mirror of
https://github.com/pgvector/pgvector.git
synced 2026-07-22 03:57:34 +08:00
Compare commits
14 Commits
| Author | SHA1 | Date | |
|---|---|---|---|
|
|
ddd873f369 | ||
|
|
8735c4782a | ||
|
|
17ec3f3852 | ||
|
|
8df12a62e7 | ||
|
|
647c9002be | ||
|
|
3e6345f02a | ||
|
|
2b3484ddcd | ||
|
|
1a6debf281 | ||
|
|
bff4d7ea68 | ||
|
|
d5b979e4fa | ||
|
|
d52426b3fa | ||
|
|
b125b2debb | ||
|
|
3d7186a17c | ||
|
|
58ba2139ce |
2
.github/workflows/build.yml
vendored
2
.github/workflows/build.yml
vendored
@@ -11,7 +11,7 @@ jobs:
|
||||
postgres: [14, 13, 12, 11, 10, 9.6]
|
||||
include:
|
||||
- os: macos-latest
|
||||
postgres: 13
|
||||
postgres: 14
|
||||
steps:
|
||||
- uses: actions/checkout@v2
|
||||
- uses: ankane/setup-postgres@v1
|
||||
|
||||
@@ -1,3 +1,11 @@
|
||||
## 0.2.2 (2022-01-15)
|
||||
|
||||
- Fixed compilation error on Mac ARM
|
||||
|
||||
## 0.2.1 (2022-01-02)
|
||||
|
||||
- Fixed `operator is not unique` error
|
||||
|
||||
## 0.2.0 (2021-10-03)
|
||||
|
||||
- Added support for Postgres 14
|
||||
|
||||
@@ -2,7 +2,7 @@
|
||||
"name": "vector",
|
||||
"abstract": "Open-source vector similarity search for Postgres",
|
||||
"description": "Supports L2 distance, inner product, and cosine distance",
|
||||
"version": "0.2.0",
|
||||
"version": "0.2.2",
|
||||
"maintainer": [
|
||||
"Andrew Kane <andrew@ankane.org>"
|
||||
],
|
||||
@@ -20,7 +20,7 @@
|
||||
"vector": {
|
||||
"file": "sql/vector.sql",
|
||||
"docfile": "README.md",
|
||||
"version": "0.2.0",
|
||||
"version": "0.2.2",
|
||||
"abstract": "Open-source vector similarity search for Postgres"
|
||||
}
|
||||
},
|
||||
|
||||
14
Makefile
14
Makefile
@@ -1,5 +1,5 @@
|
||||
EXTENSION = vector
|
||||
EXTVERSION = 0.2.0
|
||||
EXTVERSION = 0.2.2
|
||||
|
||||
MODULE_big = vector
|
||||
DATA = $(wildcard sql/*--*.sql)
|
||||
@@ -11,6 +11,13 @@ REGRESS_OPTS = --inputdir=test
|
||||
|
||||
OPTFLAGS = -march=native
|
||||
|
||||
# Mac ARM doesn't support -march=native
|
||||
ifeq ($(shell uname -s), Darwin)
|
||||
ifeq ($(shell uname -p), arm)
|
||||
OPTFLAGS =
|
||||
endif
|
||||
endif
|
||||
|
||||
# For auto-vectorization:
|
||||
# - GCC (needs -ftree-vectorize OR -O3) - https://gcc.gnu.org/projects/tree-ssa/vectorization.html
|
||||
# - Clang (could use pragma instead) - https://llvm.org/docs/Vectorizers.html
|
||||
@@ -42,3 +49,8 @@ prove_installcheck:
|
||||
dist:
|
||||
mkdir -p dist
|
||||
git archive --format zip --prefix=$(EXTENSION)-$(EXTVERSION)/ --output dist/$(EXTENSION)-$(EXTVERSION).zip master
|
||||
|
||||
.PHONY: docker
|
||||
|
||||
docker:
|
||||
docker build --pull --no-cache -t ankane/pgvector:latest .
