Compare commits

..

1 Commits

Author SHA1 Message Date
Andrew Kane
9542b5548f Check architecture bit width is consistent with Postgres 2025-03-24 19:03:40 -07:00
39 changed files with 202 additions and 635 deletions

8
.dockerignore Normal file
View File

@@ -0,0 +1,8 @@
/.git/
/dist/
/results/
/tmp_check/
/sql/vector--?.?.?.sql
regression.*
*.o
*.so

View File

@@ -8,29 +8,27 @@ jobs:
fail-fast: false fail-fast: false
matrix: matrix:
include: include:
- postgres: 19
os: ubuntu-24.04
- postgres: 18 - postgres: 18
os: ubuntu-24.04 os: ubuntu-24.04
- postgres: 17 - postgres: 17
os: ubuntu-24.04 os: ubuntu-24.04
- postgres: 16 - postgres: 16
os: ubuntu-24.04-arm os: ubuntu-22.04
- postgres: 15 - postgres: 15
os: ubuntu-22.04 os: ubuntu-22.04
- postgres: 14 - postgres: 14
os: ubuntu-22.04-arm os: ubuntu-20.04
- postgres: 13 - postgres: 13
os: ubuntu-22.04 os: ubuntu-20.04
steps: steps:
- uses: actions/checkout@v6 - uses: actions/checkout@v4
- uses: ankane/setup-postgres@v1 - uses: ankane/setup-postgres@v1
with: with:
postgres-version: ${{ matrix.postgres }} postgres-version: ${{ matrix.postgres }}
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 ${{ matrix.postgres >= 18 && '-Wno-missing-field-initializers' || '' }} PG_CFLAGS: -DUSE_ASSERT_CHECKING -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
@@ -48,18 +46,18 @@ jobs:
fail-fast: false fail-fast: false
matrix: matrix:
include: include:
- postgres: 18 - postgres: 17
os: macos-26 os: macos-15
- postgres: 14 - postgres: 14
os: macos-15-intel os: macos-13
steps: steps:
- uses: actions/checkout@v6 - uses: actions/checkout@v4
- uses: ankane/setup-postgres@v1 - uses: ankane/setup-postgres@v1
with: with:
postgres-version: ${{ matrix.postgres }} postgres-version: ${{ matrix.postgres }}
- run: make - run: make
env: env:
PG_CFLAGS: -DUSE_ASSERT_CHECKING -Wall -Wextra -Werror -Wno-unused-parameter -Wno-unknown-warning-option ${{ matrix.postgres >= 18 && '-Wno-missing-field-initializers' || '' }} PG_CFLAGS: -DUSE_ASSERT_CHECKING -Wall -Wextra -Werror -Wno-unused-parameter
- run: make install - run: make install
- run: make installcheck - run: make installcheck
- if: ${{ failure() }} - if: ${{ failure() }}
@@ -72,35 +70,27 @@ 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_2' || 'REL_14_21' }} TAG: ${{ matrix.postgres == 17 && 'REL_17_2' || 'REL_14_15' }}
- 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
- run: make clean && $(brew --prefix llvm@$LLVM_VERSION)/bin/scan-build --status-bugs make - run: make clean && $(brew --prefix llvm@$LLVM_VERSION)/bin/scan-build --status-bugs make
env: env:
LLVM_VERSION: ${{ matrix.os == 'macos-26' && 20 || 18 }} LLVM_VERSION: ${{ matrix.os == 'macos-15' && 18 || 15 }}
PG_CFLAGS: -DUSE_ASSERT_CHECKING PG_CFLAGS: -DUSE_ASSERT_CHECKING
windows: windows:
runs-on: ${{ matrix.os }} runs-on: windows-latest
if: ${{ !startsWith(github.ref_name, 'mac') }} if: ${{ !startsWith(github.ref_name, 'mac') }}
strategy:
fail-fast: false
matrix:
include:
- postgres: 17
os: windows-2025
- postgres: 14
os: windows-2022
steps: steps:
- uses: actions/checkout@v6 - uses: actions/checkout@v4
- uses: ankane/setup-postgres@v1 - uses: ankane/setup-postgres@v1
with: with:
postgres-version: ${{ matrix.postgres }} postgres-version: 14
- run: | - run: |
call "C:\Program Files\Microsoft Visual Studio\2022\Enterprise\VC\Auxiliary\Build\vcvars64.bat" && ^ call "C:\Program Files\Microsoft Visual Studio\2022\Enterprise\VC\Auxiliary\Build\vcvars64.bat" && ^
nmake /NOLOGO /F Makefile.win && ^ nmake /NOLOGO /F Makefile.win && ^
nmake /NOLOGO /F Makefile.win install && ^ nmake /NOLOGO /F Makefile.win install && ^
nmake /NOLOGO /F Makefile.win installcheck ${{ matrix.postgres != 17 && 'PG_REGRESS=$(PGROOT)\bin\pg_regress' || '' }} && ^ nmake /NOLOGO /F Makefile.win installcheck && ^
nmake /NOLOGO /F Makefile.win clean && ^ nmake /NOLOGO /F Makefile.win clean && ^
nmake /NOLOGO /F Makefile.win uninstall nmake /NOLOGO /F Makefile.win uninstall
shell: cmd shell: cmd
@@ -133,10 +123,10 @@ 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@v6 - uses: actions/checkout@v4
- uses: ankane/setup-postgres-valgrind@v1 - uses: ankane/setup-postgres-valgrind@v1
with: with:
postgres-version: 18 postgres-version: 17
check-ub: yes check-ub: yes
- run: make OPTFLAGS="" - run: make OPTFLAGS=""
- run: sudo --preserve-env=PG_CONFIG make install - run: sudo --preserve-env=PG_CONFIG make install

