Compare commits

..

1 Commits

Author SHA1 Message Date
Andrew Kane
b7c2068b76 Switched to TidStore for HNSW vacuuming for Postgres 17+ 2026-06-17 11:27:24 -07:00
12 changed files with 71 additions and 29 deletions

View File

@@ -1,4 +1,4 @@
## 0.8.3 (2026-06-17)
## 0.8.3 (unreleased)
- Fixed possible index corruption with HNSW vacuuming
- Fixed performance regression with Hamming distance and Jaccard distance with Postgres 18

View File

@@ -5,7 +5,7 @@ ARG DEBIAN_CODENAME=bookworm
FROM postgres:$PG_MAJOR-$DEBIAN_CODENAME
ARG PG_MAJOR
ADD https://github.com/pgvector/pgvector.git#v0.8.3 /tmp/pgvector
ADD https://github.com/pgvector/pgvector.git#v0.8.2 /tmp/pgvector
RUN apt-get update && \
apt-mark hold locales && \

View File

@@ -2,7 +2,7 @@
"name": "vector",
"abstract": "Open-source vector similarity search for Postgres",
"description": "Supports L2 distance, inner product, and cosine distance",
"version": "0.8.3",
"version": "0.8.2",
"maintainer": [
"Andrew Kane <andrew@ankane.org>"
],
@@ -20,7 +20,7 @@
"vector": {
"file": "sql/vector.sql",
"docfile": "README.md",
"version": "0.8.3",
"version": "0.8.2",
"abstract": "Open-source vector similarity search for Postgres"
}
},

View File

@@ -1,5 +1,5 @@
EXTENSION = vector
EXTVERSION = 0.8.3
EXTVERSION = 0.8.2
MODULE_big = vector
DATA = $(wildcard sql/*--*--*.sql)

View File

@@ -1,5 +1,5 @@
EXTENSION = vector
EXTVERSION = 0.8.3
EXTVERSION = 0.8.2
DATA_built = sql\$(EXTENSION)--$(EXTVERSION).sql
OBJS = src\bitutils.obj src\bitvec.obj src\halfutils.obj src\halfvec.obj src\hnsw.obj src\hnswbuild.obj src\hnswinsert.obj src\hnswscan.obj src\hnswutils.obj src\hnswvacuum.obj src\ivfbuild.obj src\ivfflat.obj src\ivfinsert.obj src\ivfkmeans.obj src\ivfscan.obj src\ivfutils.obj src\ivfvacuum.obj src\sparsevec.obj src\vector.obj

