Compare commits

..

1 Commits

Author SHA1 Message Date
Andrew Kane
c8e4bf1c3c Added test for arrays 2024-02-03 16:50:17 -08:00
15 changed files with 155 additions and 144 deletions

View File

@@ -28,7 +28,7 @@ jobs:
dev-files: true dev-files: true
- run: make - run: make
env: env:
PG_CFLAGS: -DUSE_ASSERT_CHECKING -Wall -Wextra -Werror -Wno-unused-parameter -Wno-sign-compare PG_CFLAGS: -Wall -Wextra -Werror -Wno-unused-parameter -Wno-sign-compare
- run: | - run: |
export PG_CONFIG=`which pg_config` export PG_CONFIG=`which pg_config`
sudo --preserve-env=PG_CONFIG make install sudo --preserve-env=PG_CONFIG make install
@@ -49,7 +49,7 @@ jobs:
postgres-version: 14 postgres-version: 14
- run: make - run: make
env: env:
PG_CFLAGS: -DUSE_ASSERT_CHECKING -Wall -Wextra -Werror -Wno-unused-parameter PG_CFLAGS: -Wall -Wextra -Werror -Wno-unused-parameter
- run: make install - run: make install
- run: make installcheck - run: make installcheck
- if: ${{ failure() }} - if: ${{ failure() }}
@@ -60,9 +60,7 @@ jobs:
wget -q https://github.com/postgres/postgres/archive/refs/tags/REL_14_10.tar.gz wget -q https://github.com/postgres/postgres/archive/refs/tags/REL_14_10.tar.gz
tar xf REL_14_10.tar.gz tar xf REL_14_10.tar.gz
- run: make prove_installcheck PROVE_FLAGS="-I ./postgres-REL_14_10/src/test/perl" PERL5LIB="/Users/runner/perl5/lib/perl5" - run: make prove_installcheck PROVE_FLAGS="-I ./postgres-REL_14_10/src/test/perl" PERL5LIB="/Users/runner/perl5/lib/perl5"
- run: make clean && /usr/local/opt/llvm@15/bin/scan-build --status-bugs make - run: make clean && /usr/local/opt/llvm@15/bin/scan-build --status-bugs make PG_CFLAGS="-DUSE_ASSERT_CHECKING"
env:
PG_CFLAGS: -DUSE_ASSERT_CHECKING
windows: windows:
runs-on: windows-latest runs-on: windows-latest
if: ${{ !startsWith(github.ref_name, 'mac') }} if: ${{ !startsWith(github.ref_name, 'mac') }}
@@ -99,15 +97,4 @@ jobs:
sudo -u postgres make installcheck sudo -u postgres make installcheck
sudo -u postgres make prove_installcheck sudo -u postgres make prove_installcheck
env: env:
PG_CFLAGS: -DUSE_ASSERT_CHECKING -Wall -Wextra -Werror -Wno-unused-parameter -Wno-sign-compare PG_CFLAGS: -Wall -Wextra -Werror -Wno-unused-parameter -Wno-sign-compare
valgrind:
if: ${{ !startsWith(github.ref_name, 'mac') && !startsWith(github.ref_name, 'windows') }}
runs-on: ubuntu-latest
steps:
- uses: actions/checkout@v4
- uses: ankane/setup-postgres-valgrind@v1
with:
postgres-version: 16
- run: make
- run: sudo --preserve-env=PG_CONFIG make install
- run: make installcheck

View File

@@ -1,12 +1,6 @@
## 0.6.2 (unreleased) ## 0.6.1 (unreleased)
- Reduced lock contention with parallel HNSW index builds
## 0.6.1 (2024-03-04)
- Fixed error with `ANALYZE` and vectors with different dimensions - Fixed error with `ANALYZE` and vectors with different dimensions
- Fixed error with `shared_preload_libraries`
- Fixed vector subtraction being marked as commutative
## 0.6.0 (2024-01-29) ## 0.6.0 (2024-01-29)

View File

@@ -1,4 +1,4 @@
Portions Copyright (c) 1996-2024, PostgreSQL Global Development Group Portions Copyright (c) 1996-2023, PostgreSQL Global Development Group
Portions Copyright (c) 1994, The Regents of the University of California Portions Copyright (c) 1994, The Regents of the University of California