View File

@@ -1,14 +1,3 @@
## 0.8.2 (2026-02-25)
- Fixed buffer overflow with parallel HNSW index build - [more info](https://github.com/pgvector/pgvector/issues/959)
- Improved `install` target on Windows
- Fixed `Index Searches` in `EXPLAIN` output for Postgres 18
## 0.8.1 (2025-09-04)
- Added support for Postgres 18 rc1
- Improved performance of `binary_quantize` function
## 0.8.0 (2024-10-30) ## 0.8.0 (2024-10-30)
- Added support for iterative index scans - Added support for iterative index scans

View File

@@ -1,11 +1,8 @@
# syntax=docker/dockerfile:1
ARG PG_MAJOR=17 ARG PG_MAJOR=17
ARG DEBIAN_CODENAME=bookworm FROM postgres:$PG_MAJOR
FROM postgres:$PG_MAJOR-$DEBIAN_CODENAME
ARG PG_MAJOR ARG PG_MAJOR
ADD https://github.com/pgvector/pgvector.git#v0.8.2 /tmp/pgvector COPY . /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) 1996-2026, PostgreSQL Global Development Group Portions Copyright (c) 1996-2025, 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": "0.8.2", "version": "0.8.0",
"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": "0.8.2", "version": "0.8.0",
"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 = 0.8.2 EXTVERSION = 0.8.0
MODULE_big = vector MODULE_big = vector
DATA = $(wildcard sql/*--*--*.sql) DATA = $(wildcard sql/*--*--*.sql)
@@ -27,11 +27,6 @@ ifneq ($(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
@@ -81,9 +76,4 @@ docker:
.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=$(PG_MAJOR) --build-arg DEBIAN_CODENAME=bookworm -t pgvector/pgvector:pg$(PG_MAJOR) -t pgvector/pgvector:pg$(PG_MAJOR)-bookworm -t pgvector/pgvector:$(EXTVERSION)-pg$(PG_MAJOR) -t pgvector/pgvector:$(EXTVERSION)-pg$(PG_MAJOR)-bookworm . docker buildx build --push --pull --no-cache --platform linux/amd64,linux/arm64 --build-arg PG_MAJOR=$(PG_MAJOR) -t pgvector/pgvector:pg$(PG_MAJOR) -t pgvector/pgvector:$(EXTVERSION)-pg$(PG_MAJOR) .
.PHONY: docker-release-trixie
docker-release-trixie:
docker buildx build --push --pull --no-cache --platform linux/amd64,linux/arm64 --build-arg PG_MAJOR=$(PG_MAJOR) --build-arg DEBIAN_CODENAME=trixie -t pgvector/pgvector:pg$(PG_MAJOR)-trixie -t pgvector/pgvector:$(EXTVERSION)-pg$(PG_MAJOR)-trixie .

