TransactionsCursor(std::uint32_t ledgerSequence, std::uint32_t transactionIndex)
+        : ledgerSequence{ledgerSequence}, transactionIndex{transactionIndex}
+    {
+    }
+
+    TransactionsCursor(std::tuple<std::uint32_t, std::uint32_t> data)
+        : ledgerSequence{std::get<0>(data)}, transactionIndex{std::get<1>(data)}
+    {
+    }
+
+    TransactionsCursor&
+    operator=(TransactionsCursor const&) = default;
+
+    bool
+    operator==(TransactionsCursor const& other) const = default;
+
+    [[nodiscard]] std::tuple<std::uint32_t, std::uint32_t>
+    asTuple() const
+    {
+        return std::make_tuple(ledgerSequence, transactionIndex);
+    }
+};
+
+/**
+ * @brief Represests a bundle of transactions with metadata and a cursor to the next page.
+ */
+struct TransactionsAndCursor
+{
+    std::vector<TransactionAndMetadata> txns;
+    std::optional<TransactionsCursor> cursor;
+};
+
+/**
+ * @brief Represents a NFToken.
+ */
+struct NFT
+{
+    ripple::uint256 tokenID;
+    std::uint32_t ledgerSequence;
+    ripple::AccountID owner;
+    Blob uri;
+    bool isBurned;
+
+    NFT() = default;
+    NFT(ripple::uint256 const& tokenID,
+        std::uint32_t ledgerSequence,
+        ripple::AccountID const& owner,
+        Blob const& uri,
+        bool isBurned)
+        : tokenID{tokenID}, ledgerSequence{ledgerSequence}, owner{owner}, uri{uri}, isBurned{isBurned}
+    {
+    }
+
+    NFT(ripple::uint256 const& tokenID, std::uint32_t ledgerSequence, ripple::AccountID const& owner, bool isBurned)
+        : NFT(tokenID, ledgerSequence, owner, {}, isBurned)
+    {
+    }
+
+    // clearly two tokens are the same if they have the same ID, but this struct stores the state of a given token at a
+    // given ledger sequence, so we also need to compare with ledgerSequence.
+    bool
+    operator==(NFT const& other) const
+    {
+        return tokenID == other.tokenID && ledgerSequence == other.ledgerSequence;
+    }
+};
+
+/**
+ * @brief Stores a range of sequences as a min and max pair.
+ */
+struct LedgerRange
+{
+    std::uint32_t minSequence = 0;
+    std::uint32_t maxSequence = 0;
+};
+
+constexpr ripple::uint256 firstKey{"0000000000000000000000000000000000000000000000000000000000000000"};
+constexpr ripple::uint256 lastKey{"FFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFF"};
+constexpr ripple::uint256 hi192{"0000000000000000000000000000000000000000000000001111111111111111"};
+
+}  // namespace data

src/data/cassandra/Concepts.h

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

src/data/cassandra/Error.h

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

src/data/cassandra/Handle.cpp

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

src/data/cassandra/Handle.h

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

src/data/cassandra/Schema.h

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

src/data/cassandra/SettingsProvider.cpp

@@ -0,0 +1,151 @@
+//------------------------------------------------------------------------------
+/*
+    This file is part of clio: https://github.com/XRPLF/clio
+    Copyright (c) 2023, the clio developers.
+
+    Permission to use, copy, modify, and distribute this software for any
+    purpose with or without fee is hereby granted, provided that the above
+    copyright notice and this permission notice appear in all copies.
+
+    THE  SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL WARRANTIES
+    WITH  REGARD  TO  THIS  SOFTWARE  INCLUDING  ALL  IMPLIED  WARRANTIES  OF
+    MERCHANTABILITY  AND  FITNESS. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR
+    ANY  SPECIAL,  DIRECT,  INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES
+    WHATSOEVER  RESULTING  FROM  LOSS  OF USE, DATA OR PROFITS, WHETHER IN AN
+    ACTION  OF  CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF
+    OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE.
+*/
+//==============================================================================
+
+#include <data/cassandra/SettingsProvider.h>
+#include <data/cassandra/impl/Cluster.h>
+#include <data/cassandra/impl/Statement.h>
+#include <util/config/Config.h>
+
+#include <boost/json.hpp>
+
+#include <fstream>
+#include <string>
+#include <thread>
+
+namespace data::cassandra {
+
+namespace detail {
+inline Settings::ContactPoints
+tag_invoke(boost::json::value_to_tag<Settings::ContactPoints>, boost::json::value const& value)
+{
+    if (not value.is_object())
+        throw std::runtime_error(
+            "Feed entire Cassandra section to parse "
+            "Settings::ContactPoints instead");
+
+    util::Config obj{value};
+    Settings::ContactPoints out;
+
+    out.contactPoints = obj.valueOrThrow<std::string>("contact_points", "`contact_points` must be a string");
+    out.port = obj.maybeValue<uint16_t>("port");
+
+    return out;
+}
+
+inline Settings::SecureConnectionBundle
+tag_invoke(boost::json::value_to_tag<Settings::SecureConnectionBundle>, boost::json::value const& value)
+{
+    if (not value.is_string())
+        throw std::runtime_error("`secure_connect_bundle` must be a string");
+    return Settings::SecureConnectionBundle{value.as_string().data()};
+}
+}  // namespace detail
+
+SettingsProvider::SettingsProvider(util::Config const& cfg, uint16_t ttl)
+    : config_{cfg}
+    , keyspace_{cfg.valueOr<std::string>("keyspace", "clio")}
+    , tablePrefix_{cfg.maybeValue<std::string>("table_prefix")}
+    , replicationFactor_{cfg.valueOr<uint16_t>("replication_factor", 3)}
+    , ttl_{ttl}
+    , settings_{parseSettings()}
+{
+}
+
+Settings
+SettingsProvider::getSettings() const
+{
+    return settings_;
+}
+
+std::optional<std::string>
+SettingsProvider::parseOptionalCertificate() const
+{
+    if (auto const certPath = config_.maybeValue<std::string>("certfile"); certPath)
+    {
+        auto const path = std::filesystem::path(*certPath);
+        std::ifstream fileStream(path.string(), std::ios::in);
+        if (!fileStream)
+        {
+            throw std::system_error(errno, std::generic_category(), "Opening certificate " + path.string());
+        }
+
+        std::string contents(std::istreambuf_iterator<char>{fileStream}, std::istreambuf_iterator<char>{});
+        if (fileStream.bad())
+        {
+            throw std::system_error(errno, std::generic_category(), "Reading certificate " + path.string());
+        }
+
+        return contents;
+    }
+
+    return std::nullopt;
+}
+
+Settings
+SettingsProvider::parseSettings() const
+{
+    auto settings = Settings::defaultSettings();
+    if (auto const bundle = config_.maybeValue<Settings::SecureConnectionBundle>("secure_connect_bundle"); bundle)
+    {
+        settings.connectionInfo = *bundle;
+    }
+    else
+    {
+        settings.connectionInfo =
+            config_.valueOrThrow<Settings::ContactPoints>("Missing contact_points in Cassandra config");
+    }
+
+    settings.threads = config_.valueOr<uint32_t>("threads", settings.threads);
+    settings.maxWriteRequestsOutstanding =
+        config_.valueOr<uint32_t>("max_write_requests_outstanding", settings.maxWriteRequestsOutstanding);
+    settings.maxReadRequestsOutstanding =
+        config_.valueOr<uint32_t>("max_read_requests_outstanding", settings.maxReadRequestsOutstanding);
+    settings.maxConnectionsPerHost =
+        config_.valueOr<uint32_t>("max_connections_per_host", settings.maxConnectionsPerHost);
+    settings.coreConnectionsPerHost =
+        config_.valueOr<uint32_t>("core_connections_per_host", settings.coreConnectionsPerHost);
+    settings.maxConcurrentRequestsThreshold = config_.valueOr<uint32_t>(
+        "max_concurrent_requests_threshold",
+        (settings.maxReadRequestsOutstanding + settings.maxWriteRequestsOutstanding) / settings.coreConnectionsPerHost);
+
+    settings.queueSizeIO = config_.maybeValue<uint32_t>("queue_size_io");
+    settings.queueSizeEvent = config_.maybeValue<uint32_t>("queue_size_event");
+    settings.writeBytesHighWatermark = config_.maybeValue<uint32_t>("write_bytes_high_water_mark");
+    settings.writeBytesLowWatermark = config_.maybeValue<uint32_t>("write_bytes_low_water_mark");
+    settings.pendingRequestsHighWatermark = config_.maybeValue<uint32_t>("pending_requests_high_water_mark");
+    settings.pendingRequestsLowWatermark = config_.maybeValue<uint32_t>("pending_requests_low_water_mark");
+    settings.maxRequestsPerFlush = config_.maybeValue<uint32_t>("max_requests_per_flush");
+    settings.maxConcurrentCreation = config_.maybeValue<uint32_t>("max_concurrent_creation");
+
+    auto const connectTimeoutSecond = config_.maybeValue<uint32_t>("connect_timeout");
+    if (connectTimeoutSecond)
+        settings.connectionTimeout = std::chrono::milliseconds{*connectTimeoutSecond * 1000};
+
+    auto const requestTimeoutSecond = config_.maybeValue<uint32_t>("request_timeout");
+    if (requestTimeoutSecond)
+        settings.requestTimeout = std::chrono::milliseconds{*requestTimeoutSecond * 1000};
+
+    settings.certificate = parseOptionalCertificate();
+    settings.username = config_.maybeValue<std::string>("username");
+    settings.password = config_.maybeValue<std::string>("password");
+
+    return settings;
+}
+
+}  // namespace data::cassandra

