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
79 changed files with 634 additions and 1280 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

View File

@@ -1,25 +1,4 @@
## 0.8.0 (unreleased)
- Added casts for arrays to `sparsevec`
- 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

@@ -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

@@ -1,9 +1,8 @@
EXTENSION = vector
EXTVERSION = 0.7.4
EXTVERSION = 0.7.0
MODULE_big = vector
DATA = $(wildcard sql/*--*--*.sql)
DATA_built = sql/$(EXTENSION)--$(EXTVERSION).sql
DATA = $(wildcard sql/*--*.sql)
OBJS = src/bitutils.o src/bitvec.o src/halfutils.o src/halfvec.o src/hnsw.o src/hnswbuild.o src/hnswinsert.o src/hnswscan.o src/hnswutils.o src/hnswvacuum.o src/ivfbuild.o src/ivfflat.o src/ivfinsert.o src/ivfkmeans.o src/ivfscan.o src/ivfutils.o src/ivfvacuum.o src/sparsevec.o src/vector.o
HEADERS = src/halfvec.h src/sparsevec.h src/vector.h
@@ -43,6 +42,8 @@ all: sql/$(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:

View File

@@ -1,7 +1,6 @@
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\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
@@ -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
@@ -46,7 +46,7 @@ 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
@@ -106,7 +106,7 @@ Insert vectors
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);
@@ -202,7 +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
@@ -473,7 +473,7 @@ SELECT * FROM items ORDER BY embedding::halfvec(3) <-> '[1,2,3]' LIMIT 5;
## 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));
@@ -551,7 +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
@@ -597,7 +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);
@@ -687,7 +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
@@ -1051,7 +1051,7 @@ This adds pgvector to the [Postgres image](https://hub.docker.com/_/postgres) (r
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=16 -t myuser/pgvector .
```

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

@@ -1,26 +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 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

@@ -782,18 +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)
@@ -811,18 +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))
@@ -180,7 +185,7 @@ float_underflow_error(void)
/*
* 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)
{
@@ -294,7 +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)
{
@@ -340,7 +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)
{
@@ -371,7 +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)
{
@@ -405,7 +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)
{
@@ -425,7 +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)
{
@@ -440,7 +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)
{
@@ -514,7 +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)
{
@@ -538,7 +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)
{
@@ -560,7 +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)
{
@@ -575,7 +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)
{
@@ -590,7 +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)
{
@@ -605,7 +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)
{
@@ -620,7 +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)
{
@@ -652,7 +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)
{
@@ -676,7 +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)
{
@@ -691,7 +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)
{
@@ -703,7 +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)
{
@@ -725,7 +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)
{
@@ -764,7 +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)
{
@@ -803,7 +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)
{
@@ -842,7 +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)
{
@@ -884,7 +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)
{
@@ -908,7 +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)
{
@@ -926,7 +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)
{
@@ -1000,7 +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)
{
@@ -1013,7 +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)
{
@@ -1026,7 +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)
{
@@ -1039,7 +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)
{
@@ -1052,7 +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)
{
@@ -1065,7 +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)
{
@@ -1078,7 +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)
{
@@ -1091,7 +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)
{
@@ -1152,7 +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)
{
@@ -1186,7 +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

@@ -9,7 +9,6 @@
#include "commands/vacuum.h"
#include "hnsw.h"
#include "miscadmin.h"
#include "utils/float.h"
#include "utils/guc.h"
#include "utils/selfuncs.h"
@@ -60,9 +59,17 @@ 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,
@@ -105,8 +112,8 @@ hnswcostestimate(PlannerInfo *root, IndexPath *path, double loop_count,
/* 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;
@@ -147,10 +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
}
/*
@@ -167,7 +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)
{
@@ -175,7 +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 */
@@ -188,24 +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)
@@ -155,13 +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;
float distance;
} HnswSearchCandidate;
pairingheap_node ph_node;
HnswCandidate *inner;
} HnswPairingHeapNode;
/* HNSW index options */
typedef struct HnswOptions
@@ -382,7 +385,7 @@ 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, FmgrInfo *procinfo, Oid collation, int m, int efConstruction, bool existing);
HnswSearchCandidate *HnswEntryCandidate(char *base, HnswElement em, Datum q, Relation rel, FmgrInfo *procinfo, Oid collation, bool loadVec);
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);

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);
@@ -371,13 +379,7 @@ UpdateNeighborsInMemory(char *base, FmgrInfo *procinfo, Oid collation, HnswEleme
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++)
{
@@ -387,6 +389,7 @@ UpdateNeighborsInMemory(char *base, FmgrInfo *procinfo, Oid collation, HnswEleme
/* 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, e, hc, lm, lc, NULL, NULL, procinfo, collation);
LWLockRelease(&neighborElement->lock);
@@ -575,13 +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;
@@ -644,7 +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;
}
@@ -680,25 +691,17 @@ 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;
@@ -1118,8 +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)
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;
@@ -76,25 +73,10 @@ 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;
@@ -202,7 +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))
if (HnswFreeOffset(index, buf, page, e, ntupSize, &nbuf, &npage, &freeOffno, &freeNeighborOffno, &newInsertPage))
{
if (nbuf != buf)
{
@@ -379,12 +361,8 @@ HnswUpdateNeighborsOnDisk(Relation index, FmgrInfo *procinfo, Oid collation, Hns
HnswElement neighborElement = HnswPtrAccess(base, hc->element);
OffsetNumber offno = neighborElement->neighborOffno;
/*
* 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.
