From 3f821794562c1926249bd67be8fa6d88bb2eb25d Mon Sep 17 00:00:00 2001 From: Cai Yudong Date: Fri, 6 Sep 2024 11:52:58 +0800 Subject: [PATCH] Update test_index_check (#817) Signed-off-by: Cai Yudong --- include/knowhere/comp/knowhere_check.h | 26 +-- include/knowhere/index/index_table.h | 39 +--- tests/ut/test_index_check.cc | 278 ++++++++++++++++++++++--- 3 files changed, 269 insertions(+), 74 deletions(-) diff --git a/include/knowhere/comp/knowhere_check.h b/include/knowhere/comp/knowhere_check.h index 41666ba31..a485b5ca4 100644 --- a/include/knowhere/comp/knowhere_check.h +++ b/include/knowhere/comp/knowhere_check.h @@ -34,7 +34,7 @@ IndexTypeAndDataTypeCheck(const std::string& index_name, VecType data_type) { } static bool -SuppportMmapIndexTypeCheck(const std::string& index_name) { +SupportMmapIndexTypeCheck(const std::string& index_name) { auto& mmap_index_table = std::get<1>(IndexFactory::StaticIndexTableInstance()); if (mmap_index_table.find(index_name) != mmap_index_table.end()) { return true; @@ -44,18 +44,18 @@ SuppportMmapIndexTypeCheck(const std::string& index_name) { } inline bool -CheckBooleanInJson(const knowhere::Json& json, std::string key, bool target) { +CheckBooleanInJson(const knowhere::Json& json, std::string key) { if (json.find(key) == json.end()) { - return false; + return true; } if (json[key].is_boolean()) { - return json[key] == target; + return json[key]; } if (json[key].is_string()) { - if (target) { - return json[key] == "true"; + if (json[key] == "true") { + return true; } else { - return json[key] == "false"; + return false; } } return false; @@ -74,16 +74,16 @@ IndexHasRawData(const knowhere::IndexType& indexType, const knowhere::MetricType static std::set no_raw_data_index_set = { IndexEnum::INDEX_FAISS_IVFPQ, IndexEnum::INDEX_FAISS_IVFSQ8, IndexEnum::INDEX_HNSW_SQ8, IndexEnum::INDEX_FAISS_GPU_IDMAP, IndexEnum::INDEX_FAISS_GPU_IVFFLAT, IndexEnum::INDEX_FAISS_GPU_IVFSQ8, - IndexEnum::INDEX_FAISS_GPU_IVFPQ, IndexEnum::INDEX_RAFT_CAGRA, IndexEnum::INDEX_GPU_CAGRA, - IndexEnum::INDEX_RAFT_IVFFLAT, IndexEnum::INDEX_GPU_IVFFLAT, IndexEnum::INDEX_RAFT_IVFPQ, - IndexEnum::INDEX_GPU_IVFPQ, + IndexEnum::INDEX_FAISS_GPU_IVFPQ, IndexEnum::INDEX_GPU_BRUTEFORCE, IndexEnum::INDEX_GPU_IVFFLAT, + IndexEnum::INDEX_GPU_IVFPQ, IndexEnum::INDEX_GPU_CAGRA, IndexEnum::INDEX_RAFT_BRUTEFORCE, + IndexEnum::INDEX_RAFT_IVFFLAT, IndexEnum::INDEX_RAFT_IVFPQ, IndexEnum::INDEX_RAFT_CAGRA, }; static std::set no_raw_data_index_alias_set = {"IVFPQ", "IVFSQ"}; static std::set conditional_hold_raw_data_index_set = { - IndexEnum::INDEX_HNSW, IndexEnum::INDEX_DISKANN, IndexEnum::INDEX_FAISS_SCANN, - IndexEnum::INDEX_FAISS_IVFSQ_CC, IndexEnum::INDEX_FAISS_IDMAP, + IndexEnum::INDEX_FAISS_IDMAP, IndexEnum::INDEX_FAISS_SCANN, IndexEnum::INDEX_FAISS_IVFSQ_CC, + IndexEnum::INDEX_HNSW, IndexEnum::INDEX_DISKANN, }; if (has_raw_data_index_set.find(indexType) != has_raw_data_index_set.end() || @@ -110,7 +110,7 @@ IndexHasRawData(const knowhere::IndexType& indexType, const knowhere::MetricType return IsMetricType(metricType, metric::L2) || IsMetricType(metricType, metric::COSINE); #endif } else if (indexType == IndexEnum::INDEX_FAISS_SCANN) { - return !CheckBooleanInJson(params, indexparam::WITH_RAW_DATA, false); + return