View File

@@ -23,7 +23,7 @@ Compile and install the extension (supports Postgres 13+)
```sh
cd /tmp
git clone --branch v0.8.3 https://github.com/pgvector/pgvector.git
git clone --branch v0.8.2 https://github.com/pgvector/pgvector.git
cd pgvector
make
make install # may need sudo
@@ -40,7 +40,7 @@ Ensure [C++ support in Visual Studio](https://learn.microsoft.com/en-us/cpp/buil
```cmd
set "PGROOT=C:\Program Files\PostgreSQL\18"
cd %TEMP%
git clone --branch v0.8.3 https://github.com/pgvector/pgvector.git
git clone --branch v0.8.2 https://github.com/pgvector/pgvector.git
cd pgvector
nmake /F Makefile.win
nmake /F Makefile.win install
@@ -1151,23 +1151,23 @@ This adds pgvector to the [Postgres image](https://hub.docker.com/_/postgres) (r
Supported tags are:
- `pg18-trixie`, `0.8.3-pg18-trixie`
- `pg18-bookworm`, `0.8.3-pg18-bookworm`, `pg18`, `0.8.3-pg18`
- `pg17-trixie`, `0.8.3-pg17-trixie`
- `pg17-bookworm`, `0.8.3-pg17-bookworm`, `pg17`, `0.8.3-pg17`
- `pg16-trixie`, `0.8.3-pg16-trixie`
- `pg16-bookworm`, `0.8.3-pg16-bookworm`, `pg16`, `0.8.3-pg16`
- `pg15-trixie`, `0.8.3-pg15-trixie`
- `pg15-bookworm`, `0.8.3-pg15-bookworm`, `pg15`, `0.8.3-pg15`
- `pg14-trixie`, `0.8.3-pg14-trixie`
- `pg14-bookworm`, `0.8.3-pg14-bookworm`, `pg14`, `0.8.3-pg14`
- `pg13-trixie`, `0.8.3-pg13-trixie`
- `pg13-bookworm`, `0.8.3-pg13-bookworm`, `pg13`, `0.8.3-pg13`
- `pg18-trixie`, `0.8.2-pg18-trixie`
- `pg18-bookworm`, `0.8.2-pg18-bookworm`, `pg18`, `0.8.2-pg18`
- `pg17-trixie`, `0.8.2-pg17-trixie`
- `pg17-bookworm`, `0.8.2-pg17-bookworm`, `pg17`, `0.8.2-pg17`
- `pg16-trixie`, `0.8.2-pg16-trixie`
- `pg16-bookworm`, `0.8.2-pg16-bookworm`, `pg16`, `0.8.2-pg16`
- `pg15-trixie`, `0.8.2-pg15-trixie`
- `pg15-bookworm`, `0.8.2-pg15-bookworm`, `pg15`, `0.8.2-pg15`
- `pg14-trixie`, `0.8.2-pg14-trixie`
- `pg14-bookworm`, `0.8.2-pg14-bookworm`, `pg14`, `0.8.2-pg14`
- `pg13-trixie`, `0.8.2-pg13-trixie`
- `pg13-bookworm`, `0.8.2-pg13-bookworm`, `pg13`, `0.8.2-pg13`
You can also build the image manually:
```sh
git clone --branch v0.8.3 https://github.com/pgvector/pgvector.git
git clone --branch v0.8.2 https://github.com/pgvector/pgvector.git
cd pgvector
docker build --pull --build-arg PG_MAJOR=18 -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 '0.8.3'" to load this file. \quit

View File

@@ -290,7 +290,11 @@ hnswhandler(PG_FUNCTION_ARGS)
.amcanparallel = false,
.amcanbuildparallel = true,
.amcaninclude = false,
#if PG_VERSION_NUM >= 170000
.amusemaintenanceworkmem = true,
#else
.amusemaintenanceworkmem = false,
#endif
.amsummarizing = false,
.amparallelvacuumoptions = VACUUM_OPTION_PARALLEL_BULKDEL,
.amkeytype = InvalidOid,
@@ -351,7 +355,11 @@ hnswhandler(PG_FUNCTION_ARGS)
amroutine->amcanbuildparallel = true;
#endif
amroutine->amcaninclude = false;
amroutine->amusemaintenanceworkmem = false; /* not used during VACUUM */
#if PG_VERSION_NUM >= 170000
amroutine->amusemaintenanceworkmem = true;
#else
amroutine->amusemaintenanceworkmem = false;
#endif
#if PG_VERSION_NUM >= 160000
amroutine->amsummarizing = false;
#endif

View File

@@ -26,6 +26,12 @@
typedef Pointer Item;
#endif
#if PG_VERSION_NUM >= 170000
#define HnswTidStore TidStore
#else
#define HnswTidStore tidhash_hash
#endif
#define HNSW_MAX_DIM 2000
#define HNSW_MAX_NNZ 1000
@@ -427,7 +433,7 @@ typedef struct HnswVacuumState
HnswSupport support;
/* Variables */