View File

@@ -2,7 +2,7 @@
"name": "vector", "name": "vector",
"abstract": "Open-source vector similarity search for Postgres", "abstract": "Open-source vector similarity search for Postgres",
"description": "Supports L2 distance, inner product, and cosine distance", "description": "Supports L2 distance, inner product, and cosine distance",
"version": "0.6.1", "version": "0.6.0",
"maintainer": [ "maintainer": [
"Andrew Kane <andrew@ankane.org>" "Andrew Kane <andrew@ankane.org>"
], ],
@@ -20,7 +20,7 @@
"vector": { "vector": {
"file": "sql/vector.sql", "file": "sql/vector.sql",
"docfile": "README.md", "docfile": "README.md",
"version": "0.6.1", "version": "0.6.0",
"abstract": "Open-source vector similarity search for Postgres" "abstract": "Open-source vector similarity search for Postgres"
} }
}, },

View File

@@ -1,5 +1,5 @@
EXTENSION = vector EXTENSION = vector
EXTVERSION = 0.6.1 EXTVERSION = 0.6.0
MODULE_big = vector MODULE_big = vector
DATA = $(wildcard sql/*--*.sql) DATA = $(wildcard sql/*--*.sql)
@@ -71,9 +71,11 @@ PG_MAJOR ?= 16
.PHONY: docker .PHONY: docker
docker: docker:
docker build --pull --no-cache --build-arg PG_MAJOR=$(PG_MAJOR) -t pgvector/pgvector:pg$(PG_MAJOR) -t pgvector/pgvector:$(EXTVERSION)-pg$(PG_MAJOR) . docker build --pull --no-cache --build-arg PG_MAJOR=$(PG_MAJOR) -t pgvector/pgvector:pg$(PG_MAJOR) .
docker build --build-arg PG_MAJOR=$(PG_MAJOR) -t pgvector/pgvector:$(EXTVERSION)-pg$(PG_MAJOR) .
.PHONY: docker-release .PHONY: docker-release
docker-release: docker-release:
docker buildx build --push --pull --no-cache --platform linux/amd64,linux/arm64 --build-arg PG_MAJOR=$(PG_MAJOR) -t pgvector/pgvector:pg$(PG_MAJOR) -t pgvector/pgvector:$(EXTVERSION)-pg$(PG_MAJOR) . docker buildx build --push --pull --no-cache --platform linux/amd64,linux/arm64 --build-arg PG_MAJOR=$(PG_MAJOR) -t pgvector/pgvector:pg$(PG_MAJOR) .
docker buildx build --push --platform linux/amd64,linux/arm64 --build-arg PG_MAJOR=$(PG_MAJOR) -t pgvector/pgvector:$(EXTVERSION)-pg$(PG_MAJOR) .

View File

@@ -1,5 +1,5 @@
EXTENSION = vector EXTENSION = vector
EXTVERSION = 0.6.1 EXTVERSION = 0.6.0
OBJS = 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\vector.obj OBJS = 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\vector.obj
HEADERS = src\vector.h HEADERS = src\vector.h