View File

@@ -1,5 +1,5 @@
EXTENSION = vector EXTENSION = vector
EXTVERSION = 0.8.2 EXTVERSION = 0.8.0
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
@@ -31,9 +31,6 @@ LIBDIR = $(PGROOT)\lib
PKGLIBDIR = $(PGROOT)\lib PKGLIBDIR = $(PGROOT)\lib
SHAREDIR = $(PGROOT)\share SHAREDIR = $(PGROOT)\share
# Use $(PGROOT)\bin\pg_regress for Postgres < 17
PG_REGRESS = $(LIBDIR)\pgxs\src\test\regress\pg_regress
CFLAGS = /nologo /I"$(INCLUDEDIR_SERVER)\port\win32_msvc" /I"$(INCLUDEDIR_SERVER)\port\win32" /I"$(INCLUDEDIR_SERVER)" /I"$(INCLUDEDIR)" CFLAGS = /nologo /I"$(INCLUDEDIR_SERVER)\port\win32_msvc" /I"$(INCLUDEDIR_SERVER)\port\win32" /I"$(INCLUDEDIR_SERVER)" /I"$(INCLUDEDIR)"
CFLAGS = $(CFLAGS) $(PG_CFLAGS) CFLAGS = $(CFLAGS) $(PG_CFLAGS)
@@ -57,11 +54,11 @@ install: all
copy $(SHLIB) "$(PKGLIBDIR)" copy $(SHLIB) "$(PKGLIBDIR)"
copy $(EXTENSION).control "$(SHAREDIR)\extension" copy $(EXTENSION).control "$(SHAREDIR)\extension"
copy sql\$(EXTENSION)--*.sql "$(SHAREDIR)\extension" copy sql\$(EXTENSION)--*.sql "$(SHAREDIR)\extension"
if not exist "$(INCLUDEDIR_SERVER)\extension\$(EXTENSION)" mkdir "$(INCLUDEDIR_SERVER)\extension\$(EXTENSION)" mkdir "$(INCLUDEDIR_SERVER)\extension\$(EXTENSION)"
for %f in ($(HEADERS)) do copy %f "$(INCLUDEDIR_SERVER)\extension\$(EXTENSION)" for %f in ($(HEADERS)) do copy %f "$(INCLUDEDIR_SERVER)\extension\$(EXTENSION)"
installcheck: installcheck:
"$(PG_REGRESS)" --bindir="$(BINDIR)" $(REGRESS_OPTS) $(REGRESS) "$(BINDIR)\pg_regress" --bindir="$(BINDIR)" $(REGRESS_OPTS) $(REGRESS)
uninstall: uninstall:
del /f "$(PKGLIBDIR)\$(SHLIB)" del /f "$(PKGLIBDIR)\$(SHLIB)"