src/data/cassandra/SettingsProvider.h

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

src/data/cassandra/Types.h

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

src/data/cassandra/impl/AsyncExecutor.h

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

src/data/cassandra/impl/Batch.cpp

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

src/data/cassandra/impl/Batch.h

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

src/data/cassandra/impl/Cluster.cpp

@@ -0,0 +1,224 @@
+//------------------------------------------------------------------------------
+/*
+    This file is part of clio: https://github.com/XRPLF/clio
+    Copyright (c) 2023, the clio developers.
+
+    Permission to use, copy, modify, and distribute this software for any
+    purpose with or without fee is hereby granted, provided that the above
+    copyright notice and this permission notice appear in all copies.
+
+    THE  SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL WARRANTIES
+    WITH  REGARD  TO  THIS  SOFTWARE  INCLUDING  ALL  IMPLIED  WARRANTIES  OF
+    MERCHANTABILITY  AND  FITNESS. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR
+    ANY  SPECIAL,  DIRECT,  INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES
+    WHATSOEVER  RESULTING  FROM  LOSS  OF USE, DATA OR PROFITS, WHETHER IN AN
+    ACTION  OF  CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF
+    OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE.
+*/
+//==============================================================================
+
+#include <data/cassandra/impl/Cluster.h>
+#include <data/cassandra/impl/SslContext.h>
+#include <data/cassandra/impl/Statement.h>
+#include <util/Expected.h>
+
+#include <fmt/core.h>
+
+#include <exception>
+#include <vector>
+
+namespace {
+static constexpr auto clusterDeleter = [](CassCluster* ptr) { cass_cluster_free(ptr); };
+
+template <class... Ts>
+struct overloadSet : Ts...
+{
+    using Ts::operator()...;
+};
+
+// explicit deduction guide (not needed as of C++20, but clang be clang)
+template <class... Ts>
+overloadSet(Ts...) -> overloadSet<Ts...>;
+};  // namespace
+
+namespace data::cassandra::detail {
+
+Cluster::Cluster(Settings const& settings) : ManagedObject{cass_cluster_new(), clusterDeleter}
+{
+    using std::to_string;
+
+    cass_cluster_set_token_aware_routing(*this, cass_true);
+    if (auto const rc = cass_cluster_set_protocol_version(*this, CASS_PROTOCOL_VERSION_V4); rc != CASS_OK)
+    {
+        throw std::runtime_error(
+            fmt::format("Error setting cassandra protocol version to v4: {}", cass_error_desc(rc)));
+    }
+
+    if (auto const rc = cass_cluster_set_num_threads_io(*this, settings.threads); rc != CASS_OK)
+    {
+        throw std::runtime_error(
+            fmt::format("Error setting cassandra io threads to {}: {}", settings.threads, cass_error_desc(rc)));
+    }
+
+    cass_log_set_level(settings.enableLog ? CASS_LOG_TRACE : CASS_LOG_DISABLED);
+    cass_cluster_set_connect_timeout(*this, settings.connectionTimeout.count());
+    cass_cluster_set_request_timeout(*this, settings.requestTimeout.count());
+
+    if (auto const rc =
+            cass_cluster_set_max_concurrent_requests_threshold(*this, settings.maxConcurrentRequestsThreshold);
+        rc != CASS_OK)
+    {
+        throw std::runtime_error(
+            fmt::format("Could not set max concurrent requests per host threshold: {}", cass_error_desc(rc)));
+    }
+
+    if (auto const rc = cass_cluster_set_max_connections_per_host(*this, settings.maxConnectionsPerHost); rc != CASS_OK)
+    {
+        throw std::runtime_error(fmt::format("Could not set max connections per host: {}", cass_error_desc(rc)));
+    }
+
+    if (auto const rc = cass_cluster_set_core_connections_per_host(*this, settings.coreConnectionsPerHost);
+        rc != CASS_OK)
+    {
+        throw std::runtime_error(fmt::format("Could not set core connections per host: {}", cass_error_desc(rc)));
+    }
+
+    auto const queueSize =
+        settings.queueSizeIO.value_or(settings.maxWriteRequestsOutstanding + settings.maxReadRequestsOutstanding);
+    if (auto const rc = cass_cluster_set_queue_size_io(*this, queueSize); rc != CASS_OK)
+    {
+        throw std::runtime_error(fmt::format("Could not set queue size for IO per host: {}", cass_error_desc(rc)));
+    }
+
+    auto apply = []<typename ValueType, typename Fn>(
+        std::optional<ValueType> const& maybeValue, Fn&& fn) requires std::is_object_v<Fn>
+    {
+        if (maybeValue)
+            std::invoke(fn, maybeValue.value());
+    };
+
+    apply(settings.queueSizeEvent, [this](auto value) {
+        if (auto const rc = cass_cluster_set_queue_size_event(*this, value); rc != CASS_OK)
+            throw std::runtime_error(
+                fmt::format("Could not set queue size for events per host: {}", cass_error_desc(rc)));
+    });
+
+    apply(settings.writeBytesHighWatermark, [this](auto value) {
+        if (auto const rc = cass_cluster_set_write_bytes_high_water_mark(*this, value); rc != CASS_OK)
+            throw std::runtime_error(fmt::format("Could not set write bytes high water_mark: {}", cass_error_desc(rc)));
+    });
+
+    apply(settings.writeBytesLowWatermark, [this](auto value) {
+        if (auto const rc = cass_cluster_set_write_bytes_low_water_mark(*this, value); rc != CASS_OK)
+            throw std::runtime_error(fmt::format("Could not set write bytes low water mark: {}", cass_error_desc(rc)));
+    });
+
+    apply(settings.pendingRequestsHighWatermark, [this](auto value) {
+        if (auto const rc = cass_cluster_set_pending_requests_high_water_mark(*this, value); rc != CASS_OK)
+            throw std::runtime_error(
+                fmt::format("Could not set pending requests high water mark: {}", cass_error_desc(rc)));
+    });
+
+    apply(settings.pendingRequestsLowWatermark, [this](auto value) {
+        if (auto const rc = cass_cluster_set_pending_requests_low_water_mark(*this, value); rc != CASS_OK)
+            throw std::runtime_error(
+                fmt::format("Could not set pending requests low water mark: {}", cass_error_desc(rc)));
+    });
+
+    apply(settings.maxRequestsPerFlush, [this](auto value) {
+        if (auto const rc = cass_cluster_set_max_requests_per_flush(*this, value); rc != CASS_OK)
+            throw std::runtime_error(fmt::format("Could not set max requests per flush: {}", cass_error_desc(rc)));
+    });
+
+    apply(settings.maxConcurrentCreation, [this](auto value) {
+        if (auto const rc = cass_cluster_set_max_concurrent_creation(*this, value); rc != CASS_OK)
+            throw std::runtime_error(fmt::format("Could not set max concurrent creation: {}", cass_error_desc(rc)));
+    });
+
+    setupConnection(settings);
+    setupCertificate(settings);
+    setupCredentials(settings);
+
+    auto valueOrDefault = []<typename T>(std::optional<T> const& maybeValue) -> std::string {
+        return maybeValue ? to_string(*maybeValue) : "default";
+    };
+
+    LOG(log_.info()) << "Threads: " << settings.threads;
+    LOG(log_.info()) << "Max concurrent requests per host: " << settings.maxConcurrentRequestsThreshold;
+    LOG(log_.info()) << "Max connections per host: " << settings.maxConnectionsPerHost;
+    LOG(log_.info()) << "Core connections per host: " << settings.coreConnectionsPerHost;
+    LOG(log_.info()) << "IO queue size: " << queueSize;
+    LOG(log_.info()) << "Event queue size: " << valueOrDefault(settings.queueSizeEvent);
+    LOG(log_.info()) << "Write bytes high watermark: " << valueOrDefault(settings.writeBytesHighWatermark);
+    LOG(log_.info()) << "Write bytes low watermark: " << valueOrDefault(settings.writeBytesLowWatermark);
+    LOG(log_.info()) << "Pending requests high watermark: " << valueOrDefault(settings.pendingRequestsHighWatermark);
+    LOG(log_.info()) << "Pending requests low watermark: " << valueOrDefault(settings.pendingRequestsLowWatermark);
+    LOG(log_.info()) << "Max requests per flush: " << valueOrDefault(settings.maxRequestsPerFlush);
+    LOG(log_.info()) << "Max concurrent creation: " << valueOrDefault(settings.maxConcurrentCreation);
+}
+
+void
+Cluster::setupConnection(Settings const& settings)
+{
+    std::visit(
+        overloadSet{
+            [this](Settings::ContactPoints const& points) { setupContactPoints(points); },
+            [this](Settings::SecureConnectionBundle const& bundle) { setupSecureBundle(bundle); }},
+        settings.connectionInfo);
+}
+
+void
+Cluster::setupContactPoints(Settings::ContactPoints const& points)
+{
+    using std::to_string;
+    auto throwErrorIfNeeded = [](CassError rc, std::string const& label, std::string const& value) {
+        if (rc != CASS_OK)
+            throw std::runtime_error(
+                fmt::format("Cassandra: Error setting {} [{}]: {}", label, value, cass_error_desc(rc)));
+    };
+
+    {
+        LOG(log_.debug()) << "Attempt connection using contact points: " << points.contactPoints;
+        auto const rc = cass_cluster_set_contact_points(*this, points.contactPoints.data());
+        throwErrorIfNeeded(rc, "contact_points", points.contactPoints);
+    }
+
+    if (points.port)
+    {
+        auto const rc = cass_cluster_set_port(*this, points.port.value());
+        throwErrorIfNeeded(rc, "port", to_string(points.port.value()));
+    }
+}
+
+void
+Cluster::setupSecureBundle(Settings::SecureConnectionBundle const& bundle)
+{
+    LOG(log_.debug()) << "Attempt connection using secure bundle";
+    if (auto const rc = cass_cluster_set_cloud_secure_connection_bundle(*this, bundle.bundle.data()); rc != CASS_OK)
+    {
+        throw std::runtime_error("Failed to connect using secure connection bundle " + bundle.bundle);
+    }
+}
+
+void
+Cluster::setupCertificate(Settings const& settings)
+{
+    if (not settings.certificate)
+        return;
+
+    LOG(log_.debug()) << "Configure SSL context";
+    SslContext context = SslContext(*settings.certificate);
+    cass_cluster_set_ssl(*this, context);
+}
+
+void
+Cluster::setupCredentials(Settings const& settings)
+{
+    if (not settings.username || not settings.password)
+        return;
+
+    LOG(log_.debug()) << "Set credentials; username: " << settings.username.value();
+    cass_cluster_set_credentials(*this, settings.username.value().c_str(), settings.password.value().c_str());
+}
+
+}  // namespace data::cassandra::detail