*/
/* Get latest neighbors since they may have changed */
/* Do not lock yet since selecting neighbors can take time */
HnswLoadNeighbors(neighborElement, index, m);
/*

View File

@@ -160,15 +160,15 @@ hnswgettuple(IndexScanDesc scan, ScanDirection dir)
so->first = false;
#if defined(HNSW_MEMORY)
elog(INFO, "memory: %zu KB", MemoryContextMemAllocated(so->tmpCtx, false) / 1024);
#if defined(HNSW_MEMORY) && PG_VERSION_NUM >= 130000
elog(INFO, "memory: %zu MB", MemoryContextMemAllocated(so->tmpCtx, false) / (1024 * 1024));
#endif
}
while (list_length(so->w) > 0)
{
char *base = NULL;
HnswSearchCandidate *hc = llast(so->w);
HnswCandidate *hc = llast(so->w);
HnswElement element = HnswPtrAccess(base, hc->element);
ItemPointer heaptid;

View File

@@ -5,7 +5,6 @@
#include "access/generic_xlog.h"
#include "catalog/pg_type.h"
#include "catalog/pg_type_d.h"
#include "common/hashfn.h"
#include "fmgr.h"
#include "hnsw.h"
#include "lib/pairingheap.h"
@@ -15,6 +14,12 @@
#include "utils/memdebug.h"
#include "utils/rel.h"
#if PG_VERSION_NUM >= 130000
#include "common/hashfn.h"
#else
#include "utils/hashutils.h"
#endif
#if PG_VERSION_NUM < 170000
static inline uint64
murmurhash64(uint64 data)
@@ -107,12 +112,6 @@ typedef union
tidhash_hash *tids;
} visited_hash;
typedef union
{
HnswElement element;
ItemPointerData indextid;
} HnswUnvisited;
/*
* Get the max number of connections in an upper layer for each element in the index
*/
@@ -301,9 +300,6 @@ HnswGetMetaPageInfo(Relation index, int *m, HnswElement * entryPoint)
page = BufferGetPage(buf);
metap = HnswPageGetMeta(page);
if (unlikely(metap->magicNumber != HNSW_MAGIC_NUMBER))
elog(ERROR, "hnsw index is not valid");
if (m != NULL)
*m = metap->m;
@@ -548,19 +544,19 @@ HnswLoadElementFromTuple(HnswElement element, HnswElementTuple etup, bool loadHe
/*
* Load an element and optionally get its distance from q
*/
static void
HnswLoadElementImpl(BlockNumber blkno, OffsetNumber offno, float *distance, Datum *q, Relation index, FmgrInfo *procinfo, Oid collation, bool loadVec, float *maxDistance, HnswElement * element)
void
HnswLoadElement(HnswElement element, float *distance, Datum *q, Relation index, FmgrInfo *procinfo, Oid collation, bool loadVec, float *maxDistance)
{
Buffer buf;
Page page;
HnswElementTuple etup;
/* Read vector */
buf = ReadBuffer(index, blkno);
buf = ReadBuffer(index, element->blkno);
LockBuffer(buf, BUFFER_LOCK_SHARE);
page = BufferGetPage(buf);
etup = (HnswElementTuple) PageGetItem(page, PageGetItemId(page, offno));
etup = (HnswElementTuple) PageGetItem(page, PageGetItemId(page, element->offno));
Assert(HnswIsElementTuple(etup));
@@ -575,32 +571,19 @@ HnswLoadElementImpl(BlockNumber blkno, OffsetNumber offno, float *distance, Datu
/* Load element */
if (distance == NULL || maxDistance == NULL || *distance < *maxDistance)
{
if (*element == NULL)
*element = HnswInitElementFromBlock(blkno, offno);
HnswLoadElementFromTuple(*element, etup, true, loadVec);
}
HnswLoadElementFromTuple(element, etup, true, loadVec);
UnlockReleaseBuffer(buf);
}
/*
* Load an element and optionally get its distance from q
*/
void
HnswLoadElement(HnswElement element, float *distance, Datum *q, Relation index, FmgrInfo *procinfo, Oid collation, bool loadVec, float *maxDistance)
{
HnswLoadElementImpl(element->blkno, element->offno, distance, q, index, procinfo, collation, loadVec, maxDistance, &element);
}
/*
* Get the distance for an element
* Get the distance for a candidate
*/
static float
GetElementDistance(char *base, HnswElement element, Datum q, FmgrInfo *procinfo, Oid collation)
GetCandidateDistance(char *base, HnswCandidate * hc, Datum q, FmgrInfo *procinfo, Oid collation)
{
Datum value = HnswGetValue(base, element);
HnswElement hce = HnswPtrAccess(base, hc->element);
Datum value = HnswGetValue(base, hce);
return DatumGetFloat8(FunctionCall2Coll(procinfo, collation, q, value));
