liuyu/pgvector
1
0
Fork 0
mirror of https://github.com/pgvector/pgvector.git synced 2026-07-04 11:40:57 +08:00
Code Issues Packages Projects Releases Wiki Activity

Added support for halfvec to IVFFlat

Browse Source
This commit is contained in:
Andrew Kane
2024-04-11 19:56:39 -07:00
parent a4531ca51f
commit 8d9400bae3
17 changed files with 439 additions and 39 deletions
Download Patch File Download Diff File Expand all files Collapse all files

26
test/expected/ivfflat_halfvec_cosine.out Normal file
View File

@@ -0,0 +1,26 @@
SET enable_seqscan = off;
CREATE TABLE t (val halfvec(3));
INSERT INTO t (val) VALUES ('[0,0,0]'), ('[1,2,3]'), ('[1,1,1]'), (NULL);
CREATE INDEX ON t USING ivfflat (val halfvec_cosine_ops) WITH (lists = 1);
INSERT INTO t (val) VALUES ('[1,2,4]');
SELECT * FROM t ORDER BY val <=> '[3,3,3]';
val
---------
[1,1,1]
[1,2,3]
[1,2,4]
(3 rows)
SELECT COUNT(*) FROM (SELECT * FROM t ORDER BY val <=> '[0,0,0]') t2;
count
-------
3
(1 row)
SELECT COUNT(*) FROM (SELECT * FROM t ORDER BY val <=> (SELECT NULL::halfvec)) t2;
count
-------
3
(1 row)
DROP TABLE t;

21
test/expected/ivfflat_halfvec_ip.out Normal file
View File

@@ -0,0 +1,21 @@
SET enable_seqscan = off;
CREATE TABLE t (val halfvec(3));
INSERT INTO t (val) VALUES ('[0,0,0]'), ('[1,2,3]'), ('[1,1,1]'), (NULL);
CREATE INDEX ON t USING ivfflat (val halfvec_ip_ops) WITH (lists = 1);
INSERT INTO t (val) VALUES ('[1,2,4]');
SELECT * FROM t ORDER BY val <#> '[3,3,3]';
val
---------
[1,2,4]
[1,2,3]
[1,1,1]
[0,0,0]
(4 rows)
SELECT COUNT(*) FROM (SELECT * FROM t ORDER BY val <#> (SELECT NULL::halfvec)) t2;
count
-------
4
(1 row)
DROP TABLE t;

36
test/expected/ivfflat_halfvec_l2.out Normal file
View File

@@ -0,0 +1,36 @@
SET enable_seqscan = off;
CREATE TABLE t (val halfvec(3));
INSERT INTO t (val) VALUES ('[0,0,0]'), ('[1,2,3]'), ('[1,1,1]'), (NULL);
CREATE INDEX ON t USING ivfflat (val halfvec_l2_ops) WITH (lists = 1);
INSERT INTO t (val) VALUES ('[1,2,4]');
SELECT * FROM t ORDER BY val <-> '[3,3,3]';
val
---------
[1,2,3]
[1,2,4]
[1,1,1]
[0,0,0]
(4 rows)
SELECT COUNT(*) FROM (SELECT * FROM t ORDER BY val <-> (SELECT NULL::halfvec)) t2;
count
-------
4
(1 row)
SELECT COUNT(*) FROM t;
count
-------
5
(1 row)
TRUNCATE t;
NOTICE: ivfflat index created with little data
DETAIL: This will cause low recall.
HINT: Drop the index until the table has more data.
SELECT * FROM t ORDER BY val <-> '[3,3,3]';
val
-----
(0 rows)
DROP TABLE t;

13
test/sql/ivfflat_halfvec_cosine.sql Normal file
View File

@@ -0,0 +1,13 @@
SET enable_seqscan = off;
CREATE TABLE t (val halfvec(3));
INSERT INTO t (val) VALUES ('[0,0,0]'), ('[1,2,3]'), ('[1,1,1]'), (NULL);
CREATE INDEX ON t USING ivfflat (val halfvec_cosine_ops) WITH (lists = 1);
INSERT INTO t (val) VALUES ('[1,2,4]');
SELECT * FROM t ORDER BY val <=> '[3,3,3]';
SELECT COUNT(*) FROM (SELECT * FROM t ORDER BY val <=> '[0,0,0]') t2;
SELECT COUNT(*) FROM (SELECT * FROM t ORDER BY val <=> (SELECT NULL::halfvec)) t2;
DROP TABLE t;

12
test/sql/ivfflat_halfvec_ip.sql Normal file
View File

@@ -0,0 +1,12 @@
SET enable_seqscan = off;
CREATE TABLE t (val halfvec(3));
INSERT INTO t (val) VALUES ('[0,0,0]'), ('[1,2,3]'), ('[1,1,1]'), (NULL);
CREATE INDEX ON t USING ivfflat (val halfvec_ip_ops) WITH (lists = 1);
INSERT INTO t (val) VALUES ('[1,2,4]');
SELECT * FROM t ORDER BY val <#> '[3,3,3]';
SELECT COUNT(*) FROM (SELECT * FROM t ORDER BY val <#> (SELECT NULL::halfvec)) t2;
DROP TABLE t;

