Compare commits

..

3 Commits

Author SHA1 Message Date
Andrew Kane
5dcdb5a75a Restored checks [skip ci] 2024-05-26 10:52:04 -04:00
Andrew Kane
2fdadebce0 Skip loading element if outside of max candidate distance 2024-05-26 09:47:15 -04:00
Andrew Kane
385f4a707e Improved performance of on-disk HNSW index builds 2024-05-26 09:19:52 -04:00
102 changed files with 1110 additions and 4813 deletions

View File

@@ -8,18 +8,18 @@ jobs:
fail-fast: false
matrix:
include:
# - postgres: 18
# os: ubuntu-24.04
- postgres: 17
os: ubuntu-24.04
os: ubuntu-22.04
- postgres: 16
os: ubuntu-22.04
- postgres: 15
os: ubuntu-22.04
- postgres: 14
os: ubuntu-20.04
os: ubuntu-22.04
- postgres: 13
os: ubuntu-20.04
- postgres: 12
os: ubuntu-20.04
steps:
- uses: actions/checkout@v4
- uses: ankane/setup-postgres@v1
@@ -49,7 +49,7 @@ jobs:
- postgres: 16
os: macos-14
- postgres: 14
os: macos-13
os: macos-12
steps:
- uses: actions/checkout@v4
- uses: ankane/setup-postgres@v1

View File

@@ -1,29 +1,4 @@
## 0.8.0 (unreleased)
- Added support for iterative index scans
- Added `intvec` type
- Added casts for arrays to `sparsevec`
- Improved cost estimation
- Improved performance of HNSW inserts and on-disk index builds
- Reduced memory usage for HNSW index scans
- Dropped support for Postgres 12
## 0.7.4 (2024-08-05)
- Fixed locking for parallel HNSW index builds
- Fixed compilation error with GCC 14 on i386 when SSE2 is not enabled
## 0.7.3 (2024-07-22)
- Fixed `failed to add index item` error with `sparsevec`
- Fixed compilation error with FreeBSD ARM
- Fixed compilation warning with MSVC and Postgres 16
## 0.7.2 (2024-06-11)
- Fixed initialization fork for indexes on unlogged tables
## 0.7.1 (2024-06-03)
## 0.7.1 (unreleased)
- Improved performance of on-disk HNSW index builds
- Fixed `undefined symbol` error with GCC 8

View File

@@ -1,4 +1,4 @@
ARG PG_MAJOR=17
ARG PG_MAJOR=16
FROM postgres:$PG_MAJOR
ARG PG_MAJOR

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": "0.7.4",
"version": "0.7.0",
"maintainer": [
"Andrew Kane <andrew@ankane.org>"
],
@@ -20,7 +20,7 @@
"vector": {
"file": "sql/vector.sql",
"docfile": "README.md",
"version": "0.7.4",
"version": "0.7.0",
"abstract": "Open-source vector similarity search for Postgres"
}
},

View File

@@ -1,11 +1,10 @@
EXTENSION = vector
EXTVERSION = 0.7.4
EXTVERSION = 0.7.0
MODULE_big = vector
DATA = $(wildcard sql/*--*--*.sql)
DATA_built = sql/$(EXTENSION)--$(EXTVERSION).sql
OBJS = src/bitutils.o src/bitvec.o src/halfutils.o src/halfvec.o src/hnsw.o src/hnswbuild.o src/hnswinsert.o src/hnswscan.o src/hnswutils.o src/hnswvacuum.o src/intvec.o src/ivfbuild.o src/ivfflat.o src/ivfinsert.o src/ivfkmeans.o src/ivfscan.o src/ivfutils.o src/ivfvacuum.o src/sparsevec.o src/vector.o
HEADERS = src/halfvec.h src/intvec.h src/sparsevec.h src/vector.h
DATA = $(wildcard sql/*--*.sql)
OBJS = src/bitutils.o src/bitvec.o src/halfutils.o src/halfvec.o src/hnsw.o src/hnswbuild.o src/hnswinsert.o src/hnswscan.o src/hnswutils.o src/hnswvacuum.o src/ivfbuild.o src/ivfflat.o src/ivfinsert.o src/ivfkmeans.o src/ivfscan.o src/ivfutils.o src/ivfvacuum.o src/sparsevec.o src/vector.o
HEADERS = src/halfvec.h src/sparsevec.h src/vector.h
TESTS = $(wildcard test/sql/*.sql)
REGRESS = $(patsubst test/sql/%.sql,%,$(TESTS))
@@ -43,6 +42,8 @@ all: sql/$(EXTENSION)--$(EXTVERSION).sql
sql/$(EXTENSION)--$(EXTVERSION).sql: sql/$(EXTENSION).sql
cp $< $@
EXTRA_CLEAN = sql/$(EXTENSION)--$(EXTVERSION).sql
PG_CONFIG ?= pg_config
PGXS := $(shell $(PG_CONFIG) --pgxs)
include $(PGXS)
@@ -52,7 +53,7 @@ ifeq ($(PROVE),)
PROVE = prove
endif
# for Postgres < 15
# for Postgres 15
PROVE_FLAGS += -I ./test/perl
prove_installcheck:
@@ -66,7 +67,7 @@ dist:
git archive --format zip --prefix=$(EXTENSION)-$(EXTVERSION)/ --output dist/$(EXTENSION)-$(EXTVERSION).zip master
# for Docker
PG_MAJOR ?= 17
PG_MAJOR ?= 16
.PHONY: docker

View File

@@ -1,9 +1,8 @@
EXTENSION = vector
EXTVERSION = 0.7.4
EXTVERSION = 0.7.0
DATA_built = sql\$(EXTENSION)--$(EXTVERSION).sql
OBJS = src\bitutils.obj src\bitvec.obj src\halfutils.obj src\halfvec.obj src\hnsw.obj src\hnswbuild.obj src\hnswinsert.obj src\hnswscan.obj src\hnswutils.obj src\hnswvacuum.obj src\intvec.obj src\ivfbuild.obj src\ivfflat.obj src\ivfinsert.obj src\ivfkmeans.obj src\ivfscan.obj src\ivfutils.obj src\ivfvacuum.obj src\sparsevec.obj src\vector.obj
HEADERS = src\halfvec.h src\intvec.h src\sparsevec.h src\vector.h
OBJS = src\bitutils.obj src\bitvec.obj src\halfutils.obj src\halfvec.obj src\hnsw.obj src\hnswbuild.obj src\hnswinsert.obj src\hnswscan.obj src\hnswutils.obj src\hnswvacuum.obj src\ivfbuild.obj src\ivfflat.obj src\ivfinsert.obj src\ivfkmeans.obj src\ivfscan.obj src\ivfutils.obj src\ivfvacuum.obj src\sparsevec.obj src\vector.obj
HEADERS = src\halfvec.h src\sparsevec.h src\vector.h
REGRESS = bit btree cast copy halfvec hnsw_bit hnsw_halfvec hnsw_sparsevec hnsw_vector ivfflat_bit ivfflat_halfvec ivfflat_vector sparsevec vector_type
REGRESS_OPTS = --inputdir=test --load-extension=$(EXTENSION)
@@ -20,6 +19,11 @@ PG_CFLAGS = $(PG_CFLAGS) $(OPTFLAGS) /O2 /fp:fast
# https://learn.microsoft.com/en-us/cpp/error-messages/tool-errors/vectorizer-and-parallelizer-messages
# PG_CFLAGS = $(PG_CFLAGS) /Qvec-report:2
all: sql\$(EXTENSION)--$(EXTVERSION).sql
sql\$(EXTENSION)--$(EXTVERSION).sql: sql\$(EXTENSION).sql
copy sql\$(EXTENSION).sql $@
# TODO use pg_config
!ifndef PGROOT
!error PGROOT is not set
@@ -39,18 +43,15 @@ SHLIB = $(EXTENSION).dll
LIBS = "$(LIBDIR)\postgres.lib"
all: $(SHLIB) $(DATA_built)
.c.obj:
$(CC) $(CFLAGS) /c $< /Fo$@
$(SHLIB): $(OBJS)
$(CC) $(CFLAGS) $(OBJS) $(LIBS) /link /DLL /OUT:$(SHLIB)
sql\$(EXTENSION)--$(EXTVERSION).sql: sql\$(EXTENSION).sql
copy sql\$(EXTENSION).sql $@
all: $(SHLIB)
install: all
install:
copy $(SHLIB) "$(PKGLIBDIR)"
copy $(EXTENSION).control "$(SHAREDIR)\extension"
copy sql\$(EXTENSION)--*.sql "$(SHAREDIR)\extension"
@@ -69,6 +70,6 @@ uninstall:
clean:
del /f $(SHLIB) $(EXTENSION).lib $(EXTENSION).exp
del /f $(DATA_built)
del /f $(OBJS)
del /f sql\$(EXTENSION)--$(EXTVERSION).sql
del /f /s /q results regression.diffs regression.out tmp_check tmp_check_iso log output_iso

