furyhawk/milvus
1
0
Fork 0
mirror of https://github.com/milvus-io/milvus.git synced 2026-07-21 02:05:41 +00:00
Code Issues Packages Projects Releases Wiki Activity

enhance: use raw scan for inverted LIKE match (#51477)

Browse Source 
pr: https://github.com/milvus-io/milvus/pull/49922

## Summary
Route generic LIKE Match operations to raw-data scan when a varchar
inverted index is present, while keeping PrefixMatch on the
inverted-index path. Update planner policy and sealed-segment coverage
for a complex LIKE pattern.

Signed-off-by: aoiasd <zhicheng.yue@zilliz.com>
This commit is contained in:
aoiasd
2026-07-17 16:42:40 +08:00
committed by GitHub
parent 0671607431
commit 6fed75383b
3 changed files with 11 additions and 9 deletions
Download Patch File Download Diff File Expand all files Collapse all files
internal/core/src/common/RegexQueryTest.cpp
+5 -4
View File
@@ -468,8 +468,9 @@ TEST_F(SealedSegmentRegexQueryTest, InnerMatchOnInvertedIndexStringField) {
ASSERT_TRUE(final[4]);
}
TEST_F(SealedSegmentRegexQueryTest, RegexQueryOnInvertedIndexStringField) {
std::string operand = "a%";
TEST_F(SealedSegmentRegexQueryTest,
MatchQueryFallsBackToRawDataWithInvertedIndex) {
std::string operand = "a%b%";
const auto& str_meta = schema->operator[](FieldName("str"));
auto column_info = test::GenColumnInfo(str_meta.get_id().get(),
proto::schema::DataType::VarChar,
@@ -491,8 +492,8 @@ TEST_F(SealedSegmentRegexQueryTest, RegexQueryOnInvertedIndexStringField) {
BitsetType final;
final = ExecuteQueryExpr(parsed, segpromote, N, MAX_TIMESTAMP);
ASSERT_FALSE(final[0]);
ASSERT_TRUE(final[1]);
ASSERT_TRUE(final[2]);
ASSERT_FALSE(final[1]);
ASSERT_FALSE(final[2]);
ASSERT_TRUE(final[3]);
ASSERT_TRUE(final[4]);
}
internal/core/src/index/InvertedIndexTantivy.h
+5 -4
View File
@@ -289,13 +289,14 @@ class InvertedIndexTantivy : public ScalarIndex<T> {
bool
ShouldUseOp(proto::plan::OpType op) const override {
// Suffix/contains LIKE and regex can be executed correctly over indexed
// terms, but for short varchar values they are often slower than raw
// scan. Keep only prefix/LIKE Match on the inverted-index path.
// Generic LIKE, suffix/contains LIKE, and regex can be executed
// correctly over indexed terms, but for short varchar values they are
// often slower than raw scan. Keep only prefix matching on the
// inverted-index path.
switch (op) {
case proto::plan::OpType::Match:
case proto::plan::OpType::PrefixMatch:
return SupportPatternMatch() || HasRawData();
case proto::plan::OpType::Match:
case proto::plan::OpType::RegexMatch:
case proto::plan::OpType::PostfixMatch:
case proto::plan::OpType::InnerMatch:
internal/core/src/index/InvertedIndexTest.cpp
+1 -1
View File
@@ -83,7 +83,7 @@ TEST(InvertedIndex, PatternMatchPlannerPolicy) {
EXPECT_TRUE(index.ShouldUseOp(proto::plan::OpType::PrefixMatch));
EXPECT_FALSE(index.ShouldUseOp(proto::plan::OpType::RegexMatch));
EXPECT_TRUE(index.ShouldUseOp(proto::plan::OpType::Equal));
EXPECT_TRUE(index.ShouldUseOp(proto::plan::OpType::Match));
EXPECT_FALSE(index.ShouldUseOp(proto::plan::OpType::Match));
EXPECT_FALSE(index.ShouldUseOp(proto::plan::OpType::InnerMatch));
EXPECT_FALSE(index.ShouldUseOp(proto::plan::OpType::PostfixMatch));