Compare commits

..

22 Commits

Author SHA1 Message Date
Andrew Kane
cab9da72c0 Version bump to 0.8.2 [skip ci] 2026-02-25 10:46:57 -08:00
Andrew Kane
d702ee31f2 Added test case for low memory HNSW index builds 2026-02-25 10:41:17 -08:00
Andrew Kane
9ed39fb94b Hardened shared memory allocator
Co-authored-by: chungkn from OneMount Group <kimngocchung.k2a@gmail.com>
2026-02-25 10:39:01 -08:00
Andrew Kane
b7e680d41a Fixed integer wraparound leading to buffer overflow with parallel HNSW index build
Reported-by: chungkn from OneMount Group <kimngocchung.k2a@gmail.com>
2026-02-25 10:35:14 -08:00
Andrew Kane
f5841f07fd Updated CI 2026-02-23 14:06:35 -08:00
Andrew Kane
35ab919bf5 Switched to statically-allocated IndexAmRoutine for Postgres 19 [skip ci] 2026-01-21 16:40:35 -08:00
Andrew Kane
5c08be0291 Updated checkout action [skip ci] 2026-01-15 20:45:56 -08:00
Jack Huang
544686feb1 Add RISC-V64 architecture support in Makefile (#948) 2026-01-14 06:11:58 -08:00
Andrew Kane
1a8ebb1129 Updated license year [skip ci] 2026-01-05 13:01:58 -08:00
Andrew Kane
aca89aec73 Switched to scanner_isspace for Postgres 17+ - resolves #942 2026-01-05 12:54:04 -08:00
Andrew Kane
c9875f5150 Simplified test case - resolves #943 2026-01-05 12:42:14 -08:00
Andrew Kane
df2c3d301f Added link to pgvector-cobol [skip ci] 2025-12-13 11:33:35 -08:00
Andrew Kane
d557f1c9b1 Added link to pgvector-tcl [skip ci] 2025-12-11 23:22:56 -08:00
Andrew Kane
3d6dd20469 Removed unneeded check 2025-12-11 16:28:45 -08:00
Andrew Kane
c711da411c Improved includes for indexes 2025-12-11 15:35:37 -08:00
Andrew Kane
3975e212cc Improved includes for ivfkmeans [skip ci] 2025-12-11 13:26:36 -08:00
Andrew Kane
2b741ef068 Improved includes for types 2025-12-11 00:35:18 -08:00
Andrew Kane
f85a7d3fd0 Keep clang-tidy happy [skip ci] 2025-12-10 20:03:26 -08:00
Andrew Kane
5c25f81789 Fixed inconsistent parameter names [skip ci] 2025-12-10 18:49:00 -08:00
Andrew Kane
d72ee71f23 Fixed compilation errors with Postgres 19 2025-12-10 14:59:29 -08:00
Andrew Kane
b7c0085afe Added instructions for APK [skip ci] 2025-12-01 22:52:23 -08:00
Andrew Kane
2b742576b4 Added links [skip ci] 2025-11-24 12:41:30 -08:00
32 changed files with 397 additions and 210 deletions

View File

@@ -23,7 +23,7 @@ jobs:
- postgres: 13 - postgres: 13
os: ubuntu-22.04 os: ubuntu-22.04
steps: steps:
- uses: actions/checkout@v5 - uses: actions/checkout@v6
- uses: ankane/setup-postgres@v1 - uses: ankane/setup-postgres@v1
with: with:
postgres-version: ${{ matrix.postgres }} postgres-version: ${{ matrix.postgres }}
@@ -53,7 +53,7 @@ jobs:
- postgres: 14 - postgres: 14
os: macos-15-intel os: macos-15-intel
steps: steps:
- uses: actions/checkout@v5 - uses: actions/checkout@v6
- uses: ankane/setup-postgres@v1 - uses: ankane/setup-postgres@v1
with: with:
postgres-version: ${{ matrix.postgres }} postgres-version: ${{ matrix.postgres }}
@@ -72,7 +72,7 @@ jobs:
tar xf $TAG.tar.gz tar xf $TAG.tar.gz
mv postgres-$TAG postgres mv postgres-$TAG postgres
env: env:
TAG: ${{ matrix.postgres == 18 && 'REL_18_0' || 'REL_14_19' }} TAG: ${{ matrix.postgres == 18 && 'REL_18_2' || 'REL_14_21' }}
- run: make prove_installcheck PROVE_FLAGS="-I ./postgres/src/test/perl -I ./test/perl" - run: make prove_installcheck PROVE_FLAGS="-I ./postgres/src/test/perl -I ./test/perl"
env: env:
PERL5LIB: /Users/runner/perl5/lib/perl5 PERL5LIB: /Users/runner/perl5/lib/perl5
@@ -92,7 +92,7 @@ jobs:
- postgres: 14 - postgres: 14
os: windows-2022 os: windows-2022
steps: steps:
- uses: actions/checkout@v5 - uses: actions/checkout@v6
- uses: ankane/setup-postgres@v1 - uses: ankane/setup-postgres@v1
with: with:
postgres-version: ${{ matrix.postgres }} postgres-version: ${{ matrix.postgres }}
@@ -133,7 +133,7 @@ jobs:
if: ${{ !startsWith(github.ref_name, 'mac') && !startsWith(github.ref_name, 'windows') }} if: ${{ !startsWith(github.ref_name, 'mac') && !startsWith(github.ref_name, 'windows') }}
runs-on: ubuntu-latest runs-on: ubuntu-latest
steps: steps:
- uses: actions/checkout@v5 - uses: actions/checkout@v6
- uses: ankane/setup-postgres-valgrind@v1 - uses: ankane/setup-postgres-valgrind@v1
with: with:
postgres-version: 18 postgres-version: 18

View File

@@ -1,5 +1,6 @@
## 0.8.2 (unreleased) ## 0.8.2 (2026-02-25)
- Fixed buffer overflow with parallel HNSW index build
- Improved `install` target on Windows - Improved `install` target on Windows
- Fixed `Index Searches` in `EXPLAIN` output for Postgres 18 - Fixed `Index Searches` in `EXPLAIN` output for Postgres 18

View File

@@ -5,7 +5,7 @@ ARG DEBIAN_CODENAME=bookworm
FROM postgres:$PG_MAJOR-$DEBIAN_CODENAME FROM postgres:$PG_MAJOR-$DEBIAN_CODENAME
ARG PG_MAJOR ARG PG_MAJOR
ADD https://github.com/pgvector/pgvector.git#v0.8.1 /tmp/pgvector ADD https://github.com/pgvector/pgvector.git#v0.8.2 /tmp/pgvector
RUN apt-get update && \ RUN apt-get update && \
apt-mark hold locales && \ apt-mark hold locales && \

View File

@@ -1,4 +1,4 @@
Portions Copyright (c) 1996-2025, PostgreSQL Global Development Group Portions Copyright (c) 1996-2026, 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": "0.8.1", "version": "0.8.2",
"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": "0.8.1", "version": "0.8.2",
"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 = 0.8.1 EXTVERSION = 0.8.2
MODULE_big = vector MODULE_big = vector
DATA = $(wildcard sql/*--*--*.sql) DATA = $(wildcard sql/*--*--*.sql)
@@ -27,6 +27,11 @@ ifneq ($(filter ppc64%, $(shell uname -m)), )
OPTFLAGS = OPTFLAGS =
endif endif
# RISC-V64 doesn't support -march=native
ifeq ($(shell uname -m), riscv64)
OPTFLAGS =
endif
# For auto-vectorization: # For auto-vectorization:
# - GCC (needs -ftree-vectorize OR -O3) - https://gcc.gnu.org/projects/tree-ssa/vectorization.html # - GCC (needs -ftree-vectorize OR -O3) - https://gcc.gnu.org/projects/tree-ssa/vectorization.html
# - Clang (could use pragma instead) - https://llvm.org/docs/Vectorizers.html # - Clang (could use pragma instead) - https://llvm.org/docs/Vectorizers.html

View File

@@ -1,5 +1,5 @@