View File

@@ -20,7 +20,7 @@ Compile and install the extension (supports Postgres 12+)
```sh ```sh
cd /tmp cd /tmp
git clone --branch v0.6.1 https://github.com/pgvector/pgvector.git git clone --branch v0.6.0 https://github.com/pgvector/pgvector.git
cd pgvector cd pgvector
make make
make install # may need sudo make install # may need sudo
@@ -28,7 +28,7 @@ make install # may need sudo
See the [installation notes](#installation-notes) if you run into issues See the [installation notes](#installation-notes) if you run into issues
You can also install it with [Docker](#docker), [Homebrew](#homebrew), [PGXN](#pgxn), [APT](#apt), [Yum](#yum), [pkg](#pkg), or [conda-forge](#conda-forge), and it comes preinstalled with [Postgres.app](#postgresapp) and many [hosted providers](#hosted-postgres). There are also instructions for [GitHub Actions](https://github.com/pgvector/setup-pgvector). You can also install it with [Docker](#docker), [Homebrew](#homebrew), [PGXN](#pgxn), [APT](#apt), [Yum](#yum), or [conda-forge](#conda-forge), and it comes preinstalled with [Postgres.app](#postgresapp) and many [hosted providers](#hosted-postgres). There are also instructions for [GitHub Actions](https://github.com/pgvector/setup-pgvector).
### Windows ### Windows
@@ -44,7 +44,7 @@ Then use `nmake` to build:
```cmd ```cmd
set "PGROOT=C:\Program Files\PostgreSQL\16" set "PGROOT=C:\Program Files\PostgreSQL\16"
git clone --branch v0.6.1 https://github.com/pgvector/pgvector.git git clone --branch v0.6.0 https://github.com/pgvector/pgvector.git
cd pgvector cd pgvector
nmake /F Makefile.win nmake /F Makefile.win
nmake /F Makefile.win install nmake /F Makefile.win install
@@ -102,12 +102,6 @@ Insert vectors
INSERT INTO items (embedding) VALUES ('[1,2,3]'), ('[4,5,6]'); 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/bulk_loading.py))
```sql
COPY items (embedding) FROM STDIN WITH (FORMAT BINARY);
```
Upsert vectors Upsert vectors
```sql ```sql
@@ -270,8 +264,6 @@ HINT: Increase maintenance_work_mem to speed up builds.
Note: Do not set `maintenance_work_mem` so high that it exhausts the memory on the server Note: Do not set `maintenance_work_mem` so high that it exhausts the memory on the server
Like other index types, its faster to create an index after loading your initial data
Starting with 0.6.0, you can also speed up index creation by increasing the number of parallel workers (2 by default) Starting with 0.6.0, you can also speed up index creation by increasing the number of parallel workers (2 by default)
```sql ```sql
@@ -581,10 +573,6 @@ or choose to store vectors inline:
ALTER TABLE items ALTER COLUMN embedding SET STORAGE PLAIN; ALTER TABLE items ALTER COLUMN embedding SET STORAGE PLAIN;
``` ```
#### Why are there less results for a query after adding an HNSW index?
Results are limited by the size of the dynamic candidate list (`hnsw.ef_search`). There may be even less results due to dead tuples or filtering conditions in the query. We recommend setting `hnsw.ef_search` to at least twice the `LIMIT` of the query. If you need more than 500 results, use an IVFFlat index instead.
#### Why are there less results for a query after adding an IVFFlat index? #### Why are there less results for a query after adding an IVFFlat index?
The index was likely created with too little data for the number of lists. Drop the index until the table has more data. The index was likely created with too little data for the number of lists. Drop the index until the table has more data.
@@ -593,8 +581,6 @@ The index was likely created with too little data for the number of lists. Drop
DROP INDEX index_name; DROP INDEX index_name;
``` ```
Results can also be limited by the number of probes (`ivfflat.probes`).
## Reference ## Reference
### Vector Type ### Vector Type
@@ -670,14 +656,6 @@ Note: Replace `16` with your Postgres server version
If compilation fails and the output includes `warning: no such sysroot directory` on Mac, reinstall Xcode Command Line Tools. If compilation fails and the output includes `warning: no such sysroot directory` on Mac, reinstall Xcode Command Line Tools.
### Portability
To compile for portability, use:
```sh
make OPTFLAGS=""
```
## Additional Installation Methods ## Additional Installation Methods
### Docker ### Docker
@@ -693,7 +671,7 @@ This adds pgvector to the [Postgres image](https://hub.docker.com/_/postgres) (r
You can also build the image manually: You can also build the image manually:
```sh ```sh
git clone --branch v0.6.1 https://github.com/pgvector/pgvector.git git clone --branch v0.6.0 https://github.com/pgvector/pgvector.git
cd pgvector cd pgvector
docker build --build-arg PG_MAJOR=16 -t myuser/pgvector . docker build --build-arg PG_MAJOR=16 -t myuser/pgvector .
``` ```
@@ -738,21 +716,6 @@ sudo dnf install pgvector_16
Note: Replace `16` with your Postgres server version Note: Replace `16` with your Postgres server version
### pkg
Install the FreeBSD package with:
```sh
pkg install postgresql15-pg_vector
```
or the port with:
```sh
cd /usr/ports/databases/pgvector
make install
```
### conda-forge ### conda-forge
With Conda Postgres, install from [conda-forge](https://anaconda.org/conda-forge/pgvector) with: With Conda Postgres, install from [conda-forge](https://anaconda.org/conda-forge/pgvector) with:

View File

@@ -1,16 +0,0 @@
-- complain if script is sourced in psql, rather than via CREATE EXTENSION
\echo Use "ALTER EXTENSION vector UPDATE TO '0.6.1'" to load this file. \quit
DROP OPERATOR - (vector, vector);
CREATE OPERATOR - (
LEFTARG = vector, RIGHTARG = vector, PROCEDURE = vector_sub
);
ALTER OPERATOR <= (vector, vector) SET (
RESTRICT = scalarlesel, JOIN = scalarlejoinsel
);
ALTER OPERATOR >= (vector, vector) SET (
RESTRICT = scalargesel, JOIN = scalargejoinsel
);

View File

@@ -180,7 +180,8 @@ CREATE OPERATOR + (
); );
CREATE OPERATOR - ( CREATE OPERATOR - (
LEFTARG = vector, RIGHTARG = vector, PROCEDURE = vector_sub LEFTARG = vector, RIGHTARG = vector, PROCEDURE = vector_sub,
COMMUTATOR = -
); );
CREATE OPERATOR * ( CREATE OPERATOR * (
@@ -194,10 +195,11 @@ CREATE OPERATOR < (
RESTRICT = scalarltsel, JOIN = scalarltjoinsel RESTRICT = scalarltsel, JOIN = scalarltjoinsel
); );
-- should use scalarlesel and scalarlejoinsel, but not supported in Postgres < 11
CREATE OPERATOR <= ( CREATE OPERATOR <= (
LEFTARG = vector, RIGHTARG = vector, PROCEDURE = vector_le, LEFTARG = vector, RIGHTARG = vector, PROCEDURE = vector_le,
COMMUTATOR = >= , NEGATOR = > , COMMUTATOR = >= , NEGATOR = > ,
RESTRICT = scalarlesel, JOIN = scalarlejoinsel RESTRICT = scalarltsel, JOIN = scalarltjoinsel
); );
CREATE OPERATOR = ( CREATE OPERATOR = (
@@ -212,10 +214,11 @@ CREATE OPERATOR <> (
RESTRICT = eqsel, JOIN = eqjoinsel RESTRICT = eqsel, JOIN = eqjoinsel
); );
-- should use scalargesel and scalargejoinsel, but not supported in Postgres < 11
CREATE OPERATOR >= ( CREATE OPERATOR >= (
LEFTARG = vector, RIGHTARG = vector, PROCEDURE = vector_ge, LEFTARG = vector, RIGHTARG = vector, PROCEDURE = vector_ge,
COMMUTATOR = <= , NEGATOR = < , COMMUTATOR = <= , NEGATOR = < ,
RESTRICT = scalargesel, JOIN = scalargejoinsel RESTRICT = scalargtsel, JOIN = scalargtjoinsel
); );
CREATE OPERATOR > ( CREATE OPERATOR > (

View File

@@ -8,7 +8,6 @@
#include "commands/progress.h" #include "commands/progress.h"
#include "commands/vacuum.h" #include "commands/vacuum.h"
#include "hnsw.h" #include "hnsw.h"
#include "miscadmin.h"
#include "utils/guc.h" #include "utils/guc.h"
#include "utils/selfuncs.h" #include "utils/selfuncs.h"
@@ -29,7 +28,7 @@ static relopt_kind hnsw_relopt_kind;
* this grows bigger, we should use a shmem_request_hook and * this grows bigger, we should use a shmem_request_hook and
* RequestAddinShmemSpace() to pre-reserve space for this. * RequestAddinShmemSpace() to pre-reserve space for this.
*/ */
void static void
HnswInitLockTranche(void) HnswInitLockTranche(void)
{ {
int *tranche_ids; int *tranche_ids;
@@ -54,8 +53,7 @@ HnswInitLockTranche(void)
void void
HnswInit(void) HnswInit(void)
{ {
if (!process_shared_preload_libraries_in_progress) HnswInitLockTranche();
HnswInitLockTranche();
hnsw_relopt_kind = add_reloption_kind(); hnsw_relopt_kind = add_reloption_kind();
add_int_reloption(hnsw_relopt_kind, "m", "Max number of connections", add_int_reloption(hnsw_relopt_kind, "m", "Max number of connections",