View File

@@ -21,7 +21,7 @@ Compile and install the extension (supports Postgres 12+)
```sh
cd /tmp
git clone --branch v0.7.4 https://github.com/pgvector/pgvector.git
git clone --branch v0.7.0 https://github.com/pgvector/pgvector.git
cd pgvector
make
make install # may need sudo
@@ -46,14 +46,12 @@ Then use `nmake` to build:
```cmd
set "PGROOT=C:\Program Files\PostgreSQL\16"
cd %TEMP%
git clone --branch v0.7.4 https://github.com/pgvector/pgvector.git
git clone --branch v0.7.0 https://github.com/pgvector/pgvector.git
cd pgvector
nmake /F Makefile.win
nmake /F Makefile.win install
```
Note: Postgres 17 is not supported yet due to an upstream issue
See the [installation notes](#installation-notes---windows) if you run into issues
You can also install it with [Docker](#docker) or [conda-forge](#conda-forge).
@@ -102,15 +100,13 @@ Or add a vector column to an existing table
ALTER TABLE items ADD COLUMN embedding vector(3);
```
Also supports [half-precision](#half-precision-vectors), [binary](#binary-vectors), and [sparse](#sparse-vectors) vectors
Insert vectors
```sql
INSERT INTO items (embedding) VALUES ('[1,2,3]'), ('[4,5,6]');
```
Or load vectors in bulk using `COPY` ([example](https://github.com/pgvector/pgvector-python/blob/master/examples/loading/example.py))
Or load vectors in bulk using `COPY` ([example](https://github.com/pgvector/pgvector-python/blob/master/examples/bulk_loading.py))
```sql
COPY items (embedding) FROM STDIN WITH (FORMAT BINARY);
@@ -149,8 +145,6 @@ Supported distance functions are:
- `<#>` - (negative) inner product
- `<=>` - cosine distance
- `<+>` - L1 distance (added in 0.7.0)
- `<~>` - Hamming distance (binary vectors, added in 0.7.0)
- `<%>` - Jaccard distance (binary vectors, added in 0.7.0)
Get the nearest neighbors to a row
@@ -208,7 +202,7 @@ You can add an index to use approximate nearest neighbor search, which trades so
Supported index types are:
- [HNSW](#hnsw)
- [HNSW](#hnsw) - added in 0.5.0
- [IVFFlat](#ivfflat)
## HNSW
@@ -223,7 +217,7 @@ L2 distance
CREATE INDEX ON items USING hnsw (embedding vector_l2_ops);
```
Note: Use `halfvec_l2_ops` for `halfvec`, `intvec_l2_ops` for `intvec`, and `sparsevec_l2_ops` for `sparsevec` (and similar with the other distance functions)
Note: Use `halfvec_l2_ops` for `halfvec` and `sparsevec_l2_ops` for `sparsevec` (and similar with the other distance functions)
Inner product
@@ -259,7 +253,6 @@ Supported types are:
- `vector` - up to 2,000 dimensions
- `halfvec` - up to 4,000 dimensions (added in 0.7.0)
- `intvec` - up to 8,000 dimensions (added in 0.8.0)
- `bit` - up to 64,000 dimensions (added in 0.7.0)
- `sparsevec` - up to 1,000 non-zero elements (added in 0.7.0)
@@ -478,19 +471,9 @@ Get the nearest neighbors
SELECT * FROM items ORDER BY embedding::halfvec(3) <-> '[1,2,3]' LIMIT 5;
```
## Integer Vectors
*Added in 0.8.0*
Use the `intvec` type to store 8-bit integer vectors
```sql
CREATE TABLE items (id bigserial PRIMARY KEY, embedding intvec(3));
```
## Binary Vectors
Use the `bit` type to store binary vectors ([example](https://github.com/pgvector/pgvector-python/blob/master/examples/imagehash/example.py))
Use the `bit` type to store binary vectors ([example](https://github.com/pgvector/pgvector-python/blob/master/examples/hash_image_search.py))
```sql
CREATE TABLE items (id bigserial PRIMARY KEY, embedding bit(3));
@@ -568,7 +551,7 @@ SELECT id, content FROM items, plainto_tsquery('hello search') query
WHERE textsearch @@ query ORDER BY ts_rank_cd(textsearch, query) DESC LIMIT 5;
```
You can use [Reciprocal Rank Fusion](https://github.com/pgvector/pgvector-python/blob/master/examples/hybrid_search/rrf.py) or a [cross-encoder](https://github.com/pgvector/pgvector-python/blob/master/examples/hybrid_search/cross_encoder.py) to combine results.
You can use [Reciprocal Rank Fusion](https://github.com/pgvector/pgvector-python/blob/master/examples/hybrid_search_rrf.py) or a [cross-encoder](https://github.com/pgvector/pgvector-python/blob/master/examples/hybrid_search.py) to combine results.
## Indexing Subvectors
@@ -614,7 +597,7 @@ Be sure to restart Postgres for changes to take effect.
### Loading
Use `COPY` for bulk loading data ([example](https://github.com/pgvector/pgvector-python/blob/master/examples/loading/example.py)).
Use `COPY` for bulk loading data ([example](https://github.com/pgvector/pgvector-python/blob/master/examples/bulk_loading.py)).
```sql
COPY items (embedding) FROM STDIN WITH (FORMAT BINARY);
@@ -704,7 +687,7 @@ Scale pgvector the same way you scale Postgres.
Scale vertically by increasing memory, CPU, and storage on a single instance. Use existing tools to [tune parameters](#tuning) and [monitor performance](#monitoring).
Scale horizontally with [replicas](https://www.postgresql.org/docs/current/hot-standby.html), or use [Citus](https://github.com/citusdata/citus) or another approach for sharding ([example](https://github.com/pgvector/pgvector-python/blob/master/examples/citus/example.py)).
Scale horizontally with [replicas](https://www.postgresql.org/docs/current/hot-standby.html), or use [Citus](https://github.com/citusdata/citus) or another approach for sharding ([example](https://github.com/pgvector/pgvector-python/blob/master/examples/citus.py)).
## Languages
@@ -951,30 +934,6 @@ Function | Description | Added
avg(halfvec) → halfvec | average | 0.7.0
sum(halfvec) → halfvec | sum | 0.7.0
### Intvec Type
Each int vector takes `dimensions + 8` bytes of storage. Each element is a single byte signed integer. Int vectors can have up to 16,000 dimensions.
### Intvec Operators
Operator | Description | Added
--- | --- | ---
<-> | Euclidean distance | 0.8.0
<#> | negative inner product | 0.8.0
<=> | cosine distance | 0.8.0
<+> | taxicab distance | 0.8.0
### Intvec Functions
Function | Description | Added
--- | --- | ---
cosine_distance(intvec, intvec) → double precision | cosine distance | 0.8.0
inner_product(intvec, intvec) → double precision | inner product | 0.8.0
l1_distance(intvec, intvec) → double precision | taxicab distance | 0.8.0
l2_distance(intvec, intvec) → double precision | Euclidean distance | 0.8.0
l2_norm(intvec) → double precision | Euclidean norm | 0.8.0
vector_dims(intvec) → integer | number of dimensions | 0.8.0
### Bit Type
Each bit vector takes `dimensions / 8 + 8` bytes of storage. See the [Postgres docs](https://www.postgresql.org/docs/current/datatype-bit.html) for more info.
@@ -1024,7 +983,7 @@ l2_normalize(sparsevec) → sparsevec | Normalize with Euclidean norm | 0.7.0
If your machine has multiple Postgres installations, specify the path to [pg_config](https://www.postgresql.org/docs/current/app-pgconfig.html) with:
```sh
export PG_CONFIG=/Library/PostgreSQL/17/bin/pg_config
export PG_CONFIG=/Library/PostgreSQL/16/bin/pg_config
```
Then re-run the installation instructions (run `make clean` before `make` if needed). If `sudo` is needed for `make install`, use:
@@ -1035,11 +994,11 @@ sudo --preserve-env=PG_CONFIG make install
A few common paths on Mac are:
- EDB installer - `/Library/PostgreSQL/17/bin/pg_config`
- Homebrew (arm64) - `/opt/homebrew/opt/postgresql@17/bin/pg_config`
- Homebrew (x86-64) - `/usr/local/opt/postgresql@17/bin/pg_config`
- EDB installer - `/Library/PostgreSQL/16/bin/pg_config`
- Homebrew (arm64) - `/opt/homebrew/opt/postgresql@16/bin/pg_config`
- Homebrew (x86-64) - `/usr/local/opt/postgresql@16/bin/pg_config`
Note: Replace `17` with your Postgres server version
Note: Replace `16` with your Postgres server version
### Missing Header
@@ -1048,10 +1007,10 @@ If compilation fails with `fatal error: postgres.h: No such file or directory`,
For Ubuntu and Debian, use:
```sh
sudo apt install postgresql-server-dev-17
sudo apt install postgresql-server-dev-16
```
Note: Replace `17` with your Postgres server version
Note: Replace `16` with your Postgres server version
### Missing SDK
@@ -1084,17 +1043,17 @@ If installation fails with `Access is denied`, re-run the installation instructi
Get the [Docker image](https://hub.docker.com/r/pgvector/pgvector) with:
```sh
docker pull pgvector/pgvector:pg17
docker pull pgvector/pgvector:pg16
```
This adds pgvector to the [Postgres image](https://hub.docker.com/_/postgres) (replace `17` with your Postgres server version, and run it the same way).
This adds pgvector to the [Postgres image](https://hub.docker.com/_/postgres) (replace `16` with your Postgres server version, and run it the same way).
You can also build the image manually:
```sh
git clone --branch v0.7.4 https://github.com/pgvector/pgvector.git
git clone --branch v0.7.0 https://github.com/pgvector/pgvector.git
cd pgvector
docker build --pull --build-arg PG_MAJOR=17 -t myuser/pgvector .
docker build --pull --build-arg PG_MAJOR=16 -t myuser/pgvector .
```
### Homebrew
@@ -1105,7 +1064,7 @@ With Homebrew Postgres, you can use:
brew install pgvector
```
Note: This only adds it to the `postgresql@17` and `postgresql@14` formulas
Note: This only adds it to the `postgresql@14` formula
### PGXN
@@ -1120,22 +1079,22 @@ pgxn install vector
Debian and Ubuntu packages are available from the [PostgreSQL APT Repository](https://wiki.postgresql.org/wiki/Apt). Follow the [setup instructions](https://wiki.postgresql.org/wiki/Apt#Quickstart) and run:
```sh
sudo apt install postgresql-17-pgvector
sudo apt install postgresql-16-pgvector
```
Note: Replace `17` with your Postgres server version
Note: Replace `16` with your Postgres server version
### Yum
RPM packages are available from the [PostgreSQL Yum Repository](https://yum.postgresql.org/). Follow the [setup instructions](https://www.postgresql.org/download/linux/redhat/) for your distribution and run:
```sh
sudo yum install pgvector_17
sudo yum install pgvector_16
# or
sudo dnf install pgvector_17
sudo dnf install pgvector_16
```
Note: Replace `17` with your Postgres server version
Note: Replace `16` with your Postgres server version
### pkg

View File

@@ -1,2 +0,0 @@
-- complain if script is sourced in psql, rather than via CREATE EXTENSION
\echo Use "ALTER EXTENSION vector UPDATE TO '0.7.1'" to load this file. \quit

View File

@@ -1,2 +0,0 @@
-- complain if script is sourced in psql, rather than via CREATE EXTENSION
\echo Use "ALTER EXTENSION vector UPDATE TO '0.7.2'" to load this file. \quit

View File

@@ -1,2 +0,0 @@
-- complain if script is sourced in psql, rather than via CREATE EXTENSION
\echo Use "ALTER EXTENSION vector UPDATE TO '0.7.3'" to load this file. \quit

View File

@@ -1,2 +0,0 @@
-- complain if script is sourced in psql, rather than via CREATE EXTENSION
\echo Use "ALTER EXTENSION vector UPDATE TO '0.7.4'" to load this file. \quit

View File

@@ -1,208 +0,0 @@
-- complain if script is sourced in psql, rather than via CREATE EXTENSION
\echo Use "ALTER EXTENSION vector UPDATE TO '0.8.0'" to load this file. \quit
CREATE FUNCTION hnsw_intvec_support(internal) RETURNS internal
AS 'MODULE_PATHNAME' LANGUAGE C;
CREATE TYPE intvec;
CREATE FUNCTION intvec_in(cstring, oid, integer) RETURNS intvec
AS 'MODULE_PATHNAME' LANGUAGE C IMMUTABLE STRICT PARALLEL SAFE;
CREATE FUNCTION intvec_out(intvec) RETURNS cstring
AS 'MODULE_PATHNAME' LANGUAGE C IMMUTABLE STRICT PARALLEL SAFE;
CREATE FUNCTION intvec_typmod_in(cstring[]) RETURNS integer
AS 'MODULE_PATHNAME' LANGUAGE C IMMUTABLE STRICT PARALLEL SAFE;
CREATE FUNCTION intvec_recv(internal, oid, integer) RETURNS intvec
AS 'MODULE_PATHNAME' LANGUAGE C IMMUTABLE STRICT PARALLEL SAFE;
CREATE FUNCTION intvec_send(intvec) RETURNS bytea
AS 'MODULE_PATHNAME' LANGUAGE C IMMUTABLE STRICT PARALLEL SAFE;
CREATE TYPE intvec (
INPUT = intvec_in,
OUTPUT = intvec_out,
TYPMOD_IN = intvec_typmod_in,
RECEIVE = intvec_recv,
SEND = intvec_send,
STORAGE = external
);
CREATE FUNCTION l2_distance(intvec, intvec) RETURNS float8
AS 'MODULE_PATHNAME', 'intvec_l2_distance' LANGUAGE C IMMUTABLE STRICT PARALLEL SAFE;
CREATE FUNCTION inner_product(intvec, intvec) RETURNS float8
AS 'MODULE_PATHNAME', 'intvec_inner_product' LANGUAGE C IMMUTABLE STRICT PARALLEL SAFE;
CREATE FUNCTION cosine_distance(intvec, intvec) RETURNS float8
AS 'MODULE_PATHNAME', 'intvec_cosine_distance' LANGUAGE C IMMUTABLE STRICT PARALLEL SAFE;
CREATE FUNCTION l1_distance(intvec, intvec) RETURNS float8
AS 'MODULE_PATHNAME', 'intvec_l1_distance' LANGUAGE C IMMUTABLE STRICT PARALLEL SAFE;
CREATE FUNCTION vector_dims(intvec) RETURNS integer
AS 'MODULE_PATHNAME', 'intvec_vector_dims' LANGUAGE C IMMUTABLE STRICT PARALLEL SAFE;
CREATE FUNCTION l2_norm(intvec) RETURNS float8
AS 'MODULE_PATHNAME', 'intvec_l2_norm' LANGUAGE C IMMUTABLE STRICT PARALLEL SAFE;
CREATE FUNCTION intvec_lt(intvec, intvec) RETURNS bool
AS 'MODULE_PATHNAME' LANGUAGE C IMMUTABLE STRICT PARALLEL SAFE;
CREATE FUNCTION intvec_le(intvec, intvec) RETURNS bool
AS 'MODULE_PATHNAME' LANGUAGE C IMMUTABLE STRICT PARALLEL SAFE;
CREATE FUNCTION intvec_eq(intvec, intvec) RETURNS bool
AS 'MODULE_PATHNAME' LANGUAGE C IMMUTABLE STRICT PARALLEL SAFE;
CREATE FUNCTION intvec_ne(intvec, intvec) RETURNS bool
AS 'MODULE_PATHNAME' LANGUAGE C IMMUTABLE STRICT PARALLEL SAFE;
CREATE FUNCTION intvec_ge(intvec, intvec) RETURNS bool
AS 'MODULE_PATHNAME' LANGUAGE C IMMUTABLE STRICT PARALLEL SAFE;
CREATE FUNCTION intvec_gt(intvec, intvec) RETURNS bool
AS 'MODULE_PATHNAME' LANGUAGE C IMMUTABLE STRICT PARALLEL SAFE;
CREATE FUNCTION intvec_cmp(intvec, intvec) RETURNS int4
AS 'MODULE_PATHNAME' LANGUAGE C IMMUTABLE STRICT PARALLEL SAFE;
CREATE FUNCTION intvec_l2_squared_distance(intvec, intvec) RETURNS float8
AS 'MODULE_PATHNAME' LANGUAGE C IMMUTABLE STRICT PARALLEL SAFE;
CREATE FUNCTION intvec_negative_inner_product(intvec, intvec) RETURNS float8
AS 'MODULE_PATHNAME' LANGUAGE C IMMUTABLE STRICT PARALLEL SAFE;
CREATE FUNCTION intvec(intvec, integer, boolean) RETURNS intvec
AS 'MODULE_PATHNAME' LANGUAGE C IMMUTABLE STRICT PARALLEL SAFE;
CREATE FUNCTION array_to_intvec(integer[], integer, boolean) RETURNS intvec
AS 'MODULE_PATHNAME' LANGUAGE C IMMUTABLE STRICT PARALLEL SAFE;
CREATE FUNCTION intvec_to_integer(intvec, integer, boolean) RETURNS integer[]
AS 'MODULE_PATHNAME' LANGUAGE C IMMUTABLE STRICT PARALLEL SAFE;
CREATE CAST (intvec AS intvec)
WITH FUNCTION intvec(intvec, integer, boolean) AS IMPLICIT;
CREATE CAST (intvec AS integer[])
WITH FUNCTION intvec_to_integer(intvec, integer, boolean) AS ASSIGNMENT;
CREATE CAST (integer[] AS intvec)
WITH FUNCTION array_to_intvec(integer[], integer, boolean) AS ASSIGNMENT;
CREATE OPERATOR <-> (
LEFTARG = intvec, RIGHTARG = intvec, PROCEDURE = l2_distance,
COMMUTATOR = '<->'
);
CREATE OPERATOR <#> (
LEFTARG = intvec, RIGHTARG = intvec, PROCEDURE = intvec_negative_inner_product,
COMMUTATOR = '<#>'
);
CREATE OPERATOR <=> (
LEFTARG = intvec, RIGHTARG = intvec, PROCEDURE = cosine_distance,
COMMUTATOR = '<=>'
);
CREATE OPERATOR <+> (
LEFTARG = intvec, RIGHTARG = intvec, PROCEDURE = l1_distance,
COMMUTATOR = '<+>'
);
CREATE OPERATOR < (
LEFTARG = intvec, RIGHTARG = intvec, PROCEDURE = intvec_lt,
COMMUTATOR = > , NEGATOR = >= ,
RESTRICT = scalarltsel, JOIN = scalarltjoinsel
);
CREATE OPERATOR <= (
LEFTARG = intvec, RIGHTARG = intvec, PROCEDURE = intvec_le,
COMMUTATOR = >= , NEGATOR = > ,
RESTRICT = scalarlesel, JOIN = scalarlejoinsel
);
CREATE OPERATOR = (
LEFTARG = intvec, RIGHTARG = intvec, PROCEDURE = intvec_eq,
COMMUTATOR = = , NEGATOR = <> ,
RESTRICT = eqsel, JOIN = eqjoinsel
);
CREATE OPERATOR <> (
LEFTARG = intvec, RIGHTARG = intvec, PROCEDURE = intvec_ne,
COMMUTATOR = <> , NEGATOR = = ,
RESTRICT = eqsel, JOIN = eqjoinsel
);
CREATE OPERATOR >= (
LEFTARG = intvec, RIGHTARG = intvec, PROCEDURE = intvec_ge,
COMMUTATOR = <= , NEGATOR = < ,
RESTRICT = scalargesel, JOIN = scalargejoinsel
);
CREATE OPERATOR > (
LEFTARG = intvec, RIGHTARG = intvec, PROCEDURE = intvec_gt,
COMMUTATOR = < , NEGATOR = <= ,
RESTRICT = scalargtsel, JOIN = scalargtjoinsel
);
CREATE OPERATOR CLASS intvec_ops
DEFAULT FOR TYPE intvec USING btree AS
OPERATOR 1 < ,
OPERATOR 2 <= ,
OPERATOR 3 = ,
OPERATOR 4 >= ,
OPERATOR 5 > ,
FUNCTION 1 intvec_cmp(intvec, intvec);
CREATE OPERATOR CLASS intvec_l2_ops
FOR TYPE intvec USING hnsw AS
OPERATOR 1 <-> (intvec, intvec) FOR ORDER BY float_ops,
FUNCTION 1 intvec_l2_squared_distance(intvec, intvec),
FUNCTION 3 hnsw_intvec_support(internal);
CREATE OPERATOR CLASS intvec_ip_ops
FOR TYPE intvec USING hnsw AS
OPERATOR 1 <#> (intvec, intvec) FOR ORDER BY float_ops,
FUNCTION 1 intvec_negative_inner_product(intvec, intvec),
FUNCTION 3 hnsw_intvec_support(internal);
CREATE OPERATOR CLASS intvec_cosine_ops
FOR TYPE intvec USING hnsw AS
OPERATOR 1 <=> (intvec, intvec) FOR ORDER BY float_ops,
FUNCTION 1 cosine_distance(intvec, intvec),
FUNCTION 2 l2_norm(intvec),
FUNCTION 3 hnsw_intvec_support(internal);
CREATE OPERATOR CLASS intvec_l1_ops
FOR TYPE intvec USING hnsw AS
OPERATOR 1 <+> (intvec, intvec) FOR ORDER BY float_ops,
FUNCTION 1 l1_distance(intvec, intvec),
FUNCTION 3 hnsw_intvec_support(internal);
CREATE FUNCTION array_to_sparsevec(integer[], integer, boolean) RETURNS sparsevec
AS 'MODULE_PATHNAME' LANGUAGE C IMMUTABLE STRICT PARALLEL SAFE;
CREATE FUNCTION array_to_sparsevec(real[], integer, boolean) RETURNS sparsevec
AS 'MODULE_PATHNAME' LANGUAGE C IMMUTABLE STRICT PARALLEL SAFE;
CREATE FUNCTION array_to_sparsevec(double precision[], integer, boolean) RETURNS sparsevec
AS 'MODULE_PATHNAME' LANGUAGE C IMMUTABLE STRICT PARALLEL SAFE;
CREATE FUNCTION array_to_sparsevec(numeric[], integer, boolean) RETURNS sparsevec
AS 'MODULE_PATHNAME' LANGUAGE C IMMUTABLE STRICT PARALLEL SAFE;
CREATE CAST (integer[] AS sparsevec)
WITH FUNCTION array_to_sparsevec(integer[], integer, boolean) AS ASSIGNMENT;
CREATE CAST (real[] AS sparsevec)
WITH FUNCTION array_to_sparsevec(real[], integer, boolean) AS ASSIGNMENT;
CREATE CAST (double precision[] AS sparsevec)
WITH FUNCTION array_to_sparsevec(double precision[], integer, boolean) AS ASSIGNMENT;
CREATE CAST (numeric[] AS sparsevec)
WITH FUNCTION array_to_sparsevec(numeric[], integer, boolean) AS ASSIGNMENT;

View File

@@ -272,9 +272,6 @@ CREATE FUNCTION ivfflat_bit_support(internal) RETURNS internal
CREATE FUNCTION hnsw_halfvec_support(internal) RETURNS internal
AS 'MODULE_PATHNAME' LANGUAGE C;
CREATE FUNCTION hnsw_intvec_support(internal) RETURNS internal
AS 'MODULE_PATHNAME' LANGUAGE C;
CREATE FUNCTION hnsw_bit_support(internal) RETURNS internal
AS 'MODULE_PATHNAME' LANGUAGE C;
@@ -650,199 +647,6 @@ CREATE OPERATOR CLASS halfvec_l1_ops
FUNCTION 1 l1_distance(halfvec, halfvec),
FUNCTION 3 hnsw_halfvec_support(internal);
-- intvec type
CREATE TYPE intvec;
CREATE FUNCTION intvec_in(cstring, oid, integer) RETURNS intvec
AS 'MODULE_PATHNAME' LANGUAGE C IMMUTABLE STRICT PARALLEL SAFE;
CREATE FUNCTION intvec_out(intvec) RETURNS cstring
AS 'MODULE_PATHNAME' LANGUAGE C IMMUTABLE STRICT PARALLEL SAFE;
CREATE FUNCTION intvec_typmod_in(cstring[]) RETURNS integer
AS 'MODULE_PATHNAME' LANGUAGE C IMMUTABLE STRICT PARALLEL SAFE;
CREATE FUNCTION intvec_recv(internal, oid, integer) RETURNS intvec
AS 'MODULE_PATHNAME' LANGUAGE C IMMUTABLE STRICT PARALLEL SAFE;
CREATE FUNCTION intvec_send(intvec) RETURNS bytea
AS 'MODULE_PATHNAME' LANGUAGE C IMMUTABLE STRICT PARALLEL SAFE;
CREATE TYPE intvec (
INPUT = intvec_in,
OUTPUT = intvec_out,
TYPMOD_IN = intvec_typmod_in,
RECEIVE = intvec_recv,
SEND = intvec_send,
STORAGE = external
);
-- intvec functions
CREATE FUNCTION l2_distance(intvec, intvec) RETURNS float8
AS 'MODULE_PATHNAME', 'intvec_l2_distance' LANGUAGE C IMMUTABLE STRICT PARALLEL SAFE;
CREATE FUNCTION inner_product(intvec, intvec) RETURNS float8
AS 'MODULE_PATHNAME', 'intvec_inner_product' LANGUAGE C IMMUTABLE STRICT PARALLEL SAFE;
CREATE FUNCTION cosine_distance(intvec, intvec) RETURNS float8
AS 'MODULE_PATHNAME', 'intvec_cosine_distance' LANGUAGE C IMMUTABLE STRICT PARALLEL SAFE;
CREATE FUNCTION l1_distance(intvec, intvec) RETURNS float8
AS 'MODULE_PATHNAME', 'intvec_l1_distance' LANGUAGE C IMMUTABLE STRICT PARALLEL SAFE;
CREATE FUNCTION vector_dims(intvec) RETURNS integer
AS 'MODULE_PATHNAME', 'intvec_vector_dims' LANGUAGE C IMMUTABLE STRICT PARALLEL SAFE;
CREATE FUNCTION l2_norm(intvec) RETURNS float8
AS 'MODULE_PATHNAME', 'intvec_l2_norm' LANGUAGE C IMMUTABLE STRICT PARALLEL SAFE;
-- intvec private functions
CREATE FUNCTION intvec_lt(intvec, intvec) RETURNS bool
AS 'MODULE_PATHNAME' LANGUAGE C IMMUTABLE STRICT PARALLEL SAFE;
CREATE FUNCTION intvec_le(intvec, intvec) RETURNS bool
AS 'MODULE_PATHNAME' LANGUAGE C IMMUTABLE STRICT PARALLEL SAFE;
CREATE FUNCTION intvec_eq(intvec, intvec) RETURNS bool
AS 'MODULE_PATHNAME' LANGUAGE C IMMUTABLE STRICT PARALLEL SAFE;
CREATE FUNCTION intvec_ne(intvec, intvec) RETURNS bool
AS 'MODULE_PATHNAME' LANGUAGE C IMMUTABLE STRICT PARALLEL SAFE;
CREATE FUNCTION intvec_ge(intvec, intvec) RETURNS bool
AS 'MODULE_PATHNAME' LANGUAGE C IMMUTABLE STRICT PARALLEL SAFE;
CREATE FUNCTION intvec_gt(intvec, intvec) RETURNS bool
AS 'MODULE_PATHNAME' LANGUAGE C IMMUTABLE STRICT PARALLEL SAFE;
CREATE FUNCTION intvec_cmp(intvec, intvec) RETURNS int4
AS 'MODULE_PATHNAME' LANGUAGE C IMMUTABLE STRICT PARALLEL SAFE;
CREATE FUNCTION intvec_l2_squared_distance(intvec, intvec) RETURNS float8
AS 'MODULE_PATHNAME' LANGUAGE C IMMUTABLE STRICT PARALLEL SAFE;
CREATE FUNCTION intvec_negative_inner_product(intvec, intvec) RETURNS float8
AS 'MODULE_PATHNAME' LANGUAGE C IMMUTABLE STRICT PARALLEL SAFE;
-- intvec cast functions
CREATE FUNCTION intvec(intvec, integer, boolean) RETURNS intvec
AS 'MODULE_PATHNAME' LANGUAGE C IMMUTABLE STRICT PARALLEL SAFE;
CREATE FUNCTION array_to_intvec(integer[], integer, boolean) RETURNS intvec
AS 'MODULE_PATHNAME' LANGUAGE C IMMUTABLE STRICT PARALLEL SAFE;
CREATE FUNCTION intvec_to_integer(intvec, integer, boolean) RETURNS integer[]
AS 'MODULE_PATHNAME' LANGUAGE C IMMUTABLE STRICT PARALLEL SAFE;
-- intvec casts
CREATE CAST (intvec AS intvec)
WITH FUNCTION intvec(intvec, integer, boolean) AS IMPLICIT;
CREATE CAST (intvec AS integer[])
WITH FUNCTION intvec_to_integer(intvec, integer, boolean) AS ASSIGNMENT;
CREATE CAST (integer[] AS intvec)
WITH FUNCTION array_to_intvec(integer[], integer, boolean) AS ASSIGNMENT;
-- intvec operators
CREATE OPERATOR <-> (
LEFTARG = intvec, RIGHTARG = intvec, PROCEDURE = l2_distance,
COMMUTATOR = '<->'
);
CREATE OPERATOR <#> (
LEFTARG = intvec, RIGHTARG = intvec, PROCEDURE = intvec_negative_inner_product,
COMMUTATOR = '<#>'
);
CREATE OPERATOR <=> (
LEFTARG = intvec, RIGHTARG = intvec, PROCEDURE = cosine_distance,
COMMUTATOR = '<=>'
);
CREATE OPERATOR <+> (
LEFTARG = intvec, RIGHTARG = intvec, PROCEDURE = l1_distance,
COMMUTATOR = '<+>'
);
CREATE OPERATOR < (
LEFTARG = intvec, RIGHTARG = intvec, PROCEDURE = intvec_lt,
COMMUTATOR = > , NEGATOR = >= ,
RESTRICT = scalarltsel, JOIN = scalarltjoinsel
);
CREATE OPERATOR <= (
LEFTARG = intvec, RIGHTARG = intvec, PROCEDURE = intvec_le,
COMMUTATOR = >= , NEGATOR = > ,
RESTRICT = scalarlesel, JOIN = scalarlejoinsel
);
CREATE OPERATOR = (
LEFTARG = intvec, RIGHTARG = intvec, PROCEDURE = intvec_eq,
COMMUTATOR = = , NEGATOR = <> ,
RESTRICT = eqsel, JOIN = eqjoinsel
);
CREATE OPERATOR <> (
LEFTARG = intvec, RIGHTARG = intvec, PROCEDURE = intvec_ne,
COMMUTATOR = <> , NEGATOR = = ,
RESTRICT = eqsel, JOIN = eqjoinsel
);
CREATE OPERATOR >= (
LEFTARG = intvec, RIGHTARG = intvec, PROCEDURE = intvec_ge,
COMMUTATOR = <= , NEGATOR = < ,
RESTRICT = scalargesel, JOIN = scalargejoinsel
);
CREATE OPERATOR > (
LEFTARG = intvec, RIGHTARG = intvec, PROCEDURE = intvec_gt,
COMMUTATOR = < , NEGATOR = <= ,
RESTRICT = scalargtsel, JOIN = scalargtjoinsel
);
-- intvec opclasses
CREATE OPERATOR CLASS intvec_ops
DEFAULT FOR TYPE intvec USING btree AS
OPERATOR 1 < ,
OPERATOR 2 <= ,
OPERATOR 3 = ,
OPERATOR 4 >= ,
OPERATOR 5 > ,
FUNCTION 1 intvec_cmp(intvec, intvec);
CREATE OPERATOR CLASS intvec_l2_ops
FOR TYPE intvec USING hnsw AS
OPERATOR 1 <-> (intvec, intvec) FOR ORDER BY float_ops,
FUNCTION 1 intvec_l2_squared_distance(intvec, intvec),
FUNCTION 3 hnsw_intvec_support(internal);
CREATE OPERATOR CLASS intvec_ip_ops
FOR TYPE intvec USING hnsw AS
OPERATOR 1 <#> (intvec, intvec) FOR ORDER BY float_ops,
FUNCTION 1 intvec_negative_inner_product(intvec, intvec),
FUNCTION 3 hnsw_intvec_support(internal);
CREATE OPERATOR CLASS intvec_cosine_ops
FOR TYPE intvec USING hnsw AS
OPERATOR 1 <=> (intvec, intvec) FOR ORDER BY float_ops,
FUNCTION 1 cosine_distance(intvec, intvec),
FUNCTION 2 l2_norm(intvec),
FUNCTION 3 hnsw_intvec_support(internal);
CREATE OPERATOR CLASS intvec_l1_ops
FOR TYPE intvec USING hnsw AS
OPERATOR 1 <+> (intvec, intvec) FOR ORDER BY float_ops,
FUNCTION 1 l1_distance(intvec, intvec),
FUNCTION 3 hnsw_intvec_support(internal);
-- bit functions
CREATE FUNCTION hamming_distance(bit, bit) RETURNS float8
@@ -978,18 +782,6 @@ CREATE FUNCTION halfvec_to_sparsevec(halfvec, integer, boolean) RETURNS sparseve
CREATE FUNCTION sparsevec_to_halfvec(sparsevec, integer, boolean) RETURNS halfvec
AS 'MODULE_PATHNAME' LANGUAGE C IMMUTABLE STRICT PARALLEL SAFE;
CREATE FUNCTION array_to_sparsevec(integer[], integer, boolean) RETURNS sparsevec
AS 'MODULE_PATHNAME' LANGUAGE C IMMUTABLE STRICT PARALLEL SAFE;
CREATE FUNCTION array_to_sparsevec(real[], integer, boolean) RETURNS sparsevec
AS 'MODULE_PATHNAME' LANGUAGE C IMMUTABLE STRICT PARALLEL SAFE;
CREATE FUNCTION array_to_sparsevec(double precision[], integer, boolean) RETURNS sparsevec
AS 'MODULE_PATHNAME' LANGUAGE C IMMUTABLE STRICT PARALLEL SAFE;
CREATE FUNCTION array_to_sparsevec(numeric[], integer, boolean) RETURNS sparsevec
AS 'MODULE_PATHNAME' LANGUAGE C IMMUTABLE STRICT PARALLEL SAFE;
-- sparsevec casts
CREATE CAST (sparsevec AS sparsevec)
@@ -1007,18 +799,6 @@ CREATE CAST (sparsevec AS halfvec)
CREATE CAST (halfvec AS sparsevec)
WITH FUNCTION halfvec_to_sparsevec(halfvec, integer, boolean) AS IMPLICIT;
CREATE CAST (integer[] AS sparsevec)
WITH FUNCTION array_to_sparsevec(integer[], integer, boolean) AS ASSIGNMENT;
CREATE CAST (real[] AS sparsevec)
WITH FUNCTION array_to_sparsevec(real[], integer, boolean) AS ASSIGNMENT;
CREATE CAST (double precision[] AS sparsevec)
WITH FUNCTION array_to_sparsevec(double precision[], integer, boolean) AS ASSIGNMENT;
CREATE CAST (numeric[] AS sparsevec)
WITH FUNCTION array_to_sparsevec(numeric[], integer, boolean) AS ASSIGNMENT;
-- sparsevec operators
CREATE OPERATOR <-> (

View File

@@ -4,8 +4,8 @@
#include "postgres.h"
/* Check version in first header */
#if PG_VERSION_NUM < 130000
#error "Requires PostgreSQL 13+"
#if PG_VERSION_NUM < 120000
#error "Requires PostgreSQL 12+"
#endif
extern uint64 (*BitHammingDistance) (uint32 bytes, unsigned char *ax, unsigned char *bx, uint64 distance);

View File

@@ -3,7 +3,6 @@
#include "bitutils.h"
#include "bitvec.h"
#include "utils/varbit.h"
#include "vector.h"
#if PG_VERSION_NUM >= 160000
#include "varatt.h"
@@ -41,7 +40,7 @@ CheckDims(VarBit *a, VarBit *b)
/*
* Get the Hamming distance between two bit vectors
*/
FUNCTION_PREFIX PG_FUNCTION_INFO_V1(hamming_distance);
PGDLLEXPORT PG_FUNCTION_INFO_V1(hamming_distance);
Datum
hamming_distance(PG_FUNCTION_ARGS)
{
@@ -56,7 +55,7 @@ hamming_distance(PG_FUNCTION_ARGS)
/*
* Get the Jaccard distance between two bit vectors
*/
FUNCTION_PREFIX PG_FUNCTION_INFO_V1(jaccard_distance);
PGDLLEXPORT PG_FUNCTION_INFO_V1(jaccard_distance);
Datum
jaccard_distance(PG_FUNCTION_ARGS)
{

View File

@@ -19,6 +19,11 @@
#include "utils/numeric.h"
#include "vector.h"
#if PG_VERSION_NUM < 130000
#define TYPALIGN_DOUBLE 'd'
#define TYPALIGN_INT 'i'
#endif
#define STATE_DIMS(x) (ARR_DIMS(x)[0] - 1)
#define CreateStateDatums(dim) palloc(sizeof(Datum) * (dim + 1))
@@ -159,10 +164,28 @@ CheckStateArray(ArrayType *statearray, const char *caller)
return (float8 *) ARR_DATA_PTR(statearray);
}
#if PG_VERSION_NUM < 120003
static pg_noinline void
float_overflow_error(void)
{
ereport(ERROR,
(errcode(ERRCODE_NUMERIC_VALUE_OUT_OF_RANGE),
errmsg("value out of range: overflow")));
}
static pg_noinline void
float_underflow_error(void)
{
ereport(ERROR,
(errcode(ERRCODE_NUMERIC_VALUE_OUT_OF_RANGE),
errmsg("value out of range: underflow")));
}
#endif
/*
* Convert textual representation to internal representation
*/
FUNCTION_PREFIX PG_FUNCTION_INFO_V1(halfvec_in);
PGDLLEXPORT PG_FUNCTION_INFO_V1(halfvec_in);
Datum
halfvec_in(PG_FUNCTION_ARGS)
{
@@ -276,7 +299,7 @@ halfvec_in(PG_FUNCTION_ARGS)
/*
* Convert internal representation to textual representation
*/
FUNCTION_PREFIX PG_FUNCTION_INFO_V1(halfvec_out);
PGDLLEXPORT PG_FUNCTION_INFO_V1(halfvec_out);
Datum
halfvec_out(PG_FUNCTION_ARGS)
{
@@ -322,7 +345,7 @@ halfvec_out(PG_FUNCTION_ARGS)
/*
* Convert type modifier
*/
FUNCTION_PREFIX PG_FUNCTION_INFO_V1(halfvec_typmod_in);
PGDLLEXPORT PG_FUNCTION_INFO_V1(halfvec_typmod_in);
Datum
halfvec_typmod_in(PG_FUNCTION_ARGS)
{
@@ -353,7 +376,7 @@ halfvec_typmod_in(PG_FUNCTION_ARGS)
/*
* Convert external binary representation to internal representation
*/
FUNCTION_PREFIX PG_FUNCTION_INFO_V1(halfvec_recv);
PGDLLEXPORT PG_FUNCTION_INFO_V1(halfvec_recv);
Datum
halfvec_recv(PG_FUNCTION_ARGS)
{
@@ -387,7 +410,7 @@ halfvec_recv(PG_FUNCTION_ARGS)
/*
* Convert internal representation to the external binary representation
*/
FUNCTION_PREFIX PG_FUNCTION_INFO_V1(halfvec_send);
PGDLLEXPORT PG_FUNCTION_INFO_V1(halfvec_send);
Datum
halfvec_send(PG_FUNCTION_ARGS)
{
@@ -407,7 +430,7 @@ halfvec_send(PG_FUNCTION_ARGS)
* Convert half vector to half vector
* This is needed to check the type modifier
*/
FUNCTION_PREFIX PG_FUNCTION_INFO_V1(halfvec);
PGDLLEXPORT PG_FUNCTION_INFO_V1(halfvec);
Datum
halfvec(PG_FUNCTION_ARGS)
{
@@ -422,7 +445,7 @@ halfvec(PG_FUNCTION_ARGS)
/*
* Convert array to half vector
*/
FUNCTION_PREFIX PG_FUNCTION_INFO_V1(array_to_halfvec);
PGDLLEXPORT PG_FUNCTION_INFO_V1(array_to_halfvec);
Datum
array_to_halfvec(PG_FUNCTION_ARGS)
{
@@ -496,7 +519,7 @@ array_to_halfvec(PG_FUNCTION_ARGS)
/*
* Convert half vector to float4[]
*/
FUNCTION_PREFIX PG_FUNCTION_INFO_V1(halfvec_to_float4);