}
@@ -608,32 +591,29 @@ GetElementDistance(char *base, HnswElement element, Datum q, FmgrInfo *procinfo,
/*
* Create a candidate for the entry point
*/
HnswSearchCandidate *
HnswCandidate *
HnswEntryCandidate(char *base, HnswElement entryPoint, Datum q, Relation index, FmgrInfo *procinfo, Oid collation, bool loadVec)
{
HnswSearchCandidate *hc = palloc(sizeof(HnswSearchCandidate));
HnswCandidate *hc = palloc(sizeof(HnswCandidate));
HnswPtrStore(base, hc->element, entryPoint);
if (index == NULL)
hc->distance = GetElementDistance(base, entryPoint, q, procinfo, collation);
hc->distance = GetCandidateDistance(base, hc, q, procinfo, collation);
else
HnswLoadElement(entryPoint, &hc->distance, &q, index, procinfo, collation, loadVec, NULL);
return hc;
}
#define HnswGetSearchCandidate(membername, ptr) pairingheap_container(HnswSearchCandidate, membername, ptr)
#define HnswGetSearchCandidateConst(membername, ptr) pairingheap_const_container(HnswSearchCandidate, membername, ptr)
/*
* Compare candidate distances
*/
static int
CompareNearestCandidates(const pairingheap_node *a, const pairingheap_node *b, void *arg)
{
if (HnswGetSearchCandidateConst(c_node, a)->distance < HnswGetSearchCandidateConst(c_node, b)->distance)
if (((const HnswPairingHeapNode *) a)->inner->distance < ((const HnswPairingHeapNode *) b)->inner->distance)
return 1;
if (HnswGetSearchCandidateConst(c_node, a)->distance > HnswGetSearchCandidateConst(c_node, b)->distance)
if (((const HnswPairingHeapNode *) a)->inner->distance > ((const HnswPairingHeapNode *) b)->inner->distance)
return -1;
return 0;
@@ -645,15 +625,27 @@ CompareNearestCandidates(const pairingheap_node *a, const pairingheap_node *b, v
static int
CompareFurthestCandidates(const pairingheap_node *a, const pairingheap_node *b, void *arg)
{
if (HnswGetSearchCandidateConst(w_node, a)->distance < HnswGetSearchCandidateConst(w_node, b)->distance)
if (((const HnswPairingHeapNode *) a)->inner->distance < ((const HnswPairingHeapNode *) b)->inner->distance)
return -1;
if (HnswGetSearchCandidateConst(w_node, a)->distance > HnswGetSearchCandidateConst(w_node, b)->distance)
if (((const HnswPairingHeapNode *) a)->inner->distance > ((const HnswPairingHeapNode *) b)->inner->distance)
return 1;
return 0;
}
/*
* Create a pairing heap node for a candidate
*/
static HnswPairingHeapNode *
CreatePairingHeapNode(HnswCandidate * c)
{
HnswPairingHeapNode *node = palloc(sizeof(HnswPairingHeapNode));
node->inner = c;
return node;
}
/*
* Init visited
*/
@@ -672,11 +664,11 @@ InitVisited(char *base, visited_hash * v, Relation index, int ef, int m)
* Add to visited
*/
static inline void
AddToVisited(char *base, visited_hash * v, HnswElementPtr elementPtr, Relation index, bool *found)
AddToVisited(char *base, visited_hash * v, HnswCandidate * hc, Relation index, bool *found)
{
if (index != NULL)
{
HnswElement element = HnswPtrAccess(base, elementPtr);
HnswElement element = HnswPtrAccess(base, hc->element);
ItemPointerData indextid;
ItemPointerSet(&indextid, element->blkno, element->offno);
@@ -684,15 +676,23 @@ AddToVisited(char *base, visited_hash * v, HnswElementPtr elementPtr, Relation i
}
else if (base != NULL)
{
HnswElement element = HnswPtrAccess(base, elementPtr);
#if PG_VERSION_NUM >= 130000
HnswElement element = HnswPtrAccess(base, hc->element);
offsethash_insert_hash(v->offsets, HnswPtrOffset(elementPtr), element->hash, found);
offsethash_insert_hash(v->offsets, HnswPtrOffset(hc->element), element->hash, found);
#else
offsethash_insert(v->offsets, HnswPtrOffset(hc->element), found);
#endif
}
else
{
HnswElement element = HnswPtrAccess(base, elementPtr);
#if PG_VERSION_NUM >= 130000
HnswElement element = HnswPtrAccess(base, hc->element);
pointerhash_insert_hash(v->pointers, (uintptr_t) HnswPtrPointer(elementPtr), element->hash, found);
