Compare commits

...

23 Commits

Author SHA1 Message Date
Andrew Kane
a6420355c5 Updated FreeBSD package name in readme [skip ci] 2026-07-10 22:17:27 -07:00
Andrew Kane
73356ecfa7 Improved check for VectorArrayInit [skip ci] 2026-07-10 16:30:35 -07:00
Andrew Kane
0e557b1d18 Updated readme [skip ci] 2026-07-10 15:16:36 -07:00
Andrew Kane
769a60884c Updated readme [skip ci] 2026-07-10 14:20:26 -07:00
Andrew Kane
8711840058 Added section on multitenancy [skip ci] 2026-07-10 14:13:13 -07:00
Andrew Kane
159b79aaad Version bump to 0.8.5 [skip ci] 2026-07-08 16:03:40 -07:00
Andrew Kane
c5a277a975 Hardened VectorArrayInit 2026-07-08 10:15:21 -07:00
Andrew Kane
124f6c61a4 Added Postgres 20 to CI [skip ci] 2026-07-07 16:15:39 -07:00
Andrew Kane
5ca52d12b6 Updated checkout action [skip ci] 2026-07-07 15:46:41 -07:00
Andrew Kane
1f61d1111b Added ubuntu-26.04 to CI [skip ci] 2026-07-07 15:46:06 -07:00
Andrew Kane
fb1b8966eb Added comment to IVFFlat vacuuming test [skip ci] 2026-07-01 17:21:26 -07:00
Andrew Kane
71ce9d3311 Added test for concurrent INSERTs, DELETEs, SELECTs, and VACUUM with IVFFlat [skip ci] 2026-07-01 17:18:34 -07:00
Andrew Kane
d19cc0d371 Improved HNSW vacuuming test [skip ci] 2026-07-01 17:14:53 -07:00
Andrew Kane
89fda3e100 Removed slow tests [skip ci] 2026-07-01 16:05:53 -07:00
Andrew Kane
30d8654b47 Added tests for HNSW index dimensions for bit type [skip ci] 2026-07-01 15:49:07 -07:00
Andrew Kane
a846385cc7 Added tests for index dimensions for bit type [skip ci] 2026-07-01 15:47:05 -07:00
Andrew Kane
98d7c4124e Added IVFFlat memory tests for bit [skip ci] 2026-07-01 15:45:18 -07:00
Andrew Kane
2a2b4a0b58 Updated changelog [skip ci] 2026-07-01 12:35:46 -07:00
Andrew Kane
f15a50387f Moved logic for calculating number of samples [skip ci] 2026-07-01 12:27:03 -07:00
Andrew Kane
971b7d7fd6 Added IVFFlat memory tests for halfvec [skip ci] 2026-07-01 12:21:08 -07:00
Andrew Kane
f51d8ed989 Improved readability of options tests [skip ci] 2026-07-01 12:17:50 -07:00
Andrew Kane
a76a18d526 Added tests for index dimensions [skip ci] 2026-07-01 12:15:27 -07:00
Andrew Kane
b383e4d191 Reduced memory usage for small tables for IVFFlat index builds - resolves #995 and resolves #996
Co-authored-by: Itai Spiegel <itai@mave.com>
2026-07-01 11:54:21 -07:00
26 changed files with 320 additions and 44 deletions

View File

@@ -8,10 +8,12 @@ jobs:
fail-fast: false
matrix:
include:
- postgres: 20
os: ubuntu-26.04
- postgres: 19
os: ubuntu-24.04
os: ubuntu-26.04
- postgres: 18
os: ubuntu-24.04
os: ubuntu-26.04-arm
- postgres: 17
os: ubuntu-24.04
- postgres: 16
@@ -23,7 +25,7 @@ jobs:
- postgres: 13
os: ubuntu-22.04
steps:
- uses: actions/checkout@v6
- uses: actions/checkout@v7
- uses: ankane/setup-postgres@v1
with:
postgres-version: ${{ matrix.postgres }}
@@ -53,7 +55,7 @@ jobs:
- postgres: 14
os: macos-15-intel
steps:
- uses: actions/checkout@v6
- uses: actions/checkout@v7
- uses: ankane/setup-postgres@v1
with:
postgres-version: ${{ matrix.postgres }}
@@ -92,7 +94,7 @@ jobs:
- postgres: 14
os: windows-2022
steps:
- uses: actions/checkout@v6
- uses: actions/checkout@v7
- uses: ankane/setup-postgres@v1
with:
postgres-version: ${{ matrix.postgres }}
@@ -133,7 +135,7 @@ jobs:
if: ${{ !startsWith(github.ref_name, 'mac') && !startsWith(github.ref_name, 'windows') }}
runs-on: ubuntu-latest
steps:
- uses: actions/checkout@v6
- uses: actions/checkout@v7
- uses: ankane/setup-postgres-valgrind@v1
with:
postgres-version: 18