|
||||
|
||||
42
README.md
42
README.md
@@ -4,7 +4,7 @@ Open-source vector similarity search for Postgres
|
||||
|
||||
```sql
|
||||
CREATE TABLE table (column vector(3));
|
||||
CREATE INDEX ON table USING ivfflat (column);
|
||||
CREATE INDEX ON table USING ivfflat (column vector_l2_ops);
|
||||
SELECT * FROM table ORDER BY column <-> '[1,2,3]' LIMIT 5;
|
||||
```
|
||||
|
||||
@@ -17,7 +17,7 @@ Supports L2 distance, inner product, and cosine distance
|
||||
Compile and install the extension (supports Postgres 9.6+)
|
||||
|
||||
```sh
|
||||
git clone --branch v0.2.0 https://github.com/ankane/pgvector.git
|
||||
git clone --branch v0.2.2 https://github.com/ankane/pgvector.git
|
||||
cd pgvector
|
||||
make
|
||||
make install # may need sudo
|
||||
@@ -62,7 +62,7 @@ Speed up queries with an approximate index. Add an index for each distance funct
|
||||
L2 distance
|
||||
|
||||
```sql
|
||||
CREATE INDEX ON table USING ivfflat (column);
|
||||
CREATE INDEX ON table USING ivfflat (column vector_l2_ops);
|
||||
```
|
||||
|
||||
Inner product
|
||||
@@ -77,16 +77,30 @@ Cosine distance
|
||||
CREATE INDEX ON table USING ivfflat (column vector_cosine_ops);
|
||||
```
|
||||
|
||||
Indexes should be created after the table has data for optimal clustering. Also, unlike typical indexes which only affect performance, you may see different results for queries after adding an approximate index.
|
||||
Indexes should be created after the table has data for optimal clustering. If the distribution of data changes significantly, you can reindex without downtime:
|
||||
|
||||
```sql
|
||||
-- Postgres 12+
|
||||
REINDEX INDEX CONCURRENTLY index_name;
|
||||
|
||||
-- Postgres < 12
|
||||
CREATE INDEX CONCURRENTLY temp_name ON table USING ivfflat (column opclass);
|
||||
DROP INDEX CONCURRENTLY index_name;
|
||||
ALTER INDEX temp_name RENAME TO index_name;
|
||||
```
|
||||
|
||||
Also, unlike typical indexes which only affect performance, you may see different results for queries after adding an approximate index.
|
||||
|
||||
### Index Options
|
||||
|
||||
Specify the number of inverted lists (100 by default)
|
||||
|
||||
```sql
|
||||
CREATE INDEX ON table USING ivfflat (column) WITH (lists = 100);
|
||||
CREATE INDEX ON table USING ivfflat (column opclass) WITH (lists = 100);
|
||||
```
|
||||
|
||||
A [good place to start](https://github.com/facebookresearch/faiss/issues/112) is `4 * sqrt(rows)`
|
||||
|
||||
### Query Options
|
||||
|
||||
Specify the number of probes (1 by default)
|
||||
@@ -111,11 +125,25 @@ COMMIT;
|
||||
Consider [partial indexes](https://www.postgresql.org/docs/current/indexes-partial.html) for queries with a `WHERE` clause
|
||||
|
||||
```sql
|
||||
CREATE INDEX ON table USING ivfflat (column) WHERE (other_column = 123);
|
||||
CREATE INDEX ON table USING ivfflat (column opclass) WHERE (other_column = 123);
|
||||
```
|
||||
|
||||
To index many different values of `other_column`, consider [partitioning](https://www.postgresql.org/docs/current/ddl-partitioning.html) on `other_column`.
|
||||
|
||||
## Performance
|
||||
|
||||
To speed up queries without an index, increase `max_parallel_workers_per_gather`.
|
||||
|
||||
```sql
|
||||
SET max_parallel_workers_per_gather = 4;
|
||||
```
|
||||
|
||||
To speed up queries with an index, increase the number of inverted lists (at the expense of recall).
|
||||
|
||||
```sql
|
||||
CREATE INDEX ON table USING ivfflat (column opclass) WITH (lists = 1000);
|
||||
```
|
||||
|
||||
## Reference
|
||||
|
||||
### Vector Type
|
||||
@@ -168,7 +196,7 @@ This adds pgvector to the [Postgres image](https://hub.docker.com/_/postgres).
|
||||
You can also build the image manually
|
||||
|
||||
```sh
|
||||
git clone --branch v0.2.0 https://github.com/ankane/pgvector.git
|
||||
git clone --branch v0.2.2 https://github.com/ankane/pgvector.git
|
||||
cd pgvector
|
||||
docker build -t pgvector .
