[Integration] Adding ART to the benchmark (#38)

* ART integration - work in progress

* art fixes

* fixing execute

* fixed art

* clean up code

* merge from master

* fixed minor bug

---------

Co-authored-by: Lucas <[email protected]>

.gitignore

@@ -189,15 +189,19 @@ cython_debug/
 .cache
 build
 
-# IDE
+# IDE - VSCode
 .vscode/
 script/infra/__pycache__
 results.db
 data/*.data
 *.bin
 
+# IDE - IntelliJ
 .idea
 cmake-build-debug/
 src/bliss/.idea/
 
+# OS - MacOS
 db_working_home
+
+

CMakeLists.txt

@@ -52,6 +52,7 @@ add_library(bliss OBJECT
     ${CMAKE_SOURCE_DIR}/src/bliss/bench_alex.h
     ${CMAKE_SOURCE_DIR}/src/bliss/bench_btree.h
     ${CMAKE_SOURCE_DIR}/src/bliss/bench_pgm.h
+    ${CMAKE_SOURCE_DIR}/src/bliss/bench_art.h
 )
 
 target_compile_features(bliss PUBLIC
@@ -65,6 +66,7 @@ target_link_libraries(bliss PUBLIC
     lipp
     tlx
     pgm
+    art
 )
 
 target_include_directories(bliss PUBLIC

external/CMakeLists.txt

@@ -72,6 +72,19 @@ endif()
 add_library(tlx INTERFACE)
 target_include_directories(tlx INTERFACE ${tlx_SOURCE_DIR}/)
 
+# Adaptive Radix Tree 
+FetchContent_Declare(
+    art
+    GIT_REPOSITORY https://github.com/BU-DiSC/ART
+    GIT_TAG main
+)
+FetchContent_GetProperties(art)
+if (NOT art_POPULATED)
+    FetchContent_Populate(art)
+endif()
+
+add_library(art INTERFACE)
+target_include_directories(art INTERFACE ${art_SOURCE_DIR})
 
 
 
@@ -86,4 +99,4 @@ if (NOT pgm_POPULATED)
 endif()
 
 add_library(pgm INTERFACE)
-target_include_directories(pgm INTERFACE ${pgm_SOURCE_DIR})
+target_include_directories(pgm INTERFACE ${pgm_SOURCE_DIR})

src/bliss/bench_art.h

@@ -0,0 +1,46 @@
+#ifndef BLISS_BENCH_ART
+#define BLISS_BENCH_ART
+
+#include <vector>
+
+#include "ART.h"
+#include "bliss/bliss_index.h"
+#include "spdlog/spdlog.h"
+
+namespace bliss {
+
+template <typename KEY_TYPE, typename VALUE_TYPE>
+class BlissARTIndex : public BlissIndex<KEY_TYPE, VALUE_TYPE> {
+   public:
+    ART::ArtNode* _index;
+    static constexpr size_t KEY_SIZE = sizeof(KEY_TYPE);
+    BlissARTIndex() { _index = nullptr; };
+
+    void bulkload(
+        std::vector<std::pair<KEY_TYPE, VALUE_TYPE>> values) override {
+        // expects the pairs to be pre-sorted before performing bulk load
+        for (const auto pair : values) {
+            put(pair.first, pair.second);
+        }
+    }
+
+    bool get(KEY_TYPE key) override {
+        uint8_t ARTkey[KEY_SIZE];
+        ART::loadKey(key, ARTkey);
+        ART::ArtNode* leaf =
+            ART::lookup(_index, ARTkey, KEY_SIZE, 0, KEY_SIZE);
+        return leaf != nullptr && ART::isLeaf(leaf);
+    }
+
+    void put(KEY_TYPE key, VALUE_TYPE value) override {
+        uint8_t ARTkey[KEY_SIZE];
+        ART::loadKey(key, ARTkey);
+        ART::insert(_index, &_index, ARTkey, 0, key, KEY_SIZE);
+    }
+
+    void end_routine() override {}
+};
+
+}  // namespace bliss
+
+#endif

src/bliss/bench_btree.h

@@ -20,7 +20,8 @@ class BlissBTreeIndex : public BlissIndex<KEY_TYPE, VALUE_TYPE> {
         this->_index.bulk_load(values.begin(), values.end());
     }
 