View File

@@ -1,7 +1,12 @@
## 0.8.5 (2026-07-08)
- Reduced memory usage for small tables for IVFFlat index builds
## 0.8.4 (2026-06-30)
- Fixed `hnsw graph not repaired` error with HNSW vacuuming
- Fixed possible error with inserts during HNSW vacuuming
- Fixed memory exceeding `maintenance_work_mem` with IVFFlat index builds
## 0.8.3 (2026-06-17)

View File

@@ -5,7 +5,7 @@ ARG DEBIAN_CODENAME=bookworm
FROM postgres:$PG_MAJOR-$DEBIAN_CODENAME
ARG PG_MAJOR
ADD https://github.com/pgvector/pgvector.git#v0.8.4 /tmp/pgvector
ADD https://github.com/pgvector/pgvector.git#v0.8.5 /tmp/pgvector
RUN apt-get update && \
apt-mark hold locales && \

View File

@@ -2,7 +2,7 @@
"name": "vector",
"abstract": "Open-source vector similarity search for Postgres",
"description": "Supports L2 distance, inner product, and cosine distance",
"version": "0.8.4",
"version": "0.8.5",
"maintainer": [
"Andrew Kane <andrew@ankane.org>"
],
@@ -20,7 +20,7 @@
"vector": {
"file": "sql/vector.sql",
"docfile": "README.md",
"version": "0.8.4",
"version": "0.8.5",
"abstract": "Open-source vector similarity search for Postgres"
}
},

View File

@@ -1,5 +1,5 @@
EXTENSION = vector
EXTVERSION = 0.8.4
EXTVERSION = 0.8.5
MODULE_big = vector
DATA = $(wildcard sql/*--*--*.sql)

View File

@@ -1,5 +1,5 @@
EXTENSION = vector
EXTVERSION = 0.8.4
EXTVERSION = 0.8.5
DATA_built = sql\$(EXTENSION)--$(EXTVERSION).sql
OBJS = src\bitutils.obj src\bitvec.obj src\halfutils.obj src\halfvec.obj src\hnsw.obj src\hnswbuild.obj src\hnswinsert.obj src\hnswscan.obj src\hnswutils.obj src\hnswvacuum.obj src\ivfbuild.obj src\ivfflat.obj src\ivfinsert.obj src\ivfkmeans.obj src\ivfscan.obj src\ivfutils.obj src\ivfvacuum.obj src\sparsevec.obj src\vector.obj