EXTENSION = vector EXTENSION = vector
EXTVERSION = 0.8.1 EXTVERSION = 0.8.2
DATA_built = sql\$(EXTENSION)--$(EXTVERSION).sql DATA_built = sql\$(EXTENSION)--$(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 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

@@ -21,7 +21,7 @@ Compile and install the extension (supports Postgres 13+)
```sh ```sh
cd /tmp cd /tmp
git clone --branch v0.8.1 https://github.com/pgvector/pgvector.git git clone --branch v0.8.2 https://github.com/pgvector/pgvector.git
cd pgvector cd pgvector
make make
make install # may need sudo make install # may need sudo
@@ -29,7 +29,7 @@ make install # may need sudo
See the [installation notes](#installation-notes---linux-and-mac) if you run into issues See the [installation notes](#installation-notes---linux-and-mac) 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), [pkg](#pkg), [APK](#apk), 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
@@ -38,7 +38,7 @@ Ensure [C++ support in Visual Studio](https://learn.microsoft.com/en-us/cpp/buil
```cmd ```cmd
set "PGROOT=C:\Program Files\PostgreSQL\18" set "PGROOT=C:\Program Files\PostgreSQL\18"
cd %TEMP% cd %TEMP%
git clone --branch v0.8.1 https://github.com/pgvector/pgvector.git git clone --branch v0.8.2 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
@@ -772,9 +772,11 @@ Use pgvector from any language with a Postgres client. You can even generate and
Language | Libraries / Examples Language | Libraries / Examples
--- | --- --- | ---
Ada | [pgvector-ada](https://github.com/pgvector/pgvector-ada) Ada | [pgvector-ada](https://github.com/pgvector/pgvector-ada)
Algol | [pgvector-algol](https://github.com/pgvector/pgvector-algol)
C | [pgvector-c](https://github.com/pgvector/pgvector-c) C | [pgvector-c](https://github.com/pgvector/pgvector-c)
C++ | [pgvector-cpp](https://github.com/pgvector/pgvector-cpp) C++ | [pgvector-cpp](https://github.com/pgvector/pgvector-cpp)
C#, F#, Visual Basic | [pgvector-dotnet](https://github.com/pgvector/pgvector-dotnet) C#, F#, Visual Basic | [pgvector-dotnet](https://github.com/pgvector/pgvector-dotnet)
COBOL | [pgvector-cobol](https://github.com/pgvector/pgvector-cobol)
Crystal | [pgvector-crystal](https://github.com/pgvector/pgvector-crystal) Crystal | [pgvector-crystal](https://github.com/pgvector/pgvector-crystal)
D | [pgvector-d](https://github.com/pgvector/pgvector-d) D | [pgvector-d](https://github.com/pgvector/pgvector-d)
Dart | [pgvector-dart](https://github.com/pgvector/pgvector-dart) Dart | [pgvector-dart](https://github.com/pgvector/pgvector-dart)
@@ -794,12 +796,15 @@ OCaml | [pgvector-ocaml](https://github.com/pgvector/pgvector-ocaml)
Pascal | [pgvector-pascal](https://github.com/pgvector/pgvector-pascal) Pascal | [pgvector-pascal](https://github.com/pgvector/pgvector-pascal)
Perl | [pgvector-perl](https://github.com/pgvector/pgvector-perl) Perl | [pgvector-perl](https://github.com/pgvector/pgvector-perl)
PHP | [pgvector-php](https://github.com/pgvector/pgvector-php) PHP | [pgvector-php](https://github.com/pgvector/pgvector-php)
Prolog | [pgvector-prolog](https://github.com/pgvector/pgvector-prolog)
Python | [pgvector-python](https://github.com/pgvector/pgvector-python) Python | [pgvector-python](https://github.com/pgvector/pgvector-python)
R | [pgvector-r](https://github.com/pgvector/pgvector-r) R | [pgvector-r](https://github.com/pgvector/pgvector-r)
Racket | [pgvector-racket](https://github.com/pgvector/pgvector-racket)
Raku | [pgvector-raku](https://github.com/pgvector/pgvector-raku) Raku | [pgvector-raku](https://github.com/pgvector/pgvector-raku)
Ruby | [pgvector-ruby](https://github.com/pgvector/pgvector-ruby), [Neighbor](https://github.com/ankane/neighbor) Ruby | [pgvector-ruby](https://github.com/pgvector/pgvector-ruby), [Neighbor](https://github.com/ankane/neighbor)
Rust | [pgvector-rust](https://github.com/pgvector/pgvector-rust) Rust | [pgvector-rust](https://github.com/pgvector/pgvector-rust)
Swift | [pgvector-swift](https://github.com/pgvector/pgvector-swift) Swift | [pgvector-swift](https://github.com/pgvector/pgvector-swift)
Tcl | [pgvector-tcl](https://github.com/pgvector/pgvector-tcl)
Zig | [pgvector-zig](https://github.com/pgvector/pgvector-zig) Zig | [pgvector-zig](https://github.com/pgvector/pgvector-zig)
## Frequently Asked Questions ## Frequently Asked Questions
@@ -1147,23 +1152,23 @@ This adds pgvector to the [Postgres image](https://hub.docker.com/_/postgres) (r
Supported tags are: Supported tags are:
- `pg18-trixie`, `0.8.1-pg18-trixie` - `pg18-trixie`, `0.8.2-pg18-trixie`
- `pg18-bookworm`, `0.8.1-pg18-bookworm`, `pg18`, `0.8.1-pg18` - `pg18-bookworm`, `0.8.2-pg18-bookworm`, `pg18`, `0.8.2-pg18`
- `pg17-trixie`, `0.8.1-pg17-trixie` - `pg17-trixie`, `0.8.2-pg17-trixie`
- `pg17-bookworm`, `0.8.1-pg17-bookworm`, `pg17`, `0.8.1-pg17` - `pg17-bookworm`, `0.8.2-pg17-bookworm`, `pg17`, `0.8.2-pg17`
- `pg16-trixie`, `0.8.1-pg16-trixie` - `pg16-trixie`, `0.8.2-pg16-trixie`
- `pg16-bookworm`, `0.8.1-pg16-bookworm`, `pg16`, `0.8.1-pg16` - `pg16-bookworm`, `0.8.2-pg16-bookworm`, `pg16`, `0.8.2-pg16`
- `pg15-trixie`, `0.8.1-pg15-trixie` - `pg15-trixie`, `0.8.2-pg15-trixie`
- `pg15-bookworm`, `0.8.1-pg15-bookworm`, `pg15`, `0.8.1-pg15` - `pg15-bookworm`, `0.8.2-pg15-bookworm`, `pg15`, `0.8.2-pg15`
- `pg14-trixie`, `0.8.1-pg14-trixie` - `pg14-trixie`, `0.8.2-pg14-trixie`
- `pg14-bookworm`, `0.8.1-pg14-bookworm`, `pg14`, `0.8.1-pg14` - `pg14-bookworm`, `0.8.2-pg14-bookworm`, `pg14`, `0.8.2-pg14`
- `pg13-trixie`, `0.8.1-pg13-trixie` - `pg13-trixie`, `0.8.2-pg13-trixie`
- `pg13-bookworm`, `0.8.1-pg13-bookworm`, `pg13`, `0.8.1-pg13` - `pg13-bookworm`, `0.8.2-pg13-bookworm`, `pg13`, `0.8.2-pg13`
You can also build the image manually: You can also build the image manually:
```sh ```sh
git clone --branch v0.8.1 https://github.com/pgvector/pgvector.git git clone --branch v0.8.2 https://github.com/pgvector/pgvector.git
cd pgvector cd pgvector
docker build --pull --build-arg PG_MAJOR=18 -t myuser/pgvector . docker build --pull --build-arg PG_MAJOR=18 -t myuser/pgvector .