src/data/cassandra/impl/Cluster.h

@@ -0,0 +1,169 @@
+//------------------------------------------------------------------------------
+/*
+    This file is part of clio: https://github.com/XRPLF/clio
+    Copyright (c) 2023, the clio developers.
+
+    Permission to use, copy, modify, and distribute this software for any
+    purpose with or without fee is hereby granted, provided that the above
+    copyright notice and this permission notice appear in all copies.
+
+    THE  SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL WARRANTIES
+    WITH  REGARD  TO  THIS  SOFTWARE  INCLUDING  ALL  IMPLIED  WARRANTIES  OF
+    MERCHANTABILITY  AND  FITNESS. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR
+    ANY  SPECIAL,  DIRECT,  INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES
+    WHATSOEVER  RESULTING  FROM  LOSS  OF USE, DATA OR PROFITS, WHETHER IN AN
+    ACTION  OF  CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF
+    OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE.
+*/
+//==============================================================================
+
+#pragma once
+
+#include <data/cassandra/impl/ManagedObject.h>
+#include <util/log/Logger.h>
+
+#include <cassandra.h>
+
+#include <chrono>
+#include <optional>
+#include <string>
+#include <string_view>
+#include <thread>
+#include <variant>
+
+namespace data::cassandra::detail {
+
+// TODO: move Settings to public interface, not detail
+
+/**
+ * @brief Bundles all cassandra settings in one place.
+ */
+struct Settings
+{
+    /**
+     * @brief Represents the configuration of contact points for cassandra.
+     */
+    struct ContactPoints
+    {
+        std::string contactPoints = "127.0.0.1";  // defaults to localhost
+        std::optional<uint16_t> port;
+    };
+
+    /**
+     * @brief Represents the configuration of a secure connection bundle.
+     */
+    struct SecureConnectionBundle
+    {
+        std::string bundle;  // no meaningful default
+    };
+
+    /** @brief Enables or disables cassandra driver logger */
+    bool enableLog = false;
+
+    /** @brief Connect timeout specified in milliseconds */
+    std::chrono::milliseconds connectionTimeout = std::chrono::milliseconds{10000};
+
+    /** @brief Request timeout specified in milliseconds */
+    std::chrono::milliseconds requestTimeout = std::chrono::milliseconds{0};  // no timeout at all
+
+    /** @brief Connection information; either ContactPoints or SecureConnectionBundle */
+    std::variant<ContactPoints, SecureConnectionBundle> connectionInfo = ContactPoints{};
+
+    /** @brief The number of threads for the driver to pool */
+    uint32_t threads = std::thread::hardware_concurrency();
+
+    /** @brief The maximum number of outstanding write requests at any given moment */
+    uint32_t maxWriteRequestsOutstanding = 10'000;
+
+    /** @brief The maximum number of outstanding read requests at any given moment */
+    uint32_t maxReadRequestsOutstanding = 100'000;
+
+    /** @brief The maximum number of connections per host */
+    uint32_t maxConnectionsPerHost = 2u;
+
+    /** @brief The number of connection per host to always have active */
+    uint32_t coreConnectionsPerHost = 2u;
+
+    /** @brief The maximum concurrent requests per connection; new connections will be created when reached */
+    uint32_t maxConcurrentRequestsThreshold =
+        (maxWriteRequestsOutstanding + maxReadRequestsOutstanding) / coreConnectionsPerHost;
+
+    /** @brief Size of the event queue */
+    std::optional<uint32_t> queueSizeEvent;
+
+    /** @brief Size of the IO queue */
+    std::optional<uint32_t> queueSizeIO;
+
+    /** @brief High watermark for bytes written */
+    std::optional<uint32_t> writeBytesHighWatermark;
+
+    /** @brief Low watermark for bytes written */
+    std::optional<uint32_t> writeBytesLowWatermark;
+
+    /** @brief High watermark for pending requests */
+    std::optional<uint32_t> pendingRequestsHighWatermark;
+
+    /** @brief Low watermark for pending requests */
+    std::optional<uint32_t> pendingRequestsLowWatermark;
+
+    /** @brief Maximum number of requests per flush */
+    std::optional<uint32_t> maxRequestsPerFlush;
+
+    /** @brief Maximum number of connections that will be created concurrently */
+    std::optional<uint32_t> maxConcurrentCreation;
+
+    /** @brief SSL certificate */
+    std::optional<std::string> certificate;  // ssl context
+
+    /** @brief Username/login */
+    std::optional<std::string> username;
+
+    /** @brief Password to match the `username` */
+    std::optional<std::string> password;
+
+    /**
+     * @brief Creates a new Settings object as a copy of the current one with overridden contact points.
+     */
+    Settings
+    withContactPoints(std::string_view contactPoints)
+    {
+        auto tmp = *this;
+        tmp.connectionInfo = ContactPoints{std::string{contactPoints}};
+        return tmp;
+    }
+
+    /**
+     * @brief Returns the default settings.
+     */
+    static Settings
+    defaultSettings()
+    {
+        return Settings();
+    }
+};
+
+class Cluster : public ManagedObject<CassCluster>
+{
+    util::Logger log_{"Backend"};
+
+public:
+    Cluster(Settings const& settings);
+
+private:
+    void
+    setupConnection(Settings const& settings);
+
+    void
+    setupContactPoints(Settings::ContactPoints const& points);
+
+    void
+    setupSecureBundle(Settings::SecureConnectionBundle const& bundle);
+
+    void
+    setupCertificate(Settings const& settings);
+
+    void
+    setupCredentials(Settings const& settings);
+};
+
+}  // namespace data::cassandra::detail