108
README.md
View File

@@ -21,7 +21,7 @@ Compile and install the extension (supports Postgres 13+)
```sh ```sh
cd /tmp cd /tmp
git clone --branch v0.8.2 https://github.com/pgvector/pgvector.git git clone --branch v0.8.0 https://github.com/pgvector/pgvector.git
cd pgvector cd pgvector
make make
make install # may need sudo make install # may need sudo
@@ -29,16 +29,24 @@ 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), [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). 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).
### Windows ### Windows
Ensure [C++ support in Visual Studio](https://learn.microsoft.com/en-us/cpp/build/building-on-the-command-line?view=msvc-170#download-and-install-the-tools) is installed and run `x64 Native Tools Command Prompt for VS [version]` as administrator. Then use `nmake` to build: Ensure [C++ support in Visual Studio](https://learn.microsoft.com/en-us/cpp/build/building-on-the-command-line?view=msvc-170#download-and-install-the-tools) is installed, and run:
```cmd ```cmd
set "PGROOT=C:\Program Files\PostgreSQL\18" call "C:\Program Files\Microsoft Visual Studio\2022\Community\VC\Auxiliary\Build\vcvars64.bat"
```
Note: The exact path will vary depending on your Visual Studio version and edition
Then use `nmake` to build:
```cmd
set "PGROOT=C:\Program Files\PostgreSQL\16"
cd %TEMP% cd %TEMP%
git clone --branch v0.8.2 https://github.com/pgvector/pgvector.git git clone --branch v0.8.0 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
@@ -310,9 +318,7 @@ You can also speed up index creation by increasing the number of parallel worker
SET max_parallel_maintenance_workers = 7; -- plus leader SET max_parallel_maintenance_workers = 7; -- plus leader
``` ```
For a large number of workers, you may need to increase `max_parallel_workers` (8 by default) For a large number of workers, you may also need to increase `max_parallel_workers` (8 by default)
The [index options](#index-options) also have a significant impact on build time (use the defaults unless seeing low recall)
### Indexing Progress ### Indexing Progress
@@ -469,6 +475,8 @@ CREATE TABLE items (embedding vector(3), category_id int) PARTITION BY LIST(cate
## Iterative Index Scans ## Iterative Index Scans
*Added in 0.8.0*
With approximate indexes, queries with filtering can return less results since filtering is applied *after* the index is scanned. Starting with 0.8.0, you can enable iterative index scans, which will automatically scan more of the index until enough results are found (or it reaches `hnsw.max_scan_tuples` or `ivfflat.max_probes`). With approximate indexes, queries with filtering can return less results since filtering is applied *after* the index is scanned. Starting with 0.8.0, you can enable iterative index scans, which will automatically scan more of the index until enough results are found (or it reaches `hnsw.max_scan_tuples` or `ivfflat.max_probes`).
Iterative scans can use strict or relaxed ordering. Iterative scans can use strict or relaxed ordering.
@@ -492,11 +500,9 @@ With relaxed ordering, you can use a [materialized CTE](https://www.postgresql.o
```sql ```sql
WITH relaxed_results AS MATERIALIZED ( WITH relaxed_results AS MATERIALIZED (
SELECT id, embedding <-> '[1,2,3]' AS distance FROM items WHERE category_id = 123 ORDER BY distance LIMIT 5 SELECT id, embedding <-> '[1,2,3]' AS distance FROM items WHERE category_id = 123 ORDER BY distance LIMIT 5
) SELECT * FROM relaxed_results ORDER BY distance + 0; ) SELECT * FROM relaxed_results ORDER BY distance;
``` ```
Note: `+ 0` is needed for Postgres 17+
For queries that filter by distance, use a materialized CTE and place the distance filter outside of it for best performance (due to the [current behavior](https://www.postgresql.org/message-id/flat/CAOdR5yGUoMQ6j7M5hNUXrySzaqZVGf_Ne%2B8fwZMRKTFxU1nbJg%40mail.gmail.com) of the Postgres executor) For queries that filter by distance, use a materialized CTE and place the distance filter outside of it for best performance (due to the [current behavior](https://www.postgresql.org/message-id/flat/CAOdR5yGUoMQ6j7M5hNUXrySzaqZVGf_Ne%2B8fwZMRKTFxU1nbJg%40mail.gmail.com) of the Postgres executor)
```sql ```sql
@@ -695,10 +701,10 @@ CREATE INDEX CONCURRENTLY ...
### Querying ### Querying
Use `EXPLAIN (ANALYZE, BUFFERS)` to debug performance. Use `EXPLAIN ANALYZE` to debug performance.