CheckBooleanInJson(params, indexparam::WITH_RAW_DATA); // INDEX_FAISS_IVFSQ_CC is not online yet } else if (indexType == IndexEnum::INDEX_FAISS_IVFSQ_CC) { return params.find(indexparam::RAW_DATA_STORE_PREFIX) != params.end(); diff --git a/include/knowhere/index/index_table.h b/include/knowhere/index/index_table.h index abb9cec3b..0004d1318 100644 --- a/include/knowhere/index/index_table.h +++ b/include/knowhere/index/index_table.h @@ -21,7 +21,8 @@ static std::set> legal_knowhere_index = { // binary ivf {IndexEnum::INDEX_FAISS_BIN_IDMAP, VecType::VECTOR_BINARY}, {IndexEnum::INDEX_FAISS_BIN_IVFFLAT, VecType::VECTOR_BINARY}, - // ivf + + // faiss index {IndexEnum::INDEX_FAISS_IDMAP, VecType::VECTOR_FLOAT}, {IndexEnum::INDEX_FAISS_IDMAP, VecType::VECTOR_FLOAT16}, {IndexEnum::INDEX_FAISS_IDMAP, VecType::VECTOR_BFLOAT16}, @@ -49,11 +50,13 @@ static std::set> legal_knowhere_index = { {IndexEnum::INDEX_FAISS_IVFSQ_CC, VecType::VECTOR_FLOAT}, {IndexEnum::INDEX_FAISS_IVFSQ_CC, VecType::VECTOR_FLOAT16}, {IndexEnum::INDEX_FAISS_IVFSQ_CC, VecType::VECTOR_BFLOAT16}, + // gpu index {IndexEnum::INDEX_GPU_BRUTEFORCE, VecType::VECTOR_FLOAT}, {IndexEnum::INDEX_GPU_IVFFLAT, VecType::VECTOR_FLOAT}, {IndexEnum::INDEX_GPU_IVFPQ, VecType::VECTOR_FLOAT}, {IndexEnum::INDEX_GPU_CAGRA, VecType::VECTOR_FLOAT}, + // hnsw {IndexEnum::INDEX_HNSW, VecType::VECTOR_FLOAT}, {IndexEnum::INDEX_HNSW, VecType::VECTOR_FLOAT16}, @@ -66,6 +69,7 @@ static std::set> legal_knowhere_index = { {IndexEnum::INDEX_HNSW_SQ8_REFINE, VecType::VECTOR_FLOAT}, {IndexEnum::INDEX_HNSW_SQ8_REFINE, VecType::VECTOR_FLOAT16}, {IndexEnum::INDEX_HNSW_SQ8_REFINE, VecType::VECTOR_BFLOAT16}, + // faiss hnsw {IndexEnum::INDEX_FAISS_HNSW_FLAT, VecType::VECTOR_FLOAT}, {IndexEnum::INDEX_FAISS_HNSW_FLAT, VecType::VECTOR_FLOAT16}, @@ -82,10 +86,12 @@ static std::set> legal_knowhere_index = { {IndexEnum::INDEX_FAISS_HNSW_PRQ, VecType::VECTOR_FLOAT}, {IndexEnum::INDEX_FAISS_HNSW_PRQ, VecType::VECTOR_FLOAT16}, {IndexEnum::INDEX_FAISS_HNSW_PRQ, VecType::VECTOR_BFLOAT16}, + // diskann {IndexEnum::INDEX_DISKANN, VecType::VECTOR_FLOAT}, {IndexEnum::INDEX_DISKANN, VecType::VECTOR_FLOAT16}, {IndexEnum::INDEX_DISKANN, VecType::VECTOR_BFLOAT16}, + // sparse index {IndexEnum::INDEX_SPARSE_INVERTED_INDEX, VecType::VECTOR_SPARSE_FLOAT}, {IndexEnum::INDEX_SPARSE_WAND, VecType::VECTOR_SPARSE_FLOAT}, @@ -95,51 +101,24 @@ static std::set legal_support_mmap_knowhere_index = { // binary ivf IndexEnum::INDEX_FAISS_BIN_IDMAP, IndexEnum::INDEX_FAISS_BIN_IVFFLAT, - // ivf - IndexEnum::INDEX_FAISS_IDMAP, - IndexEnum::INDEX_FAISS_IDMAP, - IndexEnum::INDEX_FAISS_IDMAP, + // faiss index + IndexEnum::INDEX_FAISS_IDMAP, IndexEnum::INDEX_FAISS_IVFFLAT, - IndexEnum::INDEX_FAISS_IVFFLAT, - IndexEnum::INDEX_FAISS_IVFFLAT, - - IndexEnum::INDEX_FAISS_IVFPQ, IndexEnum::INDEX_FAISS_IVFPQ, - IndexEnum::INDEX_FAISS_IVFPQ, - - IndexEnum::INDEX_FAISS_SCANN, - IndexEnum::INDEX_FAISS_SCANN, IndexEnum::INDEX_FAISS_SCANN, - - IndexEnum::INDEX_FAISS_IVFSQ8, - IndexEnum::INDEX_FAISS_IVFSQ8, IndexEnum::INDEX_FAISS_IVFSQ8, - - IndexEnum::INDEX_FAISS_IVFSQ_CC, - IndexEnum::INDEX_FAISS_IVFSQ_CC, IndexEnum::INDEX_FAISS_IVFSQ_CC, // hnsw