View File

@@ -23,7 +23,7 @@ Compile and install the extension (supports Postgres 13+)
```sh
cd /tmp
git clone --branch v0.8.4 https://github.com/pgvector/pgvector.git
git clone --branch v0.8.5 https://github.com/pgvector/pgvector.git
cd pgvector
make
make install # may need sudo
@@ -40,7 +40,7 @@ Ensure [C++ support in Visual Studio](https://learn.microsoft.com/en-us/cpp/buil
```cmd
set "PGROOT=C:\Program Files\PostgreSQL\18"
cd %TEMP%
git clone --branch v0.8.4 https://github.com/pgvector/pgvector.git
git clone --branch v0.8.5 https://github.com/pgvector/pgvector.git
cd pgvector
nmake /F Makefile.win
nmake /F Makefile.win install
@@ -465,6 +465,16 @@ If filtering by many different values, consider [partitioning](https://www.postg
CREATE TABLE items (embedding vector(3), category_id int) PARTITION BY LIST(category_id);
```
## Multitenancy
For applications with multiple tenants, sharing an approximate index between tenants means vectors from one tenant can affect recall (and speed) for other tenants.
For tenant isolation, use [list partitioning](https://www.postgresql.org/docs/current/ddl-partitioning.html) or separate tables.
```sql
CREATE TABLE items (customer_id int, embedding vector(3)) PARTITION BY LIST(customer_id);
```
## Iterative Index Scans
With approximate indexes, queries with filtering can return less results since filtering is applied *after* the index is scanned. Starting with 0.8.0, you can enable iterative index scans, which will automatically scan more of the index until enough results are found (or it reaches `hnsw.max_scan_tuples` or `ivfflat.max_probes`).
@@ -1151,23 +1161,23 @@ This adds pgvector to the [Postgres image](https://hub.docker.com/_/postgres) (r
Supported tags are:
- `pg18-trixie`, `0.8.4-pg18-trixie`
- `pg18-bookworm`, `0.8.4-pg18-bookworm`, `pg18`, `0.8.4-pg18`
- `pg17-trixie`, `0.8.4-pg17-trixie`
- `pg17-bookworm`, `0.8.4-pg17-bookworm`, `pg17`, `0.8.4-pg17`
- `pg16-trixie`, `0.8.4-pg16-trixie`
- `pg16-bookworm`, `0.8.4-pg16-bookworm`, `pg16`, `0.8.4-pg16`
- `pg15-trixie`, `0.8.4-pg15-trixie`
- `pg15-bookworm`, `0.8.4-pg15-bookworm`, `pg15`, `0.8.4-pg15`
- `pg14-trixie`, `0.8.4-pg14-trixie`
- `pg14-bookworm`, `0.8.4-pg14-bookworm`, `pg14`, `0.8.4-pg14`
- `pg13-trixie`, `0.8.4-pg13-trixie`
- `pg13-bookworm`, `0.8.4-pg13-bookworm`, `pg13`, `0.8.4-pg13`
- `pg18-trixie`, `0.8.5-pg18-trixie`
- `pg18-bookworm`, `0.8.5-pg18-bookworm`, `pg18`, `0.8.5-pg18`
- `pg17-trixie`, `0.8.5-pg17-trixie`
- `pg17-bookworm`, `0.8.5-pg17-bookworm`, `pg17`, `0.8.5-pg17`
- `pg16-trixie`, `0.8.5-pg16-trixie`
- `pg16-bookworm`, `0.8.5-pg16-bookworm`, `pg16`, `0.8.5-pg16`
- `pg15-trixie`, `0.8.5-pg15-trixie`
- `pg15-bookworm`, `0.8.5-pg15-bookworm`, `pg15`, `0.8.5-pg15`
- `pg14-trixie`, `0.8.5-pg14-trixie`
- `pg14-bookworm`, `0.8.5-pg14-bookworm`, `pg14`, `0.8.5-pg14`
- `pg13-trixie`, `0.8.5-pg13-trixie`
- `pg13-bookworm`, `0.8.5-pg13-bookworm`, `pg13`, `0.8.5-pg13`
You can also build the image manually:
```sh
git clone --branch v0.8.4 https://github.com/pgvector/pgvector.git
git clone --branch v0.8.5 https://github.com/pgvector/pgvector.git
cd pgvector
docker build --pull --build-arg PG_MAJOR=18 -t myuser/pgvector .
```
@@ -1223,7 +1233,7 @@ Note: Replace `18` with your Postgres server version
