pgvector.git / summary / log / commit / refs

commit    4e2b76e627936d950da60dd02eea97b3f2bcc2a3
Author:   Andrew Kane <andrew@ankane.org>
Date:     Wed Apr 17 00:20:48 2024 +0000

    Skip duplicate center check for bit [skip ci]


 src/ivfkmeans.c                       | 29 +++++++++++++++--------------
 test/expected/ivfflat_bit_hamming.out |  4 ++++
 test/sql/ivfflat_bit_hamming.sql      |  1 +
 3 files changed, 20 insertions(+), 14 deletions(-)

diff --git a/src/ivfkmeans.c b/src/ivfkmeans.c
index c0e5f361..3a740909 100644
--- a/src/ivfkmeans.c
+++ b/src/ivfkmeans.c
@@ -707,21 +707,22 @@ CheckCenters(Relation index, VectorArray centers, IvfflatType type)
 			elog(ERROR, "Unsupported type");
 	}
 
-	/* Ensure no duplicate centers */
-	/* Fine to sort in-place */
-	if (type == IVFFLAT_TYPE_VECTOR)
-		qsort(centers->items, centers->length, centers->itemsize, CompareVectors);
-	else if (type == IVFFLAT_TYPE_HALFVEC)
-		qsort(centers->items, centers->length, centers->itemsize, CompareHalfVectors);
-	else if (type == IVFFLAT_TYPE_BIT)
-		qsort(centers->items, centers->length, centers->itemsize, CompareBitVectors);
-	else
-		elog(ERROR, "Unsupported type");
-
-	for (int i = 1; i < centers->length; i++)
+	if (type != IVFFLAT_TYPE_BIT)
 	{
-		if (datumIsEqual(PointerGetDatum(VectorArrayGet(centers, i)), PointerGetDatum(VectorArrayGet(centers, i - 1)), false, -1))
-			elog(ERROR, "Duplicate centers detected. Please report a bug.");
+		/* Ensure no duplicate centers */
+		/* Fine to sort in-place */
+		if (type == IVFFLAT_TYPE_VECTOR)
+			qsort(centers->items, centers->length, centers->itemsize, CompareVectors);
+		else if (type == IVFFLAT_TYPE_HALFVEC)
+			qsort(centers->items, centers->length, centers->itemsize, CompareHalfVectors);
+		else
+			elog(ERROR, "Unsupported type");
+
+		for (int i = 1; i < centers->length; i++)
+		{
+			if (datumIsEqual(PointerGetDatum(VectorArrayGet(centers, i)), PointerGetDatum(VectorArrayGet(centers, i - 1)), false, -1))
+				elog(ERROR, "Duplicate centers detected. Please report a bug.");
+		}
 	}
 
 	/* Ensure no zero vectors for cosine distance */
diff --git a/test/expected/ivfflat_bit_hamming.out b/test/expected/ivfflat_bit_hamming.out
index cee8317e..43fbf71c 100644
--- a/test/expected/ivfflat_bit_hamming.out
+++ b/test/expected/ivfflat_bit_hamming.out
@@ -29,4 +29,8 @@ DETAIL:  This will cause low recall.
 HINT:  Drop the index until the table has more data.
 CREATE INDEX ON t USING ivfflat ((val::bit(64001)) bit_hamming_ops) WITH (lists = 1);
 ERROR:  column cannot have more than 64000 dimensions for ivfflat index
+CREATE INDEX ON t USING ivfflat ((val::bit(2)) bit_hamming_ops) WITH (lists = 5);
+NOTICE:  ivfflat index created with little data
+DETAIL:  This will cause low recall.
+HINT:  Drop the index until the table has more data.
 DROP TABLE t;
diff --git a/test/sql/ivfflat_bit_hamming.sql b/test/sql/ivfflat_bit_hamming.sql
index 6c697ae1..3f227675 100644
--- a/test/sql/ivfflat_bit_hamming.sql
+++ b/test/sql/ivfflat_bit_hamming.sql
@@ -16,4 +16,5 @@ CREATE TABLE t (val varbit(3));
 CREATE INDEX ON t USING ivfflat (val bit_hamming_ops) WITH (lists = 1);
 CREATE INDEX ON t USING ivfflat ((val::bit(3)) bit_hamming_ops) WITH (lists = 1);
 CREATE INDEX ON t USING ivfflat ((val::bit(64001)) bit_hamming_ops) WITH (lists = 1);
+CREATE INDEX ON t USING ivfflat ((val::bit(2)) bit_hamming_ops) WITH (lists = 5);
 DROP TABLE t;


[parent: 04af15c9d6a6]