PGDLLEXPORT PG_FUNCTION_INFO_V1(halfvec_to_float4);
Datum
halfvec_to_float4(PG_FUNCTION_ARGS)
{
@@ -520,7 +543,7 @@ halfvec_to_float4(PG_FUNCTION_ARGS)
/*
* Convert vector to half vec
*/
FUNCTION_PREFIX PG_FUNCTION_INFO_V1(vector_to_halfvec);
PGDLLEXPORT PG_FUNCTION_INFO_V1(vector_to_halfvec);
Datum
vector_to_halfvec(PG_FUNCTION_ARGS)
{
@@ -542,7 +565,7 @@ vector_to_halfvec(PG_FUNCTION_ARGS)
/*
* Get the L2 distance between half vectors
*/
FUNCTION_PREFIX PG_FUNCTION_INFO_V1(halfvec_l2_distance);
PGDLLEXPORT PG_FUNCTION_INFO_V1(halfvec_l2_distance);
Datum
halfvec_l2_distance(PG_FUNCTION_ARGS)
{
@@ -557,7 +580,7 @@ halfvec_l2_distance(PG_FUNCTION_ARGS)
/*
* Get the L2 squared distance between half vectors
*/
FUNCTION_PREFIX PG_FUNCTION_INFO_V1(halfvec_l2_squared_distance);
PGDLLEXPORT PG_FUNCTION_INFO_V1(halfvec_l2_squared_distance);
Datum
halfvec_l2_squared_distance(PG_FUNCTION_ARGS)
{
@@ -572,7 +595,7 @@ halfvec_l2_squared_distance(PG_FUNCTION_ARGS)
/*
* Get the inner product of two half vectors
*/
FUNCTION_PREFIX PG_FUNCTION_INFO_V1(halfvec_inner_product);
PGDLLEXPORT PG_FUNCTION_INFO_V1(halfvec_inner_product);
Datum
halfvec_inner_product(PG_FUNCTION_ARGS)
{
@@ -587,7 +610,7 @@ halfvec_inner_product(PG_FUNCTION_ARGS)
/*
* Get the negative inner product of two half vectors
*/
FUNCTION_PREFIX PG_FUNCTION_INFO_V1(halfvec_negative_inner_product);
PGDLLEXPORT PG_FUNCTION_INFO_V1(halfvec_negative_inner_product);
Datum
halfvec_negative_inner_product(PG_FUNCTION_ARGS)
{
@@ -602,7 +625,7 @@ halfvec_negative_inner_product(PG_FUNCTION_ARGS)
/*
* Get the cosine distance between two half vectors
*/
FUNCTION_PREFIX PG_FUNCTION_INFO_V1(halfvec_cosine_distance);
PGDLLEXPORT PG_FUNCTION_INFO_V1(halfvec_cosine_distance);
Datum
halfvec_cosine_distance(PG_FUNCTION_ARGS)
{
@@ -634,7 +657,7 @@ halfvec_cosine_distance(PG_FUNCTION_ARGS)
* Currently uses angular distance since needs to satisfy triangle inequality
* Assumes inputs are unit vectors (skips norm)
*/
FUNCTION_PREFIX PG_FUNCTION_INFO_V1(halfvec_spherical_distance);
PGDLLEXPORT PG_FUNCTION_INFO_V1(halfvec_spherical_distance);
Datum
halfvec_spherical_distance(PG_FUNCTION_ARGS)
{
@@ -658,7 +681,7 @@ halfvec_spherical_distance(PG_FUNCTION_ARGS)
/*
* Get the L1 distance between two half vectors
*/
FUNCTION_PREFIX PG_FUNCTION_INFO_V1(halfvec_l1_distance);
PGDLLEXPORT PG_FUNCTION_INFO_V1(halfvec_l1_distance);
Datum
halfvec_l1_distance(PG_FUNCTION_ARGS)
{
@@ -673,7 +696,7 @@ halfvec_l1_distance(PG_FUNCTION_ARGS)
/*
* Get the dimensions of a half vector
*/
FUNCTION_PREFIX PG_FUNCTION_INFO_V1(halfvec_vector_dims);
PGDLLEXPORT PG_FUNCTION_INFO_V1(halfvec_vector_dims);
Datum
halfvec_vector_dims(PG_FUNCTION_ARGS)
{
@@ -685,7 +708,7 @@ halfvec_vector_dims(PG_FUNCTION_ARGS)
/*
* Get the L2 norm of a half vector
*/
FUNCTION_PREFIX PG_FUNCTION_INFO_V1(halfvec_l2_norm);
PGDLLEXPORT PG_FUNCTION_INFO_V1(halfvec_l2_norm);
Datum
halfvec_l2_norm(PG_FUNCTION_ARGS)
{
@@ -707,7 +730,7 @@ halfvec_l2_norm(PG_FUNCTION_ARGS)
/*
* Normalize a half vector with the L2 norm
*/
FUNCTION_PREFIX PG_FUNCTION_INFO_V1(halfvec_l2_normalize);
PGDLLEXPORT PG_FUNCTION_INFO_V1(halfvec_l2_normalize);
Datum
halfvec_l2_normalize(PG_FUNCTION_ARGS)
{
@@ -746,7 +769,7 @@ halfvec_l2_normalize(PG_FUNCTION_ARGS)
/*
* Add half vectors
*/
FUNCTION_PREFIX PG_FUNCTION_INFO_V1(halfvec_add);
PGDLLEXPORT PG_FUNCTION_INFO_V1(halfvec_add);
Datum
halfvec_add(PG_FUNCTION_ARGS)
{
@@ -785,7 +808,7 @@ halfvec_add(PG_FUNCTION_ARGS)
/*
* Subtract half vectors
*/
FUNCTION_PREFIX PG_FUNCTION_INFO_V1(halfvec_sub);
PGDLLEXPORT PG_FUNCTION_INFO_V1(halfvec_sub);
Datum
halfvec_sub(PG_FUNCTION_ARGS)
{
@@ -824,7 +847,7 @@ halfvec_sub(PG_FUNCTION_ARGS)
/*
* Multiply half vectors
*/
FUNCTION_PREFIX PG_FUNCTION_INFO_V1(halfvec_mul);
PGDLLEXPORT PG_FUNCTION_INFO_V1(halfvec_mul);
Datum
halfvec_mul(PG_FUNCTION_ARGS)
{
@@ -866,7 +889,7 @@ halfvec_mul(PG_FUNCTION_ARGS)
/*
* Concatenate half vectors
*/
FUNCTION_PREFIX PG_FUNCTION_INFO_V1(halfvec_concat);
PGDLLEXPORT PG_FUNCTION_INFO_V1(halfvec_concat);
Datum
halfvec_concat(PG_FUNCTION_ARGS)
{
@@ -890,7 +913,7 @@ halfvec_concat(PG_FUNCTION_ARGS)
/*
* Quantize a half vector
*/
FUNCTION_PREFIX PG_FUNCTION_INFO_V1(halfvec_binary_quantize);
PGDLLEXPORT PG_FUNCTION_INFO_V1(halfvec_binary_quantize);
Datum
halfvec_binary_quantize(PG_FUNCTION_ARGS)
{
@@ -908,7 +931,7 @@ halfvec_binary_quantize(PG_FUNCTION_ARGS)
/*
* Get a subvector
*/
FUNCTION_PREFIX PG_FUNCTION_INFO_V1(halfvec_subvector);
PGDLLEXPORT PG_FUNCTION_INFO_V1(halfvec_subvector);
Datum
halfvec_subvector(PG_FUNCTION_ARGS)
{
@@ -982,7 +1005,7 @@ halfvec_cmp_internal(HalfVector * a, HalfVector * b)
/*
* Less than
*/
FUNCTION_PREFIX PG_FUNCTION_INFO_V1(halfvec_lt);
PGDLLEXPORT PG_FUNCTION_INFO_V1(halfvec_lt);
Datum
halfvec_lt(PG_FUNCTION_ARGS)
{
@@ -995,7 +1018,7 @@ halfvec_lt(PG_FUNCTION_ARGS)
/*
* Less than or equal
*/
FUNCTION_PREFIX PG_FUNCTION_INFO_V1(halfvec_le);
PGDLLEXPORT PG_FUNCTION_INFO_V1(halfvec_le);
Datum
halfvec_le(PG_FUNCTION_ARGS)
{
@@ -1008,7 +1031,7 @@ halfvec_le(PG_FUNCTION_ARGS)
/*
* Equal
*/
FUNCTION_PREFIX PG_FUNCTION_INFO_V1(halfvec_eq);
PGDLLEXPORT PG_FUNCTION_INFO_V1(halfvec_eq);
Datum
halfvec_eq(PG_FUNCTION_ARGS)
{
@@ -1021,7 +1044,7 @@ halfvec_eq(PG_FUNCTION_ARGS)
/*
* Not equal
*/
FUNCTION_PREFIX PG_FUNCTION_INFO_V1(halfvec_ne);
PGDLLEXPORT PG_FUNCTION_INFO_V1(halfvec_ne);
Datum
halfvec_ne(PG_FUNCTION_ARGS)
{
@@ -1034,7 +1057,7 @@ halfvec_ne(PG_FUNCTION_ARGS)
/*
* Greater than or equal
*/
FUNCTION_PREFIX PG_FUNCTION_INFO_V1(halfvec_ge);
PGDLLEXPORT PG_FUNCTION_INFO_V1(halfvec_ge);
Datum
halfvec_ge(PG_FUNCTION_ARGS)
{
@@ -1047,7 +1070,7 @@ halfvec_ge(PG_FUNCTION_ARGS)
/*
* Greater than
*/
FUNCTION_PREFIX PG_FUNCTION_INFO_V1(halfvec_gt);
PGDLLEXPORT PG_FUNCTION_INFO_V1(halfvec_gt);
Datum
halfvec_gt(PG_FUNCTION_ARGS)
{
@@ -1060,7 +1083,7 @@ halfvec_gt(PG_FUNCTION_ARGS)
/*
* Compare half vectors
*/
FUNCTION_PREFIX PG_FUNCTION_INFO_V1(halfvec_cmp);
PGDLLEXPORT PG_FUNCTION_INFO_V1(halfvec_cmp);
Datum
halfvec_cmp(PG_FUNCTION_ARGS)
{
@@ -1073,7 +1096,7 @@ halfvec_cmp(PG_FUNCTION_ARGS)
/*
* Accumulate half vectors
*/
FUNCTION_PREFIX PG_FUNCTION_INFO_V1(halfvec_accum);
PGDLLEXPORT PG_FUNCTION_INFO_V1(halfvec_accum);
Datum
halfvec_accum(PG_FUNCTION_ARGS)
{
@@ -1134,7 +1157,7 @@ halfvec_accum(PG_FUNCTION_ARGS)
/*
* Average half vectors
*/
FUNCTION_PREFIX PG_FUNCTION_INFO_V1(halfvec_avg);
PGDLLEXPORT PG_FUNCTION_INFO_V1(halfvec_avg);
Datum
halfvec_avg(PG_FUNCTION_ARGS)
{
@@ -1168,7 +1191,7 @@ halfvec_avg(PG_FUNCTION_ARGS)
/*
* Convert sparse vector to half vector
*/
FUNCTION_PREFIX PG_FUNCTION_INFO_V1(sparsevec_to_halfvec);
PGDLLEXPORT PG_FUNCTION_INFO_V1(sparsevec_to_halfvec);
Datum
sparsevec_to_halfvec(PG_FUNCTION_ARGS)
{

View File

@@ -38,7 +38,7 @@
/* F16C has better performance than _Float16 (on x86-64) */
#if defined(__F16C__)
#define F16C_SUPPORT
#elif defined(__FLT16_MAX__) && !defined(HALFVEC_DISPATCH) && !defined(__FreeBSD__) && (!defined(__i386__) || defined(__SSE2__))
#elif defined(__FLT16_MAX__) && !defined(HALFVEC_DISPATCH)
#define FLT16_SUPPORT
#endif

View File

@@ -9,25 +9,14 @@
#include "commands/vacuum.h"
#include "hnsw.h"
#include "miscadmin.h"
#include "utils/float.h"
#include "utils/guc.h"
#include "utils/selfuncs.h"
#include "utils/spccache.h"
#if PG_VERSION_NUM < 150000
#define MarkGUCPrefixReserved(x) EmitWarningsOnPlaceholders(x)
#endif
static const struct config_enum_entry hnsw_iterative_search_options[] = {
{"off", HNSW_ITERATIVE_SEARCH_OFF, false},
{"relaxed_order", HNSW_ITERATIVE_SEARCH_RELAXED, false},
{"strict_order", HNSW_ITERATIVE_SEARCH_STRICT, false},
{NULL, 0, false}
};
int hnsw_ef_search;
int hnsw_max_search_tuples;
int hnsw_iterative_search;
int hnsw_lock_tranche_id;
static relopt_kind hnsw_relopt_kind;
@@ -70,23 +59,22 @@ HnswInit(void)
hnsw_relopt_kind = add_reloption_kind();
add_int_reloption(hnsw_relopt_kind, "m", "Max number of connections",
HNSW_DEFAULT_M, HNSW_MIN_M, HNSW_MAX_M, AccessExclusiveLock);
HNSW_DEFAULT_M, HNSW_MIN_M, HNSW_MAX_M
#if PG_VERSION_NUM >= 130000
,AccessExclusiveLock
#endif
);
add_int_reloption(hnsw_relopt_kind, "ef_construction", "Size of the dynamic candidate list for construction",
HNSW_DEFAULT_EF_CONSTRUCTION, HNSW_MIN_EF_CONSTRUCTION, HNSW_MAX_EF_CONSTRUCTION, AccessExclusiveLock);
HNSW_DEFAULT_EF_CONSTRUCTION, HNSW_MIN_EF_CONSTRUCTION, HNSW_MAX_EF_CONSTRUCTION
#if PG_VERSION_NUM >= 130000
,AccessExclusiveLock
#endif
);
DefineCustomIntVariable("hnsw.ef_search", "Sets the size of the dynamic candidate list for search",
"Valid range is 1..1000.", &hnsw_ef_search,
HNSW_DEFAULT_EF_SEARCH, HNSW_MIN_EF_SEARCH, HNSW_MAX_EF_SEARCH, PGC_USERSET, 0, NULL, NULL, NULL);
DefineCustomEnumVariable("hnsw.iterative_search", "Sets the iterative search mode",
NULL, &hnsw_iterative_search,
HNSW_ITERATIVE_SEARCH_OFF, hnsw_iterative_search_options, PGC_USERSET, 0, NULL, NULL, NULL);
/* This is approximate and does not apply to the initial scan */
DefineCustomIntVariable("hnsw.max_search_tuples", "Sets the max number of candidates to visit for iterative search",
"-1 means no limit", &hnsw_max_search_tuples,
-1, -1, INT_MAX, PGC_USERSET, 0, NULL, NULL, NULL);
MarkGUCPrefixReserved("hnsw");
}
@@ -118,16 +106,14 @@ hnswcostestimate(PlannerInfo *root, IndexPath *path, double loop_count,
{
GenericCosts costs;
int m;
double ratio;
double startupPages;
double spc_seq_page_cost;
int entryLevel;
Relation index;
/* Never use index without order */
if (path->indexorderbys == NULL)
{
*indexStartupCost = get_float8_infinity();
*indexTotalCost = get_float8_infinity();
*indexStartupCost = DBL_MAX;
*indexTotalCost = DBL_MAX;
*indexSelectivity = 0;
*indexCorrelation = 0;
*indexPages = 0;
@@ -136,71 +122,21 @@ hnswcostestimate(PlannerInfo *root, IndexPath *path, double loop_count,
MemSet(&costs, 0, sizeof(costs));
genericcostestimate(root, path, loop_count, &costs);
index = index_open(path->indexinfo->indexoid, NoLock);
HnswGetMetaPageInfo(index, &m, NULL);
index_close(index, NoLock);
/*
* HNSW cost estimation follows a formula that accounts for the total
* number of tuples indexed combined with the parameters that most
* influence the duration of the index scan, namely: m - the number of
* tuples that are scanned in each step of the HNSW graph traversal
* ef_search - which influences the total number of steps taken at layer 0
*
* The source of the vector data can impact how many steps it takes to
* converge on the set of vectors to return to the executor. Currently, we
* use a hardcoded scaling factor (HNSWScanScalingFactor) to help
* influence that, but this could later become a configurable parameter
* based on the cost estimations.
*
* The tuple estimator formula is below:
*
* numIndexTuples = entryLevel * m + layer0TuplesMax * layer0Selectivity
*
* "entryLevel * m" represents the floor of tuples we need to scan to get
* to layer 0 (L0).
*
* "layer0TuplesMax" is the estimated total number of tuples we'd scan at
* L0 if we weren't discarding already visited tuples as part of the scan.
*
* "layer0Selectivity" estimates the percentage of tuples that are scanned
* at L0, accounting for previously visited tuples, multiplied by the
* "scalingFactor" (currently hardcoded).
*/
if (path->indexinfo->tuples > 0)
{
double scalingFactor = 0.55;
int entryLevel = (int) (log(path->indexinfo->tuples) * HnswGetMl(m));
int layer0TuplesMax = HnswGetLayerM(m, 0) * hnsw_ef_search;
double layer0Selectivity = scalingFactor * log(path->indexinfo->tuples) / (log(m) * (1 + log(hnsw_ef_search)));
/* Approximate entry level */
entryLevel = (int) -log(1.0 / path->indexinfo->tuples) * HnswGetMl(m);
ratio = (entryLevel * m + layer0TuplesMax * layer0Selectivity) / path->indexinfo->tuples;
/* TODO Improve estimate of visited tuples (currently underestimates) */
/* Account for number of tuples (or entry level), m, and ef_search */
costs.numIndexTuples = (entryLevel + 2) * m;
if (ratio > 1)
ratio = 1;
}
else
ratio = 1;
genericcostestimate(root, path, loop_count, &costs);
get_tablespace_page_costs(path->indexinfo->reltablespace, NULL, &spc_seq_page_cost);
/* Startup cost is cost before returning the first row */
costs.indexStartupCost = costs.indexTotalCost * ratio;
/* Adjust cost if needed since TOAST not included in seq scan cost */
startupPages = costs.numIndexPages * ratio;
if (startupPages > path->indexinfo->rel->pages && ratio < 0.5)
{
/* Change all page cost from random to sequential */
costs.indexStartupCost -= startupPages * (costs.spc_random_page_cost - spc_seq_page_cost);
/* Remove cost of extra pages */
costs.indexStartupCost -= (startupPages - path->indexinfo->rel->pages) * spc_seq_page_cost;
}
*indexStartupCost = costs.indexStartupCost;
/* Use total cost since most work happens before first tuple is returned */
*indexStartupCost = costs.indexTotalCost;
*indexTotalCost = costs.indexTotalCost;
*indexSelectivity = costs.indexSelectivity;
*indexCorrelation = costs.indexCorrelation;
@@ -218,10 +154,23 @@ hnswoptions(Datum reloptions, bool validate)
{"ef_construction", RELOPT_TYPE_INT, offsetof(HnswOptions, efConstruction)},
};
#if PG_VERSION_NUM >= 130000
return (bytea *) build_reloptions(reloptions, validate,
hnsw_relopt_kind,
sizeof(HnswOptions),
tab, lengthof(tab));
#else
relopt_value *options;
int numoptions;
HnswOptions *rdopts;
options = parseRelOptions(reloptions, validate, hnsw_relopt_kind, &numoptions);
rdopts = allocateReloptStruct(sizeof(HnswOptions), options, numoptions);
fillRelOptions((void *) rdopts, sizeof(HnswOptions), options, numoptions,
validate, tab, lengthof(tab));
return (bytea *) rdopts;
#endif
}
/*
@@ -238,7 +187,7 @@ hnswvalidate(Oid opclassoid)
*
* See https://www.postgresql.org/docs/current/index-api.html
*/
FUNCTION_PREFIX PG_FUNCTION_INFO_V1(hnswhandler);
PGDLLEXPORT PG_FUNCTION_INFO_V1(hnswhandler);
Datum
hnswhandler(PG_FUNCTION_ARGS)
{
@@ -246,7 +195,9 @@ hnswhandler(PG_FUNCTION_ARGS)
amroutine->amstrategies = 0;
amroutine->amsupport = 3;
#if PG_VERSION_NUM >= 130000
amroutine->amoptsprocnum = 0;
#endif
amroutine->amcanorder = false;
amroutine->amcanorderbyop = true;
amroutine->amcanbackward = false; /* can change direction mid-scan */
@@ -259,24 +210,17 @@ hnswhandler(PG_FUNCTION_ARGS)
amroutine->amclusterable = false;
amroutine->ampredlocks = false;
amroutine->amcanparallel = false;
#if PG_VERSION_NUM >= 170000
amroutine->amcanbuildparallel = true;
#endif
amroutine->amcaninclude = false;
#if PG_VERSION_NUM >= 130000
amroutine->amusemaintenanceworkmem = false; /* not used during VACUUM */
#if PG_VERSION_NUM >= 160000
amroutine->amsummarizing = false;
#endif
amroutine->amparallelvacuumoptions = VACUUM_OPTION_PARALLEL_BULKDEL;
#endif
amroutine->amkeytype = InvalidOid;
/* Interface functions */
amroutine->ambuild = hnswbuild;
amroutine->ambuildempty = hnswbuildempty;
amroutine->aminsert = hnswinsert;
#if PG_VERSION_NUM >= 170000
amroutine->aminsertcleanup = NULL;
#endif
amroutine->ambulkdelete = hnswbulkdelete;
amroutine->amvacuumcleanup = hnswvacuumcleanup;
amroutine->amcanreturn = NULL;

View File

@@ -76,6 +76,11 @@
#define SeedRandom(seed) srandom(seed)
#endif
#if PG_VERSION_NUM < 130000
#define list_delete_last(list) list_truncate(list, list_length(list) - 1)
#define list_sort(list, cmp) ((list) = list_qsort(list, cmp))
#endif
#define HnswIsElementTuple(tup) ((tup)->type == HNSW_ELEMENT_TUPLE_TYPE)
#define HnswIsNeighborTuple(tup) ((tup)->type == HNSW_NEIGHBOR_TUPLE_TYPE)
@@ -88,9 +93,6 @@
/* Ensure fits on page and in uint8 */
#define HnswGetMaxLevel(m) Min(((BLCKSZ - MAXALIGN(SizeOfPageHeaderData) - MAXALIGN(sizeof(HnswPageOpaqueData)) - offsetof(HnswNeighborTupleData, indextids) - sizeof(ItemIdData)) / (sizeof(ItemPointerData)) / (m)) - 2, 255)
#define HnswGetSearchCandidate(membername, ptr) pairingheap_container(HnswSearchCandidate, membername, ptr)
#define HnswGetSearchCandidateConst(membername, ptr) pairingheap_const_container(HnswSearchCandidate, membername, ptr)
#define HnswGetValue(base, element) PointerGetDatum(HnswPtrAccess(base, (element)->value))
#if PG_VERSION_NUM < 140005
@@ -109,17 +111,8 @@
/* Variables */
extern int hnsw_ef_search;
extern int hnsw_iterative_search;
extern int hnsw_max_search_tuples;
extern int hnsw_lock_tranche_id;
typedef enum HnswIterativeSearchMode
{
HNSW_ITERATIVE_SEARCH_OFF,
HNSW_ITERATIVE_SEARCH_RELAXED,
HNSW_ITERATIVE_SEARCH_STRICT
} HnswIterativeSearchMode;
typedef struct HnswElementData HnswElementData;
typedef struct HnswNeighborArray HnswNeighborArray;
@@ -141,7 +134,6 @@ struct HnswElementData
uint8 heaptidsLength;
uint8 level;
uint8 deleted;
uint8 version;
uint32 hash;
HnswNeighborsPtr neighbors;
BlockNumber blkno;
@@ -168,13 +160,11 @@ struct HnswNeighborArray
HnswCandidate items[FLEXIBLE_ARRAY_MEMBER];
};
typedef struct HnswSearchCandidate
typedef struct HnswPairingHeapNode
{
pairingheap_node c_node;
pairingheap_node w_node;
HnswElementPtr element;
double distance;
} HnswSearchCandidate;
pairingheap_node ph_node;
HnswCandidate *inner;
} HnswPairingHeapNode;
/* HNSW index options */
typedef struct HnswOptions
@@ -198,8 +188,8 @@ typedef struct HnswGraph
/* Allocations state */
LWLock allocatorLock;
Size memoryUsed;
Size memoryTotal;
long memoryUsed;
long memoryTotal;
/* Flushed state */
LWLock flushLock;
@@ -250,18 +240,6 @@ typedef struct HnswTypeInfo
void (*checkValue) (Pointer v);
} HnswTypeInfo;
typedef struct HnswSupport
{
FmgrInfo *procinfo;
FmgrInfo *normprocinfo;
Oid collation;
} HnswSupport;
typedef struct HnswQuery
{
Datum value;
} HnswQuery;
typedef struct HnswBuildState
{
/* Info */
@@ -281,7 +259,9 @@ typedef struct HnswBuildState
double reltuples;
/* Support functions */
HnswSupport support;
FmgrInfo *procinfo;
FmgrInfo *normprocinfo;
Oid collation;
/* Variables */
HnswGraph graphData;
@@ -329,10 +309,10 @@ typedef struct HnswElementTupleData
uint8 type;
uint8 level;
uint8 deleted;
uint8 version;
uint8 unused;
ItemPointerData heaptids[HNSW_HEAPTIDS];
ItemPointerData neighbortid;
uint16 unused;
uint16 unused2;
Vector data;
} HnswElementTupleData;
@@ -341,41 +321,24 @@ typedef HnswElementTupleData * HnswElementTuple;
typedef struct HnswNeighborTupleData
{
uint8 type;
uint8 version;
uint8 unused;
uint16 count;
ItemPointerData indextids[FLEXIBLE_ARRAY_MEMBER];
} HnswNeighborTupleData;
typedef HnswNeighborTupleData * HnswNeighborTuple;
typedef union
{
struct pointerhash_hash *pointers;
struct offsethash_hash *offsets;
struct tidhash_hash *tids;
} visited_hash;
typedef union
{
HnswElement element;
ItemPointerData indextid;
} HnswUnvisited;
typedef struct HnswScanOpaqueData
{
const HnswTypeInfo *typeInfo;
bool first;
List *w;
visited_hash v;
pairingheap *discarded;
HnswQuery q;
int m;
int64 tuples;
double previousDistance;
MemoryContext tmpCtx;
/* Support functions */
HnswSupport support;
FmgrInfo *procinfo;
FmgrInfo *normprocinfo;
Oid collation;
} HnswScanOpaqueData;
typedef HnswScanOpaqueData * HnswScanOpaque;
@@ -393,7 +356,8 @@ typedef struct HnswVacuumState
int efConstruction;
/* Support functions */
HnswSupport support;
FmgrInfo *procinfo;
Oid collation;
/* Variables */
struct tidhash_hash *deleted;
@@ -409,33 +373,30 @@ typedef struct HnswVacuumState
int HnswGetM(Relation index);
int HnswGetEfConstruction(Relation index);
FmgrInfo *HnswOptionalProcInfo(Relation index, uint16 procnum);
void HnswInitSupport(HnswSupport * support, Relation index);
Datum HnswNormValue(const HnswTypeInfo * typeInfo, Oid collation, Datum value);
bool HnswCheckNorm(HnswSupport * support, Datum value);
bool HnswCheckNorm(FmgrInfo *procinfo, Oid collation, Datum value);
Buffer HnswNewBuffer(Relation index, ForkNumber forkNum);
void HnswInitPage(Buffer buf, Page page);
void HnswInit(void);
List *HnswSearchLayer(char *base, HnswQuery * q, List *ep, int ef, int lc, Relation index, HnswSupport * support, int m, bool inserting, HnswElement skipElement, visited_hash * v, pairingheap **discarded, bool initVisited, int64 *tuples);
List *HnswSearchLayer(char *base, Datum q, List *ep, int ef, int lc, Relation index, FmgrInfo *procinfo, Oid collation, int m, bool inserting, HnswElement skipElement);
HnswElement HnswGetEntryPoint(Relation index);
void HnswGetMetaPageInfo(Relation index, int *m, HnswElement * entryPoint);
void *HnswAlloc(HnswAllocator * allocator, Size size);
HnswElement HnswInitElement(char *base, ItemPointer tid, int m, double ml, int maxLevel, HnswAllocator * alloc);
HnswElement HnswInitElementFromBlock(BlockNumber blkno, OffsetNumber offno);
void HnswFindElementNeighbors(char *base, HnswElement element, HnswElement entryPoint, Relation index, HnswSupport * support, int m, int efConstruction, bool existing);
HnswSearchCandidate *HnswEntryCandidate(char *base, HnswElement em, HnswQuery * q, Relation rel, HnswSupport * support, bool loadVec);
void HnswFindElementNeighbors(char *base, HnswElement element, HnswElement entryPoint, Relation index, FmgrInfo *procinfo, Oid collation, int m, int efConstruction, bool existing);
HnswCandidate *HnswEntryCandidate(char *base, HnswElement em, Datum q, Relation rel, FmgrInfo *procinfo, Oid collation, bool loadVec);
void HnswUpdateMetaPage(Relation index, int updateEntry, HnswElement entryPoint, BlockNumber insertPage, ForkNumber forkNum, bool building);
void HnswSetNeighborTuple(char *base, HnswNeighborTuple ntup, HnswElement e, int m);
void HnswAddHeapTid(HnswElement element, ItemPointer heaptid);
HnswNeighborArray *HnswInitNeighborArray(int lm, HnswAllocator * allocator);
void HnswInitNeighbors(char *base, HnswElement element, int m, HnswAllocator * alloc);
bool HnswInsertTupleOnDisk(Relation index, HnswSupport * support, Datum value, ItemPointer heaptid, bool building);
void HnswUpdateNeighborsOnDisk(Relation index, HnswSupport * support, HnswElement e, int m, bool checkExisting, bool building);
bool HnswInsertTupleOnDisk(Relation index, Datum value, Datum *values, bool *isnull, ItemPointer heap_tid, bool building);
void HnswUpdateNeighborsOnDisk(Relation index, FmgrInfo *procinfo, Oid collation, HnswElement e, int m, bool checkExisting, bool building);
void HnswLoadElementFromTuple(HnswElement element, HnswElementTuple etup, bool loadHeaptids, bool loadVec);
void HnswLoadElement(HnswElement element, double *distance, HnswQuery * q, Relation index, HnswSupport * support, bool loadVec, double *maxDistance);
bool HnswFormIndexValue(Datum *out, Datum *values, bool *isnull, const HnswTypeInfo * typeInfo, HnswSupport * support);
void HnswLoadElement(HnswElement element, float *distance, Datum *q, Relation index, FmgrInfo *procinfo, Oid collation, bool loadVec, float *maxDistance);
void HnswSetElementTuple(char *base, HnswElementTuple etup, HnswElement element);
void HnswUpdateConnection(char *base, HnswNeighborArray * neighbors, HnswElement newElement, float distance, int lm, int *updateIdx, Relation index, HnswSupport * support);
bool HnswLoadNeighborTids(HnswElement element, ItemPointerData *indextids, Relation index, int m, int lm, int lc);
void HnswUpdateConnection(char *base, HnswElement element, HnswCandidate * hc, int lm, int lc, int *updateIdx, Relation index, FmgrInfo *procinfo, Oid collation);
void HnswLoadNeighbors(HnswElement element, Relation index, int m);
void HnswInitLockTranche(void);
const HnswTypeInfo *HnswGetTypeInfo(Relation index);
PGDLLEXPORT void HnswParallelBuildMain(dsm_segment *seg, shm_toc *toc);

View File

@@ -60,6 +60,12 @@
#include "pgstat.h"
#endif
#if PG_VERSION_NUM >= 130000
#define CALLBACK_ITEM_POINTER ItemPointer tid
#else
#define CALLBACK_ITEM_POINTER HeapTuple hup
#endif
#if PG_VERSION_NUM >= 140000
#include "utils/backend_status.h"
#include "utils/wait_event.h"
@@ -69,6 +75,10 @@
#define PARALLEL_KEY_HNSW_AREA UINT64CONST(0xA000000000000002)
#define PARALLEL_KEY_QUERY_TEXT UINT64CONST(0xA000000000000003)
#if PG_VERSION_NUM < 130000
#define GENERATIONCHUNK_RAWSIZE (SIZEOF_SIZE_T + SIZEOF_VOID_P * 2)
#endif
/*
* Create the metapage
*/
@@ -182,9 +192,7 @@ CreateGraphPages(HnswBuildState * buildstate)
/* Initial size check */
if (etupSize > HNSW_TUPLE_ALLOC_SIZE)
ereport(ERROR,
(errcode(ERRCODE_PROGRAM_LIMIT_EXCEEDED),
errmsg("index tuple too large")));
elog(ERROR, "index tuple too large");
HnswSetElementTuple(base, etup, element);
@@ -366,18 +374,12 @@ AddElementInMemory(char *base, HnswGraph * graph, HnswElement element)
* Update neighbors
*/
static void
UpdateNeighborsInMemory(char *base, HnswSupport * support, HnswElement e, int m)
UpdateNeighborsInMemory(char *base, FmgrInfo *procinfo, Oid collation, HnswElement e, int m)
{
for (int lc = e->level; lc >= 0; lc--)
{
int lm = HnswGetLayerM(m, lc);
Size neighborsSize = HNSW_NEIGHBOR_ARRAY_SIZE(lm);
HnswNeighborArray *neighbors = palloc(neighborsSize);
/* Copy neighbors to local memory */
LWLockAcquire(&e->lock, LW_SHARED);
memcpy(neighbors, HnswGetNeighbors(base, e, lc), neighborsSize);
LWLockRelease(&e->lock);
HnswNeighborArray *neighbors = HnswGetNeighbors(base, e, lc);
for (int i = 0; i < neighbors->length; i++)
{
@@ -387,8 +389,9 @@ UpdateNeighborsInMemory(char *base, HnswSupport * support, HnswElement e, int m)
/* Keep scan-build happy on Mac x86-64 */
Assert(neighborElement);
/* Use element for lock instead of hc since hc can be replaced */
LWLockAcquire(&neighborElement->lock, LW_EXCLUSIVE);
HnswUpdateConnection(base, HnswGetNeighbors(base, neighborElement, lc), e, hc->distance, lm, NULL, NULL, support);
HnswUpdateConnection(base, e, hc, lm, lc, NULL, NULL, procinfo, collation);
LWLockRelease(&neighborElement->lock);
}
}
@@ -398,7 +401,7 @@ UpdateNeighborsInMemory(char *base, HnswSupport * support, HnswElement e, int m)
* Update graph in memory
*/
static void
UpdateGraphInMemory(HnswSupport * support, HnswElement element, int m, int efConstruction, HnswElement entryPoint, HnswBuildState * buildstate)
UpdateGraphInMemory(FmgrInfo *procinfo, Oid collation, HnswElement element, int m, int efConstruction, HnswElement entryPoint, HnswBuildState * buildstate)
{
HnswGraph *graph = buildstate->graph;
char *base = buildstate->hnswarea;
@@ -411,7 +414,7 @@ UpdateGraphInMemory(HnswSupport * support, HnswElement element, int m, int efCon
AddElementInMemory(base, graph, element);
/* Update neighbors */
UpdateNeighborsInMemory(base, support, element, m);
UpdateNeighborsInMemory(base, procinfo, collation, element, m);
/* Update entry point if needed (already have lock) */
if (entryPoint == NULL || element->level > entryPoint->level)
@@ -424,8 +427,9 @@ UpdateGraphInMemory(HnswSupport * support, HnswElement element, int m, int efCon
static void
InsertTupleInMemory(HnswBuildState * buildstate, HnswElement element)
{
FmgrInfo *procinfo = buildstate->procinfo;
Oid collation = buildstate->collation;
HnswGraph *graph = buildstate->graph;
HnswSupport *support = &buildstate->support;
HnswElement entryPoint;
LWLock *entryLock = &graph->entryLock;
LWLock *entryWaitLock = &graph->entryWaitLock;
@@ -457,10 +461,10 @@ InsertTupleInMemory(HnswBuildState * buildstate, HnswElement element)
}
/* Find neighbors for element */
HnswFindElementNeighbors(base, element, entryPoint, NULL, support, m, efConstruction, false);
HnswFindElementNeighbors(base, element, entryPoint, NULL, procinfo, collation, m, efConstruction, false);
/* Update graph in memory */
UpdateGraphInMemory(support, element, m, efConstruction, entryPoint, buildstate);
UpdateGraphInMemory(procinfo, collation, element, m, efConstruction, entryPoint, buildstate);
/* Release entry lock */
LWLockRelease(entryLock);
@@ -472,19 +476,30 @@ InsertTupleInMemory(HnswBuildState * buildstate, HnswElement element)
static bool
InsertTuple(Relation index, Datum *values, bool *isnull, ItemPointer heaptid, HnswBuildState * buildstate)
{
const HnswTypeInfo *typeInfo = buildstate->typeInfo;
HnswGraph *graph = buildstate->graph;
HnswElement element;
HnswAllocator *allocator = &buildstate->allocator;
HnswSupport *support = &buildstate->support;
Size valueSize;
Pointer valuePtr;
LWLock *flushLock = &graph->flushLock;
char *base = buildstate->hnswarea;
Datum value;
/* Form index value */
if (!HnswFormIndexValue(&value, values, isnull, buildstate->typeInfo, support))
return false;
/* Detoast once for all calls */
Datum value = PointerGetDatum(PG_DETOAST_DATUM(values[0]));
/* Check value */
if (typeInfo->checkValue != NULL)
typeInfo->checkValue(DatumGetPointer(value));
/* Normalize if needed */
if (buildstate->normprocinfo != NULL)
{
if (!HnswCheckNorm(buildstate->normprocinfo, buildstate->collation, value))
return false;
value = HnswNormValue(typeInfo, buildstate->collation, value);
}
/* Get datum size */
valueSize = VARSIZE_ANY(DatumGetPointer(value));
@@ -497,7 +512,7 @@ InsertTuple(Relation index, Datum *values, bool *isnull, ItemPointer heaptid, Hn
{
LWLockRelease(flushLock);
return HnswInsertTupleOnDisk(index, support, value, heaptid, true);
return HnswInsertTupleOnDisk(index, value, values, isnull, heaptid, true);
}
/*
@@ -529,7 +544,7 @@ InsertTuple(Relation index, Datum *values, bool *isnull, ItemPointer heaptid, Hn
LWLockRelease(flushLock);
return HnswInsertTupleOnDisk(index, support, value, heaptid, true);
return HnswInsertTupleOnDisk(index, value, values, isnull, heaptid, true);
}
/* Ok, we can proceed to allocate the element */
@@ -563,13 +578,17 @@ InsertTuple(Relation index, Datum *values, bool *isnull, ItemPointer heaptid, Hn
* Callback for table_index_build_scan
*/
static void
BuildCallback(Relation index, ItemPointer tid, Datum *values,
BuildCallback(Relation index, CALLBACK_ITEM_POINTER, Datum *values,
bool *isnull, bool tupleIsAlive, void *state)
{
HnswBuildState *buildstate = (HnswBuildState *) state;
HnswGraph *graph = buildstate->graph;
MemoryContext oldCtx;
#if PG_VERSION_NUM < 130000
ItemPointer tid = &hup->t_self;
#endif
/* Skip nulls */
if (isnull[0])
return;
@@ -595,7 +614,7 @@ BuildCallback(Relation index, ItemPointer tid, Datum *values,
* Initialize the graph
*/