IndexEnum::INDEX_HNSW, - IndexEnum::INDEX_HNSW, - IndexEnum::INDEX_HNSW, - IndexEnum::INDEX_HNSW_SQ8, - IndexEnum::INDEX_HNSW_SQ8, - IndexEnum::INDEX_HNSW_SQ8, - - IndexEnum::INDEX_HNSW_SQ8_REFINE, - IndexEnum::INDEX_HNSW_SQ8_REFINE, IndexEnum::INDEX_HNSW_SQ8_REFINE, // faiss hnsw IndexEnum::INDEX_FAISS_HNSW_FLAT, - IndexEnum::INDEX_FAISS_HNSW_SQ, - IndexEnum::INDEX_FAISS_HNSW_PQ, - IndexEnum::INDEX_FAISS_HNSW_PRQ, // sparse index diff --git a/tests/ut/test_index_check.cc b/tests/ut/test_index_check.cc index 6830a722e..a8d3f8674 100644 --- a/tests/ut/test_index_check.cc +++ b/tests/ut/test_index_check.cc @@ -18,47 +18,263 @@ #include "knowhere/comp/knowhere_check.h" #include "knowhere/index/index_factory.h" +using namespace knowhere; + +// knowhere/index/index_table.h TEST_CASE("Test index and data type check", "[IndexCheckTest]") { SECTION("Test valid") { - REQUIRE(knowhere::KnowhereCheck::IndexTypeAndDataTypeCheck(knowhere::IndexEnum::INDEX_HNSW, - knowhere::VecType::VECTOR_FLOAT) == true); - REQUIRE(knowhere::KnowhereCheck::IndexTypeAndDataTypeCheck(knowhere::IndexEnum::INDEX_HNSW, - knowhere::VecType::VECTOR_BFLOAT16) == true); - -#ifndef KNOWHERE_WITH_CARDINAL - REQUIRE(knowhere::KnowhereCheck::IndexTypeAndDataTypeCheck(knowhere::IndexEnum::INDEX_HNSW, - knowhere::VecType::VECTOR_BINARY) == false); - REQUIRE(knowhere::KnowhereCheck::IndexTypeAndDataTypeCheck(knowhere::IndexEnum::INDEX_HNSW, - knowhere::VecType::VECTOR_SPARSE_FLOAT) == false); + // binary index + CHECK(KnowhereCheck::IndexTypeAndDataTypeCheck(IndexEnum::INDEX_FAISS_BIN_IDMAP, VecType::VECTOR_BINARY)); + CHECK(KnowhereCheck::IndexTypeAndDataTypeCheck(IndexEnum::INDEX_FAISS_BIN_IVFFLAT, VecType::VECTOR_BINARY)); + + // faiss index + CHECK(KnowhereCheck::IndexTypeAndDataTypeCheck(IndexEnum::INDEX_FAISS_IDMAP, VecType::VECTOR_FLOAT)); + CHECK(KnowhereCheck::IndexTypeAndDataTypeCheck(IndexEnum::INDEX_FAISS_IDMAP, VecType::VECTOR_FLOAT16)); + CHECK(KnowhereCheck::IndexTypeAndDataTypeCheck(IndexEnum::INDEX_FAISS_IDMAP, VecType::VECTOR_BFLOAT16)); + + CHECK(KnowhereCheck::IndexTypeAndDataTypeCheck(IndexEnum::INDEX_FAISS_IVFFLAT, VecType::VECTOR_FLOAT)); + CHECK(KnowhereCheck::IndexTypeAndDataTypeCheck(IndexEnum::INDEX_FAISS_IVFFLAT, VecType::VECTOR_FLOAT16)); + CHECK(KnowhereCheck::IndexTypeAndDataTypeCheck(IndexEnum::INDEX_FAISS_IVFFLAT, VecType::VECTOR_BFLOAT16)); + + CHECK(KnowhereCheck::IndexTypeAndDataTypeCheck(IndexEnum::INDEX_FAISS_IVFFLAT_CC, VecType::VECTOR_FLOAT)); + CHECK(KnowhereCheck::IndexTypeAndDataTypeCheck(IndexEnum::INDEX_FAISS_IVFFLAT_CC, VecType::VECTOR_FLOAT16)); + CHECK(KnowhereCheck::IndexTypeAndDataTypeCheck(IndexEnum::INDEX_FAISS_IVFFLAT_CC, VecType::VECTOR_BFLOAT16)); + + CHECK(KnowhereCheck::IndexTypeAndDataTypeCheck(IndexEnum::INDEX_FAISS_IVFPQ, VecType::VECTOR_FLOAT)); + CHECK(KnowhereCheck::IndexTypeAndDataTypeCheck(IndexEnum::INDEX_FAISS_IVFPQ, VecType::VECTOR_FLOAT16)); + CHECK(KnowhereCheck::IndexTypeAndDataTypeCheck(IndexEnum::INDEX_FAISS_IVFPQ, VecType::VECTOR_BFLOAT16)); + + CHECK(KnowhereCheck::IndexTypeAndDataTypeCheck(IndexEnum::INDEX_FAISS_SCANN, VecType::VECTOR_FLOAT)); + CHECK(KnowhereCheck::IndexTypeAndDataTypeCheck(IndexEnum::INDEX_FAISS_SCANN, VecType::VECTOR_FLOAT16)); + CHECK(KnowhereCheck::IndexTypeAndDataTypeCheck(IndexEnum::INDEX_FAISS_SCANN, VecType::VECTOR_BFLOAT16)); + + CHECK(KnowhereCheck::IndexTypeAndDataTypeCheck(IndexEnum::INDEX_FAISS_IVFSQ8, VecType::VECTOR_FLOAT)); + CHECK(KnowhereCheck::IndexTypeAndDataTypeCheck(IndexEnum::INDEX_FAISS_IVFSQ8, VecType::VECTOR_FLOAT16)); + CHECK(KnowhereCheck::IndexTypeAndDataTypeCheck(IndexEnum::INDEX_FAISS_IVFSQ8, VecType::VECTOR_BFLOAT16)); + + CHECK(KnowhereCheck::IndexTypeAndDataTypeCheck(IndexEnum::INDEX_FAISS_IVFSQ_CC, VecType::VECTOR_FLOAT)); + CHECK(KnowhereCheck::IndexTypeAndDataTypeCheck(IndexEnum::INDEX_FAISS_IVFSQ_CC, VecType::VECTOR_FLOAT16)); + CHECK(KnowhereCheck::IndexTypeAndDataTypeCheck(IndexEnum::INDEX_FAISS_IVFSQ_CC, VecType::VECTOR_BFLOAT16)); + + // gpu index + CHECK(KnowhereCheck::IndexTypeAndDataTypeCheck(IndexEnum::INDEX_GPU_BRUTEFORCE, VecType::VECTOR_FLOAT)); + CHECK_FALSE(KnowhereCheck::IndexTypeAndDataTypeCheck(IndexEnum::INDEX_GPU_BRUTEFORCE, VecType::VECTOR_FLOAT16)); + CHECK_FALSE( + KnowhereCheck::IndexTypeAndDataTypeCheck(IndexEnum::INDEX_GPU_BRUTEFORCE, VecType::VECTOR_BFLOAT16)); + + CHECK(KnowhereCheck::IndexTypeAndDataTypeCheck(IndexEnum::INDEX_GPU_IVFFLAT, VecType::VECTOR_FLOAT)); + CHECK_FALSE(KnowhereCheck::IndexTypeAndDataTypeCheck(IndexEnum::INDEX_GPU_IVFFLAT, VecType::VECTOR_FLOAT16)); + CHECK_FALSE(KnowhereCheck::IndexTypeAndDataTypeCheck(IndexEnum::INDEX_GPU_IVFFLAT, VecType::VECTOR_BFLOAT16)); + + CHECK(KnowhereCheck::IndexTypeAndDataTypeCheck(IndexEnum::INDEX_GPU_IVFPQ, VecType::VECTOR_FLOAT)); + CHECK_FALSE(KnowhereCheck::IndexTypeAndDataTypeCheck(IndexEnum::INDEX_GPU_IVFPQ, VecType::VECTOR_FLOAT16)); + CHECK_FALSE(KnowhereCheck::IndexTypeAndDataTypeCheck(IndexEnum::INDEX_GPU_IVFPQ, VecType::VECTOR_BFLOAT16)); + + CHECK(KnowhereCheck::IndexTypeAndDataTypeCheck(IndexEnum::INDEX_GPU_CAGRA, VecType::VECTOR_FLOAT)); + CHECK_FALSE(KnowhereCheck::IndexTypeAndDataTypeCheck(IndexEnum::INDEX_GPU_CAGRA, VecType::VECTOR_FLOAT16)); + CHECK_FALSE(KnowhereCheck::IndexTypeAndDataTypeCheck(IndexEnum::INDEX_GPU_CAGRA, VecType::VECTOR_BFLOAT16)); + + // HNSW + CHECK(KnowhereCheck::IndexTypeAndDataTypeCheck(IndexEnum::INDEX_HNSW, VecType::VECTOR_FLOAT)); + CHECK(KnowhereCheck::IndexTypeAndDataTypeCheck(IndexEnum::INDEX_HNSW, VecType::VECTOR_FLOAT16)); + CHECK(KnowhereCheck::IndexTypeAndDataTypeCheck(IndexEnum::INDEX_HNSW, VecType::VECTOR_BFLOAT16)); + +#ifdef KNOWHERE_WITH_CARDINAL + CHECK(KnowhereCheck::IndexTypeAndDataTypeCheck(IndexEnum::INDEX_HNSW, VecType::VECTOR_BINARY)); + CHECK(KnowhereCheck::IndexTypeAndDataTypeCheck(IndexEnum::INDEX_HNSW, VecType::VECTOR_SPARSE_FLOAT)); #else - REQUIRE(knowhere::KnowhereCheck::IndexTypeAndDataTypeCheck(knowhere::IndexEnum::INDEX_HNSW, - knowhere::VecType::VECTOR_BINARY) == true); - REQUIRE(knowhere::KnowhereCheck::IndexTypeAndDataTypeCheck(knowhere::IndexEnum::INDEX_HNSW, - knowhere::VecType::VECTOR_SPARSE_FLOAT) == true); + CHECK_FALSE(KnowhereCheck::IndexTypeAndDataTypeCheck(IndexEnum::INDEX_HNSW, VecType::VECTOR_BINARY)); + CHECK_FALSE(KnowhereCheck::IndexTypeAndDataTypeCheck(IndexEnum::INDEX_HNSW, VecType::VECTOR_SPARSE_FLOAT)); #endif - REQUIRE(knowhere::KnowhereCheck::IndexTypeAndDataTypeCheck(knowhere::IndexEnum::INDEX_DISKANN, - knowhere::VecType::VECTOR_FLOAT) == true); - REQUIRE(knowhere::KnowhereCheck::IndexTypeAndDataTypeCheck(knowhere::IndexEnum::INDEX_DISKANN, - knowhere::VecType::VECTOR_FLOAT16) == true); - REQUIRE(knowhere::KnowhereCheck::IndexTypeAndDataTypeCheck(knowhere::IndexEnum::INDEX_DISKANN, - knowhere::VecType::VECTOR_BINARY) == false); - REQUIRE(knowhere::KnowhereCheck::IndexTypeAndDataTypeCheck(knowhere::IndexEnum::INDEX_DISKANN, - knowhere::VecType::VECTOR_SPARSE_FLOAT) == false); + + CHECK(KnowhereCheck::IndexTypeAndDataTypeCheck(IndexEnum::INDEX_HNSW_SQ8, VecType::VECTOR_FLOAT)); + CHECK(KnowhereCheck::IndexTypeAndDataTypeCheck(IndexEnum::INDEX_HNSW_SQ8, VecType::VECTOR_FLOAT16)); + CHECK(KnowhereCheck::IndexTypeAndDataTypeCheck(IndexEnum::INDEX_HNSW_SQ8, VecType::VECTOR_BFLOAT16)); + + CHECK(KnowhereCheck::IndexTypeAndDataTypeCheck(IndexEnum::INDEX_HNSW_SQ8_REFINE, VecType::VECTOR_FLOAT)); + CHECK(KnowhereCheck::IndexTypeAndDataTypeCheck(IndexEnum::INDEX_HNSW_SQ8_REFINE, VecType::VECTOR_FLOAT16)); + CHECK(KnowhereCheck::IndexTypeAndDataTypeCheck(IndexEnum::INDEX_HNSW_SQ8_REFINE, VecType::VECTOR_BFLOAT16)); + + // faiss hnsw + CHECK(KnowhereCheck::IndexTypeAndDataTypeCheck(IndexEnum::INDEX_FAISS_HNSW_FLAT, VecType::VECTOR_FLOAT)); + CHECK(KnowhereCheck::IndexTypeAndDataTypeCheck(IndexEnum::INDEX_FAISS_HNSW_FLAT, VecType::VECTOR_FLOAT16)); + CHECK(KnowhereCheck::IndexTypeAndDataTypeCheck(IndexEnum::INDEX_FAISS_HNSW_FLAT, VecType::VECTOR_BFLOAT16)); + + CHECK(KnowhereCheck::IndexTypeAndDataTypeCheck(IndexEnum::INDEX_FAISS_HNSW_SQ, VecType::VECTOR_FLOAT)); + CHECK(KnowhereCheck::IndexTypeAndDataTypeCheck(IndexEnum::INDEX_FAISS_HNSW_SQ, VecType::VECTOR_FLOAT16)); + CHECK(KnowhereCheck::IndexTypeAndDataTypeCheck(IndexEnum::INDEX_FAISS_HNSW_SQ, VecType::VECTOR_BFLOAT16)); + + CHECK(KnowhereCheck::IndexTypeAndDataTypeCheck(IndexEnum::INDEX_FAISS_HNSW_PQ, VecType::VECTOR_FLOAT)); + CHECK(KnowhereCheck::IndexTypeAndDataTypeCheck(IndexEnum::INDEX_FAISS_HNSW_PQ, VecType::VECTOR_FLOAT16)); + CHECK(KnowhereCheck::IndexTypeAndDataTypeCheck(IndexEnum::INDEX_FAISS_HNSW_PQ, VecType::VECTOR_BFLOAT16)); + + CHECK(KnowhereCheck::IndexTypeAndDataTypeCheck(IndexEnum::INDEX_FAISS_HNSW_PRQ, VecType::VECTOR_FLOAT)); + CHECK(KnowhereCheck::IndexTypeAndDataTypeCheck(IndexEnum::INDEX_FAISS_HNSW_PRQ, VecType::VECTOR_FLOAT16)); + CHECK(KnowhereCheck::IndexTypeAndDataTypeCheck(IndexEnum::INDEX_FAISS_HNSW_PRQ, VecType::VECTOR_BFLOAT16)); + + // diskann + CHECK(KnowhereCheck::IndexTypeAndDataTypeCheck(IndexEnum::INDEX_DISKANN, VecType::VECTOR_FLOAT)); + CHECK(KnowhereCheck::IndexTypeAndDataTypeCheck(IndexEnum::INDEX_DISKANN, VecType::VECTOR_FLOAT16)); + CHECK(KnowhereCheck::IndexTypeAndDataTypeCheck(IndexEnum::INDEX_DISKANN, VecType::VECTOR_BFLOAT16)); + + // sparse index + CHECK(KnowhereCheck::IndexTypeAndDataTypeCheck(IndexEnum::INDEX_SPARSE_INVERTED_INDEX, + VecType::VECTOR_SPARSE_FLOAT)); + CHECK(KnowhereCheck::IndexTypeAndDataTypeCheck(IndexEnum::INDEX_SPARSE_WAND, VecType::VECTOR_SPARSE_FLOAT)); } } TEST_CASE("Test support mmap index", "[IndexCheckTest]") { SECTION("Test