Install the FreeBSD package with:
```sh
pkg install postgresql17-pgvector
pkg install postgresql18-pgvector
```
or the port with:

View File

@@ -0,0 +1,2 @@
-- complain if script is sourced in psql, rather than via CREATE EXTENSION
\echo Use "ALTER EXTENSION vector UPDATE TO '0.8.5'" to load this file. \quit

View File

@@ -438,15 +438,22 @@ ComputeCenters(IvfflatBuildState * buildstate)
pgstat_progress_update_param(PROGRESS_CREATEIDX_SUBPHASE, PROGRESS_IVFFLAT_PHASE_KMEANS);
/* Target 50 samples per list, with at least 10000 samples */
/* The number of samples has a large effect on index build time */
numSamples = buildstate->lists * 50;
if (numSamples < 10000)
numSamples = 10000;
/* Skip samples for unlogged table */
if (buildstate->heap == NULL)
numSamples = 1;
else
{
int64 maxTuples = (int64) RelationGetNumberOfBlocks(buildstate->heap) * MaxHeapTuplesPerPage;
/* Target 50 samples per list, with at least 10000 samples */
/* The number of samples has a large effect on index build time */
numSamples = buildstate->lists * 50;
if (numSamples < 10000)
numSamples = 10000;
/* Save memory since will not have more than max tuples */
numSamples = Max(Min(numSamples, maxTuples), 1);
}
/* Sample rows */
buildstate->memoryUsed += VECTOR_ARRAY_SIZE(numSamples, buildstate->itemsize);

View File

@@ -23,11 +23,15 @@
VectorArray
VectorArrayInit(int maxlen, int dimensions, Size itemsize)
{
VectorArray res = palloc(sizeof(VectorArrayData));
VectorArray res;
if (maxlen < 1 || dimensions < 1 || itemsize == 0)
elog(ERROR, "cannot create vector array");
/* Ensure items are aligned to prevent UB */
itemsize = MAXALIGN(itemsize);
res = palloc(sizeof(VectorArrayData));
res->length = 0;
res->maxlen = maxlen;
res->dim = dimensions;

View File

@@ -40,7 +40,7 @@
#endif
#if PG_VERSION_NUM >= 180000
PG_MODULE_MAGIC_EXT(.name = "vector", .version = "0.8.4");
PG_MODULE_MAGIC_EXT(.name = "vector", .version = "0.8.5");
#else
PG_MODULE_MAGIC;
#endif

View File

@@ -49,3 +49,11 @@ CREATE INDEX ON t USING hnsw ((val::bit(3)) bit_hamming_ops);
CREATE INDEX ON t USING hnsw ((val::bit(64001)) bit_hamming_ops);
ERROR: column cannot have more than 64000 dimensions for hnsw index
DROP TABLE t;
-- dimensions
CREATE TABLE t (val bit(64000));
CREATE INDEX ON t USING hnsw (val bit_hamming_ops);
DROP TABLE t;
CREATE TABLE t (val bit(64001));
CREATE INDEX ON t USING hnsw (val bit_hamming_ops);
ERROR: column cannot have more than 64000 dimensions for hnsw index
DROP TABLE t;

View File

@@ -100,3 +100,11 @@ SELECT COUNT(*) FROM (SELECT * FROM t ORDER BY val <+> (SELECT NULL::halfvec)) t
(1 row)
DROP TABLE t;
-- dimensions
CREATE TABLE t (val halfvec(4000));
CREATE INDEX ON t USING hnsw (val halfvec_l2_ops);
DROP TABLE t;
CREATE TABLE t (val halfvec(4001));
CREATE INDEX ON t USING hnsw (val halfvec_l2_ops);
ERROR: column cannot have more than 4000 dimensions for hnsw index
DROP TABLE t;

View File

@@ -161,6 +161,7 @@ ERROR: value 1001 out of bounds for option "ef_construction"
DETAIL: Valid values are between "4" and "1000".
CREATE INDEX ON t USING hnsw (val vector_l2_ops) WITH (m = 16, ef_construction = 31);
ERROR: ef_construction must be greater than or equal to 2 * m
DROP TABLE t;
SHOW hnsw.ef_search;
hnsw.ef_search
----------------
@@ -198,4 +199,11 @@ SET hnsw.scan_mem_multiplier = 0;