``` ```
@@ -1229,6 +1234,14 @@ cd /usr/ports/databases/pgvector
make install make install
``` ```
### APK
Install the Alpine package with:
```sh
apk add postgresql-pgvector
```
### 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

@@ -0,0 +1,2 @@
-- complain if script is sourced in psql, rather than via CREATE EXTENSION
\echo Use "ALTER EXTENSION vector UPDATE TO '0.8.2'" to load this file. \quit

View File

@@ -2,6 +2,7 @@
#include "bitutils.h" #include "bitutils.h"
#include "bitvec.h" #include "bitvec.h"
#include "fmgr.h"
#include "utils/varbit.h" #include "utils/varbit.h"
#include "vector.h" #include "vector.h"

View File

@@ -1,5 +1,7 @@
#include "postgres.h" #include "postgres.h"
#include <math.h>
#include "halfutils.h" #include "halfutils.h"
#include "halfvec.h" #include "halfvec.h"

View File

@@ -13,12 +13,20 @@
#include "port.h" /* for strtof() */ #include "port.h" /* for strtof() */
#include "sparsevec.h" #include "sparsevec.h"
#include "utils/array.h" #include "utils/array.h"
#include "utils/builtins.h"
#include "utils/float.h" #include "utils/float.h"
#include "utils/fmgrprotos.h"
#include "utils/lsyscache.h" #include "utils/lsyscache.h"
#include "utils/numeric.h" #include "utils/varbit.h"
#include "vector.h" #include "vector.h"
#if PG_VERSION_NUM >= 160000
#include "varatt.h"
#endif
#if PG_VERSION_NUM >= 170000
#include "parser/scansup.h"
#endif
#define STATE_DIMS(x) (ARR_DIMS(x)[0] - 1) #define STATE_DIMS(x) (ARR_DIMS(x)[0] - 1)
#define CreateStateDatums(dim) palloc(sizeof(Datum) * (dim + 1)) #define CreateStateDatums(dim) palloc(sizeof(Datum) * (dim + 1))
@@ -129,9 +137,9 @@ InitHalfVector(int dim)
return result; return result;
} }
/* #if PG_VERSION_NUM >= 170000
* Check for whitespace, since array_isspace() is static #define halfvec_isspace(ch) scanner_isspace(ch)
*/ #else
static inline bool static inline bool
halfvec_isspace(char ch) halfvec_isspace(char ch)
{ {
@@ -144,6 +152,7 @@ halfvec_isspace(char ch)
return true; return true;
return false; return false;
} }
#endif
/* /*
* Check state array * Check state array

View File

@@ -1,18 +1,24 @@
#include "postgres.h" #include "postgres.h"
#include <float.h> #include <float.h>
#include <limits.h>
#include <math.h> #include <math.h>
#include "access/amapi.h" #include "access/amapi.h"
#include "access/genam.h"
#include "access/reloptions.h" #include "access/reloptions.h"
#include "commands/progress.h" #include "commands/progress.h"
#include "commands/vacuum.h" #include "commands/vacuum.h"
#include "fmgr.h"
#include "hnsw.h" #include "hnsw.h"
#include "miscadmin.h" #include "miscadmin.h"
#include "nodes/pg_list.h"
#include "utils/float.h" #include "utils/float.h"
#include "utils/guc.h" #include "utils/guc.h"
#include "utils/relcache.h"
#include "utils/selfuncs.h" #include "utils/selfuncs.h"
#include "utils/spccache.h" #include "utils/spccache.h"
#include "vector.h"
#if PG_VERSION_NUM < 150000 #if PG_VERSION_NUM < 150000
#define MarkGUCPrefixReserved(x) EmitWarningsOnPlaceholders(x) #define MarkGUCPrefixReserved(x) EmitWarningsOnPlaceholders(x)
@@ -260,6 +266,64 @@ FUNCTION_PREFIX PG_FUNCTION_INFO_V1(hnswhandler);
Datum Datum
hnswhandler(PG_FUNCTION_ARGS) hnswhandler(PG_FUNCTION_ARGS)
{ {
#if PG_VERSION_NUM >= 190000
static const IndexAmRoutine amroutine = {
.type = T_IndexAmRoutine,
.amstrategies = 0,
.amsupport = 3,
.amoptsprocnum = 0,
.amcanorder = false,
.amcanorderbyop = true,
.amcanhash = false,
.amconsistentequality = false,
.amconsistentordering = false,
.amcanbackward = false,
.amcanunique = false,
.amcanmulticol = false,
.amoptionalkey = true,
.amsearcharray = false,
.amsearchnulls = false,
.amstorage = false,
.amclusterable = false,
.ampredlocks = false,
.amcanparallel = false,
.amcanbuildparallel = true,
.amcaninclude = false,
.amusemaintenanceworkmem = false,
.amsummarizing = false,
.amparallelvacuumoptions = VACUUM_OPTION_PARALLEL_BULKDEL,
.amkeytype = InvalidOid,
.ambuild = hnswbuild,
.ambuildempty = hnswbuildempty,
.aminsert = hnswinsert,
.aminsertcleanup = NULL,
.ambulkdelete = hnswbulkdelete,
.amvacuumcleanup = hnswvacuumcleanup,
.amcanreturn = NULL,
.amcostestimate = hnswcostestimate,
.amgettreeheight = NULL,
.amoptions = hnswoptions,
.amproperty = NULL,
.ambuildphasename = hnswbuildphasename,
.amvalidate = hnswvalidate,
.amadjustmembers = NULL,
.ambeginscan = hnswbeginscan,
.amrescan = hnswrescan,
.amgettuple = hnswgettuple,
.amgetbitmap = NULL,
.amendscan = hnswendscan,
.ammarkpos = NULL,
.amrestrpos = NULL,
.amestimateparallelscan = NULL,
.aminitparallelscan = NULL,
.amparallelrescan = NULL,
.amtranslatestrategy = NULL,
.amtranslatecmptype = NULL,
};
PG_RETURN_POINTER(&amroutine);
#else
IndexAmRoutine *amroutine = makeNode(IndexAmRoutine); IndexAmRoutine *amroutine = makeNode(IndexAmRoutine);
amroutine->amstrategies = 0; amroutine->amstrategies = 0;
@@ -333,4 +397,5 @@ hnswhandler(PG_FUNCTION_ARGS)
#endif #endif
PG_RETURN_POINTER(amroutine); PG_RETURN_POINTER(amroutine);
#endif
} }

View File

@@ -3,6 +3,8 @@
#include "postgres.h" #include "postgres.h"
#include <math.h>
#include "access/genam.h" #include "access/genam.h"
#include "access/parallel.h" #include "access/parallel.h"
#include "lib/pairingheap.h" #include "lib/pairingheap.h"
@@ -12,6 +14,10 @@
#include "utils/sampling.h" #include "utils/sampling.h"
#include "vector.h" #include "vector.h"
#if PG_VERSION_NUM >= 190000
typedef Pointer Item;
#endif
#define HNSW_MAX_DIM 2000 #define HNSW_MAX_DIM 2000
#define HNSW_MAX_NNZ 1000 #define HNSW_MAX_NNZ 1000
@@ -362,13 +368,6 @@ typedef union
ItemPointerData indextid; ItemPointerData indextid;
} HnswUnvisited; } HnswUnvisited;
typedef struct HnswReadStreamData
{
HnswUnvisited *unvisited;
int unvisitedLength;
int visited;
} HnswReadStreamData;
typedef struct HnswScanOpaqueData typedef struct HnswScanOpaqueData
{ {
const HnswTypeInfo *typeInfo; const HnswTypeInfo *typeInfo;
@@ -424,14 +423,14 @@ bool HnswCheckNorm(HnswSupport * support, Datum value);
Buffer HnswNewBuffer(Relation index, ForkNumber forkNum); Buffer HnswNewBuffer(Relation index, ForkNumber forkNum);
void HnswInitPage(Buffer buf, Page page); void HnswInitPage(Buffer buf, Page page);
void HnswInit(void); void HnswInit(void);
List *HnswSearchLayer(char *base, HnswQuery * q, List *ep, int ef, int lc, Relation index, HnswSupport * support, int m, bool inserting, HnswElement skipElement, visited_hash * v, pairingheap **discarded, bool initVisited, int64 *tuples, bool maintenance); List *HnswSearchLayer(char *base, HnswQuery * q, List *ep, int ef, int lc, Relation index, HnswSupport * support, int m, bool inserting, HnswElement skipElement, visited_hash * v, pairingheap **discarded, bool initVisited, int64 *tuples);
HnswElement HnswGetEntryPoint(Relation index); HnswElement HnswGetEntryPoint(Relation index);
void HnswGetMetaPageInfo(Relation index, int *m, HnswElement * entryPoint); void HnswGetMetaPageInfo(Relation index, int *m, HnswElement * entryPoint);
void *HnswAlloc(HnswAllocator * allocator, Size size); void *HnswAlloc(HnswAllocator * allocator, Size size);
HnswElement HnswInitElement(char *base, ItemPointer tid, int m, double ml, int maxLevel, HnswAllocator * alloc); HnswElement HnswInitElement(char *base, ItemPointer tid, int m, double ml, int maxLevel, HnswAllocator * alloc);
HnswElement HnswInitElementFromBlock(BlockNumber blkno, OffsetNumber offno); HnswElement HnswInitElementFromBlock(BlockNumber blkno, OffsetNumber offno);