static void
InitGraph(HnswGraph * graph, char *base, Size memoryTotal)
InitGraph(HnswGraph * graph, char *base, long memoryTotal)
{
/* Initialize the lock tranche if needed */
HnswInitLockTranche();
@@ -632,7 +651,11 @@ HnswMemoryContextAlloc(Size size, void *state)
HnswBuildState *buildstate = (HnswBuildState *) state;
void *chunk = MemoryContextAlloc(buildstate->graphCtx, size);
#if PG_VERSION_NUM >= 130000
buildstate->graphData.memoryUsed = MemoryContextMemAllocated(buildstate->graphCtx, false);
#else
buildstate->graphData.memoryUsed += MAXALIGN(size);
#endif
return chunk;
}
@@ -668,33 +691,27 @@ InitBuildState(HnswBuildState * buildstate, Relation heap, Relation index, Index
/* Disallow varbit since require fixed dimensions */
if (TupleDescAttr(index->rd_att, 0)->atttypid == VARBITOID)
ereport(ERROR,
(errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
errmsg("type not supported for hnsw index")));
elog(ERROR, "type not supported for hnsw index");
/* Require column to have dimensions to be indexed */
if (buildstate->dimensions < 0)
ereport(ERROR,
(errcode(ERRCODE_INVALID_PARAMETER_VALUE),
errmsg("column does not have dimensions")));
elog(ERROR, "column does not have dimensions");
if (buildstate->dimensions > buildstate->typeInfo->maxDimensions)
ereport(ERROR,
(errcode(ERRCODE_PROGRAM_LIMIT_EXCEEDED),
errmsg("column cannot have more than %d dimensions for hnsw index", buildstate->typeInfo->maxDimensions)));
elog(ERROR, "column cannot have more than %d dimensions for hnsw index", buildstate->typeInfo->maxDimensions);
if (buildstate->efConstruction < 2 * buildstate->m)
ereport(ERROR,
(errcode(ERRCODE_INVALID_PARAMETER_VALUE),
errmsg("ef_construction must be greater than or equal to 2 * m")));
elog(ERROR, "ef_construction must be greater than or equal to 2 * m");
buildstate->reltuples = 0;
buildstate->indtuples = 0;
/* Get support functions */
HnswInitSupport(&buildstate->support, index);
buildstate->procinfo = index_getprocinfo(index, 1, HNSW_DISTANCE_PROC);
buildstate->normprocinfo = HnswOptionalProcInfo(index, HNSW_NORM_PROC);
buildstate->collation = index->rd_indcollation[0];
InitGraph(&buildstate->graphData, NULL, (Size) maintenance_work_mem * 1024L);
InitGraph(&buildstate->graphData, NULL, maintenance_work_mem * 1024L);
buildstate->graph = &buildstate->graphData;
buildstate->ml = HnswGetMl(buildstate->m);
buildstate->maxLevel = HnswGetMaxLevel(buildstate->m);
@@ -1104,8 +1121,8 @@ BuildIndex(Relation heap, Relation index, IndexInfo *indexInfo,
BuildGraph(buildstate, forkNum);
if (RelationNeedsWAL(index) || forkNum == INIT_FORKNUM)
log_newpage_range(index, forkNum, 0, RelationGetNumberOfBlocksInFork(index, forkNum), true);
if (RelationNeedsWAL(index))
log_newpage_range(index, forkNum, 0, RelationGetNumberOfBlocks(index), true);
FreeBuildState(buildstate);
}

View File

@@ -36,15 +36,14 @@ GetInsertPage(Relation index)
* Check for a free offset
*/
static bool
HnswFreeOffset(Relation index, Buffer buf, Page page, HnswElement element, Size etupSize, Size ntupSize, Buffer *nbuf, Page *npage, OffsetNumber *freeOffno, OffsetNumber *freeNeighborOffno, BlockNumber *newInsertPage, uint8 *tupleVersion)
HnswFreeOffset(Relation index, Buffer buf, Page page, HnswElement element, Size ntupSize, Buffer *nbuf, Page *npage, OffsetNumber *freeOffno, OffsetNumber *freeNeighborOffno, BlockNumber *newInsertPage)
{
OffsetNumber offno;
OffsetNumber maxoffno = PageGetMaxOffsetNumber(page);
for (offno = FirstOffsetNumber; offno <= maxoffno; offno = OffsetNumberNext(offno))
{
ItemId eitemid = PageGetItemId(page, offno);
HnswElementTuple etup = (HnswElementTuple) PageGetItem(page, eitemid);
HnswElementTuple etup = (HnswElementTuple) PageGetItem(page, PageGetItemId(page, offno));
/* Skip neighbor tuples */
if (!HnswIsElementTuple(etup))
@@ -55,9 +54,7 @@ HnswFreeOffset(Relation index, Buffer buf, Page page, HnswElement element, Size
BlockNumber elementPage = BufferGetBlockNumber(buf);
BlockNumber neighborPage = ItemPointerGetBlockNumber(&etup->neighbortid);
OffsetNumber neighborOffno = ItemPointerGetOffsetNumber(&etup->neighbortid);
ItemId nitemid;
Size pageFree;
Size npageFree;
ItemId itemid;
if (!BlockNumberIsValid(*newInsertPage))
*newInsertPage = elementPage;
@@ -76,29 +73,13 @@ HnswFreeOffset(Relation index, Buffer buf, Page page, HnswElement element, Size
*npage = BufferGetPage(*nbuf);
}
nitemid = PageGetItemId(*npage, neighborOffno);
itemid = PageGetItemId(*npage, neighborOffno);
/* Ensure aligned for space check */
Assert(etupSize == MAXALIGN(etupSize));
Assert(ntupSize == MAXALIGN(ntupSize));
/*
* Calculate free space individually since tuples are overwritten
* individually (in separate calls to PageIndexTupleOverwrite)
*/
pageFree = ItemIdGetLength(eitemid) + PageGetExactFreeSpace(page);
npageFree = ItemIdGetLength(nitemid);
if (neighborPage != elementPage)
npageFree += PageGetExactFreeSpace(*npage);
else if (pageFree >= etupSize)
npageFree += pageFree - etupSize;
/* Check for space */
if (pageFree >= etupSize && npageFree >= ntupSize)
/* Check for space on neighbor tuple page */
if (PageGetFreeSpace(*npage) + ItemIdGetLength(itemid) - sizeof(ItemIdData) >= ntupSize)
{
*freeOffno = offno;
*freeNeighborOffno = neighborOffno;
*tupleVersion = etup->version;
return true;
}
else if (*nbuf != buf)
@@ -154,7 +135,6 @@ AddElementOnDisk(Relation index, HnswElement e, int m, BlockNumber insertPage, B
OffsetNumber freeOffno = InvalidOffsetNumber;
OffsetNumber freeNeighborOffno = InvalidOffsetNumber;
BlockNumber newInsertPage = InvalidBlockNumber;
uint8 tupleVersion;
char *base = NULL;
/* Calculate sizes */
@@ -204,7 +184,7 @@ AddElementOnDisk(Relation index, HnswElement e, int m, BlockNumber insertPage, B
}
/* Next, try space from a deleted element */
if (HnswFreeOffset(index, buf, page, e, etupSize, ntupSize, &nbuf, &npage, &freeOffno, &freeNeighborOffno, &newInsertPage, &tupleVersion))
if (HnswFreeOffset(index, buf, page, e, ntupSize, &nbuf, &npage, &freeOffno, &freeNeighborOffno, &newInsertPage))
{
if (nbuf != buf)
{
@@ -214,10 +194,6 @@ AddElementOnDisk(Relation index, HnswElement e, int m, BlockNumber insertPage, B
npage = GenericXLogRegisterBuffer(state, nbuf, 0);
}
/* Set tuple version */
etup->version = tupleVersion;
ntup->version = tupleVersion;
break;
}
@@ -340,107 +316,6 @@ AddElementOnDisk(Relation index, HnswElement e, int m, BlockNumber insertPage, B
*updatedInsertPage = newInsertPage;
}
/*
* Load neighbors
*/
static HnswNeighborArray *
HnswLoadNeighbors(HnswElement element, Relation index, int m, int lm, int lc)
{
char *base = NULL;
HnswNeighborArray *neighbors = HnswInitNeighborArray(lm, NULL);
ItemPointerData indextids[HNSW_MAX_M * 2];
if (!HnswLoadNeighborTids(element, indextids, index, m, lm, lc))
return neighbors;
for (int i = 0; i < lm; i++)
{
ItemPointer indextid = &indextids[i];
HnswElement e;
HnswCandidate *hc;
if (!ItemPointerIsValid(indextid))
break;
e = HnswInitElementFromBlock(ItemPointerGetBlockNumber(indextid), ItemPointerGetOffsetNumber(indextid));
hc = &neighbors->items[neighbors->length++];
HnswPtrStore(base, hc->element, e);
}
return neighbors;
}
/*
* Load elements for insert
*/
static void
LoadElementsForInsert(HnswNeighborArray * neighbors, HnswQuery * q, int *idx, Relation index, HnswSupport * support)
{
char *base = NULL;
for (int i = 0; i < neighbors->length; i++)
{
HnswCandidate *hc = &neighbors->items[i];
HnswElement element = HnswPtrAccess(base, hc->element);
double distance;
HnswLoadElement(element, &distance, q, index, support, true, NULL);
hc->distance = distance;
/* Prune element if being deleted */
if (element->heaptidsLength == 0)
{
*idx = i;
break;
}
}
}
/*
* Get update index
*/
static int
GetUpdateIndex(HnswElement element, HnswElement newElement, float distance, int m, int lm, int lc, Relation index, HnswSupport * support, MemoryContext updateCtx)
{
char *base = NULL;
int idx = -1;
HnswNeighborArray *neighbors;
MemoryContext oldCtx = MemoryContextSwitchTo(updateCtx);
/*
* Get latest neighbors since they may have changed. Do not lock yet since
* selecting neighbors can take time. Could use optimistic locking to
* retry if another update occurs before getting exclusive lock.
*/
neighbors = HnswLoadNeighbors(element, index, m, lm, lc);
/*
* Could improve performance for vacuuming by checking neighbors against
* list of elements being deleted to find index. It's important to exclude
* already deleted elements for this since they can be replaced at any
* time.
*/
if (neighbors->length < lm)
idx = -2;
else
{
HnswQuery q;
q.value = HnswGetValue(base, element);
LoadElementsForInsert(neighbors, &q, &idx, index, support);
if (idx == -1)
HnswUpdateConnection(base, neighbors, newElement, distance, lm, &idx, index, support);
}
MemoryContextSwitchTo(oldCtx);
MemoryContextReset(updateCtx);
return idx;
}
/*
* Check if connection already exists
*/
@@ -461,94 +336,14 @@ ConnectionExists(HnswElement e, HnswNeighborTuple ntup, int startIdx, int lm)
return false;
}
/*
* Update neighbor
*/
static void
UpdateNeighborOnDisk(HnswElement element, HnswElement newElement, int idx, int m, int lm, int lc, Relation index, bool checkExisting, bool building)
{
Buffer buf;
Page page;
GenericXLogState *state;
HnswNeighborTuple ntup;
int startIdx;
OffsetNumber offno = element->neighborOffno;
/* Register page */
buf = ReadBuffer(index, element->neighborPage);
LockBuffer(buf, BUFFER_LOCK_EXCLUSIVE);
if (building)
{
state = NULL;
page = BufferGetPage(buf);
}
else
{
state = GenericXLogStart(index);
page = GenericXLogRegisterBuffer(state, buf, 0);
}
/* Get tuple */
ntup = (HnswNeighborTuple) PageGetItem(page, PageGetItemId(page, offno));
/* Calculate index for update */
startIdx = (element->level - lc) * m;
/* Check for existing connection */
if (checkExisting && ConnectionExists(newElement, ntup, startIdx, lm))
idx = -1;
else if (idx == -2)
{
/* Find free offset if still exists */
/* TODO Retry updating connections if not */
for (int j = 0; j < lm; j++)
{
if (!ItemPointerIsValid(&ntup->indextids[startIdx + j]))
{
idx = startIdx + j;
break;
}
}
}
else
idx += startIdx;
/* Make robust to issues */
if (idx >= 0 && idx < ntup->count)
{
ItemPointer indextid = &ntup->indextids[idx];
/* Update neighbor on the buffer */
ItemPointerSet(indextid, newElement->blkno, newElement->offno);
/* Commit */
if (building)
MarkBufferDirty(buf);
else
GenericXLogFinish(state);
}
else if (!building)
GenericXLogAbort(state);
UnlockReleaseBuffer(buf);
}
/*
* Update neighbors
*/
void
HnswUpdateNeighborsOnDisk(Relation index, HnswSupport * support, HnswElement e, int m, bool checkExisting, bool building)
HnswUpdateNeighborsOnDisk(Relation index, FmgrInfo *procinfo, Oid collation, HnswElement e, int m, bool checkExisting, bool building)
{
char *base = NULL;
/* Use separate memory context to improve performance for larger vectors */
MemoryContext updateCtx = GenerationContextCreate(CurrentMemoryContext,
"Hnsw insert update context",
#if PG_VERSION_NUM >= 150000
128 * 1024, 128 * 1024,
#endif
128 * 1024);
for (int lc = e->level; lc >= 0; lc--)
{
int lm = HnswGetLayerM(m, lc);
@@ -557,20 +352,92 @@ HnswUpdateNeighborsOnDisk(Relation index, HnswSupport * support, HnswElement e,
for (int i = 0; i < neighbors->length; i++)
{
HnswCandidate *hc = &neighbors->items[i];
Buffer buf;
Page page;
GenericXLogState *state;
HnswNeighborTuple ntup;
int idx = -1;
int startIdx;
HnswElement neighborElement = HnswPtrAccess(base, hc->element);
int idx;
OffsetNumber offno = neighborElement->neighborOffno;
idx = GetUpdateIndex(neighborElement, e, hc->distance, m, lm, lc, index, support, updateCtx);
/* Get latest neighbors since they may have changed */
/* Do not lock yet since selecting neighbors can take time */
HnswLoadNeighbors(neighborElement, index, m);
/*
* Could improve performance for vacuuming by checking neighbors
* against list of elements being deleted to find index. It's
* important to exclude already deleted elements for this since
* they can be replaced at any time.
*/
/* Select neighbors */
HnswUpdateConnection(NULL, e, hc, lm, lc, &idx, index, procinfo, collation);
/* New element was not selected as a neighbor */
if (idx == -1)
continue;
UpdateNeighborOnDisk(neighborElement, e, idx, m, lm, lc, index, checkExisting, building);
/* Register page */
buf = ReadBuffer(index, neighborElement->neighborPage);
LockBuffer(buf, BUFFER_LOCK_EXCLUSIVE);
if (building)
{
state = NULL;
page = BufferGetPage(buf);
}
else
{
state = GenericXLogStart(index);
page = GenericXLogRegisterBuffer(state, buf, 0);
}
/* Get tuple */
ntup = (HnswNeighborTuple) PageGetItem(page, PageGetItemId(page, offno));
/* Calculate index for update */
startIdx = (neighborElement->level - lc) * m;
/* Check for existing connection */
if (checkExisting && ConnectionExists(e, ntup, startIdx, lm))
idx = -1;
else if (idx == -2)
{
/* Find free offset if still exists */
/* TODO Retry updating connections if not */
for (int j = 0; j < lm; j++)
{
if (!ItemPointerIsValid(&ntup->indextids[startIdx + j]))
{
idx = startIdx + j;
break;
}
}
}
else
idx += startIdx;
/* Make robust to issues */
if (idx >= 0 && idx < ntup->count)
{
ItemPointer indextid = &ntup->indextids[idx];
/* Update neighbor on the buffer */
ItemPointerSet(indextid, e->blkno, e->offno);
/* Commit */
if (building)
MarkBufferDirty(buf);
else
GenericXLogFinish(state);
}
else if (!building)
GenericXLogAbort(state);
UnlockReleaseBuffer(buf);
}
}
MemoryContextDelete(updateCtx);
}
/*
@@ -660,7 +527,7 @@ FindDuplicateOnDisk(Relation index, HnswElement element, bool building)
* Update graph on disk
*/
static void
UpdateGraphOnDisk(Relation index, HnswSupport * support, HnswElement element, int m, int efConstruction, HnswElement entryPoint, bool building)
UpdateGraphOnDisk(Relation index, FmgrInfo *procinfo, Oid collation, HnswElement element, int m, int efConstruction, HnswElement entryPoint, bool building)
{
BlockNumber newInsertPage = InvalidBlockNumber;
@@ -676,7 +543,7 @@ UpdateGraphOnDisk(Relation index, HnswSupport * support, HnswElement element, in
HnswUpdateMetaPage(index, 0, NULL, newInsertPage, MAIN_FORKNUM, building);
/* Update neighbors */
HnswUpdateNeighborsOnDisk(index, support, element, m, false, building);
HnswUpdateNeighborsOnDisk(index, procinfo, collation, element, m, false, building);
/* Update entry point if needed */
if (entryPoint == NULL || element->level > entryPoint->level)
@@ -687,12 +554,14 @@ UpdateGraphOnDisk(Relation index, HnswSupport * support, HnswElement element, in
* Insert a tuple into the index
*/
bool
HnswInsertTupleOnDisk(Relation index, HnswSupport * support, Datum value, ItemPointer heaptid, bool building)
HnswInsertTupleOnDisk(Relation index, Datum value, Datum *values, bool *isnull, ItemPointer heap_tid, bool building)
{
HnswElement entryPoint;
HnswElement element;
int m;
int efConstruction = HnswGetEfConstruction(index);
FmgrInfo *procinfo = index_getprocinfo(index, 1, HNSW_DISTANCE_PROC);
Oid collation = index->rd_indcollation[0];
LOCKMODE lockmode = ShareLock;
char *base = NULL;
@@ -707,7 +576,7 @@ HnswInsertTupleOnDisk(Relation index, HnswSupport * support, Datum value, ItemPo
HnswGetMetaPageInfo(index, &m, &entryPoint);
/* Create an element */
element = HnswInitElement(base, heaptid, m, HnswGetMl(m), HnswGetMaxLevel(m), NULL);
element = HnswInitElement(base, heap_tid, m, HnswGetMl(m), HnswGetMaxLevel(m), NULL);
HnswPtrStore(base, element->value, DatumGetPointer(value));
/* Prevent concurrent inserts when likely updating entry point */
@@ -725,10 +594,10 @@ HnswInsertTupleOnDisk(Relation index, HnswSupport * support, Datum value, ItemPo
}
/* Find neighbors for element */
HnswFindElementNeighbors(base, element, entryPoint, index, support, m, efConstruction, false);
HnswFindElementNeighbors(base, element, entryPoint, index, procinfo, collation, m, efConstruction, false);
/* Update graph on disk */
UpdateGraphOnDisk(index, support, element, m, efConstruction, entryPoint, building);
UpdateGraphOnDisk(index, procinfo, collation, element, m, efConstruction, entryPoint, building);
/* Release lock */
UnlockPage(index, HNSW_UPDATE_LOCK, lockmode);
@@ -740,19 +609,31 @@ HnswInsertTupleOnDisk(Relation index, HnswSupport * support, Datum value, ItemPo
* Insert a tuple into the index
*/
static void
HnswInsertTuple(Relation index, Datum *values, bool *isnull, ItemPointer heaptid)
HnswInsertTuple(Relation index, Datum *values, bool *isnull, ItemPointer heap_tid)
{
Datum value;
const HnswTypeInfo *typeInfo = HnswGetTypeInfo(index);
HnswSupport support;
FmgrInfo *normprocinfo;
Oid collation = index->rd_indcollation[0];
HnswInitSupport(&support, index);
/* Detoast once for all calls */
value = PointerGetDatum(PG_DETOAST_DATUM(values[0]));
/* Form index value */
if (!HnswFormIndexValue(&value, values, isnull, typeInfo, &support))
return;
/* Check value */
if (typeInfo->checkValue != NULL)
typeInfo->checkValue(DatumGetPointer(value));
HnswInsertTupleOnDisk(index, &support, value, heaptid, false);
/* Normalize if needed */
normprocinfo = HnswOptionalProcInfo(index, HNSW_NORM_PROC);
if (normprocinfo != NULL)
{
if (!HnswCheckNorm(normprocinfo, collation, value))
return;
value = HnswNormValue(typeInfo, collation, value);
}
HnswInsertTupleOnDisk(index, value, values, isnull, heap_tid, false);
}
/*

View File

@@ -5,74 +5,39 @@
#include "pgstat.h"
#include "storage/bufmgr.h"
#include "storage/lmgr.h"
#include "utils/float.h"
#include "utils/memutils.h"
/*
* Algorithm 5 from paper
*/
static List *
GetScanItems(IndexScanDesc scan, Datum value)
GetScanItems(IndexScanDesc scan, Datum q)
{
HnswScanOpaque so = (HnswScanOpaque) scan->opaque;
Relation index = scan->indexRelation;
HnswSupport *support = &so->support;
FmgrInfo *procinfo = so->procinfo;
Oid collation = so->collation;
List *ep;
List *w;
int m;
HnswElement entryPoint;
char *base = NULL;
HnswQuery *q = &so->q;
/* Get m and entry point */
HnswGetMetaPageInfo(index, &m, &entryPoint);
q->value = value;
so->m = m;
if (entryPoint == NULL)
return NIL;
ep = list_make1(HnswEntryCandidate(base, entryPoint, q, index, support, false));
ep = list_make1(HnswEntryCandidate(base, entryPoint, q, index, procinfo, collation, false));
for (int lc = entryPoint->level; lc >= 1; lc--)
{
w = HnswSearchLayer(base, q, ep, 1, lc, index, support, m, false, NULL, NULL, NULL, true, NULL);
w = HnswSearchLayer(base, q, ep, 1, lc, index, procinfo, collation, m, false, NULL);
ep = w;
}
return HnswSearchLayer(base, q, ep, hnsw_ef_search, 0, index, support, m, false, NULL, &so->v, hnsw_iterative_search != HNSW_ITERATIVE_SEARCH_OFF ? &so->discarded : NULL, true, &so->tuples);
}
/*
* Resume scan at ground level with discarded candidates
*/
static List *
ResumeScanItems(IndexScanDesc scan)
{
HnswScanOpaque so = (HnswScanOpaque) scan->opaque;
Relation index = scan->indexRelation;
List *ep = NIL;
char *base = NULL;
int batch_size = hnsw_ef_search;
if (pairingheap_is_empty(so->discarded))
return NIL;
/* Get next batch of candidates */
for (int i = 0; i < batch_size; i++)
{
HnswSearchCandidate *sc;
if (pairingheap_is_empty(so->discarded))
break;
sc = HnswGetSearchCandidate(w_node, pairingheap_remove_first(so->discarded));
ep = lappend(ep, sc);
}
return HnswSearchLayer(base, &so->q, ep, batch_size, 0, index, &so->support, so->m, false, NULL, &so->v, &so->discarded, false, &so->tuples);
return HnswSearchLayer(base, q, ep, hnsw_ef_search, 0, index, procinfo, collation, m, false, NULL);
}
/*
@@ -95,24 +60,13 @@ GetScanValue(IndexScanDesc scan)
Assert(!VARATT_IS_EXTENDED(DatumGetPointer(value)));
/* Normalize if needed */
if (so->support.normprocinfo != NULL)
value = HnswNormValue(so->typeInfo, so->support.collation, value);
if (so->normprocinfo != NULL)
value = HnswNormValue(so->typeInfo, so->collation, value);
}
return value;
}
#if defined(HNSW_MEMORY)
/*
* Show memory usage
*/
static void
ShowMemoryUsage(HnswScanOpaque so)
{
elog(INFO, "memory: %zu KB, tuples: " INT64_FORMAT, MemoryContextMemAllocated(so->tmpCtx, false) / 1024, so->tuples);
}
#endif
/*
* Prepare for an index scan
*/
@@ -127,19 +81,14 @@ hnswbeginscan(Relation index, int nkeys, int norderbys)
so = (HnswScanOpaque) palloc(sizeof(HnswScanOpaqueData));
so->typeInfo = HnswGetTypeInfo(index);
so->first = true;
so->v.tids = NULL;
so->discarded = NULL;
/*
* Use a lower max allocation size than default to allow scanning more
* tuples for iterative search before exceeding work_mem
*/
so->tmpCtx = AllocSetContextCreate(CurrentMemoryContext,
"Hnsw scan temporary context",
0, 8 * 1024, 512 * 1024);
ALLOCSET_DEFAULT_SIZES);
/* Set support functions */
HnswInitSupport(&so->support, index);
so->procinfo = index_getprocinfo(index, 1, HNSW_DISTANCE_PROC);
so->normprocinfo = HnswOptionalProcInfo(index, HNSW_NORM_PROC);
so->collation = index->rd_indcollation[0];
scan->opaque = so;
@@ -154,15 +103,7 @@ hnswrescan(IndexScanDesc scan, ScanKey keys, int nkeys, ScanKey orderbys, int no
{
HnswScanOpaque so = (HnswScanOpaque) scan->opaque;
if (so->v.tids != NULL)
tidhash_reset(so->v.tids);
if (so->discarded != NULL)
pairingheap_reset(so->discarded);
so->first = true;
so->tuples = 0;
so->previousDistance = -get_float8_infinity();
MemoryContextReset(so->tmpCtx);
if (keys && scan->numberOfKeys > 0)
@@ -219,105 +160,27 @@ hnswgettuple(IndexScanDesc scan, ScanDirection dir)
so->first = false;
#if defined(HNSW_MEMORY)
ShowMemoryUsage(so);
#if defined(HNSW_MEMORY) && PG_VERSION_NUM >= 130000
elog(INFO, "memory: %zu MB", MemoryContextMemAllocated(so->tmpCtx, false) / (1024 * 1024));
#endif
}
for (;;)
while (list_length(so->w) > 0)
{
char *base = NULL;
HnswSearchCandidate *sc;
HnswElement element;
HnswCandidate *hc = llast(so->w);
HnswElement element = HnswPtrAccess(base, hc->element);
ItemPointer heaptid;
if (list_length(so->w) == 0)
{
if (hnsw_iterative_search == HNSW_ITERATIVE_SEARCH_OFF)
break;
/* Empty index */
if (so->discarded == NULL)
break;
/* Reached max number of tuples */
if (hnsw_max_search_tuples != -1 && so->tuples >= hnsw_max_search_tuples)
{
if (pairingheap_is_empty(so->discarded))
break;
/* Return remaining tuples */
so->w = lappend(so->w, HnswGetSearchCandidate(w_node, pairingheap_remove_first(so->discarded)));
}
/* Prevent scans from consuming too much memory */
else if (MemoryContextMemAllocated(so->tmpCtx, false) > (Size) work_mem * 1024L)
{
if (pairingheap_is_empty(so->discarded))
{
ereport(DEBUG1,
(errmsg("hnsw index scan exceeded work_mem after " INT64_FORMAT " tuples", so->tuples),
errhint("Increase work_mem to scan more tuples.")));
break;
}
/* Return remaining tuples */
so->w = lappend(so->w, HnswGetSearchCandidate(w_node, pairingheap_remove_first(so->discarded)));
}
else
{
/*
* Locking ensures when neighbors are read, the elements they
* reference will not be deleted (and replaced) during the
* iteration.
*
* Elements loaded into memory on previous iterations may have
* been deleted (and replaced), so when reading neighbors, the
* element version must be checked.
*/
LockPage(scan->indexRelation, HNSW_SCAN_LOCK, ShareLock);
so->w = ResumeScanItems(scan);
UnlockPage(scan->indexRelation, HNSW_SCAN_LOCK, ShareLock);
#if defined(HNSW_MEMORY)
ShowMemoryUsage(so);
#endif
}
if (list_length(so->w) == 0)
break;
}
sc = llast(so->w);
element = HnswPtrAccess(base, sc->element);
/* Move to next element if no valid heap TIDs */
if (element->heaptidsLength == 0)
{
so->w = list_delete_last(so->w);
/* Mark memory as free for next iteration */
if (hnsw_iterative_search != HNSW_ITERATIVE_SEARCH_OFF)
{
pfree(element);
pfree(sc);
}
continue;
}
heaptid = &element->heaptids[--element->heaptidsLength];
if (hnsw_iterative_search == HNSW_ITERATIVE_SEARCH_STRICT)
{
if (sc->distance < so->previousDistance)
continue;
so->previousDistance = sc->distance;
}
MemoryContextSwitchTo(oldCtx);
scan->xs_heaptid = *heaptid;

File diff suppressed because it is too large Load Diff

View File

@@ -184,12 +184,13 @@ static void
RepairGraphElement(HnswVacuumState * vacuumstate, HnswElement element, HnswElement entryPoint)
{
Relation index = vacuumstate->index;
HnswSupport *support = &vacuumstate->support;
Buffer buf;
Page page;
GenericXLogState *state;
int m = vacuumstate->m;
int efConstruction = vacuumstate->efConstruction;
FmgrInfo *procinfo = vacuumstate->procinfo;
Oid collation = vacuumstate->collation;
BufferAccessStrategy bas = vacuumstate->bas;
HnswNeighborTuple ntup = vacuumstate->ntup;
Size ntupSize = HNSW_NEIGHBOR_TUPLE_SIZE(element->level, m);
@@ -204,7 +205,7 @@ RepairGraphElement(HnswVacuumState * vacuumstate, HnswElement element, HnswEleme
element->heaptidsLength = 0;
/* Find neighbors for element, skipping itself */
HnswFindElementNeighbors(base, element, entryPoint, index, support, m, efConstruction, true);
HnswFindElementNeighbors(base, element, entryPoint, index, procinfo, collation, m, efConstruction, true);
/* Zero memory for each element */
MemSet(ntup, 0, HNSW_TUPLE_ALLOC_SIZE);
@@ -228,7 +229,7 @@ RepairGraphElement(HnswVacuumState * vacuumstate, HnswElement element, HnswEleme
UnlockReleaseBuffer(buf);
/* Update neighbors */
HnswUpdateNeighborsOnDisk(index, support, element, m, true, false);
HnswUpdateNeighborsOnDisk(index, procinfo, collation, element, m, true, false);
}
/*
@@ -238,7 +239,6 @@ static void
RepairGraphEntryPoint(HnswVacuumState * vacuumstate)
{
Relation index = vacuumstate->index;
HnswSupport *support = &vacuumstate->support;
HnswElement highestPoint = &vacuumstate->highestPoint;
HnswElement entryPoint;
MemoryContext oldCtx = MemoryContextSwitchTo(vacuumstate->tmpCtx);
@@ -256,7 +256,7 @@ RepairGraphEntryPoint(HnswVacuumState * vacuumstate)
LockPage(index, HNSW_UPDATE_LOCK, ShareLock);
/* Load element */
HnswLoadElement(highestPoint, NULL, NULL, index, support, true, NULL);
HnswLoadElement(highestPoint, NULL, NULL, index, vacuumstate->procinfo, vacuumstate->collation, true, NULL);
/* Repair if needed */
if (NeedsUpdated(vacuumstate, highestPoint))
@@ -294,7 +294,7 @@ RepairGraphEntryPoint(HnswVacuumState * vacuumstate)
* is outdated, this can remove connections at higher levels in
* the graph until they are repaired, but this should be fine.
*/
HnswLoadElement(entryPoint, NULL, NULL, index, support, true, NULL);
HnswLoadElement(entryPoint, NULL, NULL, index, vacuumstate->procinfo, vacuumstate->collation, true, NULL);
if (NeedsUpdated(vacuumstate, entryPoint))
{
@@ -527,14 +527,6 @@ MarkDeleted(HnswVacuumState * vacuumstate)
for (int i = 0; i < ntup->count; i++)
ItemPointerSetInvalid(&ntup->indextids[i]);
/* Increment version */
/* This is used to avoid incorrect reads for iterative scans */
/* Reserve some bits for future use */
etup->version++;
if (etup->version > 15)
etup->version = 1;
ntup->version = etup->version;
/*
* We modified the tuples in place, no need to call
* PageIndexTupleOverwrite
@@ -581,13 +573,13 @@ InitVacuumState(HnswVacuumState * vacuumstate, IndexVacuumInfo *info, IndexBulkD
vacuumstate->callback_state = callback_state;
vacuumstate->efConstruction = HnswGetEfConstruction(index);
vacuumstate->bas = GetAccessStrategy(BAS_BULKREAD);
vacuumstate->procinfo = index_getprocinfo(index, 1, HNSW_DISTANCE_PROC);
vacuumstate->collation = index->rd_indcollation[0];
vacuumstate->ntup = palloc0(HNSW_TUPLE_ALLOC_SIZE);
vacuumstate->tmpCtx = AllocSetContextCreate(CurrentMemoryContext,
"Hnsw vacuum temporary context",
ALLOCSET_DEFAULT_SIZES);
HnswInitSupport(&vacuumstate->support, index);
/* Get m from metapage */
HnswGetMetaPageInfo(index, &vacuumstate->m, NULL);

View File

@@ -1,754 +0,0 @@
#include "postgres.h"
#include <limits.h>
#include <math.h>
#include "catalog/pg_type.h"
#include "fmgr.h"
#include "intvec.h"
#include "lib/stringinfo.h"
#include "libpq/pqformat.h"
#include "utils/array.h"
#include "utils/builtins.h"
#include "utils/lsyscache.h"
/*
* Ensure same dimensions
*/
static inline void
CheckDims(IntVector * a, IntVector * b)
{
if (a->dim != b->dim)
ereport(ERROR,
(errcode(ERRCODE_DATA_EXCEPTION),
errmsg("different intvec dimensions %d and %d", a->dim, b->dim)));
}
/*
* Ensure expected dimensions
*/
static inline void
CheckExpectedDim(int32 typmod, int dim)
{
if (typmod != -1 && typmod != dim)
ereport(ERROR,
(errcode(ERRCODE_DATA_EXCEPTION),
errmsg("expected %d dimensions, not %d", typmod, dim)));
}
/*
* Ensure valid dimensions
*/
static inline void
CheckDim(int dim)
{
if (dim < 1)
ereport(ERROR,
(errcode(ERRCODE_DATA_EXCEPTION),
errmsg("intvec must have at least 1 dimension")));
if (dim > INTVEC_MAX_DIM)
ereport(ERROR,
(errcode(ERRCODE_PROGRAM_LIMIT_EXCEEDED),
errmsg("intvec cannot have more than %d dimensions", INTVEC_MAX_DIM)));
}
/*
* Ensure element in range
*/
static inline void
CheckElement(long value)
{
if (value < SCHAR_MIN || value > SCHAR_MAX)
ereport(ERROR,
(errcode(ERRCODE_NUMERIC_VALUE_OUT_OF_RANGE),
errmsg("value \"%ld\" is out of range for type intvec", value)));
}
/*
* Allocate and initialize a new int vector
*/
IntVector *
InitIntVector(int dim)
{
IntVector *result;
int size;
size = INTVEC_SIZE(dim);
result = (IntVector *) palloc0(size);
SET_VARSIZE(result, size);
result->dim = dim;
return result;
}
/*
* Check for whitespace, since array_isspace() is static
*/
static inline bool
intvec_isspace(char ch)
{
if (ch == ' ' ||
ch == '\t' ||
ch == '\n' ||
ch == '\r' ||
ch == '\v' ||
ch == '\f')
return true;
return false;
}
/*
* Convert textual representation to internal representation
*/
PGDLLEXPORT PG_FUNCTION_INFO_V1(intvec_in);
Datum
intvec_in(PG_FUNCTION_ARGS)
{
char *lit = PG_GETARG_CSTRING(0);
int32 typmod = PG_GETARG_INT32(2);
int8 x[INTVEC_MAX_DIM];
int dim = 0;
char *pt = lit;
IntVector *result;
while (intvec_isspace(*pt))
pt++;
if (*pt != '[')
ereport(ERROR,
(errcode(ERRCODE_INVALID_TEXT_REPRESENTATION),
errmsg("invalid input syntax for type intvec: \"%s\"", lit),
errdetail("Vector contents must start with \"[\".")));
pt++;
while (intvec_isspace(*pt))
pt++;
if (*pt == ']')
ereport(ERROR,
(errcode(ERRCODE_DATA_EXCEPTION),