pointerhash_insert_hash(v->pointers, (uintptr_t) HnswPtrPointer(hc->element), element->hash, found);
#else
pointerhash_insert(v->pointers, (uintptr_t) HnswPtrPointer(hc->element), found);
#endif
}
}
@@ -700,89 +700,20 @@ AddToVisited(char *base, visited_hash * v, HnswElementPtr elementPtr, Relation i
* Count element towards ef
*/
static inline bool
CountElement(HnswElement skipElement, HnswElement e)
CountElement(char *base, HnswElement skipElement, HnswCandidate * hc)
{
HnswElement e;
if (skipElement == NULL)
return true;
/* Ensure does not access heaptidsLength during in-memory build */
pg_memory_barrier();
/* Keep scan-build happy on Mac x86-64 */
Assert(e);
e = HnswPtrAccess(base, hc->element);
return e->heaptidsLength != 0;
}
/*
* Load unvisited neighbors from memory
*/
static void
HnswLoadUnvisitedFromMemory(char *base, HnswElement element, HnswUnvisited * unvisited, int *unvisitedLength, visited_hash * v, int lc, HnswNeighborArray * localNeighborhood, Size neighborhoodSize)
{
/* Get the neighborhood at layer lc */
HnswNeighborArray *neighborhood = HnswGetNeighbors(base, element, lc);
/* Copy neighborhood to local memory */
LWLockAcquire(&element->lock, LW_SHARED);
memcpy(localNeighborhood, neighborhood, neighborhoodSize);
LWLockRelease(&element->lock);
*unvisitedLength = 0;
for (int i = 0; i < localNeighborhood->length; i++)
{
HnswCandidate *hc = &localNeighborhood->items[i];
bool found;
AddToVisited(base, v, hc->element, NULL, &found);
if (!found)
unvisited[(*unvisitedLength)++].element = HnswPtrAccess(base, hc->element);
}
}
/*
* Load unvisited neighbors from disk
*/
static void
HnswLoadUnvisitedFromDisk(HnswElement element, HnswUnvisited * unvisited, int *unvisitedLength, visited_hash * v, Relation index, int m, int lm, int lc)
{
Buffer buf;
Page page;
HnswNeighborTuple ntup;
int start;
ItemPointerData indextids[HNSW_MAX_M * 2];
buf = ReadBuffer(index, element->neighborPage);
LockBuffer(buf, BUFFER_LOCK_SHARE);
page = BufferGetPage(buf);
ntup = (HnswNeighborTuple) PageGetItem(page, PageGetItemId(page, element->neighborOffno));
start = (element->level - lc) * m;
/* Copy to minimize lock time */
memcpy(&indextids, ntup->indextids + start, lm * sizeof(ItemPointerData));
UnlockReleaseBuffer(buf);
*unvisitedLength = 0;
for (int i = 0; i < lm; i++)
{
ItemPointer indextid = &indextids[i];
bool found;
if (!ItemPointerIsValid(indextid))
break;
tidhash_insert(v->tids, *indextid, &found);
if (!found)
unvisited[(*unvisitedLength)++].indextid = *indextid;
}
}
/*
* Algorithm 2 from paper
*/
@@ -795,45 +726,43 @@ HnswSearchLayer(char *base, Datum q, List *ep, int ef, int lc, Relation index, F
int wlen = 0;
visited_hash v;
ListCell *lc2;
HnswNeighborArray *localNeighborhood = NULL;
Size neighborhoodSize = 0;
int lm = HnswGetLayerM(m, lc);
HnswUnvisited *unvisited = palloc(lm * sizeof(HnswUnvisited));
int unvisitedLength;
HnswNeighborArray *neighborhoodData = NULL;
Size neighborhoodSize;
InitVisited(base, &v, index, ef, m);
/* Create local memory for neighborhood if needed */
if (index == NULL)
{
neighborhoodSize = HNSW_NEIGHBOR_ARRAY_SIZE(lm);
localNeighborhood = palloc(neighborhoodSize);
neighborhoodSize = HNSW_NEIGHBOR_ARRAY_SIZE(HnswGetLayerM(m, lc));
neighborhoodData = palloc(neighborhoodSize);
}
/* Add entry points to v, C, and W */
foreach(lc2, ep)
{
HnswSearchCandidate *hc = (HnswSearchCandidate *) lfirst(lc2);
HnswCandidate *hc = (HnswCandidate *) lfirst(lc2);
bool found;
AddToVisited(base, &v, hc->element, index, &found);
AddToVisited(base, &v, hc, index, &found);
pairingheap_add(C, &hc->c_node);
pairingheap_add(W, &hc->w_node);
pairingheap_add(C, &(CreatePairingHeapNode(hc)->ph_node));
pairingheap_add(W, &(CreatePairingHeapNode(hc)->ph_node));
/*
* Do not count elements being deleted towards ef when vacuuming. It
* would be ideal to do this for inserts as well, but this could
* affect insert performance.