valid") { - REQUIRE(knowhere::KnowhereCheck::SuppportMmapIndexTypeCheck(knowhere::IndexEnum::INDEX_HNSW) == true); - REQUIRE(knowhere::KnowhereCheck::SuppportMmapIndexTypeCheck(knowhere::IndexEnum::INDEX_SPARSE_WAND) == true); - REQUIRE(knowhere::KnowhereCheck::SuppportMmapIndexTypeCheck(knowhere::IndexEnum::INDEX_SPARSE_INVERTED_INDEX) == - true); -#ifndef KNOWHERE_WITH_CARDINAL - REQUIRE(knowhere::KnowhereCheck::SuppportMmapIndexTypeCheck(knowhere::IndexEnum::INDEX_DISKANN) == false); + CHECK(KnowhereCheck::SupportMmapIndexTypeCheck(IndexEnum::INDEX_FAISS_BIN_IDMAP)); + CHECK(KnowhereCheck::SupportMmapIndexTypeCheck(IndexEnum::INDEX_FAISS_BIN_IVFFLAT)); + + // faiss index + CHECK(KnowhereCheck::SupportMmapIndexTypeCheck(IndexEnum::INDEX_FAISS_IDMAP)); + CHECK(KnowhereCheck::SupportMmapIndexTypeCheck(IndexEnum::INDEX_FAISS_IVFFLAT)); + CHECK(KnowhereCheck::SupportMmapIndexTypeCheck(IndexEnum::INDEX_FAISS_IVFPQ)); + CHECK(KnowhereCheck::SupportMmapIndexTypeCheck(IndexEnum::INDEX_FAISS_SCANN)); + CHECK(KnowhereCheck::SupportMmapIndexTypeCheck(IndexEnum::INDEX_FAISS_IVFSQ8)); + CHECK(KnowhereCheck::SupportMmapIndexTypeCheck(IndexEnum::INDEX_FAISS_IVFSQ_CC)); + + // hnsw + CHECK(KnowhereCheck::SupportMmapIndexTypeCheck(IndexEnum::INDEX_HNSW)); + CHECK(KnowhereCheck::SupportMmapIndexTypeCheck(IndexEnum::INDEX_HNSW_SQ8)); + CHECK(KnowhereCheck::SupportMmapIndexTypeCheck(IndexEnum::INDEX_HNSW_SQ8_REFINE)); + + // faiss hnsw + CHECK(KnowhereCheck::SupportMmapIndexTypeCheck(IndexEnum::INDEX_FAISS_HNSW_FLAT)); + CHECK(KnowhereCheck::SupportMmapIndexTypeCheck(IndexEnum::INDEX_FAISS_HNSW_SQ)); + CHECK(KnowhereCheck::SupportMmapIndexTypeCheck(IndexEnum::INDEX_FAISS_HNSW_PQ)); + CHECK(KnowhereCheck::SupportMmapIndexTypeCheck(IndexEnum::INDEX_FAISS_HNSW_PRQ)); + + // sparse index + CHECK(KnowhereCheck::SupportMmapIndexTypeCheck(IndexEnum::INDEX_SPARSE_INVERTED_INDEX)); + CHECK(KnowhereCheck::SupportMmapIndexTypeCheck(IndexEnum::INDEX_SPARSE_WAND)); + +#ifdef KNOWHERE_WITH_CARDINAL + CHECK(KnowhereCheck::SuppportMmapIndexTypeCheck(IndexEnum::INDEX_DISKANN)); #else - REQUIRE(knowhere::KnowhereCheck::SuppportMmapIndexTypeCheck(knowhere::IndexEnum::INDEX_DISKANN) == true); + CHECK_FALSE(KnowhereCheck::SupportMmapIndexTypeCheck(IndexEnum::INDEX_DISKANN)); #endif - REQUIRE(knowhere::KnowhereCheck::SuppportMmapIndexTypeCheck(knowhere::IndexEnum::INDEX_FAISS_BIN_IVFFLAT) == - true); + } +} + +TEST_CASE("Test index has raw data", "[IndexHasRawData]") { + SECTION("Normal test") { + auto ver = Version::GetCurrentVersion().VersionNumber(); + + // binary index + CHECK(KnowhereCheck::IndexHasRawData(IndexEnum::INDEX_FAISS_BIN_IDMAP, metric::HAMMING, ver, {})); + CHECK(KnowhereCheck::IndexHasRawData(IndexEnum::INDEX_FAISS_BIN_IDMAP, metric::JACCARD, ver, {})); + + CHECK(KnowhereCheck::IndexHasRawData(IndexEnum::INDEX_FAISS_BIN_IVFFLAT, metric::HAMMING, ver, {})); + CHECK(KnowhereCheck::IndexHasRawData(IndexEnum::INDEX_FAISS_BIN_IVFFLAT, metric::JACCARD, ver, {})); + + // faiss index + CHECK(KnowhereCheck::IndexHasRawData(IndexEnum::INDEX_FAISS_IDMAP, metric::L2, ver, {})); + CHECK(KnowhereCheck::IndexHasRawData(IndexEnum::INDEX_FAISS_IDMAP, metric::IP, ver, {})); + CHECK(KnowhereCheck::IndexHasRawData(IndexEnum::INDEX_FAISS_IDMAP, metric::COSINE, ver, {})); + + CHECK(KnowhereCheck::IndexHasRawData(IndexEnum::INDEX_FAISS_IVFFLAT, metric::L2, ver, {})); + CHECK(KnowhereCheck::IndexHasRawData(IndexEnum::INDEX_FAISS_IVFFLAT, metric::IP, ver, {})); + CHECK(KnowhereCheck::IndexHasRawData(IndexEnum::INDEX_FAISS_IVFFLAT, metric::COSINE, ver, {})); + + CHECK(KnowhereCheck::IndexHasRawData(IndexEnum::INDEX_FAISS_IVFFLAT_CC, metric::L2, ver, {})); + CHECK(KnowhereCheck::IndexHasRawData(IndexEnum::INDEX_FAISS_IVFFLAT_CC, metric::IP, ver, {})); + CHECK(KnowhereCheck::IndexHasRawData(IndexEnum::INDEX_FAISS_IVFFLAT_CC, metric::COSINE, ver, {})); + + CHECK_FALSE(KnowhereCheck::IndexHasRawData(IndexEnum::INDEX_FAISS_IVFPQ, metric::L2, ver, {})); + CHECK_FALSE(KnowhereCheck::IndexHasRawData(IndexEnum::INDEX_FAISS_IVFPQ, metric::IP, ver, {})); + CHECK_FALSE(KnowhereCheck::IndexHasRawData(IndexEnum::INDEX_FAISS_IVFPQ, metric::COSINE, ver, {})); + + CHECK(KnowhereCheck::IndexHasRawData(IndexEnum::INDEX_FAISS_SCANN, metric::L2, ver, {})); + CHECK(KnowhereCheck::IndexHasRawData(IndexEnum::INDEX_FAISS_SCANN, metric::IP, ver, {})); + CHECK(KnowhereCheck::IndexHasRawData(IndexEnum::INDEX_FAISS_SCANN, metric::COSINE, ver, {})); + + CHECK_FALSE(KnowhereCheck::IndexHasRawData(IndexEnum::INDEX_FAISS_IVFSQ8, metric::L2, ver, {})); + CHECK_FALSE(KnowhereCheck::IndexHasRawData(IndexEnum::INDEX_FAISS_IVFSQ8, metric::IP, ver, {})); + CHECK_FALSE(KnowhereCheck::IndexHasRawData(IndexEnum::INDEX_FAISS_IVFSQ8, metric::COSINE, ver, {})); + + CHECK_FALSE(KnowhereCheck::IndexHasRawData(IndexEnum::INDEX_FAISS_IVFSQ_CC, metric::L2, ver, {})); + CHECK_FALSE(KnowhereCheck::IndexHasRawData(IndexEnum::INDEX_FAISS_IVFSQ_CC, metric::IP, ver, {})); + CHECK_FALSE(KnowhereCheck::IndexHasRawData(IndexEnum::INDEX_FAISS_IVFSQ_CC, metric::COSINE, ver, {})); + + // HNSW + CHECK(KnowhereCheck::IndexHasRawData(IndexEnum::INDEX_HNSW, metric::L2, ver, {})); + CHECK(KnowhereCheck::IndexHasRawData(IndexEnum::INDEX_HNSW, metric::IP, ver, {})); + CHECK(KnowhereCheck::IndexHasRawData(IndexEnum::INDEX_HNSW, metric::COSINE, ver, {})); + + CHECK_FALSE(KnowhereCheck::IndexHasRawData(IndexEnum::INDEX_HNSW_SQ8, metric::L2, ver, {})); + CHECK_FALSE(KnowhereCheck::IndexHasRawData(IndexEnum::INDEX_HNSW_SQ8, metric::IP, ver, {})); + CHECK_FALSE(KnowhereCheck::IndexHasRawData(IndexEnum::INDEX_HNSW_SQ8, metric::COSINE, ver, {})); + + CHECK(KnowhereCheck::IndexHasRawData(IndexEnum::INDEX_HNSW_SQ8_REFINE, metric::L2, ver, {})); + CHECK(KnowhereCheck::IndexHasRawData(IndexEnum::INDEX_HNSW_SQ8_REFINE, metric::IP, ver, {})); + CHECK(KnowhereCheck::IndexHasRawData(IndexEnum::INDEX_HNSW_SQ8_REFINE, metric::COSINE, ver, {})); + + // faiss HNSW + CHECK_FALSE(KnowhereCheck::IndexHasRawData(IndexEnum::INDEX_FAISS_HNSW_FLAT, metric::L2, ver, {})); + CHECK_FALSE(KnowhereCheck::IndexHasRawData(IndexEnum::INDEX_FAISS_HNSW_FLAT, metric::IP, ver, {})); + CHECK_FALSE(KnowhereCheck::IndexHasRawData(IndexEnum::INDEX_FAISS_HNSW_FLAT, metric::COSINE, ver, {})); + + CHECK_FALSE(KnowhereCheck::IndexHasRawData(IndexEnum::INDEX_FAISS_HNSW_SQ, metric::L2, ver, {})); + CHECK_FALSE(KnowhereCheck::IndexHasRawData(IndexEnum::INDEX_FAISS_HNSW_SQ, metric::IP, ver, {})); + CHECK_FALSE(KnowhereCheck::IndexHasRawData(IndexEnum::INDEX_FAISS_HNSW_SQ, metric::COSINE, ver, {})); + + CHECK_FALSE(KnowhereCheck::IndexHasRawData(IndexEnum::INDEX_FAISS_HNSW_