void HnswFindElementNeighbors(char *base, HnswElement element, HnswElement entryPoint, Relation index, HnswSupport * support, int m, int efConstruction, bool existing, bool maintenance); void HnswFindElementNeighbors(char *base, HnswElement element, HnswElement entryPoint, Relation index, HnswSupport * support, int m, int efConstruction, bool existing);
HnswSearchCandidate *HnswEntryCandidate(char *base, HnswElement em, HnswQuery * q, Relation rel, HnswSupport * support, bool loadVec); HnswSearchCandidate *HnswEntryCandidate(char *base, HnswElement entryPoint, HnswQuery * q, Relation index, HnswSupport * support, bool loadVec);
void HnswUpdateMetaPage(Relation index, int updateEntry, HnswElement entryPoint, BlockNumber insertPage, ForkNumber forkNum, bool building); 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 HnswSetNeighborTuple(char *base, HnswNeighborTuple ntup, HnswElement e, int m);
void HnswAddHeapTid(HnswElement element, ItemPointer heaptid); void HnswAddHeapTid(HnswElement element, ItemPointer heaptid);

View File

@@ -36,11 +36,12 @@
*/ */
#include "postgres.h" #include "postgres.h"
#include <math.h> #include "access/genam.h"
#include "access/parallel.h" #include "access/parallel.h"
#include "access/relscan.h"
#include "access/table.h" #include "access/table.h"
#include "access/tableam.h" #include "access/tableam.h"
#include "access/tupdesc.h"
#include "access/xact.h" #include "access/xact.h"
#include "access/xloginsert.h" #include "access/xloginsert.h"
#include "catalog/index.h" #include "catalog/index.h"
@@ -48,11 +49,14 @@
#include "commands/progress.h" #include "commands/progress.h"
#include "hnsw.h" #include "hnsw.h"
#include "miscadmin.h" #include "miscadmin.h"
#include "nodes/execnodes.h"
#include "optimizer/optimizer.h" #include "optimizer/optimizer.h"
#include "storage/bufmgr.h" #include "storage/bufmgr.h"
#include "tcop/tcopprot.h" #include "tcop/tcopprot.h"
#include "utils/datum.h" #include "utils/datum.h"
#include "utils/memutils.h" #include "utils/memutils.h"
#include "utils/rel.h"
#include "utils/snapmgr.h"
#if PG_VERSION_NUM >= 160000 #if PG_VERSION_NUM >= 160000
#include "varatt.h" #include "varatt.h"
@@ -73,6 +77,8 @@
#define PARALLEL_KEY_HNSW_AREA UINT64CONST(0xA000000000000002) #define PARALLEL_KEY_HNSW_AREA UINT64CONST(0xA000000000000002)
#define PARALLEL_KEY_QUERY_TEXT UINT64CONST(0xA000000000000003) #define PARALLEL_KEY_QUERY_TEXT UINT64CONST(0xA000000000000003)
#define HNSW_MAX_GRAPH_MEMORY (SIZE_MAX / 2)
/* /*
* Create the metapage * Create the metapage
*/ */
@@ -461,7 +467,7 @@ InsertTupleInMemory(HnswBuildState * buildstate, HnswElement element)
} }
/* Find neighbors for element */ /* Find neighbors for element */
HnswFindElementNeighbors(base, element, entryPoint, NULL, support, m, efConstruction, false, true); HnswFindElementNeighbors(base, element, entryPoint, NULL, support, m, efConstruction, false);
/* Update graph in memory */ /* Update graph in memory */
UpdateGraphInMemory(support, element, m, entryPoint, buildstate); UpdateGraphInMemory(support, element, m, entryPoint, buildstate);
@@ -485,6 +491,7 @@ InsertTuple(Relation index, Datum *values, bool *isnull, ItemPointer heaptid, Hn
LWLock *flushLock = &graph->flushLock; LWLock *flushLock = &graph->flushLock;
char *base = buildstate->hnswarea; char *base = buildstate->hnswarea;
Datum value; Datum value;
Size memoryMargin;
/* Form index value */ /* Form index value */
if (!HnswFormIndexValue(&value, values, isnull, buildstate->typeInfo, support)) if (!HnswFormIndexValue(&value, values, isnull, buildstate->typeInfo, support))
@@ -493,6 +500,9 @@ InsertTuple(Relation index, Datum *values, bool *isnull, ItemPointer heaptid, Hn
/* Get datum size */ /* Get datum size */
valueSize = VARSIZE_ANY(DatumGetPointer(value)); valueSize = VARSIZE_ANY(DatumGetPointer(value));
/* In a parallel build, add a margin so allocations never fail */
memoryMargin = base == NULL ? 0 : 1024 * 1024;
/* Ensure graph not flushed when inserting */ /* Ensure graph not flushed when inserting */
LWLockAcquire(flushLock, LW_SHARED); LWLockAcquire(flushLock, LW_SHARED);
@@ -514,7 +524,7 @@ InsertTuple(Relation index, Datum *values, bool *isnull, ItemPointer heaptid, Hn
* Check that we have enough memory available for the new element now that * Check that we have enough memory available for the new element now that
* we have the allocator lock, and flush pages if needed. * we have the allocator lock, and flush pages if needed.
*/ */
if (graph->memoryUsed >= graph->memoryTotal) if (graph->memoryUsed + memoryMargin >= graph->memoryTotal)
{ {
LWLockRelease(&graph->allocatorLock); LWLockRelease(&graph->allocatorLock);
@@ -549,7 +559,7 @@ InsertTuple(Relation index, Datum *values, bool *isnull, ItemPointer heaptid, Hn
/* Copy the datum */ /* Copy the datum */
memcpy(valuePtr, DatumGetPointer(value), valueSize); memcpy(valuePtr, DatumGetPointer(value), valueSize);
HnswPtrStore(base, element->value, valuePtr); HnswPtrStore(base, element->value, (char *) valuePtr);
/* Create a lock for the element */ /* Create a lock for the element */
LWLockInitialize(&element->lock, hnsw_lock_tranche_id); LWLockInitialize(&element->lock, hnsw_lock_tranche_id);
@@ -607,7 +617,7 @@ InitGraph(HnswGraph * graph, char *base, Size memoryTotal)
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;
graph->memoryTotal = memoryTotal; graph->memoryTotal = Min(memoryTotal, HNSW_MAX_GRAPH_MEMORY);
graph->flushed = false; graph->flushed = false;
graph->indtuples = 0; graph->indtuples = 0;
SpinLockInit(&graph->lock); SpinLockInit(&graph->lock);
@@ -648,9 +658,17 @@ static void *
HnswSharedMemoryAlloc(Size size, void *state) HnswSharedMemoryAlloc(Size size, void *state)
{ {
HnswBuildState *buildstate = (HnswBuildState *) state; HnswBuildState *buildstate = (HnswBuildState *) state;
void *chunk = buildstate->hnswarea + buildstate->graph->memoryUsed; Size alignedSize = MAXALIGN(size);
void *chunk;
buildstate->graph->memoryUsed += MAXALIGN(size); if (alignedSize > 1024 * 1024)
elog(ERROR, "hnsw allocation too large");
if (buildstate->graph->memoryUsed + alignedSize > buildstate->graph->memoryTotal)
elog(ERROR, "hnsw allocator out of memory");
chunk = buildstate->hnswarea + buildstate->graph->memoryUsed;
buildstate->graph->memoryUsed += alignedSize;
return chunk; return chunk;
} }
@@ -936,6 +954,8 @@ HnswBeginParallel(HnswBuildState * buildstate, bool isconcurrent, int request)
if (esthnswarea > estother) if (esthnswarea > estother)
esthnswarea -= estother; esthnswarea -= estother;
esthnswarea = Min(esthnswarea, HNSW_MAX_GRAPH_MEMORY);
shm_toc_estimate_chunk(&pcxt->estimator, esthnswarea); shm_toc_estimate_chunk(&pcxt->estimator, esthnswarea);
shm_toc_estimate_keys(&pcxt->estimator, 2); shm_toc_estimate_keys(&pcxt->estimator, 2);
@@ -978,8 +998,7 @@ HnswBeginParallel(HnswBuildState * buildstate, bool isconcurrent, int request)
snapshot); snapshot);
hnswarea = (char *) shm_toc_allocate(pcxt->toc, esthnswarea); hnswarea = (char *) shm_toc_allocate(pcxt->toc, esthnswarea);
/* Report less than allocated so never fails */ InitGraph(&hnswshared->graphData, hnswarea, esthnswarea);
InitGraph(&hnswshared->graphData, hnswarea, esthnswarea - 1024 * 1024);
/* /*
* Avoid base address for relptr for Postgres < 14.5 * Avoid base address for relptr for Postgres < 14.5

View File

@@ -1,13 +1,14 @@
#include "postgres.h" #include "postgres.h"
#include <math.h> #include "access/genam.h"
#include "access/generic_xlog.h" #include "access/generic_xlog.h"
#include "hnsw.h" #include "hnsw.h"