View File

@@ -80,7 +80,7 @@
#if PG_VERSION_NUM < 130000 #if PG_VERSION_NUM < 130000
#define list_delete_last(list) list_truncate(list, list_length(list) - 1) #define list_delete_last(list) list_truncate(list, list_length(list) - 1)
#define list_sort(list, cmp) ((list) = list_qsort(list, cmp)) #define list_sort(list, cmp) list_qsort(list, cmp)
#endif #endif
#define HnswIsElementTuple(tup) ((tup)->type == HNSW_ELEMENT_TUPLE_TYPE) #define HnswIsElementTuple(tup) ((tup)->type == HNSW_ELEMENT_TUPLE_TYPE)
@@ -185,7 +185,6 @@ typedef struct HnswGraph
/* Entry state */ /* Entry state */
LWLock entryLock; LWLock entryLock;
LWLock entryWaitLock;
HnswElementPtr entryPoint; HnswElementPtr entryPoint;
/* Allocations state */ /* Allocations state */
@@ -390,7 +389,6 @@ void HnswLoadElement(HnswElement element, float *distance, Datum *q, Relation i
void HnswSetElementTuple(char *base, HnswElementTuple etup, HnswElement element); void HnswSetElementTuple(char *base, HnswElementTuple etup, HnswElement element);
void HnswUpdateConnection(char *base, HnswElement element, HnswCandidate * hc, int lm, int lc, int *updateIdx, Relation index, FmgrInfo *procinfo, Oid collation); void HnswUpdateConnection(char *base, HnswElement element, HnswCandidate * hc, int lm, int lc, int *updateIdx, Relation index, FmgrInfo *procinfo, Oid collation);
void HnswLoadNeighbors(HnswElement element, Relation index, int m); void HnswLoadNeighbors(HnswElement element, Relation index, int m);
void HnswInitLockTranche(void);
PGDLLEXPORT void HnswParallelBuildMain(dsm_segment *seg, shm_toc *toc); PGDLLEXPORT void HnswParallelBuildMain(dsm_segment *seg, shm_toc *toc);
/* Index access methods */ /* Index access methods */