ERROR: 0 is outside the valid range for parameter "hnsw.scan_mem_multiplier" (1 .. 1000)
SET hnsw.scan_mem_multiplier = 1001;
ERROR: 1001 is outside the valid range for parameter "hnsw.scan_mem_multiplier" (1 .. 1000)
-- dimensions
CREATE TABLE t (val vector(2000));
CREATE INDEX ON t USING hnsw (val vector_l2_ops);
DROP TABLE t;
CREATE TABLE t (val vector(2001));
CREATE INDEX ON t USING hnsw (val vector_l2_ops);
ERROR: column cannot have more than 2000 dimensions for hnsw index
DROP TABLE t;

View File

@@ -35,3 +35,32 @@ 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;
-- dimensions
CREATE TABLE t (val bit(64000));
CREATE INDEX ON t USING ivfflat (val bit_hamming_ops);
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;
CREATE TABLE t (val bit(64001));
CREATE INDEX ON t USING ivfflat (val bit_hamming_ops);
ERROR: column cannot have more than 64000 dimensions for ivfflat index
DROP TABLE t;
-- memory
SET maintenance_work_mem = '1MB';
CREATE TABLE t (val bit(64000));
CREATE INDEX ON t USING ivfflat (val bit_hamming_ops);
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;
RESET maintenance_work_mem;
SET maintenance_work_mem = '29MB';
CREATE TABLE t (val bit(64000));
INSERT INTO t (val) VALUES (B'0'::bit(64000));
CREATE INDEX ON t USING ivfflat (val bit_hamming_ops);
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;
RESET maintenance_work_mem;

View File

@@ -82,3 +82,32 @@ SELECT COUNT(*) FROM (SELECT * FROM t ORDER BY val <=> (SELECT NULL::halfvec)) t
(1 row)
DROP TABLE t;
-- dimensions
CREATE TABLE t (val halfvec(4000));
CREATE INDEX ON t USING ivfflat (val halfvec_l2_ops);
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;
CREATE TABLE t (val halfvec(4001));
CREATE INDEX ON t USING ivfflat (val halfvec_l2_ops);
ERROR: column cannot have more than 4000 dimensions for ivfflat index
DROP TABLE t;
-- memory
SET maintenance_work_mem = '1MB';
CREATE TABLE t (val halfvec(4000));
CREATE INDEX ON t USING ivfflat (val halfvec_l2_ops);
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;
RESET maintenance_work_mem;
SET maintenance_work_mem = '6MB';
CREATE TABLE t (val halfvec(4000));
INSERT INTO t (val) VALUES (array_fill(0, ARRAY[4000]));
CREATE INDEX ON t USING ivfflat (val halfvec_l2_ops);
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;
RESET maintenance_work_mem;

View File

@@ -143,6 +143,7 @@ DETAIL: Valid values are between "1" and "32768".
CREATE INDEX ON t USING ivfflat (val vector_l2_ops) WITH (lists = 32769);
ERROR: value 32769 out of bounds for option "lists"
DETAIL: Valid values are between "1" and "32768".
DROP TABLE t;
SHOW ivfflat.probes;
ivfflat.probes
----------------
@@ -172,4 +173,32 @@ SET ivfflat.max_probes = 0;
ERROR: 0 is outside the valid range for parameter "ivfflat.max_probes" (1 .. 32768)
SET ivfflat.max_probes = 32769;
ERROR: 32769 is outside the valid range for parameter "ivfflat.max_probes" (1 .. 32768)
-- dimensions
CREATE TABLE t (val vector(2000));
CREATE INDEX ON t USING ivfflat (val vector_l2_ops);
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;
CREATE TABLE t (val vector(2001));
CREATE INDEX ON t USING ivfflat (val vector_l2_ops);
ERROR: column cannot have more than 2000 dimensions for ivfflat index
DROP TABLE t;
-- memory