#include "nodes/execnodes.h"
#include "storage/bufmgr.h" #include "storage/bufmgr.h"
#include "storage/lmgr.h" #include "storage/lmgr.h"
#include "utils/datum.h" #include "utils/datum.h"
#include "utils/memutils.h" #include "utils/memutils.h"
#include "utils/rel.h"
#if PG_VERSION_NUM >= 160000 #if PG_VERSION_NUM >= 160000
#include "varatt.h" #include "varatt.h"
@@ -712,7 +713,7 @@ HnswInsertTupleOnDisk(Relation index, HnswSupport * support, Datum value, ItemPo
/* Create an element */ /* Create an element */
element = HnswInitElement(base, heaptid, m, HnswGetMl(m), HnswGetMaxLevel(m), NULL); element = HnswInitElement(base, heaptid, m, HnswGetMl(m), HnswGetMaxLevel(m), NULL);
HnswPtrStore(base, element->value, DatumGetPointer(value)); HnswPtrStore(base, element->value, (char *) DatumGetPointer(value));
/* Prevent concurrent inserts when likely updating entry point */ /* Prevent concurrent inserts when likely updating entry point */
if (entryPoint == NULL || element->level > entryPoint->level) if (entryPoint == NULL || element->level > entryPoint->level)
@@ -729,7 +730,7 @@ HnswInsertTupleOnDisk(Relation index, HnswSupport * support, Datum value, ItemPo
} }
/* Find neighbors for element */ /* Find neighbors for element */
HnswFindElementNeighbors(base, element, entryPoint, index, support, m, efConstruction, false, building); HnswFindElementNeighbors(base, element, entryPoint, index, support, m, efConstruction, false);
/* Update graph on disk */ /* Update graph on disk */
UpdateGraphOnDisk(index, support, element, m, entryPoint, building); UpdateGraphOnDisk(index, support, element, m, entryPoint, building);

View File

@@ -1,12 +1,21 @@
#include "postgres.h" #include "postgres.h"
#include "access/genam.h"
#include "access/relscan.h" #include "access/relscan.h"
#include "hnsw.h" #include "hnsw.h"
#include "lib/pairingheap.h"
#include "miscadmin.h"
#include "nodes/pg_list.h"
#include "pgstat.h" #include "pgstat.h"
#include "storage/bufmgr.h"
#include "storage/lmgr.h" #include "storage/lmgr.h"
#include "utils/float.h" #include "utils/float.h"
#include "utils/memutils.h" #include "utils/memutils.h"
#include "utils/relcache.h"
#include "utils/snapmgr.h"
#if PG_VERSION_NUM >= 160000
#include "varatt.h"
#endif
/* /*
* Algorithm 5 from paper * Algorithm 5 from paper
@@ -37,11 +46,11 @@ GetScanItems(IndexScanDesc scan, Datum value)
for (int lc = entryPoint->level; lc >= 1; lc--) for (int lc = entryPoint->level; lc >= 1; lc--)
{ {
w = HnswSearchLayer(base, q, ep, 1, lc, index, support, m, false, NULL, NULL, NULL, true, NULL, false); w = HnswSearchLayer(base, q, ep, 1, lc, index, support, m, false, NULL, NULL, NULL, true, NULL);
ep = w; ep = w;
} }
return HnswSearchLayer(base, q, ep, hnsw_ef_search, 0, index, support, m, false, NULL, &so->v, hnsw_iterative_scan != HNSW_ITERATIVE_SCAN_OFF ? &so->discarded : NULL, true, &so->tuples, false); return HnswSearchLayer(base, q, ep, hnsw_ef_search, 0, index, support, m, false, NULL, &so->v, hnsw_iterative_scan != HNSW_ITERATIVE_SCAN_OFF ? &so->discarded : NULL, true, &so->tuples);
} }
/* /*
@@ -72,7 +81,7 @@ ResumeScanItems(IndexScanDesc scan)
ep = lappend(ep, sc); ep = lappend(ep, sc);
} }
return HnswSearchLayer(base, &so->q, ep, batch_size, 0, index, &so->support, so->m, false, NULL, &so->v, &so->discarded, false, &so->tuples, false); return HnswSearchLayer(base, &so->q, ep, batch_size, 0, index, &so->support, so->m, false, NULL, &so->v, &so->discarded, false, &so->tuples);
} }
/* /*

View File

@@ -2,27 +2,25 @@
#include <math.h> #include <math.h>
#include "access/genam.h"
#include "access/generic_xlog.h" #include "access/generic_xlog.h"
#include "catalog/pg_type.h"
#include "catalog/pg_type_d.h"
#include "common/hashfn.h" #include "common/hashfn.h"
#include "fmgr.h" #include "fmgr.h"
#include "hnsw.h" #include "hnsw.h"
#include "lib/pairingheap.h" #include "lib/pairingheap.h"
#include "nodes/pg_list.h"
#include "port/atomics.h"
#include "sparsevec.h" #include "sparsevec.h"
#include "storage/bufmgr.h" #include "storage/bufmgr.h"
#include "utils/datum.h" #include "utils/datum.h"
#include "utils/memdebug.h" #include "utils/memdebug.h"
#include "utils/rel.h" #include "utils/rel.h"
#include "vector.h"
#if PG_VERSION_NUM >= 160000 #if PG_VERSION_NUM >= 160000
#include "varatt.h" #include "varatt.h"
#endif #endif
#if PG_VERSION_NUM >= 170000
#include "storage/read_stream.h"
#endif
#if PG_VERSION_NUM < 170000 #if PG_VERSION_NUM < 170000
static inline uint64 static inline uint64
murmurhash64(uint64 data) murmurhash64(uint64 data)
@@ -264,7 +262,7 @@ HnswInitElement(char *base, ItemPointer heaptid, int m, double ml, int maxLevel,
HnswInitNeighbors(base, element, m, allocator); HnswInitNeighbors(base, element, m, allocator);
HnswPtrStore(base, element->value, (Pointer) NULL); HnswPtrStore(base, element->value, (char *) NULL);
return element; return element;
} }
@@ -290,7 +288,7 @@ HnswInitElementFromBlock(BlockNumber blkno, OffsetNumber offno)
element->blkno = blkno; element->blkno = blkno;
element->offno = offno; element->offno = offno;
HnswPtrStore(base, element->neighbors, (HnswNeighborArrayPtr *) NULL); HnswPtrStore(base, element->neighbors, (HnswNeighborArrayPtr *) NULL);
HnswPtrStore(base, element->value, (Pointer) NULL); HnswPtrStore(base, element->value, (char *) NULL);
return element; return element;
} }
@@ -516,7 +514,7 @@ HnswLoadElementFromTuple(HnswElement element, HnswElementTuple etup, bool loadHe
char *base = NULL; char *base = NULL;
Datum value = datumCopy(PointerGetDatum(&etup->data), false, -1); Datum value = datumCopy(PointerGetDatum(&etup->data), false, -1);
HnswPtrStore(base, element->value, DatumGetPointer(value)); HnswPtrStore(base, element->value, (char *) DatumGetPointer(value));
} }
} }
@@ -533,12 +531,14 @@ HnswGetDistance(Datum a, Datum b, HnswSupport * support)
* Load an element and optionally get its distance from q * Load an element and optionally get its distance from q
*/ */
static void static void
HnswLoadElementImpl(Buffer buf, OffsetNumber offno, double *distance, HnswQuery * q, Relation index, HnswSupport * support, bool loadVec, double *maxDistance, HnswElement * element) HnswLoadElementImpl(BlockNumber blkno, OffsetNumber offno, double *distance, HnswQuery * q, Relation index, HnswSupport * support, bool loadVec, double *maxDistance, HnswElement * element)
{ {
Buffer buf;
Page page; Page page;
HnswElementTuple etup; HnswElementTuple etup;
/* Read vector */ /* Read vector */
buf = ReadBuffer(index, blkno);
LockBuffer(buf, BUFFER_LOCK_SHARE); LockBuffer(buf, BUFFER_LOCK_SHARE);
page = BufferGetPage(buf); page = BufferGetPage(buf);
@@ -559,7 +559,7 @@ HnswLoadElementImpl(Buffer buf, OffsetNumber offno, double *distance, HnswQuery
if (distance == NULL || maxDistance == NULL || *distance < *maxDistance) if (distance == NULL || maxDistance == NULL || *distance < *maxDistance)
{ {
if (*element == NULL) if (*element == NULL)
*element = HnswInitElementFromBlock(BufferGetBlockNumber(buf), offno); *element = HnswInitElementFromBlock(blkno, offno);
HnswLoadElementFromTuple(*element, etup, true, loadVec); HnswLoadElementFromTuple(*element, etup, true, loadVec);
} }
@@ -573,9 +573,7 @@ HnswLoadElementImpl(Buffer buf, OffsetNumber offno, double *distance, HnswQuery
void void
HnswLoadElement(HnswElement element, double *distance, HnswQuery * q, Relation index, HnswSupport * support, bool loadVec, double *maxDistance) HnswLoadElement(HnswElement element, double *distance, HnswQuery * q, Relation index, HnswSupport * support, bool loadVec, double *maxDistance)