src/data/cassandra/impl/ExecutionStrategy.h

@@ -0,0 +1,459 @@
+//------------------------------------------------------------------------------
+/*
+    This file is part of clio: https://github.com/XRPLF/clio
+    Copyright (c) 2023, the clio developers.
+
+    Permission to use, copy, modify, and distribute this software for any
+    purpose with or without fee is hereby granted, provided that the above
+    copyright notice and this permission notice appear in all copies.
+
+    THE  SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL WARRANTIES
+    WITH  REGARD  TO  THIS  SOFTWARE  INCLUDING  ALL  IMPLIED  WARRANTIES  OF
+    MERCHANTABILITY  AND  FITNESS. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR
+    ANY  SPECIAL,  DIRECT,  INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES
+    WHATSOEVER  RESULTING  FROM  LOSS  OF USE, DATA OR PROFITS, WHETHER IN AN
+    ACTION  OF  CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF
+    OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE.
+*/
+//==============================================================================
+
+#pragma once
+
+#include <data/cassandra/Handle.h>
+#include <data/cassandra/Types.h>
+#include <data/cassandra/impl/AsyncExecutor.h>
+#include <util/Expected.h>
+#include <util/log/Logger.h>
+
+#include <boost/asio.hpp>
+#include <boost/asio/spawn.hpp>
+
+#include <atomic>
+#include <condition_variable>
+#include <functional>
+#include <memory>
+#include <mutex>
+#include <optional>
+#include <thread>
+
+namespace data::cassandra::detail {
+
+// TODO: this could probably be also moved out of detail and into the main cassandra namespace.
+
+/**
+ * @brief Implements async and sync querying against the cassandra DB with support for throttling.
+ *
+ * Note: A lot of the code that uses yield is repeated below.
+ * This is ok for now because we are hopefully going to be getting rid of it entirely later on.
+ */
+template <typename HandleType = Handle>
+class DefaultExecutionStrategy
+{
+    util::Logger log_{"Backend"};
+
+    std::uint32_t maxWriteRequestsOutstanding_;
+    std::atomic_uint32_t numWriteRequestsOutstanding_ = 0;
+
+    std::uint32_t maxReadRequestsOutstanding_;
+    std::atomic_uint32_t numReadRequestsOutstanding_ = 0;
+
+    std::mutex throttleMutex_;
+    std::condition_variable throttleCv_;
+
+    std::mutex syncMutex_;
+    std::condition_variable syncCv_;
+
+    boost::asio::io_context ioc_;
+    std::optional<boost::asio::io_service::work> work_;
+
+    std::reference_wrapper<HandleType const> handle_;
+    std::thread thread_;
+
+public:
+    using ResultOrErrorType = typename HandleType::ResultOrErrorType;
+    using StatementType = typename HandleType::StatementType;
+    using PreparedStatementType = typename HandleType::PreparedStatementType;
+    using FutureType = typename HandleType::FutureType;
+    using FutureWithCallbackType = typename HandleType::FutureWithCallbackType;
+    using ResultType = typename HandleType::ResultType;
+    using CompletionTokenType = boost::asio::yield_context;
+
+    /**
+     * @param settings The settings to use
+     * @param handle A handle to the cassandra database
+     */
+    DefaultExecutionStrategy(Settings const& settings, HandleType const& handle)
+        : maxWriteRequestsOutstanding_{settings.maxWriteRequestsOutstanding}
+        , maxReadRequestsOutstanding_{settings.maxReadRequestsOutstanding}
+        , work_{ioc_}
+        , handle_{std::cref(handle)}
+        , thread_{[this]() { ioc_.run(); }}
+    {
+        LOG(log_.info()) << "Max write requests outstanding is " << maxWriteRequestsOutstanding_
+                         << "; Max read requests outstanding is " << maxReadRequestsOutstanding_;
+    }
+
+    ~DefaultExecutionStrategy()
+    {
+        work_.reset();
+        ioc_.stop();
+        thread_.join();
+    }
+
+    /**
+     * @brief Wait for all async writes to finish before unblocking.
+     */
+    void
+    sync()
+    {
+        LOG(log_.debug()) << "Waiting to sync all writes...";
+        std::unique_lock<std::mutex> lck(syncMutex_);
+        syncCv_.wait(lck, [this]() { return finishedAllWriteRequests(); });
+        LOG(log_.debug()) << "Sync done.";
+    }
+
+    /**
+     * @return true if outstanding read requests allowance is exhausted; false otherwise
+     */
+    bool
+    isTooBusy() const
+    {
+        return numReadRequestsOutstanding_ >= maxReadRequestsOutstanding_;
+    }
+
+    /**
+     * @brief Blocking query execution used for writing data.
+     *
+     * Retries forever sleeping for 5 milliseconds between attempts.
+     */
+    ResultOrErrorType
+    writeSync(StatementType const& statement)
+    {
+        while (true)
+        {
+            if (auto res = handle_.get().execute(statement); res)
+            {
+                return res;
+            }
+            else
+            {
+                LOG(log_.warn()) << "Cassandra sync write error, retrying: " << res.error();
+                std::this_thread::sleep_for(std::chrono::milliseconds(5));
+            }
+        }
+    }
+
+    /**
+     * @brief Blocking query execution used for writing data.
+     *
+     * Retries forever sleeping for 5 milliseconds between attempts.
+     */
+    template <typename... Args>
+    ResultOrErrorType
+    writeSync(PreparedStatementType const& preparedStatement, Args&&... args)
+    {
+        return writeSync(preparedStatement.bind(std::forward<Args>(args)...));
+    }
+
+    /**
+     * @brief Non-blocking query execution used for writing data.
+     *
+     * Retries forever with retry policy specified by @ref AsyncExecutor
+     *
+     * @param preparedStatement Statement to prepare and execute
+     * @param args Args to bind to the prepared statement
+     * @throw DatabaseTimeout on timeout
+     */
+    template <typename... Args>
+    void
+    write(PreparedStatementType const& preparedStatement, Args&&... args)
+    {
+        auto statement = preparedStatement.bind(std::forward<Args>(args)...);
+        incrementOutstandingRequestCount();
+
+        // Note: lifetime is controlled by std::shared_from_this internally
+        AsyncExecutor<std::decay_t<decltype(statement)>, HandleType>::run(
+            ioc_, handle_, std::move(statement), [this](auto const&) { decrementOutstandingRequestCount(); });
+    }
+
+    /**
+     * @brief Non-blocking batched query execution used for writing data.
+     *
+     * Retries forever with retry policy specified by @ref AsyncExecutor.
+     *
+     * @param statements Vector of statements to execute as a batch
+     * @throw DatabaseTimeout on timeout
+     */
+    void
+    write(std::vector<StatementType>&& statements)
+    {
+        if (statements.empty())
+            return;
+
+        incrementOutstandingRequestCount();
+
+        // Note: lifetime is controlled by std::shared_from_this internally
+        AsyncExecutor<std::decay_t<decltype(statements)>, HandleType>::run(
+            ioc_, handle_, std::move(statements), [this](auto const&) { decrementOutstandingRequestCount(); });
+    }
+
+    /**
+     * @brief Coroutine-based query execution used for reading data.
+     *
+     * Retries forever until successful or throws an exception on timeout.
+     *
+     * @param token Completion token (yield_context)
+     * @param preparedStatement Statement to prepare and execute
+     * @param args Args to bind to the prepared statement
+     * @throw DatabaseTimeout on timeout
+     * @return ResultType or error wrapped in Expected
+     */
+    template <typename... Args>
+    [[maybe_unused]] ResultOrErrorType
+    read(CompletionTokenType token, PreparedStatementType const& preparedStatement, Args&&... args)
+    {
+        return read(token, preparedStatement.bind(std::forward<Args>(args)...));
+    }
+
+    /**
+     * @brief Coroutine-based query execution used for reading data.
+     *
+     * Retries forever until successful or throws an exception on timeout.
+     *
+     * @param token Completion token (yield_context)
+     * @param statements Statements to execute in a batch
+     * @throw DatabaseTimeout on timeout
+     * @return ResultType or error wrapped in Expected
+     */
+    [[maybe_unused]] ResultOrErrorType
+    read(CompletionTokenType token, std::vector<StatementType> const& statements)
+    {
+        auto const numStatements = statements.size();
+        std::optional<FutureWithCallbackType> future;
+
+        // todo: perhaps use policy instead