SET maintenance_work_mem = '1MB';
CREATE TABLE t (val vector(2000));
CREATE INDEX ON t USING ivfflat (val vector_l2_ops);
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;
RESET maintenance_work_mem;
SET maintenance_work_mem = '5MB';
CREATE TABLE t (val vector(2000));
INSERT INTO t (val) VALUES (array_fill(0, ARRAY[2000]));
CREATE INDEX ON t USING ivfflat (val vector_l2_ops);
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;
RESET maintenance_work_mem;

View File

@@ -33,3 +33,13 @@ CREATE INDEX ON t USING hnsw (val bit_hamming_ops);
CREATE INDEX ON t USING hnsw ((val::bit(3)) bit_hamming_ops);
CREATE INDEX ON t USING hnsw ((val::bit(64001)) bit_hamming_ops);
DROP TABLE t;
-- dimensions
CREATE TABLE t (val bit(64000));
CREATE INDEX ON t USING hnsw (val bit_hamming_ops);
DROP TABLE t;
CREATE TABLE t (val bit(64001));
CREATE INDEX ON t USING hnsw (val bit_hamming_ops);
DROP TABLE t;

View File

@@ -56,3 +56,13 @@ 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;
-- dimensions
CREATE TABLE t (val halfvec(4000));
CREATE INDEX ON t USING hnsw (val halfvec_l2_ops);
DROP TABLE t;
CREATE TABLE t (val halfvec(4001));
CREATE INDEX ON t USING hnsw (val halfvec_l2_ops);
DROP TABLE t;

View File

@@ -95,23 +95,28 @@ CREATE INDEX ON t USING hnsw (val vector_l2_ops) WITH (m = 101);
CREATE INDEX ON t USING hnsw (val vector_l2_ops) WITH (ef_construction = 3);
CREATE INDEX ON t USING hnsw (val vector_l2_ops) WITH (ef_construction = 1001);
CREATE INDEX ON t USING hnsw (val vector_l2_ops) WITH (m = 16, ef_construction = 31);
DROP TABLE t;
SHOW hnsw.ef_search;
SET hnsw.ef_search = 0;
SET hnsw.ef_search = 1001;
SHOW hnsw.iterative_scan;
SET hnsw.iterative_scan = on;
SHOW hnsw.max_scan_tuples;
SET hnsw.max_scan_tuples = 0;
SHOW hnsw.scan_mem_multiplier;
SET hnsw.scan_mem_multiplier = 0;
SET hnsw.scan_mem_multiplier = 1001;
-- dimensions
CREATE TABLE t (val vector(2000));
CREATE INDEX ON t USING hnsw (val vector_l2_ops);
DROP TABLE t;
CREATE TABLE t (val vector(2001));
CREATE INDEX ON t USING hnsw (val vector_l2_ops);
DROP TABLE t;

View File

@@ -21,3 +21,28 @@ 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;
-- dimensions
CREATE TABLE t (val bit(64000));
CREATE INDEX ON t USING ivfflat (val bit_hamming_ops);
DROP TABLE t;
CREATE TABLE t (val bit(64001));
CREATE INDEX ON t USING ivfflat (val bit_hamming_ops);
DROP TABLE t;
-- memory
SET maintenance_work_mem = '1MB';
CREATE TABLE t (val bit(64000));
CREATE INDEX ON t USING ivfflat (val bit_hamming_ops);
DROP TABLE t;
RESET maintenance_work_mem;
SET maintenance_work_mem = '29MB';
CREATE TABLE t (val bit(64000));
INSERT INTO t (val) VALUES (B'0'::bit(64000));
CREATE INDEX ON t USING ivfflat (val bit_hamming_ops);
DROP TABLE t;
RESET maintenance_work_mem;

View File

@@ -43,3 +43,28 @@ 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;
-- dimensions
CREATE TABLE t (val halfvec(4000));
CREATE INDEX ON t USING ivfflat (val halfvec_l2_ops);
DROP TABLE t;
CREATE TABLE t (val halfvec(4001));
CREATE INDEX ON t USING ivfflat (val halfvec_l2_ops);
DROP TABLE t;
-- memory
SET maintenance_work_mem = '1MB';
CREATE TABLE t (val halfvec(4000));
CREATE INDEX ON t USING ivfflat (val halfvec_l2_ops);
DROP TABLE t;
RESET maintenance_work_mem;
SET maintenance_work_mem = '6MB';
CREATE TABLE t (val halfvec(4000));