*/
if (CountElement(skipElement, HnswPtrAccess(base, hc->element)))
if (CountElement(base, skipElement, hc))
wlen++;
}
while (!pairingheap_is_empty(C))
{
HnswSearchCandidate *c = HnswGetSearchCandidate(c_node, pairingheap_remove_first(C));
HnswSearchCandidate *f = HnswGetSearchCandidate(w_node, pairingheap_first(W));
HnswNeighborArray *neighborhood;
HnswCandidate *c = ((HnswPairingHeapNode *) pairingheap_remove_first(C))->inner;
HnswCandidate *f = ((HnswPairingHeapNode *) pairingheap_first(W))->inner;
HnswElement cElement;
if (c->distance > f->distance)
@@ -841,67 +770,72 @@ HnswSearchLayer(char *base, Datum q, List *ep, int ef, int lc, Relation index, F
cElement = HnswPtrAccess(base, c->element);
if (HnswPtrIsNull(base, cElement->neighbors))
HnswLoadNeighbors(cElement, index, m);
/* Get the neighborhood at layer lc */
neighborhood = HnswGetNeighbors(base, cElement, lc);
/* Copy neighborhood to local memory if needed */
if (index == NULL)
HnswLoadUnvisitedFromMemory(base, cElement, unvisited, &unvisitedLength, &v, lc, localNeighborhood, neighborhoodSize);
else
HnswLoadUnvisitedFromDisk(cElement, unvisited, &unvisitedLength, &v, index, m, lm, lc);
for (int i = 0; i < unvisitedLength; i++)
{
HnswElement eElement;
HnswSearchCandidate *e;
float eDistance;
bool alwaysAdd = wlen < ef;
LWLockAcquire(&cElement->lock, LW_SHARED);
memcpy(neighborhoodData, neighborhood, neighborhoodSize);
LWLockRelease(&cElement->lock);
neighborhood = neighborhoodData;
}
f = HnswGetSearchCandidate(w_node, pairingheap_first(W));
for (int i = 0; i < neighborhood->length; i++)
{
HnswCandidate *e = &neighborhood->items[i];
bool visited;
if (index == NULL)
AddToVisited(base, &v, e, index, &visited);
if (!visited)
{
eElement = unvisited[i].element;
eDistance = GetElementDistance(base, eElement, q, procinfo, collation);
}
else
{
ItemPointer indextid = &unvisited[i].indextid;
BlockNumber blkno = ItemPointerGetBlockNumber(indextid);
OffsetNumber offno = ItemPointerGetOffsetNumber(indextid);
float eDistance;
HnswElement eElement = HnswPtrAccess(base, e->element);
/* Avoid any allocations if not adding */
eElement = NULL;
HnswLoadElementImpl(blkno, offno, &eDistance, &q, index, procinfo, collation, inserting, alwaysAdd ? NULL : &f->distance, &eElement);
f = ((HnswPairingHeapNode *) pairingheap_first(W))->inner;
if (eElement == NULL)
continue;
}
if (index == NULL)
eDistance = GetCandidateDistance(base, e, q, procinfo, collation);
else
HnswLoadElement(eElement, &eDistance, &q, index, procinfo, collation, inserting, wlen >= ef ? &f->distance : NULL);
if (!(eDistance < f->distance || alwaysAdd))
continue;
if (eDistance < f->distance || wlen < ef)
{
HnswCandidate *ec;
Assert(!eElement->deleted);
Assert(!eElement->deleted);
/* Make robust to issues */
if (eElement->level < lc)
continue;
/* Make robust to issues */
if (eElement->level < lc)
continue;
/* Create a new candidate */
e = palloc(sizeof(HnswSearchCandidate));
HnswPtrStore(base, e->element, eElement);
e->distance = eDistance;
pairingheap_add(C, &e->c_node);
pairingheap_add(W, &e->w_node);
/* Copy e */
ec = palloc(sizeof(HnswCandidate));
HnswPtrStore(base, ec->element, eElement);
ec->distance = eDistance;
/*
* Do not count elements being deleted towards ef when vacuuming.
* It would be ideal to do this for inserts as well, but this
* could affect insert performance.
*/
if (CountElement(skipElement, eElement))
{
wlen++;
pairingheap_add(C, &(CreatePairingHeapNode(ec)->ph_node));
pairingheap_add(W, &(CreatePairingHeapNode(ec)->ph_node));
/* No need to decrement wlen */
if (wlen > ef)
pairingheap_remove_first(W);
/*
* Do not count elements being deleted towards ef when
* vacuuming. It would be ideal to do this for inserts as
* well, but this could affect insert performance.
*/
if (CountElement(base, skipElement, e))
{
wlen++;
/* No need to decrement wlen */
if (wlen > ef)
pairingheap_remove_first(W);
}
}
}
}
}
@@ -909,7 +843,7 @@ HnswSearchLayer(char *base, Datum q, List *ep, int ef, int lc, Relation index, F
/* Add each element of W to w */
while (!pairingheap_is_empty(W))
{
HnswSearchCandidate *hc = HnswGetSearchCandidate(w_node, pairingheap_remove_first(W));
HnswCandidate *hc = ((HnswPairingHeapNode *) pairingheap_remove_first(W))->inner;
w = lappend(w, hc);
}
@@ -921,10 +855,17 @@ HnswSearchLayer(char *base, Datum q, List *ep, int ef, int lc, Relation index, F
* Compare candidate distances with pointer tie-breaker
*/
static int
#if PG_VERSION_NUM >= 130000
CompareCandidateDistances(const ListCell *a, const ListCell *b)
{
HnswCandidate *hca = lfirst(a);
HnswCandidate *hcb = lfirst(b);
#else
CompareCandidateDistances(const void *a, const void *b)
{
HnswCandidate *hca = lfirst(*(ListCell **) a);
HnswCandidate *hcb = lfirst(*(ListCell **) b);
#endif
if (hca->distance < hcb->distance)
return 1;
@@ -945,10 +886,17 @@ CompareCandidateDistances(const ListCell *a, const ListCell *b)
* Compare candidate distances with offset tie-breaker
*/
static int
#if PG_VERSION_NUM >= 130000
CompareCandidateDistancesOffset(const ListCell *a, const ListCell *b)
{
HnswCandidate *hca = lfirst(a);
HnswCandidate *hcb = lfirst(b);
#else
CompareCandidateDistancesOffset(const void *a, const void *b)
{
HnswCandidate *hca = lfirst(*(ListCell **) a);
HnswCandidate *hcb = lfirst(*(ListCell **) b);
#endif