+        while (true)
+        {
+            numReadRequestsOutstanding_ += numStatements;
+            // TODO: see if we can avoid using shared_ptr for self here
+            auto init = [this, &statements, &future]<typename Self>(Self& self) {
+                future.emplace(handle_.get().asyncExecute(
+                    statements, [sself = std::make_shared<Self>(std::move(self))](auto&& res) mutable {
+                        // Note: explicit work below needed on linux/gcc11
+                        auto executor = boost::asio::get_associated_executor(*sself);
+                        boost::asio::post(
+                            executor,
+                            [sself = std::move(sself),
+                             res = std::move(res),
+                             _ = boost::asio::make_work_guard(executor)]() mutable {
+                                sself->complete(std::move(res));
+                                sself.reset();
+                            });
+                    }));
+            };
+
+            auto res = boost::asio::async_compose<CompletionTokenType, void(ResultOrErrorType)>(
+                init, token, boost::asio::get_associated_executor(token));
+            numReadRequestsOutstanding_ -= numStatements;
+
+            if (res)
+            {
+                return res;
+            }
+            else
+            {
+                LOG(log_.error()) << "Failed batch read in coroutine: " << res.error();
+                throwErrorIfNeeded(res.error());
+            }
+        }
+    }
+
+    /**
+     * @brief Coroutine-based query execution used for reading data.
+     *
+     * Retries forever until successful or throws an exception on timeout.
+     *
+     * @param token Completion token (yield_context)
+     * @param statement Statement to execute
+     * @throw DatabaseTimeout on timeout
+     * @return ResultType or error wrapped in Expected
+     */
+    [[maybe_unused]] ResultOrErrorType
+    read(CompletionTokenType token, StatementType const& statement)
+    {
+        std::optional<FutureWithCallbackType> future;
+
+        // todo: perhaps use policy instead
+        while (true)
+        {
+            ++numReadRequestsOutstanding_;
+            // TODO: see if we can avoid using shared_ptr for self here
+            auto init = [this, &statement, &future]<typename Self>(Self& self) {
+                future.emplace(handle_.get().asyncExecute(
+                    statement, [sself = std::make_shared<Self>(std::move(self))](auto&&) mutable {
+                        // Note: explicit work below needed on linux/gcc11
+                        auto executor = boost::asio::get_associated_executor(*sself);
+                        boost::asio::post(
+                            executor, [sself = std::move(sself), _ = boost::asio::make_work_guard(executor)]() mutable {
+                                sself->complete();
+                                sself.reset();
+                            });
+                    }));
+            };
+
+            boost::asio::async_compose<CompletionTokenType, void()>(
+                init, token, boost::asio::get_associated_executor(token));
+            --numReadRequestsOutstanding_;
+
+            if (auto res = future->get(); res)
+            {
+                return res;
+            }
+            else
+            {
+                LOG(log_.error()) << "Failed read in coroutine: " << res.error();
+                throwErrorIfNeeded(res.error());
+            }
+        }
+    }
+
+    /**
+     * @brief Coroutine-based query execution used for reading data.
+     *
+     * Attempts to execute each statement. On any error the whole vector will be
+     * discarded and exception will be thrown.
+     *
+     * @param token Completion token (yield_context)
+     * @param statements Statements to execute
+     * @throw DatabaseTimeout on db error
+     * @return Vector of results
+     */
+    std::vector<ResultType>
+    readEach(CompletionTokenType token, std::vector<StatementType> const& statements)
+    {
+        std::atomic_bool hadError = false;
+        std::atomic_int numOutstanding = statements.size();
+        numReadRequestsOutstanding_ += statements.size();
+
+        auto futures = std::vector<FutureWithCallbackType>{};
+        futures.reserve(numOutstanding);
+
+        auto init = [this, &statements, &futures, &hadError, &numOutstanding]<typename Self>(Self& self) {
+            auto sself = std::make_shared<Self>(std::move(self));  // TODO: see if we can avoid this
+            auto executionHandler = [&hadError, &numOutstanding, sself = std::move(sself)](auto const& res) mutable {
+                if (not res)
+                    hadError = true;
+
+                // when all async operations complete unblock the result
+                if (--numOutstanding == 0)
+                {
+                    // Note: explicit work below needed on linux/gcc11
+                    auto executor = boost::asio::get_associated_executor(*sself);
+                    boost::asio::post(
+                        executor, [sself = std::move(sself), _ = boost::asio::make_work_guard(executor)]() mutable {
+                            sself->complete();
+                            sself.reset();
+                        });
+                }
+            };
+
+            std::transform(
+                std::cbegin(statements),
+                std::cend(statements),
+                std::back_inserter(futures),
+                [this, &executionHandler](auto const& statement) {
+                    return handle_.get().asyncExecute(statement, executionHandler);
+                });
+        };
+
+        boost::asio::async_compose<CompletionTokenType, void()>(
+            init, token, boost::asio::get_associated_executor(token));
+        numReadRequestsOutstanding_ -= statements.size();
+
+        if (hadError)
+            throw DatabaseTimeout{};
+
+        std::vector<ResultType> results;
+        results.reserve(futures.size());
+
+        // it's safe to call blocking get on futures here as we already waited for the coroutine to resume above.
+        std::transform(
+            std::make_move_iterator(std::begin(futures)),
+            std::make_move_iterator(std::end(futures)),
+            std::back_inserter(results),
+            [](auto&& future) {
+                auto entry = future.get();
+                auto&& res = entry.value();
+                return std::move(res);
+            });
+
+        assert(futures.size() == statements.size());
+        assert(results.size() == statements.size());
+        return results;
+    }
+
+private:
+    void
+    incrementOutstandingRequestCount()
+    {
+        {
+            std::unique_lock<std::mutex> lck(throttleMutex_);
+            if (!canAddWriteRequest())
+            {
+                LOG(log_.trace()) << "Max outstanding requests reached. "
+                                  << "Waiting for other requests to finish";
+                throttleCv_.wait(lck, [this]() { return canAddWriteRequest(); });
+            }
+        }
+        ++numWriteRequestsOutstanding_;
+    }
+
+    void
+    decrementOutstandingRequestCount()
+    {
+        // sanity check
+        if (numWriteRequestsOutstanding_ == 0)
+        {
+            assert(false);
+            throw std::runtime_error("decrementing num outstanding below 0");
+        }
+        size_t cur = (--numWriteRequestsOutstanding_);
+        {
+            // mutex lock required to prevent race condition around spurious
+            // wakeup
+            std::lock_guard lck(throttleMutex_);
+            throttleCv_.notify_one();
+        }
+        if (cur == 0)
+        {
+            // mutex lock required to prevent race condition around spurious
+            // wakeup
+            std::lock_guard lck(syncMutex_);
+            syncCv_.notify_one();
+        }
+    }
+
+    bool
+    canAddWriteRequest() const
+    {
+        return numWriteRequestsOutstanding_ < maxWriteRequestsOutstanding_;
+    }
+
+    bool
+    finishedAllWriteRequests() const
+    {
+        return numWriteRequestsOutstanding_ == 0;
+    }
+
+    void
+    throwErrorIfNeeded(CassandraError err) const
+    {
+        if (err.isTimeout())
+            throw DatabaseTimeout();
+
+        if (err.isInvalidQuery())
+            throw std::runtime_error("Invalid query");
+    }
+};
+
+}  // namespace data::cassandra::detail