errmsg("intvec must have at least 1 dimension")));
for (;;)
{
long val;
char *stringEnd;
if (dim == INTVEC_MAX_DIM)
ereport(ERROR,
(errcode(ERRCODE_PROGRAM_LIMIT_EXCEEDED),
errmsg("intvec cannot have more than %d dimensions", VECTOR_MAX_DIM)));
while (intvec_isspace(*pt))
pt++;
/* Check for empty string like float4in */
if (*pt == '\0')
ereport(ERROR,
(errcode(ERRCODE_INVALID_TEXT_REPRESENTATION),
errmsg("invalid input syntax for type intvec: \"%s\"", lit)));
errno = 0;
/* Use similar logic as int2vectorin */
val = strtol(pt, &stringEnd, 10);
if (stringEnd == pt)
ereport(ERROR,
(errcode(ERRCODE_INVALID_TEXT_REPRESENTATION),
errmsg("invalid input syntax for type intvec: \"%s\"", lit)));
/* Check for range error like float4in */
if (errno == ERANGE || val < SCHAR_MIN || val > SCHAR_MAX)
ereport(ERROR,
(errcode(ERRCODE_NUMERIC_VALUE_OUT_OF_RANGE),
errmsg("\"%s\" is out of range for type intvec", pnstrdup(pt, stringEnd - pt))));
CheckElement(val);
x[dim++] = val;
pt = stringEnd;
while (intvec_isspace(*pt))
pt++;
if (*pt == ',')
pt++;
else if (*pt == ']')
{
pt++;
break;
}
else
ereport(ERROR,
(errcode(ERRCODE_INVALID_TEXT_REPRESENTATION),
errmsg("invalid input syntax for type intvec: \"%s\"", lit)));
}
/* Only whitespace is allowed after the closing brace */
while (intvec_isspace(*pt))
pt++;
if (*pt != '\0')
ereport(ERROR,
(errcode(ERRCODE_INVALID_TEXT_REPRESENTATION),
errmsg("invalid input syntax for type intvec: \"%s\"", lit),
errdetail("Junk after closing right brace.")));
CheckDim(dim);
CheckExpectedDim(typmod, dim);
result = InitIntVector(dim);
for (int i = 0; i < dim; i++)
result->x[i] = x[i];
PG_RETURN_POINTER(result);
}
/*
* Convert internal representation to textual representation
*/
PGDLLEXPORT PG_FUNCTION_INFO_V1(intvec_out);
Datum
intvec_out(PG_FUNCTION_ARGS)
{
IntVector *vector = PG_GETARG_INTVEC_P(0);
int dim = vector->dim;
char *buf;
char *ptr;
/*
* Need:
*
* dim * 4 bytes for elements (-128 to 127)
*
* dim - 1 bytes for separator
*
* 3 bytes for [, ], and \0
*/
buf = (char *) palloc(5 * dim + 2);
ptr = buf;
*ptr = '[';
ptr++;
for (int i = 0; i < dim; i++)
{
if (i > 0)
{
*ptr = ',';
ptr++;
}
#if PG_VERSION_NUM >= 140000
ptr += pg_ltoa(vector->x[i], ptr);
#else
pg_ltoa(vector->x[i], ptr);
while (*ptr != '\0')
ptr++;
#endif
}
*ptr = ']';
ptr++;
*ptr = '\0';
PG_FREE_IF_COPY(vector, 0);
PG_RETURN_CSTRING(buf);
}
/*
* Convert type modifier
*/
PGDLLEXPORT PG_FUNCTION_INFO_V1(intvec_typmod_in);
Datum
intvec_typmod_in(PG_FUNCTION_ARGS)
{
ArrayType *ta = PG_GETARG_ARRAYTYPE_P(0);
int32 *tl;
int n;
tl = ArrayGetIntegerTypmods(ta, &n);
if (n != 1)
ereport(ERROR,
(errcode(ERRCODE_INVALID_PARAMETER_VALUE),
errmsg("invalid type modifier")));
if (*tl < 1)
ereport(ERROR,
(errcode(ERRCODE_INVALID_PARAMETER_VALUE),
errmsg("dimensions for type intvec must be at least 1")));
if (*tl > INTVEC_MAX_DIM)
ereport(ERROR,
(errcode(ERRCODE_INVALID_PARAMETER_VALUE),
errmsg("dimensions for type intvec cannot exceed %d", INTVEC_MAX_DIM)));
PG_RETURN_INT32(*tl);
}
/*
* Convert external binary representation to internal representation
*/
PGDLLEXPORT PG_FUNCTION_INFO_V1(intvec_recv);
Datum
intvec_recv(PG_FUNCTION_ARGS)
{
StringInfo buf = (StringInfo) PG_GETARG_POINTER(0);
int32 typmod = PG_GETARG_INT32(2);
IntVector *result;
int16 dim;
int16 unused;
dim = pq_getmsgint(buf, sizeof(int16));
unused = pq_getmsgint(buf, sizeof(int16));
CheckDim(dim);
CheckExpectedDim(typmod, dim);
if (unused != 0)
ereport(ERROR,
(errcode(ERRCODE_DATA_EXCEPTION),
errmsg("expected unused to be 0, not %d", unused)));
result = InitIntVector(dim);
for (int i = 0; i < dim; i++)
result->x[i] = pq_getmsgint(buf, sizeof(int8));
PG_RETURN_POINTER(result);
}
/*
* Convert internal representation to the external binary representation
*/
PGDLLEXPORT PG_FUNCTION_INFO_V1(intvec_send);
Datum
intvec_send(PG_FUNCTION_ARGS)
{
IntVector *vec = PG_GETARG_INTVEC_P(0);
StringInfoData buf;
pq_begintypsend(&buf);
pq_sendint(&buf, vec->dim, sizeof(int16));
pq_sendint(&buf, vec->unused, sizeof(int16));
for (int i = 0; i < vec->dim; i++)
pq_sendint8(&buf, vec->x[i]);
PG_RETURN_BYTEA_P(pq_endtypsend(&buf));
}
/*
* Convert int vector to int vector
* This is needed to check the type modifier
*/
PGDLLEXPORT PG_FUNCTION_INFO_V1(intvec);
Datum
intvec(PG_FUNCTION_ARGS)
{
IntVector *vec = PG_GETARG_INTVEC_P(0);
int32 typmod = PG_GETARG_INT32(1);
CheckExpectedDim(typmod, vec->dim);
PG_RETURN_POINTER(vec);
}
/*
* Convert array to intvec vector
*/
PGDLLEXPORT PG_FUNCTION_INFO_V1(array_to_intvec);
Datum
array_to_intvec(PG_FUNCTION_ARGS)
{
ArrayType *array = PG_GETARG_ARRAYTYPE_P(0);
int32 typmod = PG_GETARG_INT32(1);
IntVector *result;
int16 typlen;
bool typbyval;
char typalign;
Datum *elemsp;
int nelemsp;
if (ARR_NDIM(array) > 1)
ereport(ERROR,
(errcode(ERRCODE_DATA_EXCEPTION),
errmsg("array must be 1-D")));
if (ARR_HASNULL(array) && array_contains_nulls(array))
ereport(ERROR,
(errcode(ERRCODE_NULL_VALUE_NOT_ALLOWED),
errmsg("array must not contain nulls")));
get_typlenbyvalalign(ARR_ELEMTYPE(array), &typlen, &typbyval, &typalign);
deconstruct_array(array, ARR_ELEMTYPE(array), typlen, typbyval, typalign, &elemsp, NULL, &nelemsp);
CheckDim(nelemsp);
CheckExpectedDim(typmod, nelemsp);
result = InitIntVector(nelemsp);
if (ARR_ELEMTYPE(array) == INT4OID)
{
for (int i = 0; i < nelemsp; i++)
{
long l = DatumGetInt32(elemsp[i]);
CheckElement(l);
result->x[i] = l;
}
}
else
{
ereport(ERROR,
(errcode(ERRCODE_DATA_EXCEPTION),
errmsg("unsupported array type")));
}
/*
* Free allocation from deconstruct_array. Do not free individual elements
* when pass-by-reference since they point to original array.
*/
pfree(elemsp);
PG_RETURN_POINTER(result);
}
/*
* Convert int vector to integer[]
*/
FUNCTION_PREFIX PG_FUNCTION_INFO_V1(intvec_to_integer);
Datum
intvec_to_integer(PG_FUNCTION_ARGS)
{
IntVector *vec = PG_GETARG_INTVEC_P(0);
Datum *datums;
ArrayType *result;
datums = (Datum *) palloc(sizeof(Datum) * vec->dim);
for (int i = 0; i < vec->dim; i++)
datums[i] = Int32GetDatum((int) vec->x[i]);
result = construct_array(datums, vec->dim, INT4OID, sizeof(int32), true, TYPALIGN_INT);
pfree(datums);
PG_RETURN_POINTER(result);
}
static int
IntvecL2SquaredDistance(int dim, int8 *ax, int8 *bx)
{
int distance = 0;
/* Auto-vectorized */
for (int i = 0; i < dim; i++)
{
int diff = (int) ax[i] - (int) bx[i];
distance += diff * diff;
}
return distance;
}
/*
* Get the L2 distance between int vectors
*/
PGDLLEXPORT PG_FUNCTION_INFO_V1(intvec_l2_distance);
Datum
intvec_l2_distance(PG_FUNCTION_ARGS)
{
IntVector *a = PG_GETARG_INTVEC_P(0);
IntVector *b = PG_GETARG_INTVEC_P(1);
CheckDims(a, b);
PG_RETURN_FLOAT8(sqrt((double) IntvecL2SquaredDistance(a->dim, a->x, b->x)));
}
/*
* Get the L2 squared distance between int vectors
*/
PGDLLEXPORT PG_FUNCTION_INFO_V1(intvec_l2_squared_distance);
Datum
intvec_l2_squared_distance(PG_FUNCTION_ARGS)
{
IntVector *a = PG_GETARG_INTVEC_P(0);
IntVector *b = PG_GETARG_INTVEC_P(1);
CheckDims(a, b);
PG_RETURN_FLOAT8((double) IntvecL2SquaredDistance(a->dim, a->x, b->x));
}
static int
IntvecInnerProduct(int dim, int8 *ax, int8 *bx)
{
int distance = 0;
/* Auto-vectorized */
for (int i = 0; i < dim; i++)
distance += (int) ax[i] * (int) bx[i];
return distance;
}
/*
* Get the inner product of two int vectors
*/
PGDLLEXPORT PG_FUNCTION_INFO_V1(intvec_inner_product);
Datum
intvec_inner_product(PG_FUNCTION_ARGS)
{
IntVector *a = PG_GETARG_INTVEC_P(0);
IntVector *b = PG_GETARG_INTVEC_P(1);
CheckDims(a, b);
PG_RETURN_FLOAT8((double) IntvecInnerProduct(a->dim, a->x, b->x));
}
/*
* Get the negative inner product of two int vectors
*/
PGDLLEXPORT PG_FUNCTION_INFO_V1(intvec_negative_inner_product);
Datum
intvec_negative_inner_product(PG_FUNCTION_ARGS)
{
IntVector *a = PG_GETARG_INTVEC_P(0);
IntVector *b = PG_GETARG_INTVEC_P(1);
CheckDims(a, b);
PG_RETURN_FLOAT8((double) -IntvecInnerProduct(a->dim, a->x, b->x));
}
static double
IntvecCosineSimilarity(int dim, int8 *ax, int8 *bx)
{
int similarity = 0;
int norma = 0;
int normb = 0;
/* Auto-vectorized */
for (int i = 0; i < dim; i++)
{
int axi = ax[i];
int bxi = bx[i];
similarity += axi * bxi;
norma += axi * axi;
normb += bxi * bxi;
}
/* Use sqrt(a * b) over sqrt(a) * sqrt(b) */
return (double) similarity / sqrt((double) norma * (double) normb);
}
/*
* Get the cosine distance between two int vectors
*/
PGDLLEXPORT PG_FUNCTION_INFO_V1(intvec_cosine_distance);
Datum
intvec_cosine_distance(PG_FUNCTION_ARGS)
{
IntVector *a = PG_GETARG_INTVEC_P(0);
IntVector *b = PG_GETARG_INTVEC_P(1);
double similarity;
CheckDims(a, b);
similarity = IntvecCosineSimilarity(a->dim, a->x, b->x);
#ifdef _MSC_VER
/* /fp:fast may not propagate NaN */
if (isnan(similarity))
PG_RETURN_FLOAT8(NAN);
#endif
/* Keep in range */
if (similarity > 1)
similarity = 1;
else if (similarity < -1)
similarity = -1;
PG_RETURN_FLOAT8(1 - similarity);
}
static int
IntvecL1Distance(int dim, int8 *ax, int8 *bx)
{
int distance = 0;
/* Auto-vectorized */
for (int i = 0; i < dim; i++)
distance += abs((int) ax[i] - (int) bx[i]);
return distance;
}
/*
* Get the L1 distance between two int vectors
*/
PGDLLEXPORT PG_FUNCTION_INFO_V1(intvec_l1_distance);
Datum
intvec_l1_distance(PG_FUNCTION_ARGS)
{
IntVector *a = PG_GETARG_INTVEC_P(0);
IntVector *b = PG_GETARG_INTVEC_P(1);
CheckDims(a, b);
PG_RETURN_FLOAT8((double) IntvecL1Distance(a->dim, a->x, b->x));
}
/*
* Get the dimensions of an int vector
*/
FUNCTION_PREFIX PG_FUNCTION_INFO_V1(intvec_vector_dims);
Datum
intvec_vector_dims(PG_FUNCTION_ARGS)
{
IntVector *a = PG_GETARG_INTVEC_P(0);
PG_RETURN_INT32(a->dim);
}
/*
* Get the L2 norm of an int vector
*/
PGDLLEXPORT PG_FUNCTION_INFO_V1(intvec_l2_norm);
Datum
intvec_l2_norm(PG_FUNCTION_ARGS)
{
IntVector *a = PG_GETARG_INTVEC_P(0);
int8 *ax = a->x;
int norm = 0;
/* Auto-vectorized */
for (int i = 0; i < a->dim; i++)
norm += (int) ax[i] * (int) ax[i];
PG_RETURN_FLOAT8(sqrt((double) norm));
}
/*
* Internal helper to compare int vectors
*/
static int
intvec_cmp_internal(IntVector * a, IntVector * b)
{
int dim = Min(a->dim, b->dim);
/* Check values before dimensions to be consistent with Postgres arrays */
for (int i = 0; i < dim; i++)
{
if (a->x[i] < b->x[i])
return -1;
if (a->x[i] > b->x[i])
return 1;
}
if (a->dim < b->dim)
return -1;
if (a->dim > b->dim)
return 1;
return 0;
}
/*
* Less than
*/
FUNCTION_PREFIX PG_FUNCTION_INFO_V1(intvec_lt);
Datum
intvec_lt(PG_FUNCTION_ARGS)
{
IntVector *a = PG_GETARG_INTVEC_P(0);
IntVector *b = PG_GETARG_INTVEC_P(1);
PG_RETURN_BOOL(intvec_cmp_internal(a, b) < 0);
}
/*
* Less than or equal
*/
FUNCTION_PREFIX PG_FUNCTION_INFO_V1(intvec_le);
Datum
intvec_le(PG_FUNCTION_ARGS)
{
IntVector *a = PG_GETARG_INTVEC_P(0);
IntVector *b = PG_GETARG_INTVEC_P(1);
PG_RETURN_BOOL(intvec_cmp_internal(a, b) <= 0);
}
/*
* Equal
*/
FUNCTION_PREFIX PG_FUNCTION_INFO_V1(intvec_eq);
Datum
intvec_eq(PG_FUNCTION_ARGS)
{
IntVector *a = PG_GETARG_INTVEC_P(0);
IntVector *b = PG_GETARG_INTVEC_P(1);
PG_RETURN_BOOL(intvec_cmp_internal(a, b) == 0);
}
/*
* Not equal
*/
FUNCTION_PREFIX PG_FUNCTION_INFO_V1(intvec_ne);
Datum
intvec_ne(PG_FUNCTION_ARGS)
{
IntVector *a = PG_GETARG_INTVEC_P(0);
IntVector *b = PG_GETARG_INTVEC_P(1);
PG_RETURN_BOOL(intvec_cmp_internal(a, b) != 0);
}
/*
* Greater than or equal
*/
FUNCTION_PREFIX PG_FUNCTION_INFO_V1(intvec_ge);
Datum
intvec_ge(PG_FUNCTION_ARGS)
{
IntVector *a = PG_GETARG_INTVEC_P(0);
IntVector *b = PG_GETARG_INTVEC_P(1);
PG_RETURN_BOOL(intvec_cmp_internal(a, b) >= 0);
}
/*
* Greater than
*/
FUNCTION_PREFIX PG_FUNCTION_INFO_V1(intvec_gt);
Datum
intvec_gt(PG_FUNCTION_ARGS)
{
IntVector *a = PG_GETARG_INTVEC_P(0);
IntVector *b = PG_GETARG_INTVEC_P(1);
PG_RETURN_BOOL(intvec_cmp_internal(a, b) > 0);
}
/*
* Compare int vectors
*/
FUNCTION_PREFIX PG_FUNCTION_INFO_V1(intvec_cmp);
Datum
intvec_cmp(PG_FUNCTION_ARGS)
{
IntVector *a = PG_GETARG_INTVEC_P(0);
IntVector *b = PG_GETARG_INTVEC_P(1);
PG_RETURN_INT32(intvec_cmp_internal(a, b));
}

View File

@@ -1,23 +0,0 @@
#ifndef INTVEC_H
#define INTVEC_H
#include "vector.h"
#define INTVEC_MAX_DIM VECTOR_MAX_DIM
#define INTVEC_SIZE(_dim) (offsetof(IntVector, x) + sizeof(int8)*(_dim))
#define DatumGetIntVector(x) ((IntVector *) PG_DETOAST_DATUM(x))
#define PG_GETARG_INTVEC_P(x) DatumGetIntVector(PG_GETARG_DATUM(x))
#define PG_RETURN_INTVEC_P(x) PG_RETURN_POINTER(x)
typedef struct IntVector
{
int32 vl_len_; /* varlena header (do not touch directly!) */
int16 dim; /* number of dimensions */
int16 unused;
int8 x[FLEXIBLE_ARRAY_MEMBER];
} IntVector;
IntVector *InitIntVector(int dim);
#endif

View File

@@ -26,6 +26,12 @@
#include "pgstat.h"
#endif
#if PG_VERSION_NUM >= 130000
#define CALLBACK_ITEM_POINTER ItemPointer tid
#else
#define CALLBACK_ITEM_POINTER HeapTuple hup
#endif
#if PG_VERSION_NUM >= 140000
#include "utils/backend_status.h"
#include "utils/wait_event.h"
@@ -90,7 +96,7 @@ AddSample(Datum *values, IvfflatBuildState * buildstate)
* Callback for sampling
*/
static void
SampleCallback(Relation index, ItemPointer tid, Datum *values,
SampleCallback(Relation index, CALLBACK_ITEM_POINTER, Datum *values,
bool *isnull, bool tupleIsAlive, void *state)
{
IvfflatBuildState *buildstate = (IvfflatBuildState *) state;
@@ -201,12 +207,16 @@ AddTupleToSort(Relation index, ItemPointer tid, Datum *values, IvfflatBuildState
* Callback for table_index_build_scan
*/
static void
BuildCallback(Relation index, ItemPointer tid, Datum *values,
BuildCallback(Relation index, CALLBACK_ITEM_POINTER, Datum *values,
bool *isnull, bool tupleIsAlive, void *state)
{
IvfflatBuildState *buildstate = (IvfflatBuildState *) state;
MemoryContext oldCtx;
#if PG_VERSION_NUM < 130000
ItemPointer tid = &hup->t_self;
#endif
/* Skip nulls */
if (isnull[0])
return;
@@ -228,11 +238,11 @@ BuildCallback(Relation index, ItemPointer tid, Datum *values,
static inline void
GetNextTuple(Tuplesortstate *sortstate, TupleDesc tupdesc, TupleTableSlot *slot, IndexTuple *itup, int *list)
{
Datum value;
bool isnull;
if (tuplesort_gettupleslot(sortstate, true, false, slot, NULL))
{
Datum value;
bool isnull;
*list = DatumGetInt32(slot_getattr(slot, 1, &isnull));
value = slot_getattr(slot, 3, &isnull);
@@ -254,8 +264,8 @@ InsertTuples(Relation index, IvfflatBuildState * buildstate, ForkNumber forkNum)
IndexTuple itup = NULL; /* silence compiler warning */
int64 inserted = 0;
TupleTableSlot *slot = MakeSingleTupleTableSlot(buildstate->sortdesc, &TTSOpsMinimalTuple);
TupleDesc tupdesc = buildstate->tupdesc;
TupleTableSlot *slot = MakeSingleTupleTableSlot(buildstate->tupdesc, &TTSOpsMinimalTuple);
TupleDesc tupdesc = RelationGetDescr(index);
pgstat_progress_update_param(PROGRESS_CREATEIDX_SUBPHASE, PROGRESS_IVFFLAT_PHASE_LOAD);
@@ -319,27 +329,20 @@ InitBuildState(IvfflatBuildState * buildstate, Relation heap, Relation index, In
buildstate->index = index;
buildstate->indexInfo = indexInfo;
buildstate->typeInfo = IvfflatGetTypeInfo(index);
buildstate->tupdesc = RelationGetDescr(index);
buildstate->lists = IvfflatGetLists(index);
buildstate->dimensions = TupleDescAttr(index->rd_att, 0)->atttypmod;
/* Disallow varbit since require fixed dimensions */
if (TupleDescAttr(index->rd_att, 0)->atttypid == VARBITOID)
ereport(ERROR,
(errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
errmsg("type not supported for ivfflat index")));
elog(ERROR, "type not supported for ivfflat index");
/* Require column to have dimensions to be indexed */
if (buildstate->dimensions < 0)
ereport(ERROR,
(errcode(ERRCODE_INVALID_PARAMETER_VALUE),
errmsg("column does not have dimensions")));
elog(ERROR, "column does not have dimensions");
if (buildstate->dimensions > buildstate->typeInfo->maxDimensions)
ereport(ERROR,
(errcode(ERRCODE_PROGRAM_LIMIT_EXCEEDED),
errmsg("column cannot have more than %d dimensions for ivfflat index", buildstate->typeInfo->maxDimensions)));
elog(ERROR, "column cannot have more than %d dimensions for ivfflat index", buildstate->typeInfo->maxDimensions);
buildstate->reltuples = 0;
buildstate->indtuples = 0;
@@ -352,17 +355,15 @@ InitBuildState(IvfflatBuildState * buildstate, Relation heap, Relation index, In
/* Require more than one dimension for spherical k-means */
if (buildstate->kmeansnormprocinfo != NULL && buildstate->dimensions == 1)
ereport(ERROR,
(errcode(ERRCODE_INVALID_PARAMETER_VALUE),
errmsg("dimensions must be greater than one for this opclass")));
elog(ERROR, "dimensions must be greater than one for this opclass");
/* Create tuple description for sorting */
buildstate->sortdesc = CreateTemplateTupleDesc(3);
TupleDescInitEntry(buildstate->sortdesc, (AttrNumber) 1, "list", INT4OID, -1, 0);
TupleDescInitEntry(buildstate->sortdesc, (AttrNumber) 2, "tid", TIDOID, -1, 0);
TupleDescInitEntry(buildstate->sortdesc, (AttrNumber) 3, "vector", buildstate->tupdesc->attrs[0].atttypid, -1, 0);
buildstate->tupdesc = CreateTemplateTupleDesc(3);
TupleDescInitEntry(buildstate->tupdesc, (AttrNumber) 1, "list", INT4OID, -1, 0);
TupleDescInitEntry(buildstate->tupdesc, (AttrNumber) 2, "tid", TIDOID, -1, 0);
TupleDescInitEntry(buildstate->tupdesc, (AttrNumber) 3, "vector", RelationGetDescr(index)->attrs[0].atttypid, -1, 0);
buildstate->slot = MakeSingleTupleTableSlot(buildstate->sortdesc, &TTSOpsVirtual);
buildstate->slot = MakeSingleTupleTableSlot(buildstate->tupdesc, &TTSOpsVirtual);
buildstate->centers = VectorArrayInit(buildstate->lists, buildstate->dimensions, buildstate->typeInfo->itemSize(buildstate->dimensions));
buildstate->listInfo = palloc(sizeof(ListInfo) * buildstate->lists);
@@ -561,20 +562,6 @@ PrintKmeansMetrics(IvfflatBuildState * buildstate)
}
#endif
/*
* Initialize build sort state
*/
static Tuplesortstate *
InitBuildSortState(TupleDesc tupdesc, int memory, SortCoordinate coordinate)
{
AttrNumber attNums[] = {1};
Oid sortOperators[] = {Int4LessOperator};
Oid sortCollations[] = {InvalidOid};
bool nullsFirstFlags[] = {false};
return tuplesort_begin_heap(tupdesc, 1, attNums, sortOperators, sortCollations, nullsFirstFlags, memory, coordinate, false);
}
/*
* Within leader, wait for end of heap scan
*/
@@ -622,6 +609,12 @@ IvfflatParallelScanAndSort(IvfflatSpool * ivfspool, IvfflatShared * ivfshared, S
double reltuples;
IndexInfo *indexInfo;
/* Sort options, which must match AssignTuples */
AttrNumber attNums[] = {1};
Oid sortOperators[] = {Int4LessOperator};
Oid sortCollations[] = {InvalidOid};
bool nullsFirstFlags[] = {false};
/* Initialize local tuplesort coordination state */
coordinate = palloc0(sizeof(SortCoordinateData));
coordinate->isWorker = true;
@@ -634,7 +627,7 @@ IvfflatParallelScanAndSort(IvfflatSpool * ivfspool, IvfflatShared * ivfshared, S
InitBuildState(&buildstate, ivfspool->heap, ivfspool->index, indexInfo);
memcpy(buildstate.centers->items, ivfcenters, buildstate.centers->itemsize * buildstate.centers->maxlen);
buildstate.centers->length = buildstate.centers->maxlen;
ivfspool->sortstate = InitBuildSortState(buildstate.sortdesc, sortmem, coordinate);
ivfspool->sortstate = tuplesort_begin_heap(buildstate.tupdesc, 1, attNums, sortOperators, sortCollations, nullsFirstFlags, sortmem, coordinate, false);
buildstate.sortstate = ivfspool->sortstate;
scan = table_beginscan_parallel(ivfspool->heap,
ParallelTableScanFromIvfflatShared(ivfshared));
@@ -931,6 +924,12 @@ AssignTuples(IvfflatBuildState * buildstate)
int parallel_workers = 0;
SortCoordinate coordinate = NULL;
/* Sort options, which must match IvfflatParallelScanAndSort */
AttrNumber attNums[] = {1};
Oid sortOperators[] = {Int4LessOperator};
Oid sortCollations[] = {InvalidOid};
bool nullsFirstFlags[] = {false};
pgstat_progress_update_param(PROGRESS_CREATEIDX_SUBPHASE, PROGRESS_IVFFLAT_PHASE_ASSIGN);
/* Calculate parallel workers */
@@ -951,7 +950,7 @@ AssignTuples(IvfflatBuildState * buildstate)
}
/* Begin serial/leader tuplesort */
buildstate->sortstate = InitBuildSortState(buildstate->sortdesc, maintenance_work_mem, coordinate);
buildstate->sortstate = tuplesort_begin_heap(buildstate->tupdesc, 1, attNums, sortOperators, sortCollations, nullsFirstFlags, maintenance_work_mem, coordinate, false);
/* Add tuples to sort */
if (buildstate->heap != NULL)
@@ -1007,10 +1006,6 @@ BuildIndex(Relation heap, Relation index, IndexInfo *indexInfo,
CreateListPages(index, buildstate->centers, buildstate->dimensions, buildstate->lists, forkNum, &buildstate->listInfo);
CreateEntryPages(buildstate, forkNum);
/* Write WAL for initialization fork since GenericXLog functions do not */
if (forkNum == INIT_FORKNUM)
log_newpage_range(index, forkNum, 0, RelationGetNumberOfBlocksInFork(index, forkNum), true);
FreeBuildState(buildstate);
}

View File

@@ -7,7 +7,6 @@
#include "commands/progress.h"
#include "commands/vacuum.h"
#include "ivfflat.h"
#include "utils/float.h"
#include "utils/guc.h"
#include "utils/selfuncs.h"
#include "utils/spccache.h"
@@ -17,16 +16,8 @@
#endif
int ivfflat_probes;
int ivfflat_iterative_search;
int ivfflat_max_probes;
static relopt_kind ivfflat_relopt_kind;
static const struct config_enum_entry ivfflat_iterative_search_options[] = {
{"off", IVFFLAT_ITERATIVE_SEARCH_OFF, false},
{"relaxed_order", IVFFLAT_ITERATIVE_SEARCH_RELAXED, false},
{NULL, 0, false}
};
/*
* Initialize index options and variables
*/
@@ -35,21 +26,16 @@ IvfflatInit(void)
{
ivfflat_relopt_kind = add_reloption_kind();
add_int_reloption(ivfflat_relopt_kind, "lists", "Number of inverted lists",
IVFFLAT_DEFAULT_LISTS, IVFFLAT_MIN_LISTS, IVFFLAT_MAX_LISTS, AccessExclusiveLock);
IVFFLAT_DEFAULT_LISTS, IVFFLAT_MIN_LISTS, IVFFLAT_MAX_LISTS
#if PG_VERSION_NUM >= 130000
,AccessExclusiveLock
#endif
);
DefineCustomIntVariable("ivfflat.probes", "Sets the number of probes",
"Valid range is 1..lists.", &ivfflat_probes,
IVFFLAT_DEFAULT_PROBES, IVFFLAT_MIN_LISTS, IVFFLAT_MAX_LISTS, PGC_USERSET, 0, NULL, NULL, NULL);
DefineCustomEnumVariable("ivfflat.iterative_search", "Sets the iterative search mode",
NULL, &ivfflat_iterative_search,
IVFFLAT_ITERATIVE_SEARCH_OFF, ivfflat_iterative_search_options, PGC_USERSET, 0, NULL, NULL, NULL);
/* If this is less than probes, probes is used */
DefineCustomIntVariable("ivfflat.max_probes", "Sets the max number of probes for iterative search",
"-1 means no limit", &ivfflat_max_probes,
-1, -1, IVFFLAT_MAX_LISTS, PGC_USERSET, 0, NULL, NULL, NULL);
MarkGUCPrefixReserved("ivfflat");
}
@@ -86,16 +72,14 @@ ivfflatcostestimate(PlannerInfo *root, IndexPath *path, double loop_count,
GenericCosts costs;
int lists;
double ratio;
double sequentialRatio = 0.5;
double startupPages;
double spc_seq_page_cost;
Relation index;
/* Never use index without order */
if (path->indexorderbys == NULL)
{
*indexStartupCost = get_float8_infinity();
*indexTotalCost = get_float8_infinity();
*indexStartupCost = DBL_MAX;
*indexTotalCost = DBL_MAX;
*indexSelectivity = 0;
*indexCorrelation = 0;
*indexPages = 0;
@@ -104,8 +88,6 @@ ivfflatcostestimate(PlannerInfo *root, IndexPath *path, double loop_count,
MemSet(&costs, 0, sizeof(costs));
genericcostestimate(root, path, loop_count, &costs);
index = index_open(path->indexinfo->indexoid, NoLock);
IvfflatGetMetaPageInfo(index, &lists, NULL);
index_close(index, NoLock);
@@ -115,26 +97,41 @@ ivfflatcostestimate(PlannerInfo *root, IndexPath *path, double loop_count,
if (ratio > 1.0)
ratio = 1.0;
/*
* This gives us the subset of tuples to visit. This value is passed into
* the generic cost estimator to determine the number of pages to visit
* during the index scan.
*/
costs.numIndexTuples = path->indexinfo->tuples * ratio;
genericcostestimate(root, path, loop_count, &costs);
get_tablespace_page_costs(path->indexinfo->reltablespace, NULL, &spc_seq_page_cost);
/* Change some page cost from random to sequential */
costs.indexTotalCost -= sequentialRatio * costs.numIndexPages * (costs.spc_random_page_cost - spc_seq_page_cost);
/* Startup cost is cost before returning the first row */
costs.indexStartupCost = costs.indexTotalCost * ratio;
/* Adjust cost if needed since TOAST not included in seq scan cost */
startupPages = costs.numIndexPages * ratio;
if (startupPages > path->indexinfo->rel->pages && ratio < 0.5)
if (costs.numIndexPages > path->indexinfo->rel->pages && ratio < 0.5)
{
/* Change rest of page cost from random to sequential */
costs.indexStartupCost -= (1 - sequentialRatio) * startupPages * (costs.spc_random_page_cost - spc_seq_page_cost);
/* Change all page cost from random to sequential */
costs.indexTotalCost -= costs.numIndexPages * (costs.spc_random_page_cost - spc_seq_page_cost);
/* Remove cost of extra pages */
costs.indexStartupCost -= (startupPages - path->indexinfo->rel->pages) * spc_seq_page_cost;
costs.indexTotalCost -= (costs.numIndexPages - path->indexinfo->rel->pages) * spc_seq_page_cost;
}
else
{
/* Change some page cost from random to sequential */
costs.indexTotalCost -= 0.5 * costs.numIndexPages * (costs.spc_random_page_cost - spc_seq_page_cost);
}
*indexStartupCost = costs.indexStartupCost;
/*
* If the list selectivity is lower than what is returned from the generic
* cost estimator, use that.
*/
if (ratio < costs.indexSelectivity)
costs.indexSelectivity = ratio;
/* Use total cost since most work happens before first tuple is returned */
*indexStartupCost = costs.indexTotalCost;
*indexTotalCost = costs.indexTotalCost;
*indexSelectivity = costs.indexSelectivity;
*indexCorrelation = costs.indexCorrelation;
@@ -151,10 +148,23 @@ ivfflatoptions(Datum reloptions, bool validate)
{"lists", RELOPT_TYPE_INT, offsetof(IvfflatOptions, lists)},
};
#if PG_VERSION_NUM >= 130000
return (bytea *) build_reloptions(reloptions, validate,
ivfflat_relopt_kind,
sizeof(IvfflatOptions),
tab, lengthof(tab));
#else
relopt_value *options;
int numoptions;
IvfflatOptions *rdopts;
options = parseRelOptions(reloptions, validate, ivfflat_relopt_kind, &numoptions);
rdopts = allocateReloptStruct(sizeof(IvfflatOptions), options, numoptions);
fillRelOptions((void *) rdopts, sizeof(IvfflatOptions), options, numoptions,
validate, tab, lengthof(tab));
return (bytea *) rdopts;
#endif
}
/*
@@ -171,7 +181,7 @@ ivfflatvalidate(Oid opclassoid)
*
* See https://www.postgresql.org/docs/current/index-api.html
*/
FUNCTION_PREFIX PG_FUNCTION_INFO_V1(ivfflathandler);
PGDLLEXPORT PG_FUNCTION_INFO_V1(ivfflathandler);
Datum
ivfflathandler(PG_FUNCTION_ARGS)
{
@@ -179,7 +189,9 @@ ivfflathandler(PG_FUNCTION_ARGS)
amroutine->amstrategies = 0;
amroutine->amsupport = 5;
#if PG_VERSION_NUM >= 130000
amroutine->amoptsprocnum = 0;
#endif
amroutine->amcanorder = false;
amroutine->amcanorderbyop = true;
amroutine->amcanbackward = false; /* can change direction mid-scan */
@@ -192,24 +204,17 @@ ivfflathandler(PG_FUNCTION_ARGS)
amroutine->amclusterable = false;
amroutine->ampredlocks = false;
amroutine->amcanparallel = false;
#if PG_VERSION_NUM >= 170000
amroutine->amcanbuildparallel = true;
#endif
amroutine->amcaninclude = false;
#if PG_VERSION_NUM >= 130000
amroutine->amusemaintenanceworkmem = false; /* not used during VACUUM */
#if PG_VERSION_NUM >= 160000
amroutine->amsummarizing = false;
#endif
amroutine->amparallelvacuumoptions = VACUUM_OPTION_PARALLEL_BULKDEL;
#endif
amroutine->amkeytype = InvalidOid;
/* Interface functions */
amroutine->ambuild = ivfflatbuild;
amroutine->ambuildempty = ivfflatbuildempty;
amroutine->aminsert = ivfflatinsert;
#if PG_VERSION_NUM >= 170000
amroutine->aminsertcleanup = NULL;
#endif
amroutine->ambulkdelete = ivfflatbulkdelete;
amroutine->amvacuumcleanup = ivfflatvacuumcleanup;
amroutine->amcanreturn = NULL; /* tuple not included in heapsort */

View File

@@ -80,14 +80,6 @@
/* Variables */
extern int ivfflat_probes;
extern int ivfflat_iterative_search;
extern int ivfflat_max_probes;
typedef enum IvfflatIterativeSearchMode
{
IVFFLAT_ITERATIVE_SEARCH_OFF,
IVFFLAT_ITERATIVE_SEARCH_RELAXED
} IvfflatIterativeSearchMode;
typedef struct VectorArrayData
{
@@ -173,7 +165,6 @@ typedef struct IvfflatBuildState
Relation index;
IndexInfo *indexInfo;
const IvfflatTypeInfo *typeInfo;
TupleDesc tupdesc;
/* Settings */
int dimensions;
@@ -207,7 +198,7 @@ typedef struct IvfflatBuildState
/* Sorting */
Tuplesortstate *sortstate;
TupleDesc sortdesc;
TupleDesc tupdesc;
TupleTableSlot *slot;
/* Memory */
@@ -256,18 +247,14 @@ typedef struct IvfflatScanOpaqueData
{
const IvfflatTypeInfo *typeInfo;
int probes;
int maxProbes;
int dimensions;
bool first;
Datum value;
MemoryContext tmpCtx;
/* Sorting */
Tuplesortstate *sortstate;
TupleDesc tupdesc;
TupleTableSlot *vslot;
TupleTableSlot *mslot;
BufferAccessStrategy bas;
TupleTableSlot *slot;
bool isnull;
/* Support functions */
FmgrInfo *procinfo;
@@ -277,9 +264,7 @@ typedef struct IvfflatScanOpaqueData
/* Lists */
pairingheap *listQueue;
BlockNumber *listPages;
int listIndex;
IvfflatScanList *lists;
IvfflatScanList lists[FLEXIBLE_ARRAY_MEMBER]; /* must come last */
} IvfflatScanOpaqueData;
typedef IvfflatScanOpaqueData * IvfflatScanOpaque;

View File

@@ -94,11 +94,8 @@ InsertTuple(Relation index, Datum *values, bool *isnull, ItemPointer heap_tid, R
value = IvfflatNormValue(typeInfo, collation, value);
}
/* Ensure index is valid */
IvfflatGetMetaPageInfo(index, NULL, NULL);
/* Find the insert page - sets the page and list info */
FindInsertPage(index, &value, &insertPage, &listInfo);
FindInsertPage(index, values, &insertPage, &listInfo);
Assert(BlockNumberIsValid(insertPage));
originalInsertPage = insertPage;

View File

@@ -151,8 +151,12 @@ RandomCenters(Relation index, VectorArray centers, const IvfflatTypeInfo * typeI
static void
ShowMemoryUsage(MemoryContext context, Size estimatedSize)
{
#if PG_VERSION_NUM >= 130000
elog(INFO, "total memory: %zu MB",
MemoryContextMemAllocated(context, true) / (1024 * 1024));
#else
MemoryContextStats(context);
#endif
elog(INFO, "estimated memory: %zu MB", estimatedSize / (1024 * 1024));
}
#endif
@@ -323,7 +327,7 @@ ElkanKmeans(Relation index, VectorArray samples, VectorArray centers, const Ivff
newCenters->length = numCenters;
#ifdef IVFFLAT_MEMORY