{ {
Buffer buf = ReadBuffer(index, element->blkno); HnswLoadElementImpl(element->blkno, element->offno, distance, q, index, support, loadVec, maxDistance, &element);
HnswLoadElementImpl(buf, element->offno, distance, q, index, support, loadVec, maxDistance, &element);
} }
/* /*
@@ -815,31 +813,11 @@ HnswLoadUnvisitedFromDisk(HnswElement element, HnswUnvisited * unvisited, int *u
} }
} }
#if PG_VERSION_NUM >= 170000
/*
* Get next block number for read stream
*/
static BlockNumber
HnswReadStreamNextBlock(ReadStream *stream, void *callback_private_data, void *per_buffer_data)
{
HnswReadStreamData *streamData = callback_private_data;
OffsetNumber *offno = per_buffer_data;
HnswUnvisited *uv;
if (streamData->visited == streamData->unvisitedLength)
return InvalidBlockNumber;
uv = &streamData->unvisited[streamData->visited++];
*offno = ItemPointerGetOffsetNumber(&uv->indextid);
return ItemPointerGetBlockNumber(&uv->indextid);
}
#endif
/* /*
* Algorithm 2 from paper * Algorithm 2 from paper
*/ */
List * List *
HnswSearchLayer(char *base, HnswQuery * q, List *ep, int ef, int lc, Relation index, HnswSupport * support, int m, bool inserting, HnswElement skipElement, visited_hash * v, pairingheap **discarded, bool initVisited, int64 *tuples, bool maintenance) HnswSearchLayer(char *base, HnswQuery * q, List *ep, int ef, int lc, Relation index, HnswSupport * support, int m, bool inserting, HnswElement skipElement, visited_hash * v, pairingheap **discarded, bool initVisited, int64 *tuples)
{ {
List *w = NIL; List *w = NIL;
pairingheap *C = pairingheap_allocate(CompareNearestCandidates, NULL); pairingheap *C = pairingheap_allocate(CompareNearestCandidates, NULL);
@@ -854,27 +832,6 @@ HnswSearchLayer(char *base, HnswQuery * q, List *ep, int ef, int lc, Relation in
int unvisitedLength; int unvisitedLength;
bool inMemory = index == NULL; bool inMemory = index == NULL;
#if PG_VERSION_NUM >= 170000
HnswReadStreamData streamData;
ReadStream *stream = NULL;
if (!inMemory)
{
int flags = READ_STREAM_DEFAULT;
if (maintenance)
{
flags |= READ_STREAM_MAINTENANCE;
}
#if PG_VERSION_NUM >= 180000
flags |= READ_STREAM_USE_BATCHING;
#endif
stream = read_stream_begin_relation(flags, NULL, index, MAIN_FORKNUM, HnswReadStreamNextBlock, &streamData, sizeof(OffsetNumber));
}
#endif
if (v == NULL) if (v == NULL)
{ {
v = &vh; v = &vh;
@@ -937,23 +894,13 @@ HnswSearchLayer(char *base, HnswQuery * q, List *ep, int ef, int lc, Relation in
if (inMemory) if (inMemory)
HnswLoadUnvisitedFromMemory(base, cElement, unvisited, &unvisitedLength, v, lc, localNeighborhood, neighborhoodSize); HnswLoadUnvisitedFromMemory(base, cElement, unvisited, &unvisitedLength, v, lc, localNeighborhood, neighborhoodSize);
else else
{
HnswLoadUnvisitedFromDisk(cElement, unvisited, &unvisitedLength, v, index, m, lm, lc); HnswLoadUnvisitedFromDisk(cElement, unvisited, &unvisitedLength, v, index, m, lm, lc);
#if PG_VERSION_NUM >= 170000
read_stream_resume(stream);
streamData.unvisited = unvisited;
streamData.unvisitedLength = unvisitedLength;
streamData.visited = 0;
#endif
}
/* OK to count elements instead of tuples */ /* OK to count elements instead of tuples */
if (tuples != NULL) if (tuples != NULL)
(*tuples) += unvisitedLength; (*tuples) += unvisitedLength;
for (int i = 0;; i++) for (int i = 0; i < unvisitedLength; i++)
{ {
HnswElement eElement; HnswElement eElement;
HnswSearchCandidate *e; HnswSearchCandidate *e;
@@ -964,46 +911,24 @@ HnswSearchLayer(char *base, HnswQuery * q, List *ep, int ef, int lc, Relation in
if (inMemory) if (inMemory)
{ {
if (i == unvisitedLength)
break;
eElement = unvisited[i].element; eElement = unvisited[i].element;
eDistance = GetElementDistance(base, eElement, q, support); eDistance = GetElementDistance(base, eElement, q, support);
} }
else else
{ {
Buffer buf; ItemPointer indextid = &unvisited[i].indextid;
OffsetNumber offno; BlockNumber blkno = ItemPointerGetBlockNumber(indextid);
OffsetNumber offno = ItemPointerGetOffsetNumber(indextid);
#if PG_VERSION_NUM >= 170000
void *offnoPtr;
buf = read_stream_next_buffer(stream, &offnoPtr);
if (!BufferIsValid(buf))
break;
offno = *((OffsetNumber *) offnoPtr);
#else
ItemPointer indextid;
if (i == unvisitedLength)
break;
indextid = &unvisited[i].indextid;
buf = ReadBuffer(index, ItemPointerGetBlockNumber(indextid));
offno = ItemPointerGetOffsetNumber(indextid);
#endif
/* Avoid any allocations if not adding */ /* Avoid any allocations if not adding */
eElement = NULL; eElement = NULL;
HnswLoadElementImpl(buf, offno, &eDistance, q, index, support, inserting, alwaysAdd || discarded != NULL ? NULL : &f->distance, &eElement); HnswLoadElementImpl(blkno, offno, &eDistance, q, index, support, inserting, alwaysAdd || discarded != NULL ? NULL : &f->distance, &eElement);
if (eElement == NULL) if (eElement == NULL)
continue; continue;
} }
if (eElement == NULL || !(eDistance < f->distance || alwaysAdd)) if (!(eDistance < f->distance || alwaysAdd))
{ {
if (discarded != NULL) if (discarded != NULL)
{ {
@@ -1053,11 +978,6 @@ HnswSearchLayer(char *base, HnswQuery * q, List *ep, int ef, int lc, Relation in
w = lappend(w, sc); w = lappend(w, sc);
} }
#if PG_VERSION_NUM >= 170000
if (!inMemory)
read_stream_end(stream);
#endif
return w; return w;
} }
@@ -1353,7 +1273,7 @@ PrecomputeHash(char *base, HnswElement element)
* Algorithm 1 from paper * Algorithm 1 from paper
*/ */
void void
HnswFindElementNeighbors(char *base, HnswElement element, HnswElement entryPoint, Relation index, HnswSupport * support, int m, int efConstruction, bool existing, bool maintenance) HnswFindElementNeighbors(char *base, HnswElement element, HnswElement entryPoint, Relation index, HnswSupport * support, int m, int efConstruction, bool existing)
{ {
List *ep; List *ep;
List *w; List *w;
@@ -1380,7 +1300,7 @@ HnswFindElementNeighbors(char *base, HnswElement element, HnswElement entryPoint
/* 1st phase: greedy search to insert level */ /* 1st phase: greedy search to insert level */
for (int lc = entryLevel; lc >= level + 1; lc--) for (int lc = entryLevel; lc >= level + 1; lc--)
{ {
w = HnswSearchLayer(base, &q, ep, 1, lc, index, support, m, true, skipElement, NULL, NULL, true, NULL, maintenance); w = HnswSearchLayer(base, &q, ep, 1, lc, index, support, m, true, skipElement, NULL, NULL, true, NULL);
ep = w; ep = w;
} }
@@ -1399,7 +1319,7 @@ HnswFindElementNeighbors(char *base, HnswElement element, HnswElement entryPoint
List *lw = NIL; List *lw = NIL;
ListCell *lc2; ListCell *lc2;
w = HnswSearchLayer(base, &q, ep, efConstruction, lc, index, support, m, true, skipElement, NULL, NULL, true, NULL, maintenance); w = HnswSearchLayer(base, &q, ep, efConstruction, lc, index, support, m, true, skipElement, NULL, NULL, true, NULL);
/* Convert search candidates to candidates */ /* Convert search candidates to candidates */
foreach(lc2, w) foreach(lc2, w)

View File

@@ -1,13 +1,14 @@
#include "postgres.h" #include "postgres.h"
#include <math.h> #include "access/genam.h"
#include "access/generic_xlog.h" #include "access/generic_xlog.h"
#include "commands/vacuum.h" #include "commands/vacuum.h"
#include "hnsw.h" #include "hnsw.h"
#include "nodes/pg_list.h"
#include "storage/bufmgr.h" #include "storage/bufmgr.h"
#include "storage/lmgr.h" #include "storage/lmgr.h"
#include "utils/memutils.h" #include "utils/memutils.h"
#include "utils/rel.h"
#if PG_VERSION_NUM >= 160000 #if PG_VERSION_NUM >= 160000