src/data/cassandra/impl/Future.cpp

@@ -0,0 +1,102 @@
+//------------------------------------------------------------------------------
+/*
+    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 <data/cassandra/Error.h>
+#include <data/cassandra/impl/Future.h>
+#include <data/cassandra/impl/Result.h>
+
+#include <exception>
+#include <vector>
+
+namespace {
+static constexpr auto futureDeleter = [](CassFuture* ptr) { cass_future_free(ptr); };
+}  // namespace
+
+namespace data::cassandra::detail {
+
+/* implicit */ Future::Future(CassFuture* ptr) : ManagedObject{ptr, futureDeleter}
+{
+}
+
+MaybeError
+Future::await() const
+{
+    if (auto const rc = cass_future_error_code(*this); rc)
+    {
+        auto errMsg = [this](std::string const& label) {
+            char const* message;
+            std::size_t len;
+            cass_future_error_message(*this, &message, &len);
+            return label + ": " + std::string{message, len};
+        }(cass_error_desc(rc));
+        return Error{CassandraError{errMsg, rc}};
+    }
+    return {};
+}
+
+ResultOrError
+Future::get() const
+{
+    if (auto const rc = cass_future_error_code(*this); rc)
+    {
+        auto const errMsg = [this](std::string const& label) {
+            char const* message;
+            std::size_t len;
+            cass_future_error_message(*this, &message, &len);
+            return label + ": " + std::string{message, len};
+        }("future::get()");
+        return Error{CassandraError{errMsg, rc}};
+    }
+    else
+    {
+        return Result{cass_future_get_result(*this)};
+    }
+}
+
+void
+invokeHelper(CassFuture* ptr, void* cbPtr)
+{
+    // Note: can't use Future{ptr}.get() because double free will occur :/
+    auto* cb = static_cast<FutureWithCallback::FnType*>(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, FnType&& cb)
+    : Future{ptr}, cb_{std::make_unique<FnType>(std::move(cb))}
+{
+    // Instead of passing `this` as the userdata void*, we pass the address of
+    // the callback itself which will survive std::move of the
+    // FutureWithCallback parent. Not ideal but I have no better solution atm.
+    cass_future_set_callback(*this, &invokeHelper, cb_.get());
+}
+
+}  // namespace data::cassandra::detail

src/data/cassandra/impl/Future.h

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

src/data/cassandra/impl/ManagedObject.h

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

src/data/cassandra/impl/Result.cpp

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

src/data/cassandra/impl/Result.h

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

src/data/cassandra/impl/RetryPolicy.h

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

src/data/cassandra/impl/Session.h

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

src/data/cassandra/impl/SslContext.cpp

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

src/data/cassandra/impl/SslContext.h

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

src/data/cassandra/impl/Statement.h

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

src/data/cassandra/impl/Tuple.cpp

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

src/data/cassandra/impl/Tuple.h

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

src/etl/ETLHelpers.h

@@ -1,5 +1,25 @@
-#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.
+*/
+//==============================================================================
+
+/** @file */
+#pragma once
+
 #include <ripple/basics/base_uint.h>
 #include <condition_variable>
 #include <mutex>
@@ -7,33 +27,38 @@
 #include <queue>
 #include <sstream>
 
-/// This datastructure is used to keep track of the sequence of the most recent
-/// ledger validated by the network. There are two methods that will wait until
-/// certain conditions are met. This datastructure is able to be "stopped". When
-/// the datastructure is stopped, any threads currently waiting are unblocked.
-/// Any later calls to methods of this datastructure will not wait. Once the
-/// datastructure is stopped, the datastructure remains stopped for the rest of
-/// its lifetime.
+namespace etl {
+/**
+ * @brief This datastructure is used to keep track of the sequence of the most recent ledger validated by the network.
+ *
+ * There are two methods that will wait until certain conditions are met. This datastructure is able to be "stopped".
+ * When the datastructure is stopped, any threads currently waiting are unblocked.
+ * Any later calls to methods of this datastructure will not wait. Once the datastructure is stopped, the datastructure
+ * remains stopped for the rest of its lifetime.
+ */
 class NetworkValidatedLedgers
 {
     // max sequence validated by network
     std::optional<uint32_t> max_;
 
     mutable std::mutex m_;
-
     std::condition_variable cv_;
 