ShowMemoryUsage(MemoryContextGetParent(CurrentMemoryContext), totalSize);
ShowMemoryUsage(MemoryContextGetParent(CurrentMemoryContext));
#endif
/* Pick initial centers */

View File

@@ -10,10 +10,6 @@
#include "miscadmin.h"
#include "pgstat.h"
#include "storage/bufmgr.h"
#include "utils/memutils.h"
#define GetScanList(ptr) pairingheap_container(IvfflatScanList, ph_node, ptr)
#define GetScanListConst(ptr) pairingheap_const_container(IvfflatScanList, ph_node, ptr)
/*
* Compare list distances
@@ -21,10 +17,10 @@
static int
CompareLists(const pairingheap_node *a, const pairingheap_node *b, void *arg)
{
if (GetScanListConst(a)->distance > GetScanListConst(b)->distance)
if (((const IvfflatScanList *) a)->distance > ((const IvfflatScanList *) b)->distance)
return 1;
if (GetScanListConst(a)->distance < GetScanListConst(b)->distance)
if (((const IvfflatScanList *) a)->distance < ((const IvfflatScanList *) b)->distance)
return -1;
return 0;
@@ -62,7 +58,7 @@ GetScanLists(IndexScanDesc scan, Datum value)
/* Use procinfo from the index instead of scan key for performance */
distance = DatumGetFloat8(so->distfunc(so->procinfo, so->collation, PointerGetDatum(&list->center), value));
if (listCount < so->maxProbes)
if (listCount < so->probes)
{
IvfflatScanList *scanlist;
@@ -75,15 +71,15 @@ GetScanLists(IndexScanDesc scan, Datum value)
pairingheap_add(so->listQueue, &scanlist->ph_node);
/* Calculate max distance */
if (listCount == so->maxProbes)
maxDistance = GetScanList(pairingheap_first(so->listQueue))->distance;
if (listCount == so->probes)
maxDistance = ((IvfflatScanList *) pairingheap_first(so->listQueue))->distance;
}
else if (distance < maxDistance)
{
IvfflatScanList *scanlist;
/* Remove */
scanlist = GetScanList(pairingheap_remove_first(so->listQueue));
scanlist = (IvfflatScanList *) pairingheap_remove_first(so->listQueue);
/* Reuse */
scanlist->startPage = list->startPage;
@@ -91,7 +87,7 @@ GetScanLists(IndexScanDesc scan, Datum value)
pairingheap_add(so->listQueue, &scanlist->ph_node);
/* Update max distance */
maxDistance = GetScanList(pairingheap_first(so->listQueue))->distance;
maxDistance = ((IvfflatScanList *) pairingheap_first(so->listQueue))->distance;
}
}
@@ -99,11 +95,6 @@ GetScanLists(IndexScanDesc scan, Datum value)
UnlockReleaseBuffer(cbuf);
}
for (int i = listCount - 1; i >= 0; i--)
so->listPages[i] = GetScanList(pairingheap_remove_first(so->listQueue))->startPage;
Assert(pairingheap_is_empty(so->listQueue));
}
/*
@@ -115,15 +106,19 @@ GetScanItems(IndexScanDesc scan, Datum value)
IvfflatScanOpaque so = (IvfflatScanOpaque) scan->opaque;
TupleDesc tupdesc = RelationGetDescr(scan->indexRelation);
double tuples = 0;
TupleTableSlot *slot = so->vslot;
int batchProbes = 0;
TupleTableSlot *slot = MakeSingleTupleTableSlot(so->tupdesc, &TTSOpsVirtual);
tuplesort_reset(so->sortstate);
/*
* Reuse same set of shared buffers for scan
*
* See postgres/src/backend/storage/buffer/README for description
*/
BufferAccessStrategy bas = GetAccessStrategy(BAS_BULKREAD);
/* Search closest probes lists */
while (so->listIndex < so->maxProbes && (++batchProbes) <= so->probes)
while (!pairingheap_is_empty(so->listQueue))
{
BlockNumber searchPage = so->listPages[so->listIndex++];
BlockNumber searchPage = ((IvfflatScanList *) pairingheap_remove_first(so->listQueue))->startPage;
/* Search all entry pages for list */
while (BlockNumberIsValid(searchPage))
@@ -132,7 +127,7 @@ GetScanItems(IndexScanDesc scan, Datum value)
Page page;
OffsetNumber maxoffno;
buf = ReadBufferExtended(scan->indexRelation, MAIN_FORKNUM, searchPage, RBM_NORMAL, so->bas);
buf = ReadBufferExtended(scan->indexRelation, MAIN_FORKNUM, searchPage, RBM_NORMAL, bas);
LockBuffer(buf, BUFFER_LOCK_SHARE);
page = BufferGetPage(buf);
maxoffno = PageGetMaxOffsetNumber(page);
@@ -171,17 +166,15 @@ GetScanItems(IndexScanDesc scan, Datum value)
}
}
if (tuples < 100 && ivfflat_iterative_search == IVFFLAT_ITERATIVE_SEARCH_OFF)
FreeAccessStrategy(bas);
if (tuples < 100)
ereport(DEBUG1,
(errmsg("index scan found few tuples"),
errdetail("Index may have been created with little data."),
errhint("Recreate the index and possibly decrease lists.")));
tuplesort_performsort(so->sortstate);
#if defined(IVFFLAT_MEMORY)
elog(INFO, "memory: %zu MB", MemoryContextMemAllocated(CurrentMemoryContext, true) / (1024 * 1024));
#endif
}
/*
@@ -218,32 +211,12 @@ GetScanValue(IndexScanDesc scan)
/* Normalize if needed */
if (so->normprocinfo != NULL)
{
MemoryContext oldCtx = MemoryContextSwitchTo(so->tmpCtx);
value = IvfflatNormValue(so->typeInfo, so->collation, value);
MemoryContextSwitchTo(oldCtx);
}
}
return value;
}
/*
* Initialize scan sort state
*/
static Tuplesortstate *
InitScanSortState(TupleDesc tupdesc)
{
AttrNumber attNums[] = {1};
Oid sortOperators[] = {Float8LessOperator};
Oid sortCollations[] = {InvalidOid};
bool nullsFirstFlags[] = {false};
return tuplesort_begin_heap(tupdesc, 1, attNums, sortOperators, sortCollations, nullsFirstFlags, work_mem, NULL, false);
}
/*
* Prepare for an index scan
*/
@@ -254,41 +227,24 @@ ivfflatbeginscan(Relation index, int nkeys, int norderbys)
IvfflatScanOpaque so;
int lists;
int dimensions;
AttrNumber attNums[] = {1};
Oid sortOperators[] = {Float8LessOperator};
Oid sortCollations[] = {InvalidOid};
bool nullsFirstFlags[] = {false};
int probes = ivfflat_probes;
int maxProbes;
MemoryContext oldCtx;
scan = RelationGetIndexScan(index, nkeys, norderbys);
/* Get lists and dimensions from metapage */
IvfflatGetMetaPageInfo(index, &lists, &dimensions);
if (ivfflat_iterative_search != IVFFLAT_ITERATIVE_SEARCH_OFF)
{
maxProbes = ivfflat_max_probes;
if (maxProbes < 0)
maxProbes = lists;
else if (maxProbes < probes)
{
/* TODO Show notice */
maxProbes = probes;
}
}
else
maxProbes = probes;
if (probes > lists)
probes = lists;
if (maxProbes > lists)
maxProbes = lists;
so = (IvfflatScanOpaque) palloc(sizeof(IvfflatScanOpaqueData));
so = (IvfflatScanOpaque) palloc(offsetof(IvfflatScanOpaqueData, lists) + probes * sizeof(IvfflatScanList));
so->typeInfo = IvfflatGetTypeInfo(index);
so->first = true;
so->probes = probes;
so->maxProbes = maxProbes;
so->dimensions = dimensions;
/* Set support functions */
@@ -296,37 +252,17 @@ ivfflatbeginscan(Relation index, int nkeys, int norderbys)
so->normprocinfo = IvfflatOptionalProcInfo(index, IVFFLAT_NORM_PROC);
so->collation = index->rd_indcollation[0];
so->tmpCtx = AllocSetContextCreate(CurrentMemoryContext,
"Ivfflat scan temporary context",
ALLOCSET_DEFAULT_SIZES);
oldCtx = MemoryContextSwitchTo(so->tmpCtx);
/* Create tuple description for sorting */
so->tupdesc = CreateTemplateTupleDesc(2);
TupleDescInitEntry(so->tupdesc, (AttrNumber) 1, "distance", FLOAT8OID, -1, 0);
TupleDescInitEntry(so->tupdesc, (AttrNumber) 2, "heaptid", TIDOID, -1, 0);
/* Prep sort */
so->sortstate = InitScanSortState(so->tupdesc);
so->sortstate = tuplesort_begin_heap(so->tupdesc, 1, attNums, sortOperators, sortCollations, nullsFirstFlags, work_mem, NULL, false);
/* Need separate slots for puttuple and gettuple */
so->vslot = MakeSingleTupleTableSlot(so->tupdesc, &TTSOpsVirtual);
so->mslot = MakeSingleTupleTableSlot(so->tupdesc, &TTSOpsMinimalTuple);
/*
* Reuse same set of shared buffers for scan
*
* See postgres/src/backend/storage/buffer/README for description
*/
so->bas = GetAccessStrategy(BAS_BULKREAD);
so->slot = MakeSingleTupleTableSlot(so->tupdesc, &TTSOpsMinimalTuple);
so->listQueue = pairingheap_allocate(CompareLists, scan);
so->listPages = palloc(maxProbes * sizeof(BlockNumber));
so->listIndex = 0;
so->lists = palloc(maxProbes * sizeof(IvfflatScanList));
MemoryContextSwitchTo(oldCtx);
scan->opaque = so;
@@ -341,9 +277,13 @@ ivfflatrescan(IndexScanDesc scan, ScanKey keys, int nkeys, ScanKey orderbys, int
{
IvfflatScanOpaque so = (IvfflatScanOpaque) scan->opaque;
#if PG_VERSION_NUM >= 130000
if (!so->first)
tuplesort_reset(so->sortstate);
#endif
so->first = true;
pairingheap_reset(so->listQueue);
so->listIndex = 0;
if (keys && scan->numberOfKeys > 0)
memmove(scan->keyData, keys, scan->numberOfKeys * sizeof(ScanKeyData));
@@ -359,8 +299,6 @@ bool
ivfflatgettuple(IndexScanDesc scan, ScanDirection dir)
{
IvfflatScanOpaque so = (IvfflatScanOpaque) scan->opaque;
ItemPointer heaptid;
bool isnull;
/*
* Index can be used to scan backward, but Postgres doesn't support
@@ -388,23 +326,23 @@ ivfflatgettuple(IndexScanDesc scan, ScanDirection dir)
IvfflatBench("GetScanLists", GetScanLists(scan, value));
IvfflatBench("GetScanItems", GetScanItems(scan, value));
so->first = false;
so->value = value;
/* Clean up if we allocated a new value */
if (value != scan->orderByData->sk_argument)
pfree(DatumGetPointer(value));
}
while (!tuplesort_gettupleslot(so->sortstate, true, false, so->mslot, NULL))
if (tuplesort_gettupleslot(so->sortstate, true, false, so->slot, NULL))
{
if (so->listIndex == so->maxProbes)
return false;
ItemPointer heaptid = (ItemPointer) DatumGetPointer(slot_getattr(so->slot, 2, &so->isnull));
IvfflatBench("GetScanItems", GetScanItems(scan, so->value));
scan->xs_heaptid = *heaptid;
scan->xs_recheck = false;
scan->xs_recheckorderby = false;
return true;
}
heaptid = (ItemPointer) DatumGetPointer(slot_getattr(so->mslot, 2, &isnull));
scan->xs_heaptid = *heaptid;
scan->xs_recheck = false;
scan->xs_recheckorderby = false;
return true;
return false;
}
/*
@@ -415,11 +353,9 @@ ivfflatendscan(IndexScanDesc scan)
{
IvfflatScanOpaque so = (IvfflatScanOpaque) scan->opaque;
/* Free any temporary files */
pairingheap_free(so->listQueue);
tuplesort_end(so->sortstate);
MemoryContextDelete(so->tmpCtx);
pfree(so);
scan->opaque = NULL;
}

View File

@@ -170,11 +170,7 @@ IvfflatGetMetaPageInfo(Relation index, int *lists, int *dimensions)
page = BufferGetPage(buf);
metap = IvfflatPageGetMeta(page);
if (unlikely(metap->magicNumber != IVFFLAT_MAGIC_NUMBER))
elog(ERROR, "ivfflat index is not valid");
if (lists != NULL)
*lists = metap->lists;
*lists = metap->lists;
if (dimensions != NULL)
*dimensions = metap->dimensions;
@@ -342,7 +338,7 @@ IvfflatGetTypeInfo(Relation index)
return (const IvfflatTypeInfo *) DatumGetPointer(FunctionCall0Coll(procinfo, InvalidOid));
}
FUNCTION_PREFIX PG_FUNCTION_INFO_V1(ivfflat_halfvec_support);
PGDLLEXPORT PG_FUNCTION_INFO_V1(ivfflat_halfvec_support);
Datum
ivfflat_halfvec_support(PG_FUNCTION_ARGS)
{
@@ -357,7 +353,7 @@ ivfflat_halfvec_support(PG_FUNCTION_ARGS)
PG_RETURN_POINTER(&typeInfo);
};
FUNCTION_PREFIX PG_FUNCTION_INFO_V1(ivfflat_bit_support);
PGDLLEXPORT PG_FUNCTION_INFO_V1(ivfflat_bit_support);
Datum
ivfflat_bit_support(PG_FUNCTION_ARGS)
{

View File

@@ -26,7 +26,7 @@ ivfflatbulkdelete(IndexVacuumInfo *info, IndexBulkDeleteResult *stats,
Page cpage;
OffsetNumber coffno;
OffsetNumber cmaxoffno;
BlockNumber listPages[MaxOffsetNumber];
BlockNumber startPages[MaxOffsetNumber];
ListInfo listInfo;
cbuf = ReadBuffer(index, blkno);
@@ -40,7 +40,7 @@ ivfflatbulkdelete(IndexVacuumInfo *info, IndexBulkDeleteResult *stats,
{
IvfflatList list = (IvfflatList) PageGetItem(cpage, PageGetItemId(cpage, coffno));
listPages[coffno - FirstOffsetNumber] = list->startPage;
startPages[coffno - FirstOffsetNumber] = list->startPage;
}
listInfo.blkno = blkno;
@@ -50,7 +50,7 @@ ivfflatbulkdelete(IndexVacuumInfo *info, IndexBulkDeleteResult *stats,
for (coffno = FirstOffsetNumber; coffno <= cmaxoffno; coffno = OffsetNumberNext(coffno))
{
BlockNumber searchPage = listPages[coffno - FirstOffsetNumber];
BlockNumber searchPage = startPages[coffno - FirstOffsetNumber];
BlockNumber insertPage = InvalidBlockNumber;
/* Iterate over entry pages */

View File

@@ -3,7 +3,6 @@
#include <limits.h>
#include <math.h>
#include "catalog/pg_type.h"
#include "common/string.h"
#include "fmgr.h"
#include "halfutils.h"
@@ -12,7 +11,6 @@
#include "sparsevec.h"
#include "utils/array.h"
#include "utils/builtins.h"
#include "utils/lsyscache.h"
#include "vector.h"
#if PG_VERSION_NUM >= 120000
@@ -190,7 +188,7 @@ CompareIndices(const void *a, const void *b)
/*
* Convert textual representation to internal representation
*/
FUNCTION_PREFIX PG_FUNCTION_INFO_V1(sparsevec_in);
PGDLLEXPORT PG_FUNCTION_INFO_V1(sparsevec_in);
Datum
sparsevec_in(PG_FUNCTION_ARGS)
{
@@ -411,7 +409,7 @@ sparsevec_in(PG_FUNCTION_ARGS)
/*
* Convert internal representation to textual representation
*/
FUNCTION_PREFIX PG_FUNCTION_INFO_V1(sparsevec_out);
PGDLLEXPORT PG_FUNCTION_INFO_V1(sparsevec_out);
Datum
sparsevec_out(PG_FUNCTION_ARGS)
{
@@ -464,7 +462,7 @@ sparsevec_out(PG_FUNCTION_ARGS)
/*
* Convert type modifier
*/
FUNCTION_PREFIX PG_FUNCTION_INFO_V1(sparsevec_typmod_in);
PGDLLEXPORT PG_FUNCTION_INFO_V1(sparsevec_typmod_in);
Datum
sparsevec_typmod_in(PG_FUNCTION_ARGS)
{
@@ -495,7 +493,7 @@ sparsevec_typmod_in(PG_FUNCTION_ARGS)
/*
* Convert external binary representation to internal representation
*/
FUNCTION_PREFIX PG_FUNCTION_INFO_V1(sparsevec_recv);
PGDLLEXPORT PG_FUNCTION_INFO_V1(sparsevec_recv);
Datum
sparsevec_recv(PG_FUNCTION_ARGS)
{
@@ -547,7 +545,7 @@ sparsevec_recv(PG_FUNCTION_ARGS)
/*
* Convert internal representation to the external binary representation
*/
FUNCTION_PREFIX PG_FUNCTION_INFO_V1(sparsevec_send);
PGDLLEXPORT PG_FUNCTION_INFO_V1(sparsevec_send);
Datum
sparsevec_send(PG_FUNCTION_ARGS)
{
@@ -574,7 +572,7 @@ sparsevec_send(PG_FUNCTION_ARGS)
* Convert sparse vector to sparse vector
* This is needed to check the type modifier
*/
FUNCTION_PREFIX PG_FUNCTION_INFO_V1(sparsevec);
PGDLLEXPORT PG_FUNCTION_INFO_V1(sparsevec);
Datum
sparsevec(PG_FUNCTION_ARGS)
{
@@ -589,7 +587,7 @@ sparsevec(PG_FUNCTION_ARGS)
/*
* Convert dense vector to sparse vector
*/
FUNCTION_PREFIX PG_FUNCTION_INFO_V1(vector_to_sparsevec);
PGDLLEXPORT PG_FUNCTION_INFO_V1(vector_to_sparsevec);
Datum
vector_to_sparsevec(PG_FUNCTION_ARGS)
{
@@ -632,7 +630,7 @@ vector_to_sparsevec(PG_FUNCTION_ARGS)
/*
* Convert half vector to sparse vector
*/
FUNCTION_PREFIX PG_FUNCTION_INFO_V1(halfvec_to_sparsevec);
PGDLLEXPORT PG_FUNCTION_INFO_V1(halfvec_to_sparsevec);
Datum
halfvec_to_sparsevec(PG_FUNCTION_ARGS)
{
@@ -672,137 +670,6 @@ halfvec_to_sparsevec(PG_FUNCTION_ARGS)
PG_RETURN_POINTER(result);
}
/*
* Convert array to sparse vector
*/
FUNCTION_PREFIX PG_FUNCTION_INFO_V1(array_to_sparsevec);
Datum
array_to_sparsevec(PG_FUNCTION_ARGS)
{
ArrayType *array = PG_GETARG_ARRAYTYPE_P(0);
int32 typmod = PG_GETARG_INT32(1);
SparseVector *result;
int16 typlen;
bool typbyval;
char typalign;
Datum *elemsp;
int nelemsp;
int nnz = 0;
float *values;
int j = 0;
if (ARR_NDIM(array) > 1)
ereport(ERROR,
(errcode(ERRCODE_DATA_EXCEPTION),
errmsg("array must be 1-D")));
if (ARR_HASNULL(array) && array_contains_nulls(array))
ereport(ERROR,
(errcode(ERRCODE_NULL_VALUE_NOT_ALLOWED),
errmsg("array must not contain nulls")));
get_typlenbyvalalign(ARR_ELEMTYPE(array), &typlen, &typbyval, &typalign);
deconstruct_array(array, ARR_ELEMTYPE(array), typlen, typbyval, typalign, &elemsp, NULL, &nelemsp);
CheckDim(nelemsp);
CheckExpectedDim(typmod, nelemsp);
#ifdef _MSC_VER
/* /fp:fast may not propagate +/-Infinity or NaN */
#define IS_NOT_ZERO(v) (isnan((float) (v)) || isinf((float) (v)) || ((float) (v)) != 0)
#else
#define IS_NOT_ZERO(v) (((float) (v)) != 0)
#endif
if (ARR_ELEMTYPE(array) == INT4OID)
{
for (int i = 0; i < nelemsp; i++)
nnz += IS_NOT_ZERO(DatumGetInt32(elemsp[i]));
}
else if (ARR_ELEMTYPE(array) == FLOAT8OID)
{
for (int i = 0; i < nelemsp; i++)
nnz += IS_NOT_ZERO(DatumGetFloat8(elemsp[i]));
}
else if (ARR_ELEMTYPE(array) == FLOAT4OID)
{
for (int i = 0; i < nelemsp; i++)
nnz += IS_NOT_ZERO(DatumGetFloat4(elemsp[i]));
}
else if (ARR_ELEMTYPE(array) == NUMERICOID)
{
for (int i = 0; i < nelemsp; i++)
nnz += IS_NOT_ZERO(DirectFunctionCall1(numeric_float4, elemsp[i]));
}
else
{
ereport(ERROR,
(errcode(ERRCODE_DATA_EXCEPTION),
errmsg("unsupported array type")));
}
result = InitSparseVector(nelemsp, nnz);
values = SPARSEVEC_VALUES(result);
#define PROCESS_ARRAY_ELEM(elem) \
do { \
float v = (float) (elem); \
if (IS_NOT_ZERO(v)) { \
/* Safety check */ \
if (j >= result->nnz) \
elog(ERROR, "safety check failed"); \
result->indices[j] = i; \
values[j] = v; \
j++; \
} \
} while (0)
if (ARR_ELEMTYPE(array) == INT4OID)
{
for (int i = 0; i < nelemsp; i++)
PROCESS_ARRAY_ELEM(DatumGetInt32(elemsp[i]));
}
else if (ARR_ELEMTYPE(array) == FLOAT8OID)
{
for (int i = 0; i < nelemsp; i++)
PROCESS_ARRAY_ELEM(DatumGetFloat8(elemsp[i]));
}
else if (ARR_ELEMTYPE(array) == FLOAT4OID)
{
for (int i = 0; i < nelemsp; i++)
PROCESS_ARRAY_ELEM(DatumGetFloat4(elemsp[i]));
}
else if (ARR_ELEMTYPE(array) == NUMERICOID)
{
for (int i = 0; i < nelemsp; i++)
PROCESS_ARRAY_ELEM(DatumGetFloat4(DirectFunctionCall1(numeric_float4, elemsp[i])));
}
else
{
ereport(ERROR,
(errcode(ERRCODE_DATA_EXCEPTION),
errmsg("unsupported array type")));
}
#undef PROCESS_ARRAY_ELEM
#undef IS_NOT_ZERO
/*
* Free allocation from deconstruct_array. Do not free individual elements
* when pass-by-reference since they point to original array.
*/
pfree(elemsp);
if (j != result->nnz)
elog(ERROR, "correctness check failed");
/* Check elements */
for (int i = 0; i < result->nnz; i++)
CheckElement(values[i]);
PG_RETURN_POINTER(result);
}
/*
* Get the L2 squared distance between sparse vectors
*/
@@ -854,7 +721,7 @@ SparsevecL2SquaredDistance(SparseVector * a, SparseVector * b)
/*
* Get the L2 distance between sparse vectors
*/
FUNCTION_PREFIX PG_FUNCTION_INFO_V1(sparsevec_l2_distance);
PGDLLEXPORT PG_FUNCTION_INFO_V1(sparsevec_l2_distance);
Datum
sparsevec_l2_distance(PG_FUNCTION_ARGS)
{
@@ -870,7 +737,7 @@ sparsevec_l2_distance(PG_FUNCTION_ARGS)
* Get the L2 squared distance between sparse vectors
* This saves a sqrt calculation
*/
FUNCTION_PREFIX PG_FUNCTION_INFO_V1(sparsevec_l2_squared_distance);
PGDLLEXPORT PG_FUNCTION_INFO_V1(sparsevec_l2_squared_distance);
Datum
sparsevec_l2_squared_distance(PG_FUNCTION_ARGS)
{
@@ -921,7 +788,7 @@ SparsevecInnerProduct(SparseVector * a, SparseVector * b)
/*
* Get the inner product of two sparse vectors
*/
FUNCTION_PREFIX PG_FUNCTION_INFO_V1(sparsevec_inner_product);
PGDLLEXPORT PG_FUNCTION_INFO_V1(sparsevec_inner_product);
Datum
sparsevec_inner_product(PG_FUNCTION_ARGS)
{
@@ -936,7 +803,7 @@ sparsevec_inner_product(PG_FUNCTION_ARGS)
/*
* Get the negative inner product of two sparse vectors
*/
FUNCTION_PREFIX PG_FUNCTION_INFO_V1(sparsevec_negative_inner_product);
PGDLLEXPORT PG_FUNCTION_INFO_V1(sparsevec_negative_inner_product);
Datum
sparsevec_negative_inner_product(PG_FUNCTION_ARGS)
{
@@ -951,7 +818,7 @@ sparsevec_negative_inner_product(PG_FUNCTION_ARGS)
/*
* Get the cosine distance between two sparse vectors
*/
FUNCTION_PREFIX PG_FUNCTION_INFO_V1(sparsevec_cosine_distance);
PGDLLEXPORT PG_FUNCTION_INFO_V1(sparsevec_cosine_distance);
Datum
sparsevec_cosine_distance(PG_FUNCTION_ARGS)
{
@@ -996,7 +863,7 @@ sparsevec_cosine_distance(PG_FUNCTION_ARGS)
/*
* Get the L1 distance between two sparse vectors
*/
FUNCTION_PREFIX PG_FUNCTION_INFO_V1(sparsevec_l1_distance);
PGDLLEXPORT PG_FUNCTION_INFO_V1(sparsevec_l1_distance);
Datum
sparsevec_l1_distance(PG_FUNCTION_ARGS)
{
@@ -1045,7 +912,7 @@ sparsevec_l1_distance(PG_FUNCTION_ARGS)
/*
* Get the L2 norm of a sparse vector
*/
FUNCTION_PREFIX PG_FUNCTION_INFO_V1(sparsevec_l2_norm);
PGDLLEXPORT PG_FUNCTION_INFO_V1(sparsevec_l2_norm);
Datum
sparsevec_l2_norm(PG_FUNCTION_ARGS)
{
@@ -1063,7 +930,7 @@ sparsevec_l2_norm(PG_FUNCTION_ARGS)
/*
* Normalize a sparse vector with the L2 norm
*/
FUNCTION_PREFIX PG_FUNCTION_INFO_V1(sparsevec_l2_normalize);
PGDLLEXPORT PG_FUNCTION_INFO_V1(sparsevec_l2_normalize);
Datum
sparsevec_l2_normalize(PG_FUNCTION_ARGS)
{
@@ -1173,7 +1040,7 @@ sparsevec_cmp_internal(SparseVector * a, SparseVector * b)
/*
* Less than
*/
FUNCTION_PREFIX PG_FUNCTION_INFO_V1(sparsevec_lt);
PGDLLEXPORT PG_FUNCTION_INFO_V1(sparsevec_lt);
Datum
sparsevec_lt(PG_FUNCTION_ARGS)
{
@@ -1186,7 +1053,7 @@ sparsevec_lt(PG_FUNCTION_ARGS)
/*
* Less than or equal
*/
FUNCTION_PREFIX PG_FUNCTION_INFO_V1(sparsevec_le);
PGDLLEXPORT PG_FUNCTION_INFO_V1(sparsevec_le);
Datum
sparsevec_le(PG_FUNCTION_ARGS)
{
@@ -1199,7 +1066,7 @@ sparsevec_le(PG_FUNCTION_ARGS)
/*
* Equal
*/
FUNCTION_PREFIX PG_FUNCTION_INFO_V1(sparsevec_eq);
PGDLLEXPORT PG_FUNCTION_INFO_V1(sparsevec_eq);
Datum
sparsevec_eq(PG_FUNCTION_ARGS)
{
@@ -1212,7 +1079,7 @@ sparsevec_eq(PG_FUNCTION_ARGS)
/*
* Not equal
*/
FUNCTION_PREFIX PG_FUNCTION_INFO_V1(sparsevec_ne);
PGDLLEXPORT PG_FUNCTION_INFO_V1(sparsevec_ne);
Datum
sparsevec_ne(PG_FUNCTION_ARGS)
{
@@ -1225,7 +1092,7 @@ sparsevec_ne(PG_FUNCTION_ARGS)
/*
* Greater than or equal
*/
FUNCTION_PREFIX PG_FUNCTION_INFO_V1(sparsevec_ge);
PGDLLEXPORT PG_FUNCTION_INFO_V1(sparsevec_ge);
Datum
sparsevec_ge(PG_FUNCTION_ARGS)
{
@@ -1238,7 +1105,7 @@ sparsevec_ge(PG_FUNCTION_ARGS)
/*
* Greater than
*/
FUNCTION_PREFIX PG_FUNCTION_INFO_V1(sparsevec_gt);
PGDLLEXPORT PG_FUNCTION_INFO_V1(sparsevec_gt);
Datum
sparsevec_gt(PG_FUNCTION_ARGS)
{
@@ -1251,7 +1118,7 @@ sparsevec_gt(PG_FUNCTION_ARGS)
/*
* Compare sparse vectors
*/
FUNCTION_PREFIX PG_FUNCTION_INFO_V1(sparsevec_cmp);
PGDLLEXPORT PG_FUNCTION_INFO_V1(sparsevec_cmp);
Datum
sparsevec_cmp(PG_FUNCTION_ARGS)
{

View File

@@ -26,6 +26,11 @@
#include "varatt.h"
#endif
#if PG_VERSION_NUM < 130000
#define TYPALIGN_DOUBLE 'd'
#define TYPALIGN_INT 'i'
#endif
#define STATE_DIMS(x) (ARR_DIMS(x)[0] - 1)
#define CreateStateDatums(dim) palloc(sizeof(Datum) * (dim + 1))
@@ -155,10 +160,28 @@ CheckStateArray(ArrayType *statearray, const char *caller)
return (float8 *) ARR_DATA_PTR(statearray);
}
#if PG_VERSION_NUM < 120003
static pg_noinline void
float_overflow_error(void)
{
ereport(ERROR,
(errcode(ERRCODE_NUMERIC_VALUE_OUT_OF_RANGE),
errmsg("value out of range: overflow")));
}
static pg_noinline void
float_underflow_error(void)
{
ereport(ERROR,
(errcode(ERRCODE_NUMERIC_VALUE_OUT_OF_RANGE),
errmsg("value out of range: underflow")));
}
#endif
/*
* Convert textual representation to internal representation
*/
FUNCTION_PREFIX PG_FUNCTION_INFO_V1(vector_in);
PGDLLEXPORT PG_FUNCTION_INFO_V1(vector_in);
Datum
vector_in(PG_FUNCTION_ARGS)
{
@@ -271,7 +294,7 @@ vector_in(PG_FUNCTION_ARGS)
/*
* Convert internal representation to textual representation
*/
FUNCTION_PREFIX PG_FUNCTION_INFO_V1(vector_out);
PGDLLEXPORT PG_FUNCTION_INFO_V1(vector_out);
Datum
vector_out(PG_FUNCTION_ARGS)
{
@@ -325,7 +348,7 @@ PrintVector(char *msg, Vector * vector)
/*
* Convert type modifier
*/
FUNCTION_PREFIX PG_FUNCTION_INFO_V1(vector_typmod_in);
PGDLLEXPORT PG_FUNCTION_INFO_V1(vector_typmod_in);
Datum
vector_typmod_in(PG_FUNCTION_ARGS)
{
@@ -356,7 +379,7 @@ vector_typmod_in(PG_FUNCTION_ARGS)
/*
* Convert external binary representation to internal representation
*/
FUNCTION_PREFIX PG_FUNCTION_INFO_V1(vector_recv);
PGDLLEXPORT PG_FUNCTION_INFO_V1(vector_recv);
Datum
vector_recv(PG_FUNCTION_ARGS)
{
@@ -390,7 +413,7 @@ vector_recv(PG_FUNCTION_ARGS)
/*
* Convert internal representation to the external binary representation
*/
FUNCTION_PREFIX PG_FUNCTION_INFO_V1(vector_send);
PGDLLEXPORT PG_FUNCTION_INFO_V1(vector_send);
Datum
vector_send(PG_FUNCTION_ARGS)
{
@@ -410,7 +433,7 @@ vector_send(PG_FUNCTION_ARGS)
* Convert vector to vector
* This is needed to check the type modifier
*/
FUNCTION_PREFIX PG_FUNCTION_INFO_V1(vector);
PGDLLEXPORT PG_FUNCTION_INFO_V1(vector);
Datum
vector(PG_FUNCTION_ARGS)
{
@@ -425,7 +448,7 @@ vector(PG_FUNCTION_ARGS)
/*
* Convert array to vector
*/
FUNCTION_PREFIX PG_FUNCTION_INFO_V1(array_to_vector);
PGDLLEXPORT PG_FUNCTION_INFO_V1(array_to_vector);
Datum
array_to_vector(PG_FUNCTION_ARGS)
{
@@ -499,7 +522,7 @@ array_to_vector(PG_FUNCTION_ARGS)
/*
* Convert vector to float4[]
*/
FUNCTION_PREFIX PG_FUNCTION_INFO_V1(vector_to_float4);
PGDLLEXPORT PG_FUNCTION_INFO_V1(vector_to_float4);
Datum
vector_to_float4(PG_FUNCTION_ARGS)
{
@@ -523,7 +546,7 @@ vector_to_float4(PG_FUNCTION_ARGS)
/*
* Convert half vector to vector
*/
FUNCTION_PREFIX PG_FUNCTION_INFO_V1(halfvec_to_vector);
PGDLLEXPORT PG_FUNCTION_INFO_V1(halfvec_to_vector);
Datum
halfvec_to_vector(PG_FUNCTION_ARGS)
{
@@ -561,7 +584,7 @@ VectorL2SquaredDistance(int dim, float *ax, float *bx)
/*
* Get the L2 distance between vectors
*/
FUNCTION_PREFIX PG_FUNCTION_INFO_V1(l2_distance);
PGDLLEXPORT PG_FUNCTION_INFO_V1(l2_distance);
Datum
l2_distance(PG_FUNCTION_ARGS)
{
@@ -577,7 +600,7 @@ l2_distance(PG_FUNCTION_ARGS)
* Get the L2 squared distance between vectors
* This saves a sqrt calculation
*/
FUNCTION_PREFIX PG_FUNCTION_INFO_V1(vector_l2_squared_distance);
PGDLLEXPORT PG_FUNCTION_INFO_V1(vector_l2_squared_distance);
Datum
vector_l2_squared_distance(PG_FUNCTION_ARGS)
{
@@ -604,7 +627,7 @@ VectorInnerProduct(int dim, float *ax, float *bx)
/*
* Get the inner product of two vectors
*/
FUNCTION_PREFIX PG_FUNCTION_INFO_V1(inner_product);
PGDLLEXPORT PG_FUNCTION_INFO_V1(inner_product);
Datum
inner_product(PG_FUNCTION_ARGS)
{
@@ -619,7 +642,7 @@ inner_product(PG_FUNCTION_ARGS)
/*
* Get the negative inner product of two vectors
*/
FUNCTION_PREFIX PG_FUNCTION_INFO_V1(vector_negative_inner_product);
PGDLLEXPORT PG_FUNCTION_INFO_V1(vector_negative_inner_product);
Datum
vector_negative_inner_product(PG_FUNCTION_ARGS)
{
@@ -653,7 +676,7 @@ VectorCosineSimilarity(int dim, float *ax, float *bx)
/*
* Get the cosine distance between two vectors
*/
FUNCTION_PREFIX PG_FUNCTION_INFO_V1(cosine_distance);
PGDLLEXPORT PG_FUNCTION_INFO_V1(cosine_distance);
Datum
cosine_distance(PG_FUNCTION_ARGS)
{
@@ -685,7 +708,7 @@ cosine_distance(PG_FUNCTION_ARGS)
* Currently uses angular distance since needs to satisfy triangle inequality
* Assumes inputs are unit vectors (skips norm)
*/
FUNCTION_PREFIX PG_FUNCTION_INFO_V1(vector_spherical_distance);
PGDLLEXPORT PG_FUNCTION_INFO_V1(vector_spherical_distance);
Datum
vector_spherical_distance(PG_FUNCTION_ARGS)
{
@@ -722,7 +745,7 @@ VectorL1Distance(int dim, float *ax, float *bx)
/*
* Get the L1 distance between two vectors
*/
FUNCTION_PREFIX PG_FUNCTION_INFO_V1(l1_distance);
PGDLLEXPORT PG_FUNCTION_INFO_V1(l1_distance);
Datum
l1_distance(PG_FUNCTION_ARGS)
{
@@ -737,7 +760,7 @@ l1_distance(PG_FUNCTION_ARGS)
/*
* Get the dimensions of a vector
*/
FUNCTION_PREFIX PG_FUNCTION_INFO_V1(vector_dims);
PGDLLEXPORT PG_FUNCTION_INFO_V1(vector_dims);
Datum
vector_dims(PG_FUNCTION_ARGS)
{
@@ -749,7 +772,7 @@ vector_dims(PG_FUNCTION_ARGS)
/*
* Get the L2 norm of a vector
*/
FUNCTION_PREFIX PG_FUNCTION_INFO_V1(vector_norm);
PGDLLEXPORT PG_FUNCTION_INFO_V1(vector_norm);
Datum
vector_norm(PG_FUNCTION_ARGS)
{
@@ -767,7 +790,7 @@ vector_norm(PG_FUNCTION_ARGS)
/*
* Normalize a vector with the L2 norm
*/
FUNCTION_PREFIX PG_FUNCTION_INFO_V1(l2_normalize);
PGDLLEXPORT PG_FUNCTION_INFO_V1(l2_normalize);
Datum
l2_normalize(PG_FUNCTION_ARGS)
{
@@ -806,7 +829,7 @@ l2_normalize(PG_FUNCTION_ARGS)
/*
* Add vectors
*/
FUNCTION_PREFIX PG_FUNCTION_INFO_V1(vector_add);
PGDLLEXPORT PG_FUNCTION_INFO_V1(vector_add);
Datum
vector_add(PG_FUNCTION_ARGS)
{
@@ -839,7 +862,7 @@ vector_add(PG_FUNCTION_ARGS)
/*
* Subtract vectors
*/
FUNCTION_PREFIX PG_FUNCTION_INFO_V1(vector_sub);
PGDLLEXPORT PG_FUNCTION_INFO_V1(vector_sub);
Datum
vector_sub(PG_FUNCTION_ARGS)
{
@@ -872,7 +895,7 @@ vector_sub(PG_FUNCTION_ARGS)
/*
* Multiply vectors
*/
FUNCTION_PREFIX PG_FUNCTION_INFO_V1(vector_mul);
PGDLLEXPORT PG_FUNCTION_INFO_V1(vector_mul);
Datum
vector_mul(PG_FUNCTION_ARGS)
{
@@ -908,7 +931,7 @@ vector_mul(PG_FUNCTION_ARGS)
/*
* Concatenate vectors
*/
FUNCTION_PREFIX PG_FUNCTION_INFO_V1(vector_concat);
PGDLLEXPORT PG_FUNCTION_INFO_V1(vector_concat);
Datum
vector_concat(PG_FUNCTION_ARGS)
{
@@ -932,7 +955,7 @@ vector_concat(PG_FUNCTION_ARGS)
/*
* Quantize a vector
*/
FUNCTION_PREFIX PG_FUNCTION_INFO_V1(binary_quantize);
PGDLLEXPORT PG_FUNCTION_INFO_V1(binary_quantize);
Datum
binary_quantize(PG_FUNCTION_ARGS)
{
@@ -950,7 +973,7 @@ binary_quantize(PG_FUNCTION_ARGS)
/*
* Get a subvector
*/
FUNCTION_PREFIX PG_FUNCTION_INFO_V1(subvector);
PGDLLEXPORT PG_FUNCTION_INFO_V1(subvector);
Datum
subvector(PG_FUNCTION_ARGS)
{
@@ -1024,7 +1047,7 @@ vector_cmp_internal(Vector * a, Vector * b)
/*
* Less than
*/
FUNCTION_PREFIX PG_FUNCTION_INFO_V1(vector_lt);
PGDLLEXPORT PG_FUNCTION_INFO_V1(vector_lt);
Datum
vector_lt(PG_FUNCTION_ARGS)
{
@@ -1037,7 +1060,7 @@ vector_lt(PG_FUNCTION_ARGS)
/*
* Less than or equal
*/
FUNCTION_PREFIX PG_FUNCTION_INFO_V1(vector_le);
PGDLLEXPORT PG_FUNCTION_INFO_V1(vector_le);
Datum
vector_le(PG_FUNCTION_ARGS)
{
@@ -1050,7 +1073,7 @@ vector_le(PG_FUNCTION_ARGS)
/*
* Equal
*/
FUNCTION_PREFIX PG_FUNCTION_INFO_V1(vector_eq);
PGDLLEXPORT PG_FUNCTION_INFO_V1(vector_eq);
Datum
vector_eq(PG_FUNCTION_ARGS)
{
@@ -1063,7 +1086,7 @@ vector_eq(PG_FUNCTION_ARGS)
/*
* Not equal
*/
FUNCTION_PREFIX PG_FUNCTION_INFO_V1(vector_ne);
PGDLLEXPORT PG_FUNCTION_INFO_V1(vector_ne);
Datum
vector_ne(PG_FUNCTION_ARGS)
{
@@ -1076,7 +1099,7 @@ vector_ne(PG_FUNCTION_ARGS)
/*
* Greater than or equal
*/
FUNCTION_PREFIX PG_FUNCTION_INFO_V1(vector_ge);
PGDLLEXPORT PG_FUNCTION_INFO_V1(vector_ge);
Datum
vector_ge(PG_FUNCTION_ARGS)
{
@@ -1089,7 +1112,7 @@ vector_ge(PG_FUNCTION_ARGS)
/*
* Greater than
*/
FUNCTION_PREFIX PG_FUNCTION_INFO_V1(vector_gt);
PGDLLEXPORT PG_FUNCTION_INFO_V1(vector_gt);
Datum
vector_gt(PG_FUNCTION_ARGS)
{
@@ -1102,7 +1125,7 @@ vector_gt(PG_FUNCTION_ARGS)
/*
* Compare vectors
*/
FUNCTION_PREFIX PG_FUNCTION_INFO_V1(vector_cmp);
PGDLLEXPORT PG_FUNCTION_INFO_V1(vector_cmp);
Datum
vector_cmp(PG_FUNCTION_ARGS)
{
@@ -1115,7 +1138,7 @@ vector_cmp(PG_FUNCTION_ARGS)
/*
* Accumulate vectors
*/
FUNCTION_PREFIX PG_FUNCTION_INFO_V1(vector_accum);
PGDLLEXPORT PG_FUNCTION_INFO_V1(vector_accum);
Datum
vector_accum(PG_FUNCTION_ARGS)
{
@@ -1176,7 +1199,7 @@ vector_accum(PG_FUNCTION_ARGS)
/*
* Combine vectors or half vectors (also used for halfvec_combine)
*/
FUNCTION_PREFIX PG_FUNCTION_INFO_V1(vector_combine);
PGDLLEXPORT PG_FUNCTION_INFO_V1(vector_combine);
Datum
vector_combine(PG_FUNCTION_ARGS)
{
@@ -1247,7 +1270,7 @@ vector_combine(PG_FUNCTION_ARGS)
/*
* Average vectors
*/