INSERT INTO t (val) VALUES (array_fill(0, ARRAY[4000]));
CREATE INDEX ON t USING ivfflat (val halfvec_l2_ops);
DROP TABLE t;
RESET maintenance_work_mem;

View File

@@ -81,19 +81,40 @@ DROP TABLE t;
CREATE TABLE t (val vector(3));
CREATE INDEX ON t USING ivfflat (val vector_l2_ops) WITH (lists = 0);
CREATE INDEX ON t USING ivfflat (val vector_l2_ops) WITH (lists = 32769);
DROP TABLE t;
SHOW ivfflat.probes;
SET ivfflat.probes = 0;
SET ivfflat.probes = 32769;
SHOW ivfflat.iterative_scan;
SET ivfflat.iterative_scan = on;
SHOW ivfflat.max_probes;
SET ivfflat.max_probes = 0;
SET ivfflat.max_probes = 32769;
-- dimensions
CREATE TABLE t (val vector(2000));
CREATE INDEX ON t USING ivfflat (val vector_l2_ops);
DROP TABLE t;
CREATE TABLE t (val vector(2001));
CREATE INDEX ON t USING ivfflat (val vector_l2_ops);
DROP TABLE t;
-- memory
SET maintenance_work_mem = '1MB';
CREATE TABLE t (val vector(2000));
CREATE INDEX ON t USING ivfflat (val vector_l2_ops);
DROP TABLE t;
RESET maintenance_work_mem;
SET maintenance_work_mem = '5MB';
CREATE TABLE t (val vector(2000));
INSERT INTO t (val) VALUES (array_fill(0, ARRAY[2000]));
CREATE INDEX ON t USING ivfflat (val vector_l2_ops);
DROP TABLE t;
RESET maintenance_work_mem;

View File

@@ -31,7 +31,7 @@ $node->pgbench(
{
"047_hnsw_vacuum_insert_insert\@500" => "INSERT INTO tst (v) VALUES (ARRAY[$array_sql]);",
"047_hnsw_vacuum_insert_delete\@500" => "DELETE FROM tst WHERE i = (SELECT i FROM tst LIMIT 1);",
"047_hnsw_vacuum_insert_select\@20" => "SELECT i FROM tst ORDER BY v <-> '[0,0,0]' LIMIT 10;",
"047_hnsw_vacuum_insert_select\@20" => "SELECT i FROM tst ORDER BY v <-> (SELECT ARRAY[$array_sql]::vector) LIMIT 10;",
"047_hnsw_vacuum_insert_vacuum\@1" => "VACUUM tst;"
}
);

View File

@@ -0,0 +1,39 @@
use strict;
use warnings FATAL => 'all';
use PostgreSQL::Test::Cluster;
use PostgreSQL::Test::Utils;
use Test::More;
my $dim = 3;
my $array_sql = join(",", ('random()') x $dim);
# Initialize node
my $node = PostgreSQL::Test::Cluster->new('node');
$node->init;
$node->start;
# Create table and index
$node->safe_psql("postgres", "CREATE EXTENSION vector;");
$node->safe_psql("postgres", "CREATE TABLE tst (i serial, v vector($dim));");
$node->safe_psql("postgres", "ALTER TABLE tst SET (autovacuum_enabled = false);");
$node->safe_psql("postgres",
"INSERT INTO tst (v) SELECT ARRAY[$array_sql] FROM generate_series(1, 1000) i;"
);
$node->safe_psql("postgres", "CREATE INDEX ON tst USING ivfflat (v vector_l2_ops) WITH (lists = 10);");
# Test no errors
$node->pgbench(
"--no-vacuum --client=5 --transactions=1500",
0,
[qr{actually processed}],
[qr{^$}],
"concurrent INSERTs, DELETEs, SELECTs, and VACUUM",
{
"048_ivfflat_vacuum_insert_insert\@500" => "INSERT INTO tst (v) VALUES (ARRAY[$array_sql]);",
"048_ivfflat_vacuum_insert_delete\@500" => "DELETE FROM tst WHERE i = (SELECT i FROM tst LIMIT 1);",
"048_ivfflat_vacuum_insert_select\@500" => "SELECT i FROM tst ORDER BY v <-> (SELECT ARRAY[$array_sql]::vector) LIMIT 10;",
"048_ivfflat_vacuum_insert_vacuum\@1" => "VACUUM tst;"
}
);
done_testing();

View File

@@ -1,4 +1,4 @@
comment = 'vector data type and ivfflat and hnsw access methods'
default_version = '0.8.4'
default_version = '0.8.5'
module_pathname = '$libdir/vector'
relocatable = true