View File

@@ -176,7 +176,7 @@ CreateGraphPages(HnswBuildState * buildstate)
Size etupSize; Size etupSize;
Size ntupSize; Size ntupSize;
Size combinedSize; Size combinedSize;
Pointer valuePtr = HnswPtrAccess(base, element->value); void *valuePtr = HnswPtrAccess(base, element->value);
/* Update iterator */ /* Update iterator */
iter = element->next; iter = element->next;
@@ -431,15 +431,10 @@ InsertTupleInMemory(HnswBuildState * buildstate, HnswElement element)
HnswGraph *graph = buildstate->graph; HnswGraph *graph = buildstate->graph;
HnswElement entryPoint; HnswElement entryPoint;
LWLock *entryLock = &graph->entryLock; LWLock *entryLock = &graph->entryLock;
LWLock *entryWaitLock = &graph->entryWaitLock;
int efConstruction = buildstate->efConstruction; int efConstruction = buildstate->efConstruction;
int m = buildstate->m; int m = buildstate->m;
char *base = buildstate->hnswarea; char *base = buildstate->hnswarea;
/* Wait if another process needs exclusive lock */
LWLockAcquire(entryWaitLock, LW_EXCLUSIVE);
LWLockRelease(entryWaitLock);
/* Get entry point */ /* Get entry point */
LWLockAcquire(entryLock, LW_SHARED); LWLockAcquire(entryLock, LW_SHARED);
entryPoint = HnswPtrAccess(base, graph->entryPoint); entryPoint = HnswPtrAccess(base, graph->entryPoint);
@@ -451,9 +446,7 @@ InsertTupleInMemory(HnswBuildState * buildstate, HnswElement element)
LWLockRelease(entryLock); LWLockRelease(entryLock);
/* Get exclusive lock */ /* Get exclusive lock */
LWLockAcquire(entryWaitLock, LW_EXCLUSIVE);
LWLockAcquire(entryLock, LW_EXCLUSIVE); LWLockAcquire(entryLock, LW_EXCLUSIVE);
LWLockRelease(entryWaitLock);
/* Get latest entry point after lock is acquired */ /* Get latest entry point after lock is acquired */
entryPoint = HnswPtrAccess(base, graph->entryPoint); entryPoint = HnswPtrAccess(base, graph->entryPoint);
@@ -608,9 +601,6 @@ BuildCallback(Relation index, CALLBACK_ITEM_POINTER, Datum *values,
static void static void
InitGraph(HnswGraph * graph, char *base, long memoryTotal) InitGraph(HnswGraph * graph, char *base, long memoryTotal)
{ {
/* Initialize the lock tranche if needed */
HnswInitLockTranche();
HnswPtrStore(base, graph->head, (HnswElement) NULL); HnswPtrStore(base, graph->head, (HnswElement) NULL);
HnswPtrStore(base, graph->entryPoint, (HnswElement) NULL); HnswPtrStore(base, graph->entryPoint, (HnswElement) NULL);
graph->memoryUsed = 0; graph->memoryUsed = 0;
@@ -619,7 +609,6 @@ InitGraph(HnswGraph * graph, char *base, long memoryTotal)
graph->indtuples = 0; graph->indtuples = 0;
SpinLockInit(&graph->lock); SpinLockInit(&graph->lock);
LWLockInitialize(&graph->entryLock, hnsw_lock_tranche_id); LWLockInitialize(&graph->entryLock, hnsw_lock_tranche_id);
LWLockInitialize(&graph->entryWaitLock, hnsw_lock_tranche_id);
LWLockInitialize(&graph->allocatorLock, hnsw_lock_tranche_id); LWLockInitialize(&graph->allocatorLock, hnsw_lock_tranche_id);
LWLockInitialize(&graph->flushLock, hnsw_lock_tranche_id); LWLockInitialize(&graph->flushLock, hnsw_lock_tranche_id);
} }
@@ -680,10 +669,6 @@ InitBuildState(HnswBuildState * buildstate, Relation heap, Relation index, Index
buildstate->efConstruction = HnswGetEfConstruction(index); buildstate->efConstruction = HnswGetEfConstruction(index);
buildstate->dimensions = TupleDescAttr(index->rd_att, 0)->atttypmod; buildstate->dimensions = TupleDescAttr(index->rd_att, 0)->atttypmod;
/* Require column to have dimensions to be indexed */
if (buildstate->dimensions < 0)
elog(ERROR, "column does not have dimensions");
if (buildstate->dimensions > HNSW_MAX_DIM) if (buildstate->dimensions > HNSW_MAX_DIM)
elog(ERROR, "column cannot have more than %d dimensions for hnsw index", HNSW_MAX_DIM); elog(ERROR, "column cannot have more than %d dimensions for hnsw index", HNSW_MAX_DIM);
@@ -985,14 +970,6 @@ HnswBeginParallel(HnswBuildState * buildstate, bool isconcurrent, int request)
/* Report less than allocated so never fails */ /* Report less than allocated so never fails */
InitGraph(&hnswshared->graphData, hnswarea, esthnswarea - 1024 * 1024); InitGraph(&hnswshared->graphData, hnswarea, esthnswarea - 1024 * 1024);
/*
* Avoid base address for relptr for Postgres < 14.5
* https://github.com/postgres/postgres/commit/7201cd18627afc64850537806da7f22150d1a83b
*/
#if PG_VERSION_NUM < 140005
hnswshared->graphData.memoryUsed += MAXALIGN(1);
#endif
shm_toc_insert(pcxt->toc, PARALLEL_KEY_HNSW_SHARED, hnswshared); shm_toc_insert(pcxt->toc, PARALLEL_KEY_HNSW_SHARED, hnswshared);
shm_toc_insert(pcxt->toc, PARALLEL_KEY_HNSW_AREA, hnswarea); shm_toc_insert(pcxt->toc, PARALLEL_KEY_HNSW_AREA, hnswarea);