if (hca->distance < hcb->distance)
return 1;
@@ -1158,7 +1106,7 @@ HnswUpdateConnection(char *base, HnswElement element, HnswCandidate * hc, int lm
if (HnswPtrIsNull(base, hc3Element->value))
HnswLoadElement(hc3Element, &hc3->distance, &q, index, procinfo, collation, true, NULL);
else
hc3->distance = GetElementDistance(base, hc3Element, q, procinfo, collation);
hc3->distance = GetCandidateDistance(base, hc3, q, procinfo, collation);
/* Prune element if being deleted */
if (hc3Element->heaptidsLength == 0)
@@ -1230,6 +1178,7 @@ RemoveElements(char *base, List *w, HnswElement skipElement)
return w2;
}
#if PG_VERSION_NUM >= 130000
/*
* Precompute hash
*/
@@ -1245,6 +1194,7 @@ PrecomputeHash(char *base, HnswElement element)
else
element->hash = hash_offset(HnswPtrOffset(ptr));
}
#endif
/*
* Algorithm 1 from paper
@@ -1259,9 +1209,11 @@ HnswFindElementNeighbors(char *base, HnswElement element, HnswElement entryPoint
Datum q = HnswGetValue(base, element);
HnswElement skipElement = existing ? element : NULL;
#if PG_VERSION_NUM >= 130000
/* Precompute hash */
if (index == NULL)
PrecomputeHash(base, element);
#endif
/* No neighbors if no entry point */
if (entryPoint == NULL)
@@ -1290,27 +1242,16 @@ HnswFindElementNeighbors(char *base, HnswElement element, HnswElement entryPoint
{
int lm = HnswGetLayerM(m, lc);
List *neighbors;
List *lw = NIL;
ListCell *lc2;
List *lw;
w = HnswSearchLayer(base, q, ep, efConstruction, lc, index, procinfo, collation, m, true, skipElement);
/* Convert search candidates to candidates */
foreach(lc2, w)
{
HnswSearchCandidate *sc = lfirst(lc2);
HnswCandidate *hc = palloc(sizeof(HnswCandidate));
hc->element = sc->element;
hc->distance = sc->distance;
lw = lappend(lw, hc);
}
/* Elements being deleted or skipped can help with search */
/* but should be removed before selecting neighbors */
if (index != NULL)
lw = RemoveElements(base, lw, skipElement);
lw = RemoveElements(base, w, skipElement);
else
lw = w;
/*
* Candidates are sorted, but not deterministically. Could set
@@ -1335,9 +1276,7 @@ SparsevecCheckValue(Pointer v)
SparseVector *vec = (SparseVector *) v;
if (vec->nnz > HNSW_MAX_NNZ)
ereport(ERROR,
(errcode(ERRCODE_PROGRAM_LIMIT_EXCEEDED),
errmsg("sparsevec cannot have more than %d non-zero elements for hnsw index", HNSW_MAX_NNZ)));
elog(ERROR, "sparsevec cannot have more than %d non-zero elements for hnsw index", HNSW_MAX_NNZ);
}
/*
@@ -1362,7 +1301,7 @@ HnswGetTypeInfo(Relation index)
return (const HnswTypeInfo *) DatumGetPointer(FunctionCall0Coll(procinfo, InvalidOid));
}
FUNCTION_PREFIX PG_FUNCTION_INFO_V1(hnsw_halfvec_support);
PGDLLEXPORT PG_FUNCTION_INFO_V1(hnsw_halfvec_support);
Datum
hnsw_halfvec_support(PG_FUNCTION_ARGS)
{
@@ -1375,7 +1314,7 @@ hnsw_halfvec_support(PG_FUNCTION_ARGS)
PG_RETURN_POINTER(&typeInfo);
};
FUNCTION_PREFIX PG_FUNCTION_INFO_V1(hnsw_bit_support);
PGDLLEXPORT PG_FUNCTION_INFO_V1(hnsw_bit_support);
Datum
hnsw_bit_support(PG_FUNCTION_ARGS)
{
@@ -1388,7 +1327,7 @@ hnsw_bit_support(PG_FUNCTION_ARGS)
PG_RETURN_POINTER(&typeInfo);
};
FUNCTION_PREFIX PG_FUNCTION_INFO_V1(hnsw_sparsevec_support);
PGDLLEXPORT PG_FUNCTION_INFO_V1(hnsw_sparsevec_support);
Datum
hnsw_sparsevec_support(PG_FUNCTION_ARGS)
{

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;
@@ -325,20 +335,14 @@ InitBuildState(IvfflatBuildState * buildstate, Relation heap, Relation index, In
/* 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;
@@ -351,9 +355,7 @@ 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->tupdesc = CreateTemplateTupleDesc(3);
@@ -560,20 +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
*/
@@ -621,6 +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;
@@ -633,7 +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.tupdesc, 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));
@@ -930,6 +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 */
@@ -950,7 +950,7 @@ AssignTuples(IvfflatBuildState * buildstate)
}
/* Begin serial/leader tuplesort */
buildstate->sortstate = InitBuildSortState(buildstate->tupdesc, 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)
@@ -1006,10 +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"
@@ -17,8 +16,6 @@
#endif
int ivfflat_probes;
int ivfflat_max_probes;
bool ivfflat_streaming;
static relopt_kind ivfflat_relopt_kind;
/*
@@ -29,20 +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);
DefineCustomIntVariable("ivfflat.max_probes", "Sets the max number of probes for iterative scans",
NULL, &ivfflat_max_probes,
-1, IVFFLAT_MIN_LISTS, IVFFLAT_MAX_LISTS, PGC_USERSET, 0, NULL, NULL, NULL);
DefineCustomBoolVariable("ivfflat.streaming", "Use streaming mode",
NULL, &ivfflat_streaming,
IVFFLAT_DEFAULT_STREAMING, PGC_USERSET, 0, NULL, NULL, NULL);
MarkGUCPrefixReserved("ivfflat");
}
@@ -67,35 +60,6 @@ ivfflatbuildphasename(int64 phasenum)
}
}
/*
* Estimate the number of probes for iterative scans
*/
static int
EstimateProbes(PlannerInfo *root, IndexPath *path, int lists)
{
double selectivity = 1;
ListCell *lc;
double tuplesPerList;
/* Cannot estimate without limit */
/* limit_tuples includes offset */
if (root->limit_tuples < 0)
return 0;
/* Get the selectivity of non-index conditions */
foreach(lc, path->indexinfo->indrestrictinfo)
{
RestrictInfo *rinfo = lfirst(lc);
/* Skip DEFAULT_INEQ_SEL since it may be a distance filter */