struct tidhash_hash *deleted;
struct HnswTidStore *deleted;
BufferAccessStrategy bas;
HnswNeighborTuple ntup;
HnswElementData highestPoint;

View File

@@ -10,6 +10,12 @@
#include "utils/memutils.h"
#include "utils/rel.h"
#if PG_VERSION_NUM >= 170000
#include "access/tidstore.h"
#include "miscadmin.h"
#include "postmaster/autovacuum.h"
#endif
#if PG_VERSION_NUM >= 160000
#include "varatt.h"
#endif
@@ -22,9 +28,13 @@
* Check if deleted list contains an index TID
*/
static bool
DeletedContains(tidhash_hash * deleted, ItemPointer indextid)
DeletedContains(HnswTidStore * deleted, ItemPointer indextid)
{
#if PG_VERSION_NUM >= 170000
return TidStoreIsMember(deleted, indextid);
#else
return tidhash_lookup(deleted, *indextid) != NULL;
#endif
}
/*
@@ -60,6 +70,10 @@ RemoveHeapTids(HnswVacuumState * vacuumstate)
OffsetNumber offno;
OffsetNumber maxoffno;
bool updated = false;
#if PG_VERSION_NUM >= 170000
OffsetNumber deletedoffs[MaxOffsetNumber];
int ndeletedoffs = 0;
#endif
vacuum_delay_point();
@@ -113,6 +127,9 @@ RemoveHeapTids(HnswVacuumState * vacuumstate)
if (!ItemPointerIsValid(&etup->heaptids[0]))
{
#if PG_VERSION_NUM >= 170000
deletedoffs[ndeletedoffs++] = offno;
#else
ItemPointerData ip;
bool found;
@@ -121,6 +138,7 @@ RemoveHeapTids(HnswVacuumState * vacuumstate)
tidhash_insert(vacuumstate->deleted, ip, &found);
Assert(!found);
#endif
}
else if (etup->level > highestLevel)
{
@@ -149,6 +167,10 @@ RemoveHeapTids(HnswVacuumState * vacuumstate)
}
}
#if PG_VERSION_NUM >= 170000
TidStoreSetBlockOffsets(vacuumstate->deleted, blkno, deletedoffs, ndeletedoffs);
#endif
blkno = HnswPageGetOpaque(page)->nextblkno;
if (updated)
@@ -482,7 +504,7 @@ RepairGraph(HnswVacuumState * vacuumstate)
MemoryContextReset(vacuumstate->tmpCtx);
#ifdef HNSW_VACUUM_PROGRESS
if (!BlockNumberIsValid(blkno) || (blkno - HNSW_HEAD_BLKNO) % 1000 == 0)
if (!BlockNumberIsValid(blkno) || ((blkno - HNSW_HEAD_BLKNO) % 1000 == 0 && blkno != HNSW_HEAD_BLKNO))
{
BlockNumber totalBlocks = RelationGetNumberOfBlocks(index);
BlockNumber currentBlocks = BlockNumberIsValid(blkno) ? blkno : totalBlocks;
@@ -739,7 +761,11 @@ InitVacuumState(HnswVacuumState * vacuumstate, IndexVacuumInfo *info, IndexBulkD
HnswGetMetaPageInfo(index, &vacuumstate->m, NULL);
/* Create hash table */
#if PG_VERSION_NUM >= 170000
vacuumstate->deleted = TidStoreCreateLocal((AmAutoVacuumWorkerProcess() && autovacuum_work_mem != -1) ? autovacuum_work_mem : maintenance_work_mem, true);
#else
vacuumstate->deleted = tidhash_create(CurrentMemoryContext, 256, NULL);
#endif
}
/*
@@ -748,7 +774,11 @@ InitVacuumState(HnswVacuumState * vacuumstate, IndexVacuumInfo *info, IndexBulkD
static void
FreeVacuumState(HnswVacuumState * vacuumstate)
{
#if PG_VERSION_NUM >= 170000
TidStoreDestroy(vacuumstate->deleted);
#else
tidhash_destroy(vacuumstate->deleted);
#endif
FreeAccessStrategy(vacuumstate->bas);
pfree(vacuumstate->ntup);
MemoryContextDelete(vacuumstate->tmpCtx);

View File

@@ -40,7 +40,7 @@
#endif
#if PG_VERSION_NUM >= 180000
PG_MODULE_MAGIC_EXT(.name = "vector", .version = "0.8.3");
PG_MODULE_MAGIC_EXT(.name = "vector", .version = "0.8.2");
#else
PG_MODULE_MAGIC;
#endif

View File

@@ -1,4 +1,4 @@
comment = 'vector data type and ivfflat and hnsw access methods'
default_version = '0.8.3'
default_version = '0.8.2'
module_pathname = '$libdir/vector'
relocatable = true