#include "varatt.h" #include "varatt.h"
@@ -178,7 +179,12 @@ NeedsUpdated(HnswVacuumState * vacuumstate, HnswElement element)
/* Also update if layer 0 is not full */ /* Also update if layer 0 is not full */
/* This could indicate too many candidates being deleted during insert */ /* This could indicate too many candidates being deleted during insert */
if (!needsUpdated) if (!needsUpdated)
{
/* Keep clang-tidy happy */
Assert(ntup->count > 0);
needsUpdated = !ItemPointerIsValid(&ntup->indextids[ntup->count - 1]); needsUpdated = !ItemPointerIsValid(&ntup->indextids[ntup->count - 1]);
}
UnlockReleaseBuffer(buf); UnlockReleaseBuffer(buf);
@@ -212,7 +218,7 @@ RepairGraphElement(HnswVacuumState * vacuumstate, HnswElement element, HnswEleme
element->heaptidsLength = 0; element->heaptidsLength = 0;
/* Find neighbors for element, skipping itself */ /* Find neighbors for element, skipping itself */
HnswFindElementNeighbors(base, element, entryPoint, index, support, m, efConstruction, true, true); HnswFindElementNeighbors(base, element, entryPoint, index, support, m, efConstruction, true);
/* Zero memory for each element */ /* Zero memory for each element */
MemSet(ntup, 0, HNSW_TUPLE_ALLOC_SIZE); MemSet(ntup, 0, HNSW_TUPLE_ALLOC_SIZE);
@@ -528,8 +534,9 @@ MarkDeleted(HnswVacuumState * vacuumstate)
ntup = (HnswNeighborTuple) PageGetItem(npage, PageGetItemId(npage, neighborOffno)); ntup = (HnswNeighborTuple) PageGetItem(npage, PageGetItemId(npage, neighborOffno));
/* Overwrite element */ /* Overwrite element */
/* Use memset instead of MemSet to keep clang-tidy happy */
etup->deleted = 1; etup->deleted = 1;
MemSet(&etup->data, 0, VARSIZE_ANY(&etup->data)); memset(&etup->data, 0, VARSIZE_ANY(&etup->data));
/* Overwrite neighbors */ /* Overwrite neighbors */
for (int i = 0; i < ntup->count; i++) for (int i = 0; i < ntup->count; i++)

View File

@@ -2,23 +2,32 @@
#include <float.h> #include <float.h>
#include "access/genam.h"
#include "access/generic_xlog.h"
#include "access/itup.h"
#include "access/relscan.h"
#include "access/table.h" #include "access/table.h"
#include "access/tableam.h" #include "access/tableam.h"
#include "access/tupdesc.h"
#include "access/parallel.h" #include "access/parallel.h"
#include "access/xact.h" #include "access/xact.h"
#include "bitvec.h" #include "access/xloginsert.h"
#include "catalog/index.h" #include "catalog/index.h"
#include "catalog/pg_operator_d.h" #include "catalog/pg_operator_d.h"
#include "catalog/pg_type_d.h" #include "catalog/pg_type_d.h"
#include "commands/progress.h" #include "commands/progress.h"
#include "halfvec.h" #include "fmgr.h"
#include "ivfflat.h" #include "ivfflat.h"
#include "miscadmin.h" #include "miscadmin.h"
#include "nodes/execnodes.h"
#include "optimizer/optimizer.h" #include "optimizer/optimizer.h"
#include "storage/bufmgr.h" #include "storage/bufmgr.h"
#include "tcop/tcopprot.h" #include "tcop/tcopprot.h"
#include "utils/memutils.h" #include "utils/memutils.h"
#include "vector.h" #include "utils/rel.h"
#include "utils/sampling.h"
#include "utils/snapmgr.h"
#include "utils/tuplesort.h"
#if PG_VERSION_NUM >= 160000 #if PG_VERSION_NUM >= 160000
#include "varatt.h" #include "varatt.h"

View File

@@ -3,14 +3,19 @@
#include <float.h> #include <float.h>
#include "access/amapi.h" #include "access/amapi.h"
#include "access/genam.h"
#include "access/reloptions.h" #include "access/reloptions.h"
#include "commands/progress.h" #include "commands/progress.h"
#include "commands/vacuum.h" #include "commands/vacuum.h"
#include "fmgr.h"
#include "ivfflat.h" #include "ivfflat.h"
#include "nodes/pg_list.h"
#include "utils/float.h" #include "utils/float.h"
#include "utils/guc.h" #include "utils/guc.h"
#include "utils/relcache.h"
#include "utils/selfuncs.h" #include "utils/selfuncs.h"
#include "utils/spccache.h" #include "utils/spccache.h"
#include "vector.h"
#if PG_VERSION_NUM < 150000 #if PG_VERSION_NUM < 150000
#define MarkGUCPrefixReserved(x) EmitWarningsOnPlaceholders(x) #define MarkGUCPrefixReserved(x) EmitWarningsOnPlaceholders(x)
@@ -179,6 +184,64 @@ FUNCTION_PREFIX PG_FUNCTION_INFO_V1(ivfflathandler);
Datum Datum
ivfflathandler(PG_FUNCTION_ARGS) ivfflathandler(PG_FUNCTION_ARGS)
{ {
#if PG_VERSION_NUM >= 190000
static const IndexAmRoutine amroutine = {
.type = T_IndexAmRoutine,
.amstrategies = 0,
.amsupport = 5,
.amoptsprocnum = 0,
.amcanorder = false,
.amcanorderbyop = true,
.amcanhash = false,
.amconsistentequality = false,
.amconsistentordering = false,
.amcanbackward = false,
.amcanunique = false,
.amcanmulticol = false,
.amoptionalkey = true,
.amsearcharray = false,
.amsearchnulls = false,
.amstorage = false,
.amclusterable = false,
.ampredlocks = false,
.amcanparallel = false,
.amcanbuildparallel = true,
.amcaninclude = false,
.amusemaintenanceworkmem = false,
.amsummarizing = false,
.amparallelvacuumoptions = VACUUM_OPTION_PARALLEL_BULKDEL,
.amkeytype = InvalidOid,
.ambuild = ivfflatbuild,
.ambuildempty = ivfflatbuildempty,
.aminsert = ivfflatinsert,
.aminsertcleanup = NULL,
.ambulkdelete = ivfflatbulkdelete,
.amvacuumcleanup = ivfflatvacuumcleanup,
.amcanreturn = NULL,
.amcostestimate = ivfflatcostestimate,
.amgettreeheight = NULL,
.amoptions = ivfflatoptions,
.amproperty = NULL,
.ambuildphasename = ivfflatbuildphasename,
.amvalidate = ivfflatvalidate,
.amadjustmembers = NULL,
.ambeginscan = ivfflatbeginscan,
.amrescan = ivfflatrescan,
.amgettuple = ivfflatgettuple,
.amgetbitmap = NULL,
.amendscan = ivfflatendscan,
.ammarkpos = NULL,
.amrestrpos = NULL,
.amestimateparallelscan = NULL,
.aminitparallelscan = NULL,
.amparallelrescan = NULL,
.amtranslatestrategy = NULL,
.amtranslatecmptype = NULL,
};
PG_RETURN_POINTER(&amroutine);
#else
IndexAmRoutine *amroutine = makeNode(IndexAmRoutine); IndexAmRoutine *amroutine = makeNode(IndexAmRoutine);
amroutine->amstrategies = 0; amroutine->amstrategies = 0;
@@ -252,4 +315,5 @@ ivfflathandler(PG_FUNCTION_ARGS)
#endif #endif
PG_RETURN_POINTER(amroutine); PG_RETURN_POINTER(amroutine);
#endif
} }

View File

@@ -25,6 +25,10 @@
#include "portability/instr_time.h" #include "portability/instr_time.h"
#endif #endif
#if PG_VERSION_NUM >= 190000
typedef Pointer Item;
#endif
#define IVFFLAT_MAX_DIM 2000 #define IVFFLAT_MAX_DIM 2000
/* Support functions */ /* Support functions */

View File

@@ -2,11 +2,16 @@
#include <float.h> #include <float.h>
#include "access/genam.h"
#include "access/generic_xlog.h" #include "access/generic_xlog.h"
#include "access/itup.h"
#include "fmgr.h"
#include "ivfflat.h" #include "ivfflat.h"
#include "nodes/execnodes.h"
#include "storage/bufmgr.h" #include "storage/bufmgr.h"
#include "storage/lmgr.h" #include "storage/lmgr.h"
#include "utils/memutils.h" #include "utils/memutils.h"
#include "utils/rel.h"
/* /*
* Find the list that minimizes the distance function * Find the list that minimizes the distance function

View File

@@ -1,17 +1,15 @@
#include "postgres.h" #include "postgres.h"
#include <float.h> #include <float.h>
#include <limits.h>
#include <math.h> #include <math.h>
#include "bitvec.h" #include "access/genam.h"
#include "halfutils.h" #include "fmgr.h"
#include "halfvec.h"
#include "ivfflat.h" #include "ivfflat.h"
#include "miscadmin.h" #include "miscadmin.h"
#include "utils/builtins.h"
#include "utils/datum.h"
#include "utils/memutils.h" #include "utils/memutils.h"
#include "vector.h" #include "utils/relcache.h"
#if PG_VERSION_NUM >= 160000 #if PG_VERSION_NUM >= 160000
#include "varatt.h" #include "varatt.h"

View File

@@ -2,15 +2,26 @@