|
||||
```
|
||||
|
||||
19
sql/vector--0.2.0--0.2.1.sql
Normal file
19
sql/vector--0.2.0--0.2.1.sql
Normal file
@@ -0,0 +1,19 @@
|
||||
-- complain if script is sourced in psql, rather than via CREATE EXTENSION
|
||||
\echo Use "ALTER EXTENSION vector UPDATE TO '0.2.1'" to load this file. \quit
|
||||
|
||||
DROP CAST (integer[] AS vector);
|
||||
DROP CAST (real[] AS vector);
|
||||
DROP CAST (double precision[] AS vector);
|
||||
DROP CAST (numeric[] AS vector);
|
||||
|
||||
CREATE CAST (integer[] AS vector)
|
||||
WITH FUNCTION array_to_vector(integer[], integer, boolean) AS ASSIGNMENT;
|
||||
|
||||
CREATE CAST (real[] AS vector)
|
||||
WITH FUNCTION array_to_vector(real[], integer, boolean) AS ASSIGNMENT;
|
||||
|
||||
CREATE CAST (double precision[] AS vector)
|
||||
WITH FUNCTION array_to_vector(double precision[], integer, boolean) AS ASSIGNMENT;
|
||||
|
||||
CREATE CAST (numeric[] AS vector)
|
||||
WITH FUNCTION array_to_vector(numeric[], integer, boolean) AS ASSIGNMENT;
|
||||
2
sql/vector--0.2.1--0.2.2.sql
Normal file
2
sql/vector--0.2.1--0.2.2.sql
Normal file
@@ -0,0 +1,2 @@
|
||||
-- complain if script is sourced in psql, rather than via CREATE EXTENSION
|
||||
\echo Use "ALTER EXTENSION vector UPDATE TO '0.2.2'" to load this file. \quit
|
||||
@@ -108,21 +108,21 @@ CREATE FUNCTION vector_to_float4(vector, integer, boolean) RETURNS real[]
|
||||
CREATE CAST (vector AS vector)
|
||||
WITH FUNCTION vector(vector, integer, boolean) AS IMPLICIT;
|
||||
|
||||
CREATE CAST (integer[] AS vector)
|
||||
WITH FUNCTION array_to_vector(integer[], integer, boolean) AS IMPLICIT;
|
||||
|
||||
CREATE CAST (real[] AS vector)
|
||||
WITH FUNCTION array_to_vector(real[], integer, boolean) AS IMPLICIT;
|
||||
|
||||
CREATE CAST (double precision[] AS vector)
|
||||
WITH FUNCTION array_to_vector(double precision[], integer, boolean) AS IMPLICIT;
|
||||
|
||||
CREATE CAST (numeric[] AS vector)
|
||||
WITH FUNCTION array_to_vector(numeric[], integer, boolean) AS IMPLICIT;
|
||||
|
||||
CREATE CAST (vector AS real[])
|
||||
WITH FUNCTION vector_to_float4(vector, integer, boolean) AS IMPLICIT;
|
||||
|
||||
CREATE CAST (integer[] AS vector)
|
||||
WITH FUNCTION array_to_vector(integer[], integer, boolean) AS ASSIGNMENT;
|
||||
|
||||
CREATE CAST (real[] AS vector)
|
||||
WITH FUNCTION array_to_vector(real[], integer, boolean) AS ASSIGNMENT;
|
||||
|
||||
CREATE CAST (double precision[] AS vector)
|
||||
WITH FUNCTION array_to_vector(double precision[], integer, boolean) AS ASSIGNMENT;
|
||||
|
||||
CREATE CAST (numeric[] AS vector)
|
||||
WITH FUNCTION array_to_vector(numeric[], integer, boolean) AS ASSIGNMENT;
|
||||
|
||||
-- operators
|
||||
|
||||
CREATE OPERATOR <-> (
|
||||
|
||||
@@ -42,3 +42,10 @@ SELECT '[1,2,3]'::vector::real[];
|
||||
|
||||
SELECT array_agg(n)::vector FROM generate_series(1, 1025) n;
|
||||
ERROR: vector cannot have more than 1024 dimensions
|
||||
-- ensure no error
|
||||
SELECT ARRAY[1,2,3] = ARRAY[1,2,3];
|
||||
?column?
|
||||
----------
|
||||
t
|
||||
(1 row)
|
||||
|
||||
|
||||
@@ -12,3 +12,6 @@ SELECT '{-Infinity}'::real[]::vector;
|
||||
SELECT '{}'::real[]::vector;
|
||||
SELECT '[1,2,3]'::vector::real[];
|
||||
SELECT array_agg(n)::vector FROM generate_series(1, 1025) n;
|
||||
|
||||
-- ensure no error
|
||||
SELECT ARRAY[1,2,3] = ARRAY[1,2,3];
|
||||
|
||||
@@ -1,4 +1,4 @@
|
||||
comment = 'vector data type and ivfflat access method'
|
||||
default_version = '0.2.0'
|
||||
default_version = '0.2.2'
|
||||
module_pathname = '$libdir/vector'
|
||||
relocatable = true
|
||||
|
||||
Reference in New Issue
Block a user