-    bool stopping_ = false;
-
 public:
+    /**
+     * @brief A factory function for NetworkValidatedLedgers.
+     */
     static std::shared_ptr<NetworkValidatedLedgers>
     make_ValidatedLedgers()
     {
         return std::make_shared<NetworkValidatedLedgers>();
     }
 
-    /// Notify the datastructure that idx has been validated by the network
-    /// @param idx sequence validated by network
+    /**
+     * @brief Notify the datastructure that idx has been validated by the network.
+     *
+     * @param idx Sequence validated by network
+     */
     void
     push(uint32_t idx)
     {
@@ -43,10 +68,13 @@ public:
         cv_.notify_all();
     }
 
-    /// Get most recently validated sequence. If no ledgers are known to have
-    /// been validated, this function waits until the next ledger is validated
-    /// @return sequence of most recently validated ledger. empty optional if
-    /// the datastructure has been stopped
+    /**
+     * @brief Get most recently validated sequence.
+     *
+     * If no ledgers are known to have been validated, this function waits until the next ledger is validated
+     *
+     * @return Sequence of most recently validated ledger. empty optional if the datastructure has been stopped
+     */
     std::optional<uint32_t>
     getMostRecent()
     {
@@ -55,19 +83,18 @@ public:
         return max_;
     }
 
-    /// Waits for the sequence to be validated by the network
-    /// @param sequence to wait for
-    /// @return true if sequence was validated, false otherwise
-    /// a return value of false means the datastructure has been stopped
+    /**
+     * @brief Waits for the sequence to be validated by the network.
+     *
+     * @param sequence The sequence to wait for
+     * @return true if sequence was validated, false otherwise a return value of false means the datastructure has been
+     * stopped
+     */
     bool
-    waitUntilValidatedByNetwork(
-        uint32_t sequence,
-        std::optional<uint32_t> maxWaitMs = {})
+    waitUntilValidatedByNetwork(uint32_t sequence, std::optional<uint32_t> maxWaitMs = {})
     {
         std::unique_lock lck(m_);
-        auto pred = [sequence, this]() -> bool {
-            return (max_ && sequence <= *max_);
-        };
+        auto pred = [sequence, this]() -> bool { return (max_ && sequence <= *max_); };
         if (maxWaitMs)
             cv_.wait_for(lck, std::chrono::milliseconds(*maxWaitMs));
         else
@@ -76,10 +103,13 @@ public:
     }
 };
 
-/// Generic thread-safe queue with an optional maximum size
-/// Note, we can't use a lockfree queue here, since we need the ability to wait
-/// for an element to be added or removed from the queue. These waits are
-/// blocking calls.
+// TODO: does the note make sense? lockfree queues provide the same blocking behaviour just without mutex, don't they?
+/**
+ * @brief Generic thread-safe queue with a max capacity.
+ *
+ * @note (original note) We can't use a lockfree queue here, since we need the ability to wait for an element to be
+ * added or removed from the queue. These waits are blocking calls.
+ */
 template <class T>
 class ThreadSafeQueue
 {
@@ -87,75 +117,96 @@ class ThreadSafeQueue
 
     mutable std::mutex m_;
     std::condition_variable cv_;
-    std::optional<uint32_t> maxSize_;
+    uint32_t maxSize_;
 
 public:
-    /// @param maxSize maximum size of the queue. Calls that would cause the
-    /// queue to exceed this size will block until free space is available
+    /**
+     * @brief Create an instance of the queue.
+     *
+     * @param maxSize maximum size of the queue. Calls that would cause the queue to exceed this size will block until
+     * free space is available.
+     */
     ThreadSafeQueue(uint32_t maxSize) : maxSize_(maxSize)
     {
     }
 
-    /// Create a queue with no maximum size
-    ThreadSafeQueue() = default;
-
-    /// @param elt element to push onto queue
-    /// if maxSize is set, this method will block until free space is available
+    /**
+     * @brief Push element onto the queue.
+     *
+     * Note: This method will block until free space is available.
+     *
+     * @param elt Element to push onto queue
+     */
     void
     push(T const& elt)
     {
         std::unique_lock lck(m_);
-        // if queue has a max size, wait until not full
-        if (maxSize_)
-            cv_.wait(lck, [this]() { return queue_.size() <= *maxSize_; });
+        cv_.wait(lck, [this]() { return queue_.size() <= maxSize_; });
         queue_.push(elt);
         cv_.notify_all();
     }
 
-    /// @param elt element to push onto queue. elt is moved from
-    /// if maxSize is set, this method will block until free space is available
+    /**
+     * @brief Push element onto the queue.
+     *
+     * Note: This method will block until free space is available
+     *
+     * @param elt Element to push onto queue. Ownership is transferred
+     */
     void
     push(T&& elt)
     {
         std::unique_lock lck(m_);
-        // if queue has a max size, wait until not full
-        if (maxSize_)
-            cv_.wait(lck, [this]() { return queue_.size() <= *maxSize_; });
+        cv_.wait(lck, [this]() { return queue_.size() <= maxSize_; });
         queue_.push(std::move(elt));
         cv_.notify_all();
     }
 
-    /// @return element popped from queue. Will block until queue is non-empty
+    /**
+     * @brief Pop element from the queue.
+     *
+     * Note: Will block until queue is non-empty.
+     *
+     * @return Element popped from queue
+     */
     T
     pop()
     {
         std::unique_lock lck(m_);
         cv_.wait(lck, [this]() { return !queue_.empty(); });
+
         T ret = std::move(queue_.front());
         queue_.pop();
-        // if queue has a max size, unblock any possible pushers
-        if (maxSize_)
-            cv_.notify_all();
+
+        cv_.notify_all();
         return ret;
     }
-    /// @return element popped from queue. Will block until queue is non-empty
+
+    /**
+     * @brief Attempt to pop an element.
+     *
+     * @return Element popped from queue or empty optional if queue was empty
+     */
     std::optional<T>
     tryPop()
     {
-        std::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 +224,4 @@ getMarkers(size_t numMarkers)
     }
     return markers;
 }
-
-#endif  // RIPPLE_APP_REPORTING_ETLHELPERS_H_INCLUDED
+}  // namespace etl