View File

@@ -860,15 +860,12 @@ HnswSearchLayer(char *base, Datum q, List *ep, int ef, int lc, Relation index, F
static int static int
#if PG_VERSION_NUM >= 130000 #if PG_VERSION_NUM >= 130000
CompareCandidateDistances(const ListCell *a, const ListCell *b) CompareCandidateDistances(const ListCell *a, const ListCell *b)
{
HnswCandidate *hca = lfirst(a);
HnswCandidate *hcb = lfirst(b);
#else #else
CompareCandidateDistances(const void *a, const void *b) CompareCandidateDistances(const void *a, const void *b)
{
HnswCandidate *hca = lfirst(*(ListCell **) a);
HnswCandidate *hcb = lfirst(*(ListCell **) b);
#endif #endif
{
HnswCandidate *hca = lfirst((ListCell *) a);
HnswCandidate *hcb = lfirst((ListCell *) b);
if (hca->distance < hcb->distance) if (hca->distance < hcb->distance)
return 1; return 1;
@@ -891,15 +888,12 @@ CompareCandidateDistances(const void *a, const void *b)
static int static int
#if PG_VERSION_NUM >= 130000 #if PG_VERSION_NUM >= 130000
CompareCandidateDistancesOffset(const ListCell *a, const ListCell *b) CompareCandidateDistancesOffset(const ListCell *a, const ListCell *b)
{
HnswCandidate *hca = lfirst(a);
HnswCandidate *hcb = lfirst(b);
#else #else
CompareCandidateDistancesOffset(const void *a, const void *b) CompareCandidateDistancesOffset(const void *a, const void *b)
{
HnswCandidate *hca = lfirst(*(ListCell **) a);
HnswCandidate *hcb = lfirst(*(ListCell **) b);
#endif #endif
{
HnswCandidate *hca = lfirst((ListCell *) a);
HnswCandidate *hcb = lfirst((ListCell *) b);
if (hca->distance < hcb->distance) if (hca->distance < hcb->distance)
return 1; return 1;
@@ -958,9 +952,7 @@ SelectNeighbors(char *base, List *c, int lm, int lc, FmgrInfo *procinfo, Oid col
{ {
List *r = NIL; List *r = NIL;
List *w = list_copy(c); List *w = list_copy(c);
HnswCandidate **wd; pairingheap *wd;
int wdlen = 0;
int wdoff = 0;
HnswNeighborArray *neighbors = HnswGetNeighbors(base, e2, lc); HnswNeighborArray *neighbors = HnswGetNeighbors(base, e2, lc);
bool mustCalculate = !neighbors->closerSet; bool mustCalculate = !neighbors->closerSet;
List *added = NIL; List *added = NIL;
@@ -969,7 +961,7 @@ SelectNeighbors(char *base, List *c, int lm, int lc, FmgrInfo *procinfo, Oid col
if (list_length(w) <= lm) if (list_length(w) <= lm)
return w; return w;
wd = palloc(sizeof(HnswCandidate *) * list_length(w)); wd = pairingheap_allocate(CompareNearestCandidates, NULL);
/* Ensure order of candidates is deterministic for closer caching */ /* Ensure order of candidates is deterministic for closer caching */
if (sortCandidates) if (sortCandidates)
@@ -1035,21 +1027,21 @@ SelectNeighbors(char *base, List *c, int lm, int lc, FmgrInfo *procinfo, Oid col
if (e->closer) if (e->closer)
r = lappend(r, e); r = lappend(r, e);
else else
wd[wdlen++] = e; pairingheap_add(wd, &(CreatePairingHeapNode(e)->ph_node));
} }
/* Cached value can only be used in future if sorted deterministically */ /* Cached value can only be used in future if sorted deterministically */
neighbors->closerSet = sortCandidates; neighbors->closerSet = sortCandidates;
/* Keep pruned connections */ /* Keep pruned connections */
while (wdoff < wdlen && list_length(r) < lm) while (!pairingheap_is_empty(wd) && list_length(r) < lm)
r = lappend(r, wd[wdoff++]); r = lappend(r, ((HnswPairingHeapNode *) pairingheap_remove_first(wd))->inner);
/* Return pruned for update connections */ /* Return pruned for update connections */
if (pruned != NULL) if (pruned != NULL)
{ {
if (wdoff < wdlen) if (!pairingheap_is_empty(wd))
*pruned = wd[wdoff]; *pruned = ((HnswPairingHeapNode *) pairingheap_first(wd))->inner;
else else
*pruned = linitial(w); *pruned = linitial(w);
} }