if (rinfo->norm_selec >= 0 && rinfo->norm_selec <= 1 && rinfo->norm_selec != (Selectivity) DEFAULT_INEQ_SEL)
selectivity *= rinfo->norm_selec;
}
tuplesPerList = path->indexinfo->tuples / (double) lists;
return root->limit_tuples / (tuplesPerList * selectivity);
}
/*
* Estimate the cost of an index scan
*/
@@ -107,7 +71,6 @@ ivfflatcostestimate(PlannerInfo *root, IndexPath *path, double loop_count,
{
GenericCosts costs;
int lists;
int probes;
double ratio;
double spc_seq_page_cost;
Relation index;
@@ -115,8 +78,8 @@ ivfflatcostestimate(PlannerInfo *root, IndexPath *path, double loop_count,
/* 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;
@@ -129,17 +92,8 @@ ivfflatcostestimate(PlannerInfo *root, IndexPath *path, double loop_count,
IvfflatGetMetaPageInfo(index, &lists, NULL);
index_close(index, NoLock);
probes = ivfflat_probes;
if (ivfflat_streaming)
{
probes = Max(probes, EstimateProbes(root, path, lists));
if (ivfflat_max_probes != -1)
probes = Min(probes, ivfflat_max_probes);
}
/* Get the ratio of lists that we need to visit */
ratio = ((double) probes) / lists;
ratio = ((double) ivfflat_probes) / lists;
if (ratio > 1.0)
ratio = 1.0;
@@ -194,10 +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
}
/*
@@ -214,7 +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)
{
@@ -222,7 +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 */
@@ -235,24 +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

@@ -43,7 +43,6 @@
#define IVFFLAT_MIN_LISTS 1
#define IVFFLAT_MAX_LISTS 32768
#define IVFFLAT_DEFAULT_PROBES 1
#define IVFFLAT_DEFAULT_STREAMING false
/* Build phases */
/* PROGRESS_CREATEIDX_SUBPHASE_INITIALIZE is 1 */
@@ -81,8 +80,6 @@
/* Variables */
extern int ivfflat_probes;
extern int ivfflat_max_probes;
extern bool ivfflat_streaming;
typedef struct VectorArrayData
{
@@ -250,17 +247,14 @@ typedef struct IvfflatScanOpaqueData
{
const IvfflatTypeInfo *typeInfo;
int probes;
int maxProbes;
int dimensions;
bool first;
Datum value;
/* Sorting */
Tuplesortstate *sortstate;
TupleDesc tupdesc;
TupleTableSlot *vslot;
TupleTableSlot *mslot;
BufferAccessStrategy bas;
TupleTableSlot *slot;
bool isnull;
/* Support functions */
FmgrInfo *procinfo;
@@ -270,8 +264,6 @@ typedef struct IvfflatScanOpaqueData
/* Lists */
pairingheap *listQueue;
BlockNumber *startPages;
int currentIndex;
IvfflatScanList lists[FLEXIBLE_ARRAY_MEMBER]; /* must come last */
} IvfflatScanOpaqueData;

View File

@@ -94,9 +94,6 @@ 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, values, &insertPage, &listInfo);
Assert(BlockNumberIsValid(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

@@ -11,23 +11,16 @@
#include "pgstat.h"
#include "storage/bufmgr.h"
#ifdef IVFFLAT_MEMORY
#include "utils/memutils.h"
#endif
#define GetScanList(ptr) pairingheap_container(IvfflatScanList, ph_node, ptr)
#define GetScanListConst(ptr) pairingheap_const_container(IvfflatScanList, ph_node, ptr)
/*
* Compare list distances
*/
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;
@@ -65,7 +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;
@@ -79,14 +72,14 @@ GetScanLists(IndexScanDesc scan, Datum value)
/* Calculate max distance */
if (listCount == so->probes)
maxDistance = GetScanList(pairingheap_first(so->listQueue))->distance;
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;
@@ -94,7 +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;
}
}
@@ -102,11 +95,6 @@ GetScanLists(IndexScanDesc scan, Datum value)
UnlockReleaseBuffer(cbuf);
}
for (int i = listCount - 1; i >= 0; i--)
so->startPages[i] = GetScanList(pairingheap_remove_first(so->listQueue))->startPage;
Assert(pairingheap_is_empty(so->listQueue));
}
/*
@@ -118,15 +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->currentIndex < so->maxProbes && (++batchProbes) <= so->probes)
while (!pairingheap_is_empty(so->listQueue))
{
BlockNumber searchPage = so->startPages[so->currentIndex++];
BlockNumber searchPage = ((IvfflatScanList *) pairingheap_remove_first(so->listQueue))->startPage;
/* Search all entry pages for list */
while (BlockNumberIsValid(searchPage))
@@ -135,7 +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);
@@ -174,17 +166,15 @@ GetScanItems(IndexScanDesc scan, Datum value)
}
}
if (tuples < 100 && !ivfflat_streaming)
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
}
/*
@@ -227,20 +217,6 @@ GetScanValue(IndexScanDesc scan)
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
*/
@@ -251,8 +227,11 @@ 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;
scan = RelationGetIndexScan(index, nkeys, norderbys);
@@ -262,21 +241,10 @@ ivfflatbeginscan(Relation index, int nkeys, int norderbys)
if (probes > lists)
probes = lists;
if (ivfflat_streaming)
{
if (ivfflat_max_probes == -1)
maxProbes = lists;
else
maxProbes = Min(ivfflat_max_probes, lists);
}
else
maxProbes = probes;
so = (IvfflatScanOpaque) palloc(offsetof(IvfflatScanOpaqueData, lists) + maxProbes * sizeof(IvfflatScanList));
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 */
@@ -290,22 +258,11 @@ ivfflatbeginscan(Relation index, int nkeys, int norderbys)
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->startPages = palloc(maxProbes * sizeof(BlockNumber));
so->currentIndex = 0;
scan->opaque = so;
@@ -320,9 +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->currentIndex = 0;