src/etl/ETLService.cpp

@@ -0,0 +1,269 @@
+//------------------------------------------------------------------------------
+/*
+    This file is part of clio: https://github.com/XRPLF/clio
+    Copyright (c) 2023, the clio developers.
+
+    Permission to use, copy, modify, and distribute this software for any
+    purpose with or without fee is hereby granted, provided that the above
+    copyright notice and this permission notice appear in all copies.
+
+    THE  SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL WARRANTIES
+    WITH  REGARD  TO  THIS  SOFTWARE  INCLUDING  ALL  IMPLIED  WARRANTIES  OF
+    MERCHANTABILITY  AND  FITNESS. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR
+    ANY  SPECIAL,  DIRECT,  INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES
+    WHATSOEVER  RESULTING  FROM  LOSS  OF USE, DATA OR PROFITS, WHETHER IN AN
+    ACTION  OF  CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF
+    OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE.
+*/
+//==============================================================================
+
+#include <etl/ETLService.h>
+
+#include <ripple/protocol/LedgerHeader.h>
+
+namespace etl {
+// Database must be populated when this starts
+std::optional<uint32_t>
+ETLService::runETLPipeline(uint32_t startSequence, uint32_t numExtractors)
+{
+    if (finishSequence_ && startSequence > *finishSequence_)
+        return {};
+
+    LOG(log_.debug()) << "Starting etl pipeline";
+    state_.isWriting = true;
+
+    auto rng = backend_->hardFetchLedgerRangeNoThrow();
+    if (!rng || rng->maxSequence < startSequence - 1)
+    {
+        assert(false);
+        throw std::runtime_error("runETLPipeline: parent ledger is null");
+    }
+
+    auto const begin = std::chrono::system_clock::now();
+    auto extractors = std::vector<std::unique_ptr<ExtractorType>>{};
+    auto pipe = DataPipeType{numExtractors, startSequence};
+
+    for (auto i = 0u; i < numExtractors; ++i)
+        extractors.push_back(std::make_unique<ExtractorType>(
+            pipe, networkValidatedLedgers_, ledgerFetcher_, startSequence + i, finishSequence_, state_));
+
+    auto transformer = TransformerType{pipe, backend_, ledgerLoader_, ledgerPublisher_, startSequence, state_};
+    transformer.waitTillFinished();  // suspend current thread until exit condition is met
+    pipe.cleanup();                  // TODO: this should probably happen automatically using destructor
+
+    // wait for all of the extractors to stop
+    for (auto& t : extractors)
+        t->waitTillFinished();
+
+    auto const end = std::chrono::system_clock::now();
+    auto const lastPublishedSeq = ledgerPublisher_.getLastPublishedSequence();
+    LOG(log_.debug()) << "Extracted and wrote " << lastPublishedSeq.value_or(startSequence) - startSequence << " in "
+                      << ((end - begin).count()) / 1000000000.0;
+
+    state_.isWriting = false;
+
+    LOG(log_.debug()) << "Stopping etl pipeline";
+    return lastPublishedSeq;
+}
+
+// Main loop of ETL.
+// The software begins monitoring the ledgers that are validated by the nework.
+// The member networkValidatedLedgers_ keeps track of the sequences of ledgers validated by the network.
+// Whenever a ledger is validated by the network, the software looks for that ledger in the database. Once the ledger is
+// found in the database, the software publishes that ledger to the ledgers stream. If a network validated ledger is not
+// found in the database after a certain amount of time, then the software attempts to take over responsibility of the
+// ETL process, where it writes new ledgers to the database. The software will relinquish control of the ETL process if
+// it detects that another process has taken over ETL.
+void
+ETLService::monitor()
+{
+    auto rng = backend_->hardFetchLedgerRangeNoThrow();
+    if (!rng)
+    {
+        LOG(log_.info()) << "Database is empty. Will download a ledger from the network.";
+        std::optional<ripple::LedgerHeader> ledger;
+
+        try
+        {
+            if (startSequence_)
+            {
+                LOG(log_.info()) << "ledger sequence specified in config. "
+                                 << "Will begin ETL process starting with ledger " << *startSequence_;
+                ledger = ledgerLoader_.loadInitialLedger(*startSequence_);
+            }
+            else
+            {
+                LOG(log_.info()) << "Waiting for next ledger to be validated by network...";
+                std::optional<uint32_t> mostRecentValidated = networkValidatedLedgers_->getMostRecent();
+
+                if (mostRecentValidated)
+                {
+                    LOG(log_.info()) << "Ledger " << *mostRecentValidated << " has been validated. Downloading...";
+                    ledger = ledgerLoader_.loadInitialLedger(*mostRecentValidated);
+                }
+                else
+                {
+                    LOG(log_.info()) << "The wait for the next validated ledger has been aborted. Exiting monitor loop";
+                    return;
+                }
+            }
+        }
+        catch (std::runtime_error const& e)
+        {
+            setAmendmentBlocked();
+
+            log_.fatal()
+                << "Failed to load initial ledger, Exiting monitor loop: " << e.what()
+                << " Possible cause: The ETL node is not compatible with the version of the rippled lib Clio is using.";
+            return;
+        }
+
+        if (ledger)
+        {
+            rng = backend_->hardFetchLedgerRangeNoThrow();
+        }
+        else
+        {
+            LOG(log_.error()) << "Failed to load initial ledger. Exiting monitor loop";
+            return;
+        }
+    }
+    else
+    {
+        if (startSequence_)
+            LOG(log_.warn()) << "start sequence specified but db is already populated";
+
+        LOG(log_.info()) << "Database already populated. Picking up from the tip of history";
+        cacheLoader_.load(rng->maxSequence);
+    }
+
+    assert(rng);
+    uint32_t nextSequence = rng->maxSequence + 1;
+
+    LOG(log_.debug()) << "Database is populated. "
+                      << "Starting monitor loop. sequence = " << nextSequence;
+
+    while (true)
+    {
+        if (auto rng = backend_->hardFetchLedgerRangeNoThrow(); rng && rng->maxSequence >= nextSequence)
+        {
+            ledgerPublisher_.publish(nextSequence, {});
+            ++nextSequence;
+        }
+        else if (networkValidatedLedgers_->waitUntilValidatedByNetwork(nextSequence, 1000))
+        {
+            LOG(log_.info()) << "Ledger with sequence = " << nextSequence << " has been validated by the network. "
+                             << "Attempting to find in database and publish";
+
+            // Attempt to take over responsibility of ETL writer after 10 failed
+            // attempts to publish the ledger. publishLedger() fails if the
+            // ledger that has been validated by the network is not found in the
+            // database after the specified number of attempts. publishLedger()
+            // waits one second between each attempt to read the ledger from the
+            // database
+            constexpr size_t timeoutSeconds = 10;
+            bool success = ledgerPublisher_.publish(nextSequence, timeoutSeconds);
+
+            if (!success)
+            {
+                LOG(log_.warn()) << "Failed to publish ledger with sequence = " << nextSequence << " . Beginning ETL";
+
+                // returns the most recent sequence published empty optional if no sequence was published
+                std::optional<uint32_t> lastPublished = runETLPipeline(nextSequence, extractorThreads_);
+                LOG(log_.info()) << "Aborting ETL. Falling back to publishing";
+
+                // if no ledger was published, don't increment nextSequence
+                if (lastPublished)
+                    nextSequence = *lastPublished + 1;
+            }
+            else
+            {
+                ++nextSequence;
+            }
+        }
+    }
+}
+
+void
+ETLService::monitorReadOnly()
+{
+    LOG(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(log_.info()) << "Starting reporting etl";
+    state_.isStopping = false;
+
+    doWork();
+}
+
+void
+ETLService::doWork()
+{
+    worker_ = std::thread([this]() {
+        beast::setCurrentThreadName("ETLService worker");
+
+        if (state_.isReadOnly)
+            monitorReadOnly();
+        else
+            monitor();
+    });
+}
+
+ETLService::ETLService(
+    util::Config const& config,
+    boost::asio::io_context& ioc,
+    std::shared_ptr<BackendInterface> backend,
+    std::shared_ptr<SubscriptionManagerType> subscriptions,
+    std::shared_ptr<LoadBalancerType> balancer,
+    std::shared_ptr<NetworkValidatedLedgersType> ledgers)
+    : backend_(backend)
+    , loadBalancer_(balancer)
+    , networkValidatedLedgers_(ledgers)
+    , cacheLoader_(config, ioc, backend, backend->cache())
+    , ledgerFetcher_(backend, balancer)
+    , ledgerLoader_(backend, balancer, ledgerFetcher_, state_)
+    , ledgerPublisher_(ioc, backend, subscriptions, state_)
+{
+    startSequence_ = config.maybeValue<uint32_t>("start_sequence");
+    finishSequence_ = config.maybeValue<uint32_t>("finish_sequence");
+    state_.isReadOnly = config.valueOr("read_only", state_.isReadOnly);
+    extractorThreads_ = config.valueOr<uint32_t>("extractor_threads", extractorThreads_);
+    txnThreshold_ = config.valueOr<size_t>("txn_threshold", txnThreshold_);
+}
+}  // namespace etl