FUNCTION_PREFIX PG_FUNCTION_INFO_V1(vector_avg);
PGDLLEXPORT PG_FUNCTION_INFO_V1(vector_avg);
Datum
vector_avg(PG_FUNCTION_ARGS)
{
@@ -1281,7 +1304,7 @@ vector_avg(PG_FUNCTION_ARGS)
/*
* Convert sparse vector to dense vector
*/
FUNCTION_PREFIX PG_FUNCTION_INFO_V1(sparsevec_to_vector);
PGDLLEXPORT PG_FUNCTION_INFO_V1(sparsevec_to_vector);
Datum
sparsevec_to_vector(PG_FUNCTION_ARGS)
{

View File

@@ -20,11 +20,4 @@ Vector *InitVector(int dim);
void PrintVector(char *msg, Vector * vector);
int vector_cmp_internal(Vector * a, Vector * b);
/* TODO Move to better place */
#if PG_VERSION_NUM >= 160000
#define FUNCTION_PREFIX
#else
#define FUNCTION_PREFIX PGDLLEXPORT
#endif
#endif

View File

@@ -140,34 +140,6 @@ SELECT '{1e-8,-1e-8}'::real[]::halfvec;
[0,-0]
(1 row)
SELECT '[1,2,3]'::intvec::integer[];
int4
---------
{1,2,3}
(1 row)
SELECT '{1,2,3}'::integer[]::intvec;
intvec
---------
[1,2,3]
(1 row)
SELECT '{1,2,3}'::integer[]::intvec(3);
intvec
---------
[1,2,3]
(1 row)
SELECT '{1,2,3}'::integer[]::intvec(2);
ERROR: expected 2 dimensions, not 3
SELECT '{127,-128}'::integer[]::intvec;
intvec
------------
[127,-128]
(1 row)
SELECT '{128,-129}'::integer[]::intvec;
ERROR: value "128" is out of range for type intvec
SELECT '[0,1.5,0,3.5,0]'::vector::sparsevec;
sparsevec
-----------------
@@ -236,62 +208,6 @@ SELECT '{1:1e-8}/1'::sparsevec::halfvec;
[0]
(1 row)
SELECT ARRAY[1,0,2,0,3,0]::sparsevec;
array
-----------------
{1:1,3:2,5:3}/6
(1 row)
SELECT ARRAY[1.0,0.0,2.0,0.0,3.0,0.0]::sparsevec;
array
-----------------
{1:1,3:2,5:3}/6
(1 row)
SELECT ARRAY[1,0,2,0,3,0]::float4[]::sparsevec;
array
-----------------
{1:1,3:2,5:3}/6
(1 row)
SELECT ARRAY[1,0,2,0,3,0]::float8[]::sparsevec;
array
-----------------
{1:1,3:2,5:3}/6
(1 row)
SELECT ARRAY[1,0,2,0,3,0]::numeric[]::sparsevec;
array
-----------------
{1:1,3:2,5:3}/6
(1 row)
SELECT '{1,0,2,0,3,0}'::real[]::sparsevec;
sparsevec
-----------------
{1:1,3:2,5:3}/6
(1 row)
SELECT '{1,0,2,0,3,0}'::real[]::sparsevec(6);
sparsevec
-----------------
{1:1,3:2,5:3}/6
(1 row)
SELECT '{1,0,2,0,3,0}'::real[]::sparsevec(5);
ERROR: expected 5 dimensions, not 6
SELECT '{NULL}'::real[]::sparsevec;
ERROR: array must not contain nulls
SELECT '{NaN}'::real[]::sparsevec;
ERROR: NaN not allowed in sparsevec
SELECT '{Infinity}'::real[]::sparsevec;
ERROR: infinite value not allowed in sparsevec
SELECT '{-Infinity}'::real[]::sparsevec;
ERROR: infinite value not allowed in sparsevec
SELECT '{}'::real[]::sparsevec;
ERROR: sparsevec must have at least 1 dimension
SELECT '{{1}}'::real[]::sparsevec;
ERROR: array must be 1-D
SELECT array_agg(n)::vector FROM generate_series(1, 16001) n;
ERROR: vector cannot have more than 16000 dimensions
SELECT array_to_vector(array_agg(n), 16001, false) FROM generate_series(1, 16001) n;

View File

@@ -30,23 +30,6 @@ SELECT * FROM t2 ORDER BY val;
(4 rows)
DROP TABLE t;
DROP TABLE t2;
-- intvec
CREATE TABLE t (val intvec(3));
INSERT INTO t (val) VALUES ('[0,0,0]'), ('[1,2,3]'), ('[1,1,1]'), (NULL);
CREATE TABLE t2 (val intvec(3));
\copy t TO 'results/intvec.bin' WITH (FORMAT binary)
\copy t2 FROM 'results/intvec.bin' WITH (FORMAT binary)
SELECT * FROM t2 ORDER BY val;
val
---------
[0,0,0]
[1,1,1]
[1,2,3]
(4 rows)
DROP TABLE t;
DROP TABLE t2;
-- sparsevec

View File

@@ -1,102 +0,0 @@
SET enable_seqscan = off;
-- L2
CREATE TABLE t (val intvec(3));
INSERT INTO t (val) VALUES ('[0,0,0]'), ('[1,2,3]'), ('[1,1,1]'), (NULL);
CREATE INDEX ON t USING hnsw (val intvec_l2_ops);
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::intvec)) t2;
count
-------
4
(1 row)
SELECT COUNT(*) FROM t;
count
-------
5
(1 row)
TRUNCATE t;
SELECT * FROM t ORDER BY val <-> '[3,3,3]';
val
-----
(0 rows)
DROP TABLE t;
-- inner product
CREATE TABLE t (val intvec(3));
INSERT INTO t (val) VALUES ('[0,0,0]'), ('[1,2,3]'), ('[1,1,1]'), (NULL);
CREATE INDEX ON t USING hnsw (val intvec_ip_ops);
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::intvec)) t2;
count
-------
4
(1 row)
DROP TABLE t;
-- cosine
CREATE TABLE t (val intvec(3));
INSERT INTO t (val) VALUES ('[0,0,0]'), ('[1,2,3]'), ('[1,1,1]'), (NULL);
CREATE INDEX ON t USING hnsw (val intvec_cosine_ops);
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::intvec)) t2;
count
-------
3
(1 row)
DROP TABLE t;
-- L1
CREATE TABLE t (val intvec(3));
INSERT INTO t (val) VALUES ('[0,0,0]'), ('[1,2,3]'), ('[1,1,1]'), (NULL);
CREATE INDEX ON t USING hnsw (val intvec_l1_ops);
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::intvec)) t2;
count
-------
4
(1 row)
DROP TABLE t;

View File

@@ -99,32 +99,6 @@ SELECT COUNT(*) FROM (SELECT * FROM t ORDER BY val <+> (SELECT NULL::vector)) t2
4
(1 row)
DROP TABLE t;
-- iterative
CREATE TABLE t (val vector(3));
INSERT INTO t (val) VALUES ('[0,0,0]'), ('[1,2,3]'), ('[1,1,1]'), (NULL);
CREATE INDEX ON t USING hnsw (val vector_l2_ops);
SET hnsw.iterative_search = strict_order;
SET hnsw.ef_search = 1;
SELECT * FROM t ORDER BY val <-> '[3,3,3]';
val
---------
[1,2,3]
[1,1,1]
[0,0,0]
(3 rows)
SET hnsw.iterative_search = relaxed_order;
SELECT * FROM t ORDER BY val <-> '[3,3,3]';
val
---------
[1,2,3]
[1,1,1]
[0,0,0]
(3 rows)
RESET hnsw.iterative_search;
RESET hnsw.ef_search;
DROP TABLE t;
-- unlogged
CREATE UNLOGGED TABLE t (val vector(3));
@@ -165,21 +139,4 @@ SET hnsw.ef_search = 0;
ERROR: 0 is outside the valid range for parameter "hnsw.ef_search" (1 .. 1000)
SET hnsw.ef_search = 1001;
ERROR: 1001 is outside the valid range for parameter "hnsw.ef_search" (1 .. 1000)
SHOW hnsw.iterative_search;
hnsw.iterative_search
-----------------------
off
(1 row)
SET hnsw.iterative_search = on;
ERROR: invalid value for parameter "hnsw.iterative_search": "on"
HINT: Available values: off, relaxed_order, strict_order.
SHOW hnsw.max_search_tuples;
hnsw.max_search_tuples
------------------------
-1
(1 row)
SET hnsw.max_search_tuples = -2;
ERROR: -2 is outside the valid range for parameter "hnsw.max_search_tuples" (-1 .. 2147483647)
DROP TABLE t;

View File

@@ -1,328 +0,0 @@
SELECT '[1,2,3]'::intvec;
intvec
---------
[1,2,3]
(1 row)
SELECT '[-1,-2,-3]'::intvec;
intvec
------------
[-1,-2,-3]
(1 row)
SELECT ' [ 1, 2 , 3 ] '::intvec;
intvec
---------
[1,2,3]
(1 row)
SELECT '[1.23456]'::intvec;
ERROR: invalid input syntax for type intvec: "[1.23456]"
LINE 1: SELECT '[1.23456]'::intvec;
^
SELECT '[hello,1]'::intvec;
ERROR: invalid input syntax for type intvec: "[hello,1]"
LINE 1: SELECT '[hello,1]'::intvec;
^
SELECT '[127,-128]'::intvec;
intvec
------------
[127,-128]
(1 row)
SELECT '[128,-129]'::intvec;
ERROR: "128" is out of range for type intvec
LINE 1: SELECT '[128,-129]'::intvec;
^
SELECT '[1,2,3'::intvec;
ERROR: invalid input syntax for type intvec: "[1,2,3"
LINE 1: SELECT '[1,2,3'::intvec;
^
SELECT '[1,2,3]9'::intvec;
ERROR: invalid input syntax for type intvec: "[1,2,3]9"
LINE 1: SELECT '[1,2,3]9'::intvec;
^
DETAIL: Junk after closing right brace.
SELECT '1,2,3'::intvec;
ERROR: invalid input syntax for type intvec: "1,2,3"
LINE 1: SELECT '1,2,3'::intvec;
^
DETAIL: Vector contents must start with "[".
SELECT ''::intvec;
ERROR: invalid input syntax for type intvec: ""
LINE 1: SELECT ''::intvec;
^
DETAIL: Vector contents must start with "[".
SELECT '['::intvec;
ERROR: invalid input syntax for type intvec: "["
LINE 1: SELECT '['::intvec;
^
SELECT '[,'::intvec;
ERROR: invalid input syntax for type intvec: "[,"
LINE 1: SELECT '[,'::intvec;
^
SELECT '[]'::intvec;
ERROR: intvec must have at least 1 dimension
LINE 1: SELECT '[]'::intvec;
^
SELECT '[1,]'::intvec;
ERROR: invalid input syntax for type intvec: "[1,]"
LINE 1: SELECT '[1,]'::intvec;
^
SELECT '[1a]'::intvec;
ERROR: invalid input syntax for type intvec: "[1a]"
LINE 1: SELECT '[1a]'::intvec;
^
SELECT '[1,,3]'::intvec;
ERROR: invalid input syntax for type intvec: "[1,,3]"
LINE 1: SELECT '[1,,3]'::intvec;
^
SELECT '[1, ,3]'::intvec;
ERROR: invalid input syntax for type intvec: "[1, ,3]"
LINE 1: SELECT '[1, ,3]'::intvec;
^
SELECT '[1,2,3]'::intvec(3);
intvec
---------
[1,2,3]
(1 row)
SELECT '[1,2,3]'::intvec(2);
ERROR: expected 2 dimensions, not 3
SELECT '[1,2,3]'::intvec(3, 2);
ERROR: invalid type modifier
LINE 1: SELECT '[1,2,3]'::intvec(3, 2);
^
SELECT '[1,2,3]'::intvec('a');
ERROR: invalid input syntax for type integer: "a"
LINE 1: SELECT '[1,2,3]'::intvec('a');
^
SELECT '[1,2,3]'::intvec(0);
ERROR: dimensions for type intvec must be at least 1
LINE 1: SELECT '[1,2,3]'::intvec(0);
^
SELECT '[1,2,3]'::intvec(16001);
ERROR: dimensions for type intvec cannot exceed 16000
LINE 1: SELECT '[1,2,3]'::intvec(16001);
^
SELECT unnest('{"[1,2,3]", "[4,5,6]"}'::intvec[]);
unnest
---------
[1,2,3]
[4,5,6]
(2 rows)
SELECT '{"[1,2,3]"}'::intvec(2)[];
ERROR: expected 2 dimensions, not 3
SELECT '[1,2,3]'::intvec < '[1,2,3]';
?column?
----------
f
(1 row)
SELECT '[1,2,3]'::intvec < '[1,2]';
?column?
----------
f
(1 row)
SELECT '[1,2,3]'::intvec <= '[1,2,3]';
?column?
----------
t
(1 row)
SELECT '[1,2,3]'::intvec <= '[1,2]';
?column?
----------
f
(1 row)
SELECT '[1,2,3]'::intvec = '[1,2,3]';
?column?
----------
t
(1 row)
SELECT '[1,2,3]'::intvec = '[1,2]';
?column?
----------
f
(1 row)
SELECT '[1,2,3]'::intvec != '[1,2,3]';
?column?
----------
f
(1 row)
SELECT '[1,2,3]'::intvec != '[1,2]';
?column?
----------
t
(1 row)
SELECT '[1,2,3]'::intvec >= '[1,2,3]';
?column?
----------
t
(1 row)
SELECT '[1,2,3]'::intvec >= '[1,2]';
?column?
----------
t
(1 row)
SELECT '[1,2,3]'::intvec > '[1,2,3]';
?column?
----------
f
(1 row)
SELECT '[1,2,3]'::intvec > '[1,2]';
?column?
----------
t
(1 row)
SELECT intvec_cmp('[1,2,3]', '[1,2,3]');
intvec_cmp
------------
0
(1 row)
SELECT intvec_cmp('[1,2,3]', '[0,0,0]');
intvec_cmp
------------
1
(1 row)
SELECT intvec_cmp('[0,0,0]', '[1,2,3]');
intvec_cmp
------------
-1
(1 row)
SELECT intvec_cmp('[1,2]', '[1,2,3]');
intvec_cmp
------------
-1
(1 row)
SELECT intvec_cmp('[1,2,3]', '[1,2]');
intvec_cmp
------------
1
(1 row)
SELECT intvec_cmp('[1,2]', '[2,3,4]');
intvec_cmp
------------
-1
(1 row)
SELECT intvec_cmp('[2,3]', '[1,2,3]');
intvec_cmp
------------
1
(1 row)
SELECT vector_dims('[1,2,3]'::intvec);
vector_dims
-------------
3
(1 row)
SELECT l2_distance('[0,0]'::intvec, '[3,4]');
l2_distance
-------------
5
(1 row)
SELECT l2_distance('[0,0]'::intvec, '[0,1]');
l2_distance
-------------
1
(1 row)
SELECT l2_distance('[1,2]'::intvec, '[3]');
ERROR: different intvec dimensions 2 and 1
SELECT '[0,0]'::intvec <-> '[3,4]';
?column?
----------
5
(1 row)
SELECT inner_product('[1,2]'::intvec, '[3,4]');
inner_product
---------------
11
(1 row)
SELECT inner_product('[1,2]'::intvec, '[3]');
ERROR: different intvec dimensions 2 and 1
SELECT inner_product('[127]'::intvec, '[127]');
inner_product
---------------
16129
(1 row)
SELECT '[1,2]'::intvec <#> '[3,4]';
?column?
----------
-11
(1 row)
SELECT cosine_distance('[1,2]'::intvec, '[2,4]');
cosine_distance
-----------------
0
(1 row)
SELECT cosine_distance('[1,2]'::intvec, '[0,0]');
cosine_distance
-----------------
NaN
(1 row)
SELECT cosine_distance('[1,1]'::intvec, '[1,1]');
cosine_distance
-----------------
0
(1 row)
SELECT cosine_distance('[1,0]'::intvec, '[0,2]');
cosine_distance
-----------------
1
(1 row)
SELECT cosine_distance('[1,1]'::intvec, '[-1,-1]');
cosine_distance
-----------------
2
(1 row)
SELECT cosine_distance('[1,2]'::intvec, '[3]');
ERROR: different intvec dimensions 2 and 1
SELECT '[1,2]'::intvec <=> '[2,4]';
?column?
----------
0
(1 row)
SELECT l1_distance('[0,0]'::intvec, '[3,4]');
l1_distance
-------------
7
(1 row)
SELECT l1_distance('[0,0]'::intvec, '[0,1]');
l1_distance
-------------
1
(1 row)
SELECT l1_distance('[1,2]'::intvec, '[3]');
ERROR: different intvec dimensions 2 and 1

View File

@@ -81,44 +81,6 @@ SELECT COUNT(*) FROM (SELECT * FROM t ORDER BY val <=> (SELECT NULL::vector)) t2
3
(1 row)
DROP TABLE t;
-- iterative
CREATE TABLE t (val vector(3));
INSERT INTO t (val) VALUES ('[0,0,0]'), ('[1,2,3]'), ('[1,1,1]'), (NULL);
CREATE INDEX ON t USING ivfflat (val vector_l2_ops) WITH (lists = 3);
SET ivfflat.iterative_search = relaxed_order;
SELECT * FROM t ORDER BY val <-> '[3,3,3]';
val
---------
[1,2,3]
[1,1,1]
[0,0,0]
(3 rows)
SET ivfflat.max_probes = 0;
SELECT * FROM t ORDER BY val <-> '[3,3,3]';
val
---------
[1,2,3]
(1 row)
SET ivfflat.max_probes = 1;
SELECT * FROM t ORDER BY val <-> '[3,3,3]';
val
---------
[1,2,3]
(1 row)
SET ivfflat.max_probes = 2;
SELECT * FROM t ORDER BY val <-> '[3,3,3]';
val
---------
[1,2,3]
[1,1,1]
(2 rows)
RESET ivfflat.iterative_search;
RESET ivfflat.max_probes;
DROP TABLE t;
-- unlogged
CREATE UNLOGGED TABLE t (val vector(3));
@@ -147,27 +109,4 @@ SHOW ivfflat.probes;
1
(1 row)
SET ivfflat.probes = 0;
ERROR: 0 is outside the valid range for parameter "ivfflat.probes" (1 .. 32768)
SET ivfflat.probes = 32769;
ERROR: 32769 is outside the valid range for parameter "ivfflat.probes" (1 .. 32768)
SHOW ivfflat.iterative_search;
ivfflat.iterative_search
--------------------------
off
(1 row)
SET ivfflat.iterative_search = on;
ERROR: invalid value for parameter "ivfflat.iterative_search": "on"
HINT: Available values: off, relaxed_order.
SHOW ivfflat.max_probes;
ivfflat.max_probes
--------------------
-1
(1 row)
SET ivfflat.max_probes = -2;
ERROR: -2 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)
DROP TABLE t;

View File

@@ -1,11 +0,0 @@
package PostgreSQL::Test::Cluster;
use PostgresNode;
sub new
{
my ($class, $name) = @_;
return get_new_node($name);
}
1;

View File

@@ -1,5 +0,0 @@
package PostgreSQL::Test::Utils;
use TestLib;
1;

View File

@@ -0,0 +1,8 @@
use PostgreSQL::Test::Cluster;
sub get_new_node
{
return PostgreSQL::Test::Cluster->new(@_);
}
1;

3
test/perl/TestLib.pm Normal file
View File

@@ -0,0 +1,3 @@
use PostgreSQL::Test::Utils;
1;

View File

@@ -38,14 +38,6 @@ SELECT '{1,2,3}'::real[]::halfvec(2);
SELECT '{65520,-65520}'::real[]::halfvec;
SELECT '{1e-8,-1e-8}'::real[]::halfvec;
SELECT '[1,2,3]'::intvec::integer[];
SELECT '{1,2,3}'::integer[]::intvec;
SELECT '{1,2,3}'::integer[]::intvec(3);
SELECT '{1,2,3}'::integer[]::intvec(2);
SELECT '{127,-128}'::integer[]::intvec;
SELECT '{128,-129}'::integer[]::intvec;
SELECT '[0,1.5,0,3.5,0]'::vector::sparsevec;
SELECT '[0,1.5,0,3.5,0]'::vector::sparsevec(5);
SELECT '[0,1.5,0,3.5,0]'::vector::sparsevec(4);
@@ -66,22 +58,6 @@ SELECT '{}/16001'::sparsevec::halfvec;
SELECT '{1:65520}/1'::sparsevec::halfvec;
SELECT '{1:1e-8}/1'::sparsevec::halfvec;
SELECT ARRAY[1,0,2,0,3,0]::sparsevec;
SELECT ARRAY[1.0,0.0,2.0,0.0,3.0,0.0]::sparsevec;
SELECT ARRAY[1,0,2,0,3,0]::float4[]::sparsevec;
SELECT ARRAY[1,0,2,0,3,0]::float8[]::sparsevec;
SELECT ARRAY[1,0,2,0,3,0]::numeric[]::sparsevec;
SELECT '{1,0,2,0,3,0}'::real[]::sparsevec;
SELECT '{1,0,2,0,3,0}'::real[]::sparsevec(6);
SELECT '{1,0,2,0,3,0}'::real[]::sparsevec(5);
SELECT '{NULL}'::real[]::sparsevec;
SELECT '{NaN}'::real[]::sparsevec;
SELECT '{Infinity}'::real[]::sparsevec;
SELECT '{-Infinity}'::real[]::sparsevec;
SELECT '{}'::real[]::sparsevec;
SELECT '{{1}}'::real[]::sparsevec;
SELECT array_agg(n)::vector FROM generate_series(1, 16001) n;
SELECT array_to_vector(array_agg(n), 16001, false) FROM generate_series(1, 16001) n;

View File

@@ -28,21 +28,6 @@ SELECT * FROM t2 ORDER BY val;
DROP TABLE t;
DROP TABLE t2;
-- intvec
CREATE TABLE t (val intvec(3));
INSERT INTO t (val) VALUES ('[0,0,0]'), ('[1,2,3]'), ('[1,1,1]'), (NULL);
CREATE TABLE t2 (val intvec(3));
\copy t TO 'results/intvec.bin' WITH (FORMAT binary)
\copy t2 FROM 'results/intvec.bin' WITH (FORMAT binary)
SELECT * FROM t2 ORDER BY val;
DROP TABLE t;
DROP TABLE t2;
-- sparsevec
CREATE TABLE t (val sparsevec(3));

View File

@@ -1,58 +0,0 @@
SET enable_seqscan = off;
-- L2
CREATE TABLE t (val intvec(3));
INSERT INTO t (val) VALUES ('[0,0,0]'), ('[1,2,3]'), ('[1,1,1]'), (NULL);
CREATE INDEX ON t USING hnsw (val intvec_l2_ops);
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::intvec)) t2;
SELECT COUNT(*) FROM t;
TRUNCATE t;
SELECT * FROM t ORDER BY val <-> '[3,3,3]';
DROP TABLE t;
-- inner product
CREATE TABLE t (val intvec(3));
INSERT INTO t (val) VALUES ('[0,0,0]'), ('[1,2,3]'), ('[1,1,1]'), (NULL);
CREATE INDEX ON t USING hnsw (val intvec_ip_ops);
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::intvec)) t2;
DROP TABLE t;
-- cosine
CREATE TABLE t (val intvec(3));
INSERT INTO t (val) VALUES ('[0,0,0]'), ('[1,2,3]'), ('[1,1,1]'), (NULL);
CREATE INDEX ON t USING hnsw (val intvec_cosine_ops);
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::intvec)) t2;
DROP TABLE t;
-- L1
CREATE TABLE t (val intvec(3));
INSERT INTO t (val) VALUES ('[0,0,0]'), ('[1,2,3]'), ('[1,1,1]'), (NULL);
CREATE INDEX ON t USING hnsw (val intvec_l1_ops);
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::intvec)) t2;
DROP TABLE t;

View File

@@ -57,23 +57,6 @@ SELECT COUNT(*) FROM (SELECT * FROM t ORDER BY val <+> (SELECT NULL::vector)) t2
DROP TABLE t;
-- iterative
CREATE TABLE t (val vector(3));
INSERT INTO t (val) VALUES ('[0,0,0]'), ('[1,2,3]'), ('[1,1,1]'), (NULL);
CREATE INDEX ON t USING hnsw (val vector_l2_ops);
SET hnsw.iterative_search = strict_order;
SET hnsw.ef_search = 1;
SELECT * FROM t ORDER BY val <-> '[3,3,3]';
SET hnsw.iterative_search = relaxed_order;
SELECT * FROM t ORDER BY val <-> '[3,3,3]';
RESET hnsw.iterative_search;
RESET hnsw.ef_search;
DROP TABLE t;
-- unlogged
CREATE UNLOGGED TABLE t (val vector(3));
@@ -98,12 +81,4 @@ SHOW hnsw.ef_search;
SET hnsw.ef_search = 0;
SET hnsw.ef_search = 1001;
SHOW hnsw.iterative_search;
SET hnsw.iterative_search = on;
SHOW hnsw.max_search_tuples;
SET hnsw.max_search_tuples = -2;
DROP TABLE t;

View File

@@ -1,73 +0,0 @@
SELECT '[1,2,3]'::intvec;
SELECT '[-1,-2,-3]'::intvec;
SELECT ' [ 1, 2 , 3 ] '::intvec;
SELECT '[1.23456]'::intvec;
SELECT '[hello,1]'::intvec;
SELECT '[127,-128]'::intvec;
SELECT '[128,-129]'::intvec;
SELECT '[1,2,3'::intvec;
SELECT '[1,2,3]9'::intvec;
SELECT '1,2,3'::intvec;
SELECT ''::intvec;
SELECT '['::intvec;
SELECT '[,'::intvec;
SELECT '[]'::intvec;
SELECT '[1,]'::intvec;
SELECT '[1a]'::intvec;
SELECT '[1,,3]'::intvec;
SELECT '[1, ,3]'::intvec;
SELECT '[1,2,3]'::intvec(3);
SELECT '[1,2,3]'::intvec(2);
SELECT '[1,2,3]'::intvec(3, 2);
SELECT '[1,2,3]'::intvec('a');
SELECT '[1,2,3]'::intvec(0);
SELECT '[1,2,3]'::intvec(16001);
SELECT unnest('{"[1,2,3]", "[4,5,6]"}'::intvec[]);
SELECT '{"[1,2,3]"}'::intvec(2)[];
SELECT '[1,2,3]'::intvec < '[1,2,3]';
SELECT '[1,2,3]'::intvec < '[1,2]';
SELECT '[1,2,3]'::intvec <= '[1,2,3]';
SELECT '[1,2,3]'::intvec <= '[1,2]';
SELECT '[1,2,3]'::intvec = '[1,2,3]';
SELECT '[1,2,3]'::intvec = '[1,2]';
SELECT '[1,2,3]'::intvec != '[1,2,3]';
SELECT '[1,2,3]'::intvec != '[1,2]';
SELECT '[1,2,3]'::intvec >= '[1,2,3]';
SELECT '[1,2,3]'::intvec >= '[1,2]';
SELECT '[1,2,3]'::intvec > '[1,2,3]';
SELECT '[1,2,3]'::intvec > '[1,2]';
SELECT intvec_cmp('[1,2,3]', '[1,2,3]');
SELECT intvec_cmp('[1,2,3]', '[0,0,0]');
SELECT intvec_cmp('[0,0,0]', '[1,2,3]');
SELECT intvec_cmp('[1,2]', '[1,2,3]');
SELECT intvec_cmp('[1,2,3]', '[1,2]');
SELECT intvec_cmp('[1,2]', '[2,3,4]');
SELECT intvec_cmp('[2,3]', '[1,2,3]');
SELECT vector_dims('[1,2,3]'::intvec);
SELECT l2_distance('[0,0]'::intvec, '[3,4]');
SELECT l2_distance('[0,0]'::intvec, '[0,1]');
SELECT l2_distance('[1,2]'::intvec, '[3]');
SELECT '[0,0]'::intvec <-> '[3,4]';
SELECT inner_product('[1,2]'::intvec, '[3,4]');
SELECT inner_product('[1,2]'::intvec, '[3]');
SELECT inner_product('[127]'::intvec, '[127]');
SELECT '[1,2]'::intvec <#> '[3,4]';
SELECT cosine_distance('[1,2]'::intvec, '[2,4]');
SELECT cosine_distance('[1,2]'::intvec, '[0,0]');
SELECT cosine_distance('[1,1]'::intvec, '[1,1]');
SELECT cosine_distance('[1,0]'::intvec, '[0,2]');
SELECT cosine_distance('[1,1]'::intvec, '[-1,-1]');
SELECT cosine_distance('[1,2]'::intvec, '[3]');
SELECT '[1,2]'::intvec <=> '[2,4]';
SELECT l1_distance('[0,0]'::intvec, '[3,4]');
SELECT l1_distance('[0,0]'::intvec, '[0,1]');
SELECT l1_distance('[1,2]'::intvec, '[3]');

View File

@@ -44,28 +44,6 @@ SELECT COUNT(*) FROM (SELECT * FROM t ORDER BY val <=> (SELECT NULL::vector)) t2
DROP TABLE t;
-- iterative
CREATE TABLE t (val vector(3));
INSERT INTO t (val) VALUES ('[0,0,0]'), ('[1,2,3]'), ('[1,1,1]'), (NULL);
CREATE INDEX ON t USING ivfflat (val vector_l2_ops) WITH (lists = 3);
SET ivfflat.iterative_search = relaxed_order;
SELECT * FROM t ORDER BY val <-> '[3,3,3]';
SET ivfflat.max_probes = 0;
SELECT * FROM t ORDER BY val <-> '[3,3,3]';
SET ivfflat.max_probes = 1;
SELECT * FROM t ORDER BY val <-> '[3,3,3]';
SET ivfflat.max_probes = 2;
SELECT * FROM t ORDER BY val <-> '[3,3,3]';
RESET ivfflat.iterative_search;
RESET ivfflat.max_probes;
DROP TABLE t;
-- unlogged
CREATE UNLOGGED TABLE t (val vector(3));
@@ -84,16 +62,4 @@ CREATE INDEX ON t USING ivfflat (val vector_l2_ops) WITH (lists = 32769);
SHOW ivfflat.probes;
SET ivfflat.probes = 0;
SET ivfflat.probes = 32769;
SHOW ivfflat.iterative_search;
SET ivfflat.iterative_search = on;
SHOW ivfflat.max_probes;
SET ivfflat.max_probes = -2;
SET ivfflat.max_probes = 32769;
DROP TABLE t;

View File

@@ -2,9 +2,9 @@
# Test generic xlog record work for ivfflat index replication.
use strict;
use warnings FATAL => 'all';
use PostgreSQL::Test::Cluster;
use PostgreSQL::Test::Utils;
use warnings;
use PostgresNode;
use TestLib;
use Test::More;
my $dim = 32;
@@ -49,7 +49,7 @@ sub test_index_replay
my $array_sql = join(",", ('random()') x $dim);
# Initialize primary node
$node_primary = PostgreSQL::Test::Cluster->new('primary');
$node_primary = get_new_node('primary');
$node_primary->init(allows_streaming => 1);
if ($dim > 32)
{
@@ -67,7 +67,7 @@ my $backup_name = 'my_backup';
$node_primary->backup($backup_name);
# Create streaming replica linking to primary
$node_replica = PostgreSQL::Test::Cluster->new('replica');
$node_replica = get_new_node('replica');
$node_replica->init_from_backup($node_primary, $backup_name, has_streaming => 1);
$node_replica->start;

View File

@@ -1,15 +1,21 @@
use strict;
use warnings FATAL => 'all';
use PostgreSQL::Test::Cluster;
use PostgreSQL::Test::Utils;
use warnings;
use PostgresNode;
use TestLib;
use Test::More;
my $dim = 3;
my $array_sql = join(",", ('random()') x $dim);
my @r = ();
for (1 .. $dim)
{
my $v = int(rand(1000)) + 1;
push(@r, "i % $v");
}
my $array_sql = join(", ", @r);
# Initialize node
my $node = PostgreSQL::Test::Cluster->new('node');
my $node = get_new_node('node');
$node->init;
$node->start;
@@ -17,20 +23,19 @@ $node->start;
$node->safe_psql("postgres", "CREATE EXTENSION vector;");
$node->safe_psql("postgres", "CREATE TABLE tst (i int4, v vector($dim));");
$node->safe_psql("postgres",
"INSERT INTO tst SELECT i, ARRAY[$array_sql] FROM generate_series(1, 100000) i;"
"INSERT INTO tst SELECT i % 10, ARRAY[$array_sql] FROM generate_series(1, 100000) i;"
);
$node->safe_psql("postgres", "CREATE INDEX ON tst USING ivfflat (v vector_l2_ops);");
# Get size
my $size = $node->safe_psql("postgres", "SELECT pg_total_relation_size('tst_v_idx');");
# Store values
$node->safe_psql("postgres", "CREATE TABLE tmp AS SELECT * FROM tst;");
# Delete all, vacuum, and insert same data
$node->safe_psql("postgres", "DELETE FROM tst;");
$node->safe_psql("postgres", "VACUUM tst;");
$node->safe_psql("postgres", "INSERT INTO tst SELECT * FROM tmp;");
$node->safe_psql("postgres",
"INSERT INTO tst SELECT i % 10, ARRAY[$array_sql] FROM generate_series(1, 100000) i;"
);
# Check size
my $new_size = $node->safe_psql("postgres", "SELECT pg_total_relation_size('tst_v_idx');");

View File

@@ -1,7 +1,7 @@
use strict;
use warnings FATAL => 'all';
use PostgreSQL::Test::Cluster;
use PostgreSQL::Test::Utils;
use warnings;
use PostgresNode;
use TestLib;
use Test::More;
my $node;
@@ -49,7 +49,7 @@ sub test_recall
}
# Initialize node
$node = PostgreSQL::Test::Cluster->new('node');
$node = get_new_node('node');
$node->init;
$node->start;

View File

@@ -1,7 +1,7 @@
use strict;
use warnings FATAL => 'all';
use PostgreSQL::Test::Cluster;
use PostgreSQL::Test::Utils;
use warnings;
use PostgresNode;
use TestLib;
use Test::More;
my $node;
@@ -48,7 +48,7 @@ sub test_recall
}
# Initialize node
$node = PostgreSQL::Test::Cluster->new('node');
$node = get_new_node('node');
$node->init;
$node->start;

View File

@@ -1,11 +1,11 @@
use strict;
use warnings FATAL => 'all';
use PostgreSQL::Test::Cluster;
use PostgreSQL::Test::Utils;
use warnings;
use PostgresNode;
use TestLib;
use Test::More;
# Initialize node
my $node = PostgreSQL::Test::Cluster->new('node');
my $node = get_new_node('node');
$node->init;
$node->start;

View File

@@ -1,11 +1,11 @@
use strict;
use warnings FATAL => 'all';
use PostgreSQL::Test::Cluster;
use PostgreSQL::Test::Utils;
use warnings;
use PostgresNode;
use TestLib;
use Test::More;
# Initialize node
my $node = PostgreSQL::Test::Cluster->new('node');
my $node = get_new_node('node');
$node->init;
$node->start;

View File

@@ -1,7 +1,7 @@
use strict;
use warnings FATAL => 'all';
use PostgreSQL::Test::Cluster;
use PostgreSQL::Test::Utils;
use warnings;
use PostgresNode;
use TestLib;
use Test::More;
my $dim = 768;
@@ -9,7 +9,7 @@ my $dim = 768;
my $array_sql = join(",", ('random()') x $dim);
# Initialize node
my $node = PostgreSQL::Test::Cluster->new('node');
my $node = get_new_node('node');
$node->init;
$node->start;

View File

@@ -1,11 +1,11 @@
use strict;
use warnings FATAL => 'all';
use PostgreSQL::Test::Cluster;
use PostgreSQL::Test::Utils;
use warnings;
use PostgresNode;
use TestLib;
use Test::More;
# Initialize node
my $node = PostgreSQL::Test::Cluster->new('node');
my $node = get_new_node('node');
$node->init;
$node->start;

View File

@@ -1,7 +1,7 @@
use strict;
use warnings FATAL => 'all';
use PostgreSQL::Test::Cluster;
use PostgreSQL::Test::Utils;
use warnings;
use PostgresNode;
use TestLib;
use Test::More;
my $dim = 3;
@@ -11,7 +11,7 @@ my $limit = 20;
my $array_sql = join(",", ('random()') x $dim);
# Initialize node
my $node = PostgreSQL::Test::Cluster->new('node');
my $node = get_new_node('node');
$node->init;
$node->start;
@@ -94,7 +94,8 @@ like($explain, qr/Seq Scan/);
$explain = $node->safe_psql("postgres", qq(
EXPLAIN ANALYZE SELECT i FROM tst WHERE v <-> '$query' < 1 ORDER BY v <-> '$query';
));
like($explain, qr/Seq Scan/);
# TODO Do not use index
like($explain, qr/Index Scan using idx/);
# Test attribute index
$node->safe_psql("postgres", "CREATE INDEX attribute_idx ON tst (c);");
@@ -109,6 +110,7 @@ $node->safe_psql("postgres", "CREATE INDEX partial_idx ON tst USING ivfflat (v v
$explain = $node->safe_psql("postgres", qq(
EXPLAIN ANALYZE SELECT i FROM tst WHERE c = $c ORDER BY v <-> '$query' LIMIT $limit;
));
like($explain, qr/Index Scan using partial_idx/);
# TODO Use partial index
like($explain, qr/Index Scan using idx/);
done_testing();

View File

@@ -2,9 +2,9 @@
# Test generic xlog record work for hnsw index replication.
use strict;
use warnings FATAL => 'all';
use PostgreSQL::Test::Cluster;
use PostgreSQL::Test::Utils;
use warnings;
use PostgresNode;
use TestLib;
use Test::More;
my $dim = 32;
@@ -49,7 +49,7 @@ sub test_index_replay
my $array_sql = join(",", ('random()') x $dim);
# Initialize primary node
$node_primary = PostgreSQL::Test::Cluster->new('primary');
$node_primary = get_new_node('primary');
$node_primary->init(allows_streaming => 1);
if ($dim > 32)
{
@@ -67,7 +67,7 @@ my $backup_name = 'my_backup';
$node_primary->backup($backup_name);
# Create streaming replica linking to primary
$node_replica = PostgreSQL::Test::Cluster->new('replica');
$node_replica = get_new_node('replica');
$node_replica->init_from_backup($node_primary, $backup_name, has_streaming => 1);
$node_replica->start;

View File

@@ -1,7 +1,7 @@
use strict;
use warnings FATAL => 'all';
use PostgreSQL::Test::Cluster;
use PostgreSQL::Test::Utils;
use warnings;
use PostgresNode;
use TestLib;
use Test::More;
my $dim = 3;
@@ -15,7 +15,7 @@ for (1 .. $dim)
my $array_sql = join(", ", @r);
# Initialize node
my $node = PostgreSQL::Test::Cluster->new('node');
my $node = get_new_node('node');
$node->init;
$node->start;

View File

@@ -1,7 +1,7 @@
use strict;
use warnings FATAL => 'all';
use PostgreSQL::Test::Cluster;
use PostgreSQL::Test::Utils;
use warnings;
use PostgresNode;
use TestLib;
use Test::More;
my $node;
@@ -47,7 +47,7 @@ sub test_recall
}
# Initialize node
$node = PostgreSQL::Test::Cluster->new('node');
$node = get_new_node('node');
$node->init;
$node->start;

View File

@@ -1,7 +1,7 @@
use strict;
use warnings FATAL => 'all';
use PostgreSQL::Test::Cluster;
use PostgreSQL::Test::Utils;
use warnings;
use PostgresNode;
use TestLib;
use Test::More;
my $node;
@@ -47,7 +47,7 @@ sub test_recall
}
# Initialize node
$node = PostgreSQL::Test::Cluster->new('node');
$node = get_new_node('node');
$node->init;
$node->start;

View File

@@ -1,7 +1,7 @@
use strict;
use warnings FATAL => 'all';
use PostgreSQL::Test::Cluster;
use PostgreSQL::Test::Utils;
use warnings;
use PostgresNode;
use TestLib;
use Test::More;
my $node;
@@ -48,7 +48,7 @@ sub test_recall
}
# Initialize node
$node = PostgreSQL::Test::Cluster->new('node');
$node = get_new_node('node');
$node->init;
$node->start;

View File

@@ -1,11 +1,11 @@
use strict;
use warnings FATAL => 'all';
use PostgreSQL::Test::Cluster;
use PostgreSQL::Test::Utils;
use warnings;
use PostgresNode;
use TestLib;
use Test::More;
# Initialize node
my $node = PostgreSQL::Test::Cluster->new('node');
my $node = get_new_node('node');
$node->init;
$node->start;

View File

@@ -1,7 +1,7 @@
use strict;
use warnings FATAL => 'all';
use PostgreSQL::Test::Cluster;
use PostgreSQL::Test::Utils;
use warnings;
use PostgresNode;
use TestLib;
use Test::More;
# Ensures elements and neighbors on both same and different pages
@@ -10,7 +10,7 @@ my $dim = 1900;
my $array_sql = join(",", ('random()') x $dim);
# Initialize node
my $node = PostgreSQL::Test::Cluster->new('node');
my $node = get_new_node('node');
$node->init;
$node->start;

View File

@@ -1,7 +1,7 @@
use strict;