if (keys && scan->numberOfKeys > 0)
memmove(scan->keyData, keys, scan->numberOfKeys * sizeof(ScanKeyData));
@@ -338,8 +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
@@ -367,25 +326,23 @@ ivfflatgettuple(IndexScanDesc scan, ScanDirection dir)
IvfflatBench("GetScanLists", GetScanLists(scan, value));
IvfflatBench("GetScanItems", GetScanItems(scan, value));
so->first = false;
so->value = value;
/* TODO clean up if we allocated a new 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->currentIndex == 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;
}
/*
@@ -397,12 +354,7 @@ ivfflatendscan(IndexScanDesc scan)
IvfflatScanOpaque so = (IvfflatScanOpaque) scan->opaque;
pairingheap_free(so->listQueue);
pfree(so->startPages);
tuplesort_end(so->sortstate);
FreeAccessStrategy(so->bas);
FreeTupleDesc(so->tupdesc);
/* TODO Free vslot and mslot without freeing TupleDesc */
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

@@ -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))
@@ -176,7 +181,7 @@ float_underflow_error(void)
/*
* 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)
{
@@ -289,7 +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)
{
@@ -343,7 +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)
{
@@ -374,7 +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)
{
@@ -408,7 +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)
{
@@ -428,7 +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)
{
@@ -443,7 +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)
{
@@ -517,7 +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)
{
@@ -541,7 +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)
{
@@ -579,7 +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)
{
@@ -595,7 +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)
{
@@ -622,7 +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)
{
@@ -637,7 +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)
{
@@ -671,7 +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)
{
@@ -703,7 +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)
{
@@ -740,7 +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)
{
@@ -755,7 +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)
{
@@ -767,7 +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)
{
@@ -785,7 +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)
{
@@ -824,7 +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)
{
@@ -857,7 +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)
{
@@ -890,7 +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)
{
@@ -926,7 +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)
{
@@ -950,7 +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)
{
@@ -968,7 +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)
{
@@ -1042,7 +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)
{
@@ -1055,7 +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)
{
@@ -1068,7 +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)
{
@@ -1081,7 +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)
{
@@ -1094,7 +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)
{
@@ -1107,7 +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)
{
@@ -1120,7 +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)
{
@@ -1133,7 +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)
{
@@ -1194,7 +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)
{
@@ -1265,7 +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)
{
@@ -1299,7 +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

@@ -208,62 +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

@@ -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

@@ -58,22 +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

@@ -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

@@ -1,7 +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;
@@ -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;

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;
@@ -11,7 +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;

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;

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;

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

@@ -1,11 +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

@@ -1,42 +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

@@ -1,31 +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;");
$node->safe_psql("postgres", "CREATE TABLE tst (i int4, v vector($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 = $node->safe_psql("postgres", qq(
SET enable_seqscan = off;
SET ivfflat.probes = 10;
SET ivfflat.streaming = on;
SELECT COUNT(*) FROM (SELECT v FROM tst WHERE i % 10000 = 0 ORDER BY v <-> (SELECT v FROM tst LIMIT 1) LIMIT 11) t;
));
is($count, 10);
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.streaming = on;
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.streaming = on;
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);
}
# 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, 0.58, $operator);
test_recall($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

@@ -1,4 +1,4 @@
comment = 'vector data type and ivfflat and hnsw access methods'
default_version = '0.7.4'
default_version = '0.7.0'
module_pathname = '$libdir/vector'
relocatable = true