113
test/t/020_hnsw_array.pl Normal file
View File

@@ -0,0 +1,113 @@
use strict;
use warnings;
use PostgresNode;
use TestLib;
use Test::More;
my $node;
my @queries = ();
my @expected;
my $limit = 20;
sub test_recall
{
my ($min, $operator) = @_;
my $correct = 0;
my $total = 0;
my $explain = $node->safe_psql("postgres", qq(
SET enable_seqscan = off;
EXPLAIN ANALYZE SELECT i FROM tst ORDER BY v $operator '$queries[0]' LIMIT $limit;
));
like($explain, qr/Index Scan/);
for my $i (0 .. $#queries)
{
my $actual = $node->safe_psql("postgres", qq(
SET enable_seqscan = off;
SELECT i FROM tst ORDER BY v $operator '$queries[$i]' LIMIT $limit;
));
my @actual_ids = split("\n", $actual);
my %actual_set = map { $_ => 1 } @actual_ids;
my @expected_ids = split("\n", $expected[$i]);
foreach (@expected_ids)
{
if (exists($actual_set{$_}))
{
$correct++;
}
$total++;
}
}
cmp_ok($correct / $total, ">=", $min, $operator);
}
# Initialize node
$node = get_new_node('node');
$node->init;
$node->start;
# Create table
$node->safe_psql("postgres", "CREATE EXTENSION vector;");
$node->safe_psql("postgres", "CREATE TABLE tst (i int4, v float4[3]);");
$node->safe_psql("postgres",
"INSERT INTO tst SELECT i, ARRAY[random(), random(), random()] FROM generate_series(1, 10000) i;"
);
$node->safe_psql("postgres", qq(
CREATE FUNCTION float4_l2_distance(float4[], float4[]) RETURNS float8
AS 'BEGIN RETURN l2_distance(\$1::vector, \$2::vector); END;'
LANGUAGE plpgsql IMMUTABLE STRICT PARALLEL SAFE;
CREATE FUNCTION float4_l2_squared_distance(float4[], float4[]) RETURNS float8
AS 'BEGIN RETURN vector_l2_squared_distance(\$1::vector, \$2::vector); END;'
LANGUAGE plpgsql IMMUTABLE STRICT PARALLEL SAFE;
CREATE OPERATOR <-> (
LEFTARG = float4[], RIGHTARG = float4[], PROCEDURE = float4_l2_distance,
COMMUTATOR = '<->'
);
CREATE OPERATOR CLASS float4_l2_ops
FOR TYPE float4[] USING hnsw AS
OPERATOR 1 <-> (float4[], float4[]) FOR ORDER BY float_ops,
FUNCTION 1 float4_l2_squared_distance(float4[], float4[]);
));
# 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 = ("float4_l2_ops");
for my $i (0 .. $#operators)
{
my $operator = $operators[$i];
my $opclass = $opclasses[$i];
# Get exact results
@expected = ();
foreach (@queries)
{
my $res = $node->safe_psql("postgres", "SELECT i FROM tst ORDER BY v $operator '$_' LIMIT $limit;");
push(@expected, $res);
}
# Add index
$node->safe_psql("postgres", "CREATE INDEX ON tst USING hnsw (v $opclass);");
my $min = $operator eq "<#>" ? 0.80 : 0.99;
test_recall($min, $operator);
}
done_testing();

View File

@@ -1,4 +1,4 @@
comment = 'vector data type and ivfflat and hnsw access methods' comment = 'vector data type and ivfflat and hnsw access methods'
default_version = '0.6.1' default_version = '0.6.0'
module_pathname = '$libdir/vector' module_pathname = '$libdir/vector'
relocatable = true relocatable = true