#include <float.h> #include <float.h>
#include "access/genam.h"
#include "access/itup.h"
#include "access/relscan.h" #include "access/relscan.h"
#include "access/tupdesc.h"
#include "catalog/pg_operator_d.h" #include "catalog/pg_operator_d.h"
#include "catalog/pg_type_d.h" #include "catalog/pg_type_d.h"
#include "fmgr.h"
#include "lib/pairingheap.h" #include "lib/pairingheap.h"
#include "ivfflat.h" #include "ivfflat.h"
#include "miscadmin.h" #include "miscadmin.h"
#include "pgstat.h" #include "pgstat.h"
#include "storage/bufmgr.h" #include "storage/bufmgr.h"
#include "utils/memutils.h" #include "utils/memutils.h"
#include "utils/rel.h"
#include "utils/snapmgr.h"
#include "utils/tuplesort.h"
#if PG_VERSION_NUM >= 160000
#include "varatt.h"
#endif
#define GetScanList(ptr) pairingheap_container(IvfflatScanList, ph_node, ptr) #define GetScanList(ptr) pairingheap_container(IvfflatScanList, ph_node, ptr)
#define GetScanListConst(ptr) pairingheap_const_container(IvfflatScanList, ph_node, ptr) #define GetScanListConst(ptr) pairingheap_const_container(IvfflatScanList, ph_node, ptr)

View File

@@ -1,13 +1,19 @@
#include "postgres.h" #include "postgres.h"
#include "access/genam.h"
#include "access/generic_xlog.h" #include "access/generic_xlog.h"
#include "bitvec.h"
#include "catalog/pg_type.h"
#include "fmgr.h" #include "fmgr.h"
#include "halfutils.h" #include "halfutils.h"
#include "halfvec.h" #include "halfvec.h"
#include "ivfflat.h" #include "ivfflat.h"
#include "storage/bufmgr.h" #include "storage/bufmgr.h"
#include "utils/relcache.h"
#include "utils/varbit.h"
#include "vector.h"
#if PG_VERSION_NUM >= 160000
#include "varatt.h"
#endif
/* /*
* Allocate a vector array * Allocate a vector array

View File

@@ -1,9 +1,12 @@
#include "postgres.h" #include "postgres.h"
#include "access/genam.h"
#include "access/generic_xlog.h" #include "access/generic_xlog.h"
#include "access/itup.h"
#include "commands/vacuum.h" #include "commands/vacuum.h"
#include "ivfflat.h" #include "ivfflat.h"
#include "storage/bufmgr.h" #include "storage/bufmgr.h"
#include "utils/relcache.h"
#if PG_VERSION_NUM >= 180000 #if PG_VERSION_NUM >= 180000
#define vacuum_delay_point() vacuum_delay_point(false) #define vacuum_delay_point() vacuum_delay_point(false)

View File

@@ -5,18 +5,27 @@
#include "catalog/pg_type.h" #include "catalog/pg_type.h"
#include "common/shortest_dec.h" #include "common/shortest_dec.h"
#include "common/string.h"
#include "fmgr.h" #include "fmgr.h"
#include "halfutils.h" #include "halfutils.h"
#include "halfvec.h" #include "halfvec.h"
#include "lib/stringinfo.h"
#include "libpq/pqformat.h" #include "libpq/pqformat.h"
#include "sparsevec.h" #include "sparsevec.h"
#include "utils/array.h" #include "utils/array.h"
#include "utils/builtins.h" #include "utils/builtins.h"
#include "utils/float.h" #include "utils/float.h"
#include "utils/fmgrprotos.h"
#include "utils/lsyscache.h" #include "utils/lsyscache.h"
#include "vector.h" #include "vector.h"
#if PG_VERSION_NUM >= 160000
#include "varatt.h"
#endif
#if PG_VERSION_NUM >= 170000
#include "parser/scansup.h"
#endif
typedef struct SparseInputElement typedef struct SparseInputElement
{ {
int32 index; int32 index;
@@ -150,9 +159,9 @@ InitSparseVector(int dim, int nnz)
return result; return result;
} }
/* #if PG_VERSION_NUM >= 170000
* Check for whitespace, since array_isspace() is static #define sparsevec_isspace(ch) scanner_isspace(ch)
*/ #else
static inline bool static inline bool
sparsevec_isspace(char ch) sparsevec_isspace(char ch)
{ {
@@ -165,6 +174,7 @@ sparsevec_isspace(char ch)
return true; return true;
return false; return false;
} }
#endif
/* /*
* Compare indices * Compare indices

View File

@@ -16,16 +16,20 @@
#include "port.h" /* for strtof() */ #include "port.h" /* for strtof() */
#include "sparsevec.h" #include "sparsevec.h"
#include "utils/array.h" #include "utils/array.h"
#include "utils/builtins.h"
#include "utils/float.h" #include "utils/float.h"
#include "utils/fmgrprotos.h"
#include "utils/lsyscache.h" #include "utils/lsyscache.h"
#include "utils/numeric.h" #include "utils/varbit.h"
#include "vector.h" #include "vector.h"
#if PG_VERSION_NUM >= 160000 #if PG_VERSION_NUM >= 160000
#include "varatt.h" #include "varatt.h"
#endif #endif
#if PG_VERSION_NUM >= 170000
#include "parser/scansup.h"
#endif
#define STATE_DIMS(x) (ARR_DIMS(x)[0] - 1) #define STATE_DIMS(x) (ARR_DIMS(x)[0] - 1)
#define CreateStateDatums(dim) palloc(sizeof(Datum) * (dim + 1)) #define CreateStateDatums(dim) palloc(sizeof(Datum) * (dim + 1))
@@ -36,7 +40,7 @@
#endif #endif
#if PG_VERSION_NUM >= 180000 #if PG_VERSION_NUM >= 180000
PG_MODULE_MAGIC_EXT(.name = "vector",.version = "0.8.1"); PG_MODULE_MAGIC_EXT(.name = "vector",.version = "0.8.2");
#else #else
PG_MODULE_MAGIC; PG_MODULE_MAGIC;
#endif #endif
@@ -129,9 +133,9 @@ InitVector(int dim)
return result; return result;
} }
/* #if PG_VERSION_NUM >= 170000
* Check for whitespace, since array_isspace() is static #define vector_isspace(ch) scanner_isspace(ch)
*/ #else
static inline bool static inline bool
vector_isspace(char ch) vector_isspace(char ch)
{ {
@@ -144,6 +148,7 @@ vector_isspace(char ch)
return true; return true;
return false; return false;
} }
#endif
/* /*
* Check state array * Check state array

View File

@@ -16,29 +16,19 @@ $node->safe_psql("postgres",
"INSERT INTO tst SELECT i, ARRAY[random(), random(), random()] FROM generate_series(1, 100000) i;" "INSERT INTO tst SELECT i, ARRAY[random(), random(), random()] FROM generate_series(1, 100000) i;"
); );
# Check each index type # Add index
my @operators = ("<->", "<#>", "<=>"); $node->safe_psql("postgres", "CREATE INDEX ON tst USING ivfflat (v vector_l2_ops);");
my @opclasses = ("vector_l2_ops", "vector_ip_ops", "vector_cosine_ops");
for my $i (0 .. $#operators) # Test 100% recall
for (1 .. 20)
{ {
my $operator = $operators[$i]; my $id = int(rand() * 100000);
my $opclass = $opclasses[$i]; my $query = $node->safe_psql("postgres", "SELECT v FROM tst WHERE i = $id;");
my $res = $node->safe_psql("postgres", qq(
# Add index SET enable_seqscan = off;
$node->safe_psql("postgres", "CREATE INDEX ON tst USING ivfflat (v $opclass);"); SELECT v FROM tst ORDER BY v <-> '$query' LIMIT 1;
));
# Test 100% recall is($res, $query);
for (1 .. 20)
{
my $id = int(rand() * 100000);
my $query = $node->safe_psql("postgres", "SELECT v FROM tst WHERE i = $id;");
my $res = $node->safe_psql("postgres", qq(
SET enable_seqscan = off;
SELECT v FROM tst ORDER BY v <-> '$query' LIMIT 1;
));
is($res, $query);
}
} }
done_testing(); done_testing();

View File

@@ -0,0 +1,29 @@
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
$node->safe_psql("postgres", "CREATE EXTENSION vector;");
$node->safe_psql("postgres", "CREATE TABLE tst (v vector(3));");
$node->safe_psql("postgres",
"INSERT INTO tst SELECT ARRAY[random(), random(), random()] FROM generate_series(1, 1000) i;"
);
my ($ret, $stdout, $stderr) = $node->psql("postgres", qq(
SET client_min_messages = DEBUG;
SET maintenance_work_mem = '3073kB';
ALTER TABLE tst SET (parallel_workers = 1);
CREATE INDEX ON tst USING hnsw (v vector_l2_ops);
));
is($ret, 0, $stderr);
like($stderr, qr/using \d+ parallel workers/);
like($stderr, qr/hnsw graph no longer fits into maintenance_work_mem after 0 tuples/);
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 = '0.8.1' default_version = '0.8.2'
module_pathname = '$libdir/vector' module_pathname = '$libdir/vector'
relocatable = true relocatable = true