PQ, metric::L2, ver, {})); + CHECK_FALSE(KnowhereCheck::IndexHasRawData(IndexEnum::INDEX_FAISS_HNSW_PQ, metric::IP, ver, {})); + CHECK_FALSE(KnowhereCheck::IndexHasRawData(IndexEnum::INDEX_FAISS_HNSW_PQ, metric::COSINE, ver, {})); + + CHECK_FALSE(KnowhereCheck::IndexHasRawData(IndexEnum::INDEX_FAISS_HNSW_PRQ, metric::L2, ver, {})); + CHECK_FALSE(KnowhereCheck::IndexHasRawData(IndexEnum::INDEX_FAISS_HNSW_PRQ, metric::IP, ver, {})); + CHECK_FALSE(KnowhereCheck::IndexHasRawData(IndexEnum::INDEX_FAISS_HNSW_PRQ, metric::COSINE, ver, {})); + + // diskann + CHECK(KnowhereCheck::IndexHasRawData(IndexEnum::INDEX_DISKANN, metric::L2, ver, {})); + CHECK_FALSE(KnowhereCheck::IndexHasRawData(IndexEnum::INDEX_DISKANN, metric::IP, ver, {})); + CHECK(KnowhereCheck::IndexHasRawData(IndexEnum::INDEX_DISKANN, metric::COSINE, ver, {})); + + // gpu index + CHECK_FALSE(KnowhereCheck::IndexHasRawData(IndexEnum::INDEX_GPU_BRUTEFORCE, metric::L2, ver, {})); + CHECK_FALSE(KnowhereCheck::IndexHasRawData(IndexEnum::INDEX_GPU_BRUTEFORCE, metric::IP, ver, {})); + CHECK_FALSE(KnowhereCheck::IndexHasRawData(IndexEnum::INDEX_GPU_BRUTEFORCE, metric::COSINE, ver, {})); + + CHECK_FALSE(KnowhereCheck::IndexHasRawData(IndexEnum::INDEX_GPU_IVFFLAT, metric::L2, ver, {})); + CHECK_FALSE(KnowhereCheck::IndexHasRawData(IndexEnum::INDEX_GPU_IVFFLAT, metric::IP, ver, {})); + CHECK_FALSE(KnowhereCheck::IndexHasRawData(IndexEnum::INDEX_GPU_IVFFLAT, metric::COSINE, ver, {})); + + CHECK_FALSE(KnowhereCheck::IndexHasRawData(IndexEnum::INDEX_GPU_IVFPQ, metric::L2, ver, {})); + CHECK_FALSE(KnowhereCheck::IndexHasRawData(IndexEnum::INDEX_GPU_IVFPQ, metric::IP, ver, {})); + CHECK_FALSE(KnowhereCheck::IndexHasRawData(IndexEnum::INDEX_GPU_IVFPQ, metric::COSINE, ver, {})); + + CHECK_FALSE(KnowhereCheck::IndexHasRawData(IndexEnum::INDEX_GPU_CAGRA, metric::L2, ver, {})); + CHECK_FALSE(KnowhereCheck::IndexHasRawData(IndexEnum::INDEX_GPU_CAGRA, metric::IP, ver, {})); + CHECK_FALSE(KnowhereCheck::IndexHasRawData(IndexEnum::INDEX_GPU_CAGRA, metric::COSINE, ver, {})); + + CHECK(KnowhereCheck::IndexHasRawData(IndexEnum::INDEX_SPARSE_INVERTED_INDEX, metric::IP, ver, {})); + CHECK(KnowhereCheck::IndexHasRawData(IndexEnum::INDEX_SPARSE_WAND, metric::IP, ver, {})); + } + + SECTION("Special test") { + auto min_ver = Version::GetMinimalVersion().VersionNumber(); + auto ver = Version::GetCurrentVersion().VersionNumber(); + + knowhere::Json json = { + {indexparam::WITH_RAW_DATA, true}, + }; + + CHECK_FALSE(KnowhereCheck::IndexHasRawData(IndexEnum::INDEX_FAISS_IDMAP, metric::COSINE, min_ver, {})); + + CHECK(KnowhereCheck::IndexHasRawData(IndexEnum::INDEX_FAISS_SCANN, metric::L2, ver, json)); + + json[indexparam::WITH_RAW_DATA] = false; + CHECK_FALSE(KnowhereCheck::IndexHasRawData(IndexEnum::INDEX_FAISS_SCANN, metric::L2, ver, json)); + json[indexparam::WITH_RAW_DATA] = "true"; + CHECK(KnowhereCheck::IndexHasRawData(IndexEnum::INDEX_FAISS_SCANN, metric::L2, ver, json)); + json[indexparam::WITH_RAW_DATA] = "false"; + CHECK_FALSE(KnowhereCheck::IndexHasRawData(IndexEnum::INDEX_FAISS_SCANN, metric::L2, ver, json)); + json[indexparam::WITH_RAW_DATA] = 1; + CHECK_FALSE(KnowhereCheck::IndexHasRawData(IndexEnum::INDEX_FAISS_SCANN, metric::L2, ver, json)); } }