-    bool get(KEY_TYPE key) override { return this->_index.exists(key); }
+    bool get(KEY_TYPE key) override {
+        return this->_index.exists(key); }
 
     void put(KEY_TYPE key, VALUE_TYPE value) override {
         this->_index.insert(std::make_pair(key, value));

src/bliss_bench.cpp

@@ -9,6 +9,7 @@
 
 #include "bliss/bench_pgm.h"
 #include "bliss/bench_alex.h"
+#include "bliss/bench_art.h"
 #include "bliss/bench_btree.h"
 #include "bliss/bench_lipp.h"
 #include "bliss/bliss_index.h"
@@ -170,6 +171,8 @@ int main(int argc, char *argv[]) {
         index.reset(new bliss::BlissLippIndex<key_type, value_type>());
     } else if (config.index == "btree") {
         index.reset(new bliss::BlissBTreeIndex<key_type, value_type>());
+    } else if (config.index == "art") {
+        index.reset(new bliss::BlissARTIndex<key_type, value_type>());
     } else if (config.index == "pgm") {
         index.reset(new bliss::BlissPGMIndex<key_type, value_type>());
     } else {

tests/CMakeLists.txt

@@ -16,4 +16,5 @@ target_include_directories(bliss_test_infra PUBLIC
 
 add_subdirectory(test_alex)
 add_subdirectory(test_lipp)
-add_subdirectory(test_btree)
+add_subdirectory(test_btree)
+add_subdirectory(test_art)

tests/bliss_index_tests.h

@@ -1,21 +1,22 @@
 #ifndef BLISS_INDEX_TESTS_H
 #define BLISS_INDEX_TESTS_H
+#include "include/pgm/pgm_index_dynamic.hpp"
 #include <alex.h>
 #include <gtest/gtest.h>
 #include <lipp.h>
 #include <spdlog/common.h>
-#include "include/pgm/pgm_index_dynamic.hpp"
 
+#include <algorithm>
 #include <cxxopts.hpp>
 #include <iostream>
-#include <string>
-#include <algorithm>
 #include <random>
+#include <string>
 
-#include "bliss/bench_pgm.h"
 #include "bliss/bench_alex.h"
+#include "bliss/bench_art.h"
 #include "bliss/bench_btree.h"
 #include "bliss/bench_lipp.h"
+#include "bliss/bench_pgm.h"
 #include "bliss/bliss_index.h"
 #include "bliss/util/args.h"
 #include "bliss/util/config.h"
@@ -28,20 +29,25 @@ using namespace bliss::utils;
 using key_type = unsigned long;
 using value_type = unsigned long;
 
-class BlissIndexTest : public testing::Test {
-   protected:
+class BlissIndexTest : public testing::Test
+{
+  protected:
     std::unique_ptr<bliss::BlissIndex<key_type, value_type>> index;
-    std::string indexes[4] = {"alex", "lipp", "btree", "pgm"};
+    std::string indexes[5] = {"alex", "lipp", "btree", "pgm", "art"};
     int num_keys = 100000;
 
-    void SetUp() {}
+    void SetUp()
+    {
+    }
 
-    void GenerateData(std::vector<key_type> &data, int num_keys,
-                      bool sorted = true) {
-        for (int i = 0; i < num_keys; i++) {
+    void GenerateData(std::vector<key_type> &data, int num_keys, bool sorted = true)
+    {
+        for (int i = 0; i < num_keys; i++)
+        {
             data.push_back(i);
         }
-        if (!sorted) {
+        if (!sorted)
+        {
             std::random_device rd;
             std::mt19937 g(rd());
             std::shuffle(data.begin(), data.end(), g);

tests/test_art/CMakeLists.txt

@@ -0,0 +1,9 @@
+get_filename_component(EXEC ${CMAKE_CURRENT_SOURCE_DIR} NAME)
+file(GLOB_RECURSE CPP_TESTS "*_tests.cpp")
+add_executable(${EXEC} ${CPP_TESTS})
+target_link_libraries(${EXEC} PRIVATE 
+bliss
+bliss_test_infra 
+GTest::gtest_main)
+include(GoogleTest)
+gtest_discover_tests(${EXEC})

tests/test_art/art_tests.cpp

@@ -0,0 +1,42 @@
+#include "bliss_index_tests.h"
+
+class ArtTest : public BlissIndexTest {};
+
+
+TEST_F(ArtTest, TestArt_Sanity) {
+    index.reset(new bliss::BlissARTIndex<key_type, key_type>());
+    std::vector<key_type> data;
+    int key = 100'000;
+    int value = 123'456;
+    index->put(key, value);
+    EXPECT_TRUE(index->get(key));
+}
+
+
+TEST_F(ArtTest, TestArt_Sorted) {
+    index.reset(new bliss::BlissARTIndex<key_type, key_type>());
+    std::vector<key_type> data;
+    GenerateData(data, num_keys);
+
+    auto insert_start = data.begin();
+    auto insert_end = data.end();
+    executor::execute_inserts(*index, insert_start, insert_end);
+
+    for (auto key : data) {
+        EXPECT_TRUE(index->get(key));
+    }
+}
+
+TEST_F(ArtTest, TestArt_Random) {
+    index.reset(new bliss::BlissARTIndex<key_type, key_type>());
+    std::vector<key_type> data;
+    GenerateData(data, num_keys, false);
+
+    auto insert_start = data.begin();
+    auto insert_end = data.end();
+    executor::execute_inserts(*index, insert_start, insert_end);
+
+    for (auto key : data) {
+        EXPECT_TRUE(index->get(key));
+    }
+}