use warnings FATAL => 'all';
use PostgreSQL::Test::Cluster;
use PostgreSQL::Test::Utils;
use warnings;
use PostgresNode;
use TestLib;
use Test::More;
my $dim = 3;
@@ -11,20 +11,16 @@ my $limit = 20;
my $array_sql = join(",", ('random()') x $dim);
# Initialize node
my $node = PostgreSQL::Test::Cluster->new('node');
my $node = get_new_node('node');
$node->init;
$node->start;
# Create table and index
$node->safe_psql("postgres", "CREATE EXTENSION vector;");
$node->safe_psql("postgres", "CREATE TABLE tst (i int4, v vector($dim), c int4, t text);");
$node->safe_psql("postgres", "CREATE TABLE cat (i int4 PRIMARY KEY, t text, b boolean);");
$node->safe_psql("postgres",
"INSERT INTO tst SELECT i, ARRAY[$array_sql], i % $nc, 'test ' || i FROM generate_series(1, 10000) i;"
);
$node->safe_psql("postgres",
"INSERT INTO cat SELECT i, 'cat ' || i, i % 5 = 0 FROM generate_series(1, $nc) i;"
);
$node->safe_psql("postgres", "CREATE INDEX idx ON tst USING hnsw (v vector_l2_ops);");
$node->safe_psql("postgres", "ANALYZE tst;");
@@ -41,7 +37,8 @@ my $c = int(rand() * $nc);
my $explain = $node->safe_psql("postgres", qq(
EXPLAIN ANALYZE SELECT i FROM tst WHERE c = $c ORDER BY v <-> '$query' LIMIT $limit;
));
like($explain, qr/Seq Scan/);
# TODO Do not use index
like($explain, qr/Index Scan using idx/);
# Test attribute filtering with few rows removed
$explain = $node->safe_psql("postgres", qq(
@@ -59,7 +56,8 @@ like($explain, qr/Index Scan using idx/);
$explain = $node->safe_psql("postgres", qq(
EXPLAIN ANALYZE SELECT i FROM tst WHERE c < 1 ORDER BY v <-> '$query' LIMIT $limit;
));
like($explain, qr/Seq Scan/);
# TODO Do not use index
like($explain, qr/Index Scan using idx/);
# Test attribute filtering with few rows removed like
$explain = $node->safe_psql("postgres", qq(
@@ -98,25 +96,13 @@ $explain = $node->safe_psql("postgres", qq(
));
like($explain, qr/Seq Scan/);
# Test join
$explain = $node->safe_psql("postgres", qq(
EXPLAIN ANALYZE SELECT cat.t FROM cat INNER JOIN tst ON cat.i = tst.c ORDER BY v <-> '$query' LIMIT $limit;
));
like($explain, qr/Index Scan using idx/);
# Test join with attribute filtering
$explain = $node->safe_psql("postgres", qq(
EXPLAIN ANALYZE SELECT cat.t FROM cat INNER JOIN tst ON cat.i = tst.c WHERE cat.b = 't' ORDER BY v <-> '$query' LIMIT $limit;
));
like($explain, qr/Index Scan using idx/);
# Test attribute index
$node->safe_psql("postgres", "CREATE INDEX attribute_idx ON tst (c);");
$explain = $node->safe_psql("postgres", qq(
EXPLAIN ANALYZE SELECT i FROM tst WHERE c = $c ORDER BY v <-> '$query' LIMIT $limit;
));
# Use attribute index
like($explain, qr/Bitmap Index Scan on attribute_idx/);
# TODO Use attribute index
like($explain, qr/Index Scan using idx/);
# Test partial index
$node->safe_psql("postgres", "CREATE INDEX partial_idx ON tst USING hnsw (v vector_l2_ops) WHERE (c = $c);");

View File

@@ -1,11 +1,11 @@
use strict;
use warnings FATAL => 'all';
use PostgreSQL::Test::Cluster;
use PostgreSQL::Test::Utils;
use warnings;
use PostgresNode;
use TestLib;
use Test::More;
# Initialize node
my $node = PostgreSQL::Test::Cluster->new('node');
my $node = get_new_node('node');
$node->init;
$node->start;
@@ -53,7 +53,7 @@ sub test_aggregate
else
{
# Does not raise overflow error in this instance due to loss of precision
is($res, "[24576,24576,49152]");
is($res, "[24576,24576,49152]")
}
}

View File

@@ -1,13 +1,13 @@
use strict;
use warnings FATAL => 'all';
use PostgreSQL::Test::Cluster;
use PostgreSQL::Test::Utils;
use warnings;
use PostgresNode;
use TestLib;
use Test::More;
my $dim = 1024;
# Initialize node
my $node = PostgreSQL::Test::Cluster->new('node');
my $node = get_new_node('node');
$node->init;
$node->start;

View File

@@ -1,7 +1,7 @@
use strict;
use warnings FATAL => 'all';
use PostgreSQL::Test::Cluster;
use PostgreSQL::Test::Utils;
use warnings;
use PostgresNode;
use TestLib;
use Test::More;
my $node;
@@ -51,7 +51,7 @@ sub test_recall
}
# Initialize node
$node = PostgreSQL::Test::Cluster->new('node');
$node = get_new_node('node');
$node->init;
$node->start;

View File

@@ -1,7 +1,7 @@
use strict;
use warnings FATAL => 'all';
use PostgreSQL::Test::Cluster;
use PostgreSQL::Test::Utils;
use warnings;
use PostgresNode;
use TestLib;
use Test::More;
my $node;
@@ -51,7 +51,7 @@ sub test_recall
}
# Initialize node
$node = PostgreSQL::Test::Cluster->new('node');
$node = get_new_node('node');
$node->init;
$node->start;

View File

@@ -1,7 +1,7 @@
use strict;
use warnings FATAL => 'all';
use PostgreSQL::Test::Cluster;
use PostgreSQL::Test::Utils;
use warnings;
use PostgresNode;
use TestLib;
use Test::More;
my $node;
@@ -51,7 +51,7 @@ sub test_recall
}
# Initialize node
$node = PostgreSQL::Test::Cluster->new('node');
$node = get_new_node('node');
$node->init;
$node->start;

View File

@@ -1,11 +1,11 @@
use strict;
use warnings FATAL => 'all';
use PostgreSQL::Test::Cluster;
use PostgreSQL::Test::Utils;
use warnings;
use PostgresNode;
use TestLib;
use Test::More;
# Initialize node
my $node = PostgreSQL::Test::Cluster->new('node');
my $node = get_new_node('node');
$node->init;
$node->start;

View File

@@ -1,7 +1,7 @@
use strict;
use warnings FATAL => 'all';
use PostgreSQL::Test::Cluster;
use PostgreSQL::Test::Utils;
use warnings;
use PostgresNode;
use TestLib;
use Test::More;
my $node;
@@ -48,7 +48,7 @@ sub test_recall
}
# Initialize node
$node = PostgreSQL::Test::Cluster->new('node');
$node = get_new_node('node');
$node->init;
$node->start;

View File

@@ -1,7 +1,7 @@
use strict;
use warnings FATAL => 'all';
use PostgreSQL::Test::Cluster;
use PostgreSQL::Test::Utils;
use warnings;
use PostgresNode;
use TestLib;
use Test::More;
my $node;
@@ -48,7 +48,7 @@ sub test_recall
}
# Initialize node
$node = PostgreSQL::Test::Cluster->new('node');
$node = get_new_node('node');
$node->init;
$node->start;

View File

@@ -1,7 +1,7 @@
use strict;
use warnings FATAL => 'all';
use PostgreSQL::Test::Cluster;
use PostgreSQL::Test::Utils;
use warnings;
use PostgresNode;
use TestLib;
use Test::More;
my $node;
@@ -48,7 +48,7 @@ sub test_recall
}
# Initialize node
$node = PostgreSQL::Test::Cluster->new('node');
$node = get_new_node('node');
$node->init;
$node->start;

View File

@@ -1,11 +1,11 @@
use strict;
use warnings FATAL => 'all';
use PostgreSQL::Test::Cluster;
use PostgreSQL::Test::Utils;
use warnings;
use PostgresNode;
use TestLib;
use Test::More;
# Initialize node
my $node = PostgreSQL::Test::Cluster->new('node');
my $node = get_new_node('node');
$node->init;
$node->start;

View File

@@ -1,7 +1,7 @@
use strict;
use warnings FATAL => 'all';
use PostgreSQL::Test::Cluster;
use PostgreSQL::Test::Utils;
use warnings;
use PostgresNode;
use TestLib;
use Test::More;
my $node;
@@ -47,7 +47,7 @@ sub test_recall
}
# Initialize node
$node = PostgreSQL::Test::Cluster->new('node');
$node = get_new_node('node');
$node->init;
$node->start;

View File

@@ -1,7 +1,7 @@
use strict;
use warnings FATAL => 'all';
use PostgreSQL::Test::Cluster;
use PostgreSQL::Test::Utils;
use warnings;
use PostgresNode;
use TestLib;
use Test::More;
my $node;
@@ -47,7 +47,7 @@ sub test_recall
}
# Initialize node
$node = PostgreSQL::Test::Cluster->new('node');
$node = get_new_node('node');
$node->init;
$node->start;

View File

@@ -1,7 +1,7 @@
use strict;
use warnings FATAL => 'all';
use PostgreSQL::Test::Cluster;
use PostgreSQL::Test::Utils;
use warnings;
use PostgresNode;
use TestLib;
use Test::More;
my $node;
@@ -48,7 +48,7 @@ sub test_recall
}
# Initialize node
$node = PostgreSQL::Test::Cluster->new('node');
$node = get_new_node('node');
$node->init;
$node->start;

View File

@@ -1,11 +1,11 @@
use strict;
use warnings FATAL => 'all';
use PostgreSQL::Test::Cluster;
use PostgreSQL::Test::Utils;
use warnings;
use PostgresNode;
use TestLib;
use Test::More;
# Initialize node
my $node = PostgreSQL::Test::Cluster->new('node');
my $node = get_new_node('node');
$node->init;
$node->start;

View File

@@ -1,7 +1,7 @@
use strict;
use warnings FATAL => 'all';
use PostgreSQL::Test::Cluster;
use PostgreSQL::Test::Utils;
use warnings;
use PostgresNode;
use TestLib;
use Test::More;
my $node;
@@ -51,7 +51,7 @@ sub test_recall
}
# Initialize node
$node = PostgreSQL::Test::Cluster->new('node');
$node = get_new_node('node');
$node->init;
$node->start;

View File

@@ -1,14 +1,14 @@
use strict;
use warnings FATAL => 'all';
use PostgreSQL::Test::Cluster;
use PostgreSQL::Test::Utils;
use warnings;
use PostgresNode;
use TestLib;
use Test::More;
my $node;
my $array_sql = join(",", ('floor(random() * 2)::int - 1') x 3);
# Initialize node
$node = PostgreSQL::Test::Cluster->new('node');
$node = get_new_node('node');
$node->init;
$node->start;
@@ -40,10 +40,6 @@ for (1 .. 50)
$actual = $node->safe_psql("postgres", "SELECT halfvec_cmp(v::halfvec, '$query'::real[]::halfvec) FROM tst");
is($expected, $actual);
# Test intvec
$actual = $node->safe_psql("postgres", "SELECT intvec_cmp(v::integer[]::intvec, '$query'::integer[]::intvec) FROM tst");
is($expected, $actual);
# Test sparsevec
$actual = $node->safe_psql("postgres", "SELECT sparsevec_cmp(v::vector::sparsevec, '$query'::real[]::vector::sparsevec) FROM tst");
is($expected, $actual);

View File

@@ -1,7 +1,7 @@
use strict;
use warnings FATAL => 'all';
use PostgreSQL::Test::Cluster;
use PostgreSQL::Test::Utils;
use warnings;
use PostgresNode;
use TestLib;
use Test::More;
my $node;
@@ -10,7 +10,7 @@ my $dim = 5;
my $array_sql = join(",", ('floor(random() * 4)::int - 2') x $dim);
# Initialize node
$node = PostgreSQL::Test::Cluster->new('node');
$node = get_new_node('node');
$node->init;
$node->start;
@@ -45,10 +45,6 @@ for my $function (@functions)
my $actual = $node->safe_psql("postgres", "SELECT $function(v::halfvec, '$query'::vector::halfvec) FROM tst");
is($expected, $actual, "halfvec $function");
# Test intvec
$actual = $node->safe_psql("postgres", "SELECT $function(v::real[]::integer[]::intvec, '$query'::vector::real[]::integer[]::intvec) FROM tst");
is($expected, $actual, "intvec $function");
# Test sparsevec
$actual = $node->safe_psql("postgres", "SELECT $function(v::sparsevec, '$query'::vector::sparsevec) FROM tst");
is($expected, $actual, "sparsevec $function");

View File

@@ -1,7 +1,7 @@
use strict;
use warnings FATAL => 'all';
use PostgreSQL::Test::Cluster;
use PostgreSQL::Test::Utils;
use warnings;
use PostgresNode;
use TestLib;
use Test::More;
my $node;
@@ -51,7 +51,7 @@ sub test_recall
}
# Initialize node
$node = PostgreSQL::Test::Cluster->new('node');
$node = get_new_node('node');
$node->init;
$node->start;

View File

@@ -1,11 +1,11 @@
use strict;
use warnings FATAL => 'all';
use PostgreSQL::Test::Cluster;
use PostgreSQL::Test::Utils;
use warnings;
use PostgresNode;
use TestLib;
use Test::More;
# Initialize node
my $node = PostgreSQL::Test::Cluster->new('node');
my $node = get_new_node('node');
$node->init;
$node->start;

View File

@@ -1,19 +1,19 @@
use strict;
use warnings FATAL => 'all';
use PostgreSQL::Test::Cluster;
use PostgreSQL::Test::Utils;
use warnings;
use PostgresNode;
use TestLib;
use Test::More;
# Initialize node
my $node = PostgreSQL::Test::Cluster->new('node');
my $node = get_new_node('node');
$node->init;
$node->start;
# Create extension
$node->safe_psql("postgres", "CREATE EXTENSION vector;");
my @types = ("vector", "halfvec", "intvec", "sparsevec");
my @inputs = ("[1.23,4.56,7.89]", "[1.23,4.56,7.89]", "[1,2,3]", "{1:1.23,2:4.56,3:7.89}/3");
my @types = ("vector", "halfvec", "sparsevec");
my @inputs = ("[1.23,4.56,7.89]", "[1.23,4.56,7.89]", "{1:1.23,2:4.56,3:7.89}/3");
my @subs = (" ", " ", ",", ":", "-", "1", "9", "\0", "2147483648", "-2147483649");
for my $i (0 .. $#types)

View File

@@ -1,42 +0,0 @@
use strict;
use warnings FATAL => 'all';
use PostgreSQL::Test::Cluster;
use PostgreSQL::Test::Utils;
use Test::More;
# 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 sparsevec(100000));");
$node->safe_psql("postgres", "CREATE INDEX ON tst USING hnsw (v sparsevec_l2_ops);");
for (1 .. 3)
{
for (1 .. 100)
{
my @elements;
my %indices;
for (1 .. int(rand() * 100))
{
my $index = int(rand() * (100000 - 1)) + 1;
if (!exists($indices{$index}))
{
my $value = rand();
push(@elements, "$index:$value");
$indices{$index} = 1;
}
}
my $embedding = "{" . join(",", @elements) . "}/100000";
$node->safe_psql("postgres", "INSERT INTO tst (v) VALUES ('$embedding');");
}
$node->safe_psql("postgres", "DELETE FROM tst WHERE i % 2 = 0;");
$node->safe_psql("postgres", "VACUUM tst;");
is(1, 1);
}
done_testing();

View File

@@ -1,60 +0,0 @@
use strict;
use warnings FATAL => 'all';
use PostgreSQL::Test::Cluster;
use PostgreSQL::Test::Utils;
use Test::More;
my @dims = (384, 1536);
my $limit = 10;
# Initialize node
my $node = PostgreSQL::Test::Cluster->new('node');
$node->init;
$node->start;
$node->safe_psql("postgres", "CREATE EXTENSION vector;");
for my $dim (@dims)
{
my $array_sql = join(",", ('random()') x $dim);
# Create table and index
$node->safe_psql("postgres", "CREATE TABLE tst (i int4, v vector($dim));");
$node->safe_psql("postgres",
"INSERT INTO tst SELECT i, ARRAY[$array_sql] FROM generate_series(1, 2000) i;"
);
$node->safe_psql("postgres", "CREATE INDEX idx ON tst USING hnsw (v vector_l2_ops);");
$node->safe_psql("postgres", "ANALYZE tst;");
# Generate query
my @r = ();
for (1 .. $dim)
{
push(@r, rand());
}
my $query = "[" . join(",", @r) . "]";
my $explain = $node->safe_psql("postgres", qq(
EXPLAIN ANALYZE SELECT i FROM tst ORDER BY v <-> '$query' LIMIT $limit;
));
like($explain, qr/Index Scan using idx/);
# 3x the rows are needed for distance filters
# since the planner uses DEFAULT_INEQ_SEL for the selectivity (should be 1)
# Recreate index for performance
$node->safe_psql("postgres", "DROP INDEX idx;");
$node->safe_psql("postgres",
"INSERT INTO tst SELECT i, ARRAY[$array_sql] FROM generate_series(2001, 6000) i;"
);
$node->safe_psql("postgres", "CREATE INDEX idx ON tst USING hnsw (v vector_l2_ops);");
$node->safe_psql("postgres", "ANALYZE tst;");
$explain = $node->safe_psql("postgres", qq(
EXPLAIN ANALYZE SELECT i FROM tst WHERE v <-> '$query' < 1 ORDER BY v <-> '$query' LIMIT $limit;
));
like($explain, qr/Index Scan using idx/);
$node->safe_psql("postgres", "DROP TABLE tst;");
}
done_testing();

View File

@@ -1,50 +0,0 @@
use strict;
use warnings FATAL => 'all';
use PostgreSQL::Test::Cluster;
use PostgreSQL::Test::Utils;
use Test::More;
my @dims = (384, 1536);
my $limit = 10;
# Initialize node
my $node = PostgreSQL::Test::Cluster->new('node');
$node->init;
$node->start;
$node->safe_psql("postgres", "CREATE EXTENSION vector;");
for my $dim (@dims)
{
my $array_sql = join(",", ('random()') x $dim);
# Create table and index
$node->safe_psql("postgres", "CREATE TABLE tst (i int4, v vector($dim));");
$node->safe_psql("postgres",
"INSERT INTO tst SELECT i, ARRAY[$array_sql] FROM generate_series(1, 5000) i;"
);
$node->safe_psql("postgres", "CREATE INDEX idx ON tst USING ivfflat (v vector_l2_ops) WITH (lists = 5);");
$node->safe_psql("postgres", "ANALYZE tst;");
# Generate query
my @r = ();
for (1 .. $dim)
{
push(@r, rand());
}
my $query = "[" . join(",", @r) . "]";
my $explain = $node->safe_psql("postgres", qq(
EXPLAIN ANALYZE SELECT i FROM tst ORDER BY v <-> '$query' LIMIT $limit;
));
like($explain, qr/Index Scan using idx/);
$explain = $node->safe_psql("postgres", qq(
EXPLAIN ANALYZE SELECT i FROM tst WHERE v <-> '$query' < 1 ORDER BY v <-> '$query' LIMIT $limit;
));
like($explain, qr/Index Scan using idx/);
$node->safe_psql("postgres", "DROP TABLE tst;");
}
done_testing();

View File

@@ -1,54 +0,0 @@
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
$node->safe_psql("postgres", "CREATE EXTENSION vector;");
$node->safe_psql("postgres", "CREATE TABLE tst (i int4 PRIMARY KEY, v vector($dim));");
$node->safe_psql("postgres",
"INSERT INTO tst SELECT i, ARRAY[$array_sql] FROM generate_series(1, 100000) i;"
);
$node->safe_psql("postgres", "CREATE INDEX ON tst USING ivfflat (v vector_l2_ops);");
my $count = $node->safe_psql("postgres", qq(
SET enable_seqscan = off;
SET ivfflat.probes = 10;
SET ivfflat.iterative_search = relaxed_order;
SELECT COUNT(*) FROM (SELECT v FROM tst WHERE i % 10000 = 0 ORDER BY v <-> (SELECT v FROM tst LIMIT 1) LIMIT 11) t;
));
is($count, 10);
foreach ((30, 50, 70))
{
my $max_probes = $_;
my $expected = $max_probes / 10;
my $sum = 0;
for my $i (1 .. 20)
{
$count = $node->safe_psql("postgres", qq(
SET enable_seqscan = off;
SET ivfflat.probes = 10;
SET ivfflat.iterative_search = relaxed_order;
SET ivfflat.max_probes = $max_probes;
SELECT COUNT(*) FROM (SELECT v FROM tst WHERE i % 10000 = 0 ORDER BY v <-> (SELECT v FROM tst WHERE i = $i) LIMIT 11) t;
));
$sum += $count;
}
my $avg = $sum / 20;
cmp_ok($avg, '>', $expected - 2);
cmp_ok($avg, '<', $expected + 2);
}
done_testing();

View File

@@ -1,125 +0,0 @@
use strict;
use warnings FATAL => 'all';
use PostgreSQL::Test::Cluster;
use PostgreSQL::Test::Utils;
use Test::More;
my $node;
my @queries = ();
my @expected;
my $limit = 20;
my @cs = (100, 1000);
sub test_recall
{
my ($c, $probes, $min, $operator) = @_;
my $correct = 0;
my $total = 0;
my $explain = $node->safe_psql("postgres", qq(
SET enable_seqscan = off;
SET ivfflat.probes = $probes;
SET ivfflat.iterative_search = relaxed_order;
EXPLAIN ANALYZE SELECT i FROM tst WHERE i % $c = 0 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;
SET ivfflat.iterative_search = relaxed_order;
SELECT i FROM tst WHERE i % $c = 0 ORDER BY v $operator '$queries[$i]' LIMIT $limit;
));
my @actual_ids = split("\n", $actual);
my @expected_ids = split("\n", $expected[$i]);
my %expected_set = map { $_ => 1 } @expected_ids;
foreach (@actual_ids)
{
if (exists($expected_set{$_}))
{
$correct++;
}
}
$total += $limit;
}
cmp_ok($correct / $total, ">=", $min, "$operator $c");
}
# Initialize node
$node = PostgreSQL::Test::Cluster->new('node');
$node->init;
$node->start;
# Create table
$node->safe_psql("postgres", "CREATE EXTENSION vector;");
$node->safe_psql("postgres", "CREATE TABLE tst (i int4, v vector(3));");
$node->safe_psql("postgres",
"INSERT INTO tst SELECT i, ARRAY[random(), random(), random()] FROM generate_series(1, 100000) i;"
);
# Generate queries
for (1 .. 20)
{
my $r1 = rand();
my $r2 = rand();
my $r3 = rand();
push(@queries, "[$r1,$r2,$r3]");
}
# Check each index type
my @operators = ("<->", "<=>");
my @opclasses = ("vector_l2_ops", "vector_cosine_ops");
for my $i (0 .. $#operators)
{
my $operator = $operators[$i];
my $opclass = $opclasses[$i];
$node->safe_psql("postgres", "CREATE INDEX idx ON tst USING ivfflat (v $opclass);");
foreach (@cs)
{
my $c = $_;
# Get exact results
@expected = ();
foreach (@queries)
{
my $res = $node->safe_psql("postgres", qq(
SET enable_indexscan = off;
WITH top AS (
SELECT v $operator '$_' AS distance FROM tst WHERE i % $c = 0 ORDER BY distance LIMIT $limit
)
SELECT i FROM tst WHERE (v $operator '$_') <= (SELECT MAX(distance) FROM top)
));
push(@expected, $res);
}
if ($c == 100)
{
test_recall($c, 1, 0.57, $operator);
test_recall($c, 10, 0.98, $operator);
}
else
{
if ($operator eq "<->")
{
test_recall($c, 1, 0.80, $operator);
}
else
{
test_recall($c, 1, 0.88, $operator);
}
}
}
$node->safe_psql("postgres", "DROP INDEX idx;");
}
done_testing();

View File

@@ -1,67 +0,0 @@
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
$node->safe_psql("postgres", "CREATE EXTENSION vector;");
$node->safe_psql("postgres", "CREATE TABLE tst (i int4 PRIMARY KEY, v vector($dim));");
$node->safe_psql("postgres",
"INSERT INTO tst SELECT i, ARRAY[$array_sql] FROM generate_series(1, 100000) i;"
);
$node->safe_psql("postgres", qq(
SET maintenance_work_mem = '128MB';
SET max_parallel_maintenance_workers = 2;
CREATE INDEX ON tst USING hnsw (v vector_l2_ops)
));
my $count = $node->safe_psql("postgres", qq(
SET enable_seqscan = off;
SET hnsw.iterative_search = relaxed_order;
SET work_mem = '8MB';
SELECT COUNT(*) FROM (SELECT v FROM tst WHERE i % 10000 = 0 ORDER BY v <-> (SELECT v FROM tst LIMIT 1) LIMIT 11) t;
));
is($count, 10);
foreach ((30000, 50000, 70000))
{
my $max_tuples = $_;
my $expected = $max_tuples / 10000;
my $sum = 0;
for my $i (1 .. 20)
{
$count = $node->safe_psql("postgres", qq(
SET enable_seqscan = off;
SET hnsw.iterative_search = relaxed_order;
SET hnsw.max_search_tuples = $max_tuples;
SET work_mem = '8MB';
SELECT COUNT(*) FROM (SELECT v FROM tst WHERE i % 10000 = 0 ORDER BY v <-> (SELECT v FROM tst WHERE i = $i) LIMIT 11) t;
));
$sum += $count;
}
my $avg = $sum / 20;
cmp_ok($avg, '>', $expected - 2);
cmp_ok($avg, '<', $expected + 2);
}
my ($ret, $stdout, $stderr) = $node->psql("postgres", qq(
SET enable_seqscan = off;
SET hnsw.iterative_search = relaxed_order;
SET client_min_messages = debug1;
SET work_mem = '2MB';
SELECT COUNT(*) FROM (SELECT v FROM tst WHERE i % 10000 = 0 ORDER BY v <-> (SELECT v FROM tst LIMIT 1) LIMIT 11) t;
));
like($stderr, qr/hnsw index scan exceeded work_mem after \d+ tuples/);
done_testing();

View File

@@ -1,118 +0,0 @@
use strict;
use warnings FATAL => 'all';
use PostgreSQL::Test::Cluster;
use PostgreSQL::Test::Utils;
use Test::More;
my $node;
my @queries = ();
my @expected;
my $limit = 20;
my $dim = 3;
my $array_sql = join(",", ('random()') x $dim);
my @cs = (50, 500);
sub test_recall
{
my ($c, $ef_search, $min, $operator, $mode) = @_;
my $correct = 0;
my $total = 0;
my $explain = $node->safe_psql("postgres", qq(
SET enable_seqscan = off;
SET hnsw.ef_search = $ef_search;
SET hnsw.iterative_search = $mode;
EXPLAIN ANALYZE SELECT i FROM tst WHERE i % $c = 0 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 hnsw.ef_search = $ef_search;
SET hnsw.iterative_search = $mode;
SELECT i FROM tst WHERE i % $c = 0 ORDER BY v $operator '$queries[$i]' LIMIT $limit;
));
my @actual_ids = split("\n", $actual);
my @expected_ids = split("\n", $expected[$i]);
my %expected_set = map { $_ => 1 } @expected_ids;
foreach (@actual_ids)
{
if (exists($expected_set{$_}))
{
$correct++;
}
}
$total += $limit;
}
cmp_ok($correct / $total, ">=", $min, "$operator $mode $c");
}
# Initialize node
$node = PostgreSQL::Test::Cluster->new('node');
$node->init;
$node->start;
# Create table
$node->safe_psql("postgres", "CREATE EXTENSION vector;");
$node->safe_psql("postgres", "CREATE TABLE tst (i int4, v vector($dim));");
$node->safe_psql("postgres",
"INSERT INTO tst SELECT i, ARRAY[$array_sql] FROM generate_series(1, 50000) 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 = ("vector_l2_ops", "vector_cosine_ops");
for my $i (0 .. $#operators)
{
my $operator = $operators[$i];
my $opclass = $opclasses[$i];
$node->safe_psql("postgres", qq(
SET maintenance_work_mem = '128MB';
CREATE INDEX idx ON tst USING hnsw (v $opclass);
));
foreach (@cs)
{
my $c = $_;
# Get exact results
@expected = ();
foreach (@queries)
{
my $res = $node->safe_psql("postgres", qq(
SET enable_indexscan = off;
WITH top AS (
SELECT v $operator '$_' AS distance FROM tst WHERE i % $c = 0 ORDER BY distance LIMIT $limit
)
SELECT i FROM tst WHERE (v $operator '$_') <= (SELECT MAX(distance) FROM top)
));
push(@expected, $res);
}
test_recall($c, 40, 0.99, $operator, "strict_order");
test_recall($c, 40, 0.99, $operator, "relaxed_order");
}
$node->safe_psql("postgres", "DROP INDEX idx;");
}
done_testing();

View File

@@ -1,132 +0,0 @@
use strict;
use warnings FATAL => 'all';
use PostgreSQL::Test::Cluster;
use PostgreSQL::Test::Utils;
use Test::More;
my $node;
my @queries = ();
my @expected;
my $limit = 20;
my $dim = 10;
my $array_sql = join(",", ('(random() * 255)::int - 128') x $dim);
sub test_recall
{
my ($min, $operator) = @_;
my $correct = 0;
my $total = 0;
my $explain = $node->safe_psql("postgres", qq(
SET enable_seqscan = off;
EXPLAIN ANALYZE SELECT i FROM tst ORDER BY v $operator '$queries[0]' LIMIT $limit;
));
like($explain, qr/Index Scan/);
for my $i (0 .. $#queries)
{
my $actual = $node->safe_psql("postgres", qq(
SET enable_seqscan = off;
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 = PostgreSQL::Test::Cluster->new('node');
$node->init;
$node->start;
# Create table
$node->safe_psql("postgres", "CREATE EXTENSION vector;");
$node->safe_psql("postgres", "CREATE TABLE tst (i int4, v intvec($dim));");
$node->safe_psql("postgres",
"INSERT INTO tst SELECT i, ARRAY[$array_sql] FROM generate_series(1, 10000) i;"
);
# Generate queries
for (1 .. 20)
{
my @r = ();
for (1 .. $dim)
{
push(@r, int(rand(256)) - 128);
}
push(@queries, "[" . join(",", @r) . "]");
}
# Check each index type
my @operators = ("<->", "<#>", "<=>", "<+>");
my @opclasses = ("intvec_l2_ops", "intvec_ip_ops", "intvec_cosine_ops", "intvec_l1_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 hnsw (v $opclass);
));
# Test approximate results
my $min = 0.98;
test_recall($min, $operator);
$node->safe_psql("postgres", "DROP INDEX idx;");
# Build index in parallel in memory
my ($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 hnsw (v $opclass);
));
is($ret, 0, $stderr);
like($stderr, qr/using \d+ parallel workers/);
# Test approximate results
test_recall($min, $operator);
$node->safe_psql("postgres", "DROP INDEX idx;");
# Build index in parallel on disk
# Set parallel_workers on table to use workers with low maintenance_work_mem
($ret, $stdout, $stderr) = $node->psql("postgres", qq(
ALTER TABLE tst SET (parallel_workers = 2);
SET client_min_messages = DEBUG;
SET maintenance_work_mem = '4MB';
CREATE INDEX idx ON tst USING hnsw (v $opclass);
ALTER TABLE tst RESET (parallel_workers);
));
is($ret, 0, $stderr);
like($stderr, qr/using \d+ parallel workers/);
like($stderr, qr/hnsw graph no longer fits into maintenance_work_mem/);
$node->safe_psql("postgres", "DROP INDEX idx;");
}
done_testing();

View File

@@ -1,113 +0,0 @@
use strict;
use warnings FATAL => 'all';
use PostgreSQL::Test::Cluster;
use PostgreSQL::Test::Utils;
use Test::More;
my $node;
my @queries = ();
my @expected;
my $limit = 20;
my $dim = 10;
my $array_sql = join(",", ('(random() * 255)::int - 128') x $dim);
sub test_recall
{
my ($min, $operator) = @_;
my $correct = 0;
my $total = 0;
my $explain = $node->safe_psql("postgres", qq(
SET enable_seqscan = off;
EXPLAIN ANALYZE SELECT i FROM tst ORDER BY v $operator '$queries[0]' LIMIT $limit;
));
like($explain, qr/Index Scan/);
for my $i (0 .. $#queries)
{
my $actual = $node->safe_psql("postgres", qq(
SET enable_seqscan = off;
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 = PostgreSQL::Test::Cluster->new('node');
$node->init;
$node->start;
# Create table
$node->safe_psql("postgres", "CREATE EXTENSION vector;");
$node->safe_psql("postgres", "CREATE TABLE tst (i serial, v intvec($dim));");
# Generate queries
for (1 .. 20)
{
my @r = ();
for (1 .. $dim)
{
push(@r, int(rand(256)) - 128);
}
push(@queries, "[" . join(",", @r) . "]");
}
# Check each index type
my @operators = ("<->", "<#>", "<=>", "<+>");
my @opclasses = ("intvec_l2_ops", "intvec_ip_ops", "intvec_cosine_ops", "intvec_l1_ops");
for my $i (0 .. $#operators)
{
my $operator = $operators[$i];
my $opclass = $opclasses[$i];
# Add index
$node->safe_psql("postgres", "CREATE INDEX idx ON tst USING hnsw (v $opclass);");
# Use concurrent inserts
$node->pgbench(
"--no-vacuum --client=10 --transactions=1000",
0,
[qr{actually processed}],
[qr{^$}],
"concurrent INSERTs",
{
"040_hnsw_intvec_insert_recall_$opclass" => "INSERT INTO tst (v) VALUES (ARRAY[$array_sql]);"
}
);
# Get exact results
@expected = ();
foreach (@queries)
{
my $res = $node->safe_psql("postgres", qq(
SET enable_indexscan = off;
SELECT i FROM tst ORDER BY v $operator '$_' LIMIT $limit;
));
push(@expected, $res);
}
# Test approximate results
my $min = 0.98;
test_recall($min, $operator);
$node->safe_psql("postgres", "DROP INDEX idx;");
$node->safe_psql("postgres", "TRUNCATE tst;");
}
done_testing();

View File

@@ -1,101 +0,0 @@
use strict;
use warnings FATAL => 'all';
use PostgreSQL::Test::Cluster;
use PostgreSQL::Test::Utils;
use Test::More;
my $node;
my @queries = ();
my @expected;
my $limit = 20;
my $dim = 10;
my $array_sql = join(",", ('(random() * 255)::int - 128') x $dim);
sub test_recall
{
my ($min, $ef_search, $test_name) = @_;
my $correct = 0;
my $total = 0;
my $explain = $node->safe_psql("postgres", qq(
SET enable_seqscan = off;
SET hnsw.ef_search = $ef_search;
EXPLAIN ANALYZE SELECT i FROM tst ORDER BY v <-> '$queries[0]' LIMIT $limit;
));
like($explain, qr/Index Scan/);
for my $i (0 .. $#queries)
{
my $actual = $node->safe_psql("postgres", qq(
SET enable_seqscan = off;
SET hnsw.ef_search = $ef_search;
SELECT i FROM tst ORDER BY v <-> '$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, $test_name);
}
# Initialize node
$node = PostgreSQL::Test::Cluster->new('node');
$node->init;
$node->start;
# Create table
$node->safe_psql("postgres", "CREATE EXTENSION vector;");
$node->safe_psql("postgres", "CREATE TABLE tst (i int4, v intvec($dim));");
$node->safe_psql("postgres", "ALTER TABLE tst SET (autovacuum_enabled = false);");
$node->safe_psql("postgres",
"INSERT INTO tst SELECT i, ARRAY[$array_sql] FROM generate_series(1, 10000) i;"
);
# Add index
$node->safe_psql("postgres", "CREATE INDEX ON tst USING hnsw (v intvec_l2_ops) WITH (m = 4, ef_construction = 8);");
# Delete data
$node->safe_psql("postgres", "DELETE FROM tst WHERE i > 2500;");
# Generate queries
for (1 .. 20)
{
my @r = ();
for (1 .. $dim)
{
push(@r, int(rand(256)) - 128);
}
push(@queries, "[" . join(",", @r) . "]");
}
# Get exact results
@expected = ();
foreach (@queries)
{
my $res = $node->safe_psql("postgres", qq(
SET enable_indexscan = off;
SELECT i FROM tst ORDER BY v <-> '$_' LIMIT $limit;
));
push(@expected, $res);
}
test_recall(0.18, $limit, "before vacuum");
test_recall(0.84, 100, "before vacuum");
# TODO Test concurrent inserts with vacuum
$node->safe_psql("postgres", "VACUUM tst;");
test_recall(0.84, $limit, "after vacuum");
done_testing();

Some files were not shown because too many files have changed in this diff Show More