16
test/sql/ivfflat_halfvec_l2.sql Normal file
View File

@@ -0,0 +1,16 @@
SET enable_seqscan = off;
CREATE TABLE t (val halfvec(3));
INSERT INTO t (val) VALUES ('[0,0,0]'), ('[1,2,3]'), ('[1,1,1]'), (NULL);
CREATE INDEX ON t USING ivfflat (val halfvec_l2_ops) WITH (lists = 1);
INSERT INTO t (val) VALUES ('[1,2,4]');
SELECT * FROM t ORDER BY val <-> '[3,3,3]';
SELECT COUNT(*) FROM (SELECT * FROM t ORDER BY val <-> (SELECT NULL::halfvec)) t2;
SELECT COUNT(*) FROM t;
TRUNCATE t;
SELECT * FROM t ORDER BY val <-> '[3,3,3]';
DROP TABLE t;

132
test/t/032_ivfflat_halfvec_build_recall.pl Normal file
View File

@@ -0,0 +1,132 @@
use strict;
use warnings;
use PostgresNode;
use TestLib;
use Test::More;
my $node;
my @queries = ();
my @expected;
my $limit = 20;
my $dim = 10;
my $array_sql = join(",", ('random()') x $dim);
sub test_recall
{
my ($probes, $min, $operator) = @_;
my $correct = 0;
my $total = 0;
my $explain = $node->safe_psql("postgres", qq(
SET enable_seqscan = off;
SET ivfflat.probes = $probes;
EXPLAIN ANALYZE SELECT i FROM tst ORDER BY v $operator '$queries[0]' LIMIT $limit;
));
like($explain, qr/Index Scan using idx on tst/);
for my $i (0 .. $#queries)
{
my $actual = $node->safe_psql("postgres", qq(
SET enable_seqscan = off;
SET ivfflat.probes = $probes;
SELECT i FROM tst ORDER BY v $operator '$queries[$i]' LIMIT $limit;
));
my @actual_ids = split("\n", $actual);
my %actual_set = map { $_ => 1 } @actual_ids;
my @expected_ids = split("\n", $expected[$i]);
foreach (@expected_ids)
{
if (exists($actual_set{$_}))
{
$correct++;
}
$total++;
}
}
cmp_ok($correct / $total, ">=", $min, $operator);
}
# Initialize node
$node = get_new_node('node');
$node->init;
$node->start;
# Create table
$node->safe_psql("postgres", "CREATE EXTENSION vector;");
$node->safe_psql("postgres", "CREATE TABLE tst (i int4, v halfvec($dim));");
$node->safe_psql("postgres",
"INSERT INTO tst SELECT i, ARRAY[$array_sql] FROM generate_series(1, 100000) i;"
);
# Generate queries
for (1 .. 20)
{
my @r = ();
for (1 .. $dim)
{
push(@r, rand());
}
push(@queries, "[" . join(",", @r) . "]");
}
# Check each index type
my @operators = ("<->", "<#>", "<=>");
my @opclasses = ("halfvec_l2_ops", "halfvec_ip_ops", "halfvec_cosine_ops");
for my $i (0 .. $#operators)
{
my $operator = $operators[$i];
my $opclass = $opclasses[$i];
# Get exact results
@expected = ();
foreach (@queries)
{
my $res = $node->safe_psql("postgres", "SELECT i FROM tst ORDER BY v $operator '$_' LIMIT $limit;");
push(@expected, $res);
}
# Build index serially
$node->safe_psql("postgres", qq(
SET max_parallel_maintenance_workers = 0;
CREATE INDEX idx ON tst USING ivfflat (v $opclass);
));
# Test approximate results
if ($operator ne "<#>")
{
# TODO Fix test (uniform random vectors all have similar inner product)
test_recall(1, 0.4, $operator);
test_recall(10, 0.95, $operator);
}
# Account for equal distances
test_recall(100, 0.9925, $operator);
$node->safe_psql("postgres", "DROP INDEX idx;");
# Build index in parallel
my ($ret, $stdout, $stderr) = $node->psql("postgres", qq(
SET client_min_messages = DEBUG;
SET min_parallel_table_scan_size = 1;
CREATE INDEX idx ON tst USING ivfflat (v $opclass);
));
is($ret, 0, $stderr);
like($stderr, qr/using \d+ parallel workers/);
# Test approximate results
if ($operator ne "<#>")
{
# TODO Fix test (uniform random vectors all have similar inner product)
test_recall(1, 0.4, $operator);
test_recall(10, 0.95, $operator);
}
# Account for equal distances
test_recall(100, 0.9925, $operator);
$node->safe_psql("postgres", "DROP INDEX idx;");
}
done_testing();