```sql ```sql
EXPLAIN (ANALYZE, BUFFERS) SELECT * FROM items ORDER BY embedding <-> '[3,1,2]' LIMIT 5; EXPLAIN ANALYZE SELECT * FROM items ORDER BY embedding <-> '[3,1,2]' LIMIT 5;
``` ```
#### Exact Search #### Exact Search
@@ -748,6 +754,8 @@ SELECT query, calls, ROUND((total_plan_time + total_exec_time) / calls) AS avg_t
FROM pg_stat_statements ORDER BY total_plan_time + total_exec_time DESC LIMIT 20; FROM pg_stat_statements ORDER BY total_plan_time + total_exec_time DESC LIMIT 20;
``` ```
Note: Replace `total_plan_time + total_exec_time` with `total_time` for Postgres < 13
Monitor recall by comparing results from approximate search with exact search. Monitor recall by comparing results from approximate search with exact search.
```sql ```sql
@@ -771,12 +779,9 @@ 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)
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)
@@ -788,23 +793,19 @@ Go | [pgvector-go](https://github.com/pgvector/pgvector-go)
Haskell | [pgvector-haskell](https://github.com/pgvector/pgvector-haskell) Haskell | [pgvector-haskell](https://github.com/pgvector/pgvector-haskell)
Java, Kotlin, Groovy, Scala | [pgvector-java](https://github.com/pgvector/pgvector-java) Java, Kotlin, Groovy, Scala | [pgvector-java](https://github.com/pgvector/pgvector-java)
JavaScript, TypeScript | [pgvector-node](https://github.com/pgvector/pgvector-node) JavaScript, TypeScript | [pgvector-node](https://github.com/pgvector/pgvector-node)
Julia | [Pgvector.jl](https://github.com/pgvector/Pgvector.jl) Julia | [pgvector-julia](https://github.com/pgvector/pgvector-julia)
Lisp | [pgvector-lisp](https://github.com/pgvector/pgvector-lisp) Lisp | [pgvector-lisp](https://github.com/pgvector/pgvector-lisp)
Lua | [pgvector-lua](https://github.com/pgvector/pgvector-lua) Lua | [pgvector-lua](https://github.com/pgvector/pgvector-lua)
Nim | [pgvector-nim](https://github.com/pgvector/pgvector-nim) Nim | [pgvector-nim](https://github.com/pgvector/pgvector-nim)
OCaml | [pgvector-ocaml](https://github.com/pgvector/pgvector-ocaml) OCaml | [pgvector-ocaml](https://github.com/pgvector/pgvector-ocaml)
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
@@ -819,11 +820,11 @@ Yes, pgvector uses the write-ahead log (WAL), which allows for replication and p
#### What if I want to index vectors with more than 2,000 dimensions? #### What if I want to index vectors with more than 2,000 dimensions?
You can use [half-precision vectors](#half-precision-vectors) or [half-precision indexing](#half-precision-indexing) to index up to 4,000 dimensions or [binary quantization](#binary-quantization) to index up to 64,000 dimensions. Other options are [indexing subvectors](#indexing-subvectors) (for models that support it) or [dimensionality reduction](https://en.wikipedia.org/wiki/Dimensionality_reduction). You can use [half-precision indexing](#half-precision-indexing) to index up to 4,000 dimensions or [binary quantization](#binary-quantization) to index up to 64,000 dimensions. Another option is [dimensionality reduction](https://en.wikipedia.org/wiki/Dimensionality_reduction).
#### Can I store vectors with different dimensions in the same column? #### Can I store vectors with different dimensions in the same column?
You can use `vector` as the type (instead of `vector(n)`). You can use `vector` as the type (instead of `vector(3)`).
```sql ```sql
CREATE TABLE embeddings (model_id bigint, item_id bigint, embedding vector, PRIMARY KEY (model_id, item_id)); CREATE TABLE embeddings (model_id bigint, item_id bigint, embedding vector, PRIMARY KEY (model_id, item_id));
@@ -1071,7 +1072,7 @@ l2_normalize(sparsevec) → sparsevec | Normalize with Euclidean norm | 0.7.0
If your machine has multiple Postgres installations, specify the path to [pg_config](https://www.postgresql.org/docs/current/app-pgconfig.html) with: If your machine has multiple Postgres installations, specify the path to [pg_config](https://www.postgresql.org/docs/current/app-pgconfig.html) with:
```sh ```sh
export PG_CONFIG=/Library/PostgreSQL/18/bin/pg_config export PG_CONFIG=/Library/PostgreSQL/17/bin/pg_config
``` ```
Then re-run the installation instructions (run `make clean` before `make` if needed). If `sudo` is needed for `make install`, use: Then re-run the installation instructions (run `make clean` before `make` if needed). If `sudo` is needed for `make install`, use:
@@ -1082,11 +1083,11 @@ sudo --preserve-env=PG_CONFIG make install
A few common paths on Mac are: A few common paths on Mac are:
- EDB installer - `/Library/PostgreSQL/18/bin/pg_config` - EDB installer - `/Library/PostgreSQL/17/bin/pg_config`
- Homebrew (arm64) - `/opt/homebrew/opt/postgresql@18/bin/pg_config` - Homebrew (arm64) - `/opt/homebrew/opt/postgresql@17/bin/pg_config`
- Homebrew (x86-64) - `/usr/local/opt/postgresql@18/bin/pg_config` - Homebrew (x86-64) - `/usr/local/opt/postgresql@17/bin/pg_config`
Note: Replace `18` with your Postgres server version Note: Replace `17` with your Postgres server version
### Missing Header ### Missing Header
@@ -1095,10 +1096,10 @@ If compilation fails with `fatal error: postgres.h: No such file or directory`,
For Ubuntu and Debian, use: For Ubuntu and Debian, use:
```sh ```sh
sudo apt install postgresql-server-dev-18 sudo apt install postgresql-server-dev-17
``` ```
Note: Replace `18` with your Postgres server version Note: Replace `17` with your Postgres server version
### Missing SDK ### Missing SDK
@@ -1128,7 +1129,7 @@ If compilation fails with `Cannot open include file: 'postgres.h': No such file
### Mismatched Architecture ### Mismatched Architecture
If compilation fails with `error C2196: case value '4' already used`, make sure youre using the `x64 Native Tools Command Prompt`. Then run `nmake /F Makefile.win clean` and re-run the installation instructions. If compilation fails with `error C2196: case value '4' already used`, make sure `vcvars64.bat` was called. Then run `nmake /F Makefile.win clean` and re-run the installation instructions.
### Missing Symbol ### Missing Symbol
@@ -1145,32 +1146,17 @@ If installation fails with `Access is denied`, re-run the installation instructi
Get the [Docker image](https://hub.docker.com/r/pgvector/pgvector) with: Get the [Docker image](https://hub.docker.com/r/pgvector/pgvector) with:
```sh ```sh
docker pull pgvector/pgvector:pg18-trixie docker pull pgvector/pgvector:pg17
``` ```
This adds pgvector to the [Postgres image](https://hub.docker.com/_/postgres) (replace `18` with your Postgres server version, and run it the same way). This adds pgvector to the [Postgres image](https://hub.docker.com/_/postgres) (replace `17` with your Postgres server version, and run it the same way).
Supported tags are:
- `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: You can also build the image manually:
```sh ```sh
git clone --branch v0.8.2 https://github.com/pgvector/pgvector.git git clone --branch v0.8.0 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=17 -t myuser/pgvector .
``` ```
If you increase `maintenance_work_mem`, make sure `--shm-size` is at least that size to avoid an error with parallel HNSW index builds. If you increase `maintenance_work_mem`, make sure `--shm-size` is at least that size to avoid an error with parallel HNSW index builds.
@@ -1187,7 +1173,7 @@ With Homebrew Postgres, you can use:
brew install pgvector brew install pgvector
``` ```
Note: This only adds it to the `postgresql@18` and `postgresql@17` formulas Note: This only adds it to the `postgresql@17` and `postgresql@14` formulas
### PGXN ### PGXN
@@ -1202,29 +1188,29 @@ pgxn install vector
Debian and Ubuntu packages are available from the [PostgreSQL APT Repository](https://wiki.postgresql.org/wiki/Apt). Follow the [setup instructions](https://wiki.postgresql.org/wiki/Apt#Quickstart) and run: Debian and Ubuntu packages are available from the [PostgreSQL APT Repository](https://wiki.postgresql.org/wiki/Apt). Follow the [setup instructions](https://wiki.postgresql.org/wiki/Apt#Quickstart) and run:
```sh ```sh
sudo apt install postgresql-18-pgvector sudo apt install postgresql-17-pgvector
``` ```
Note: Replace `18` with your Postgres server version Note: Replace `17` with your Postgres server version
### Yum ### Yum
RPM packages are available from the [PostgreSQL Yum Repository](https://yum.postgresql.org/). Follow the [setup instructions](https://www.postgresql.org/download/linux/redhat/) for your distribution and run: RPM packages are available from the [PostgreSQL Yum Repository](https://yum.postgresql.org/). Follow the [setup instructions](https://www.postgresql.org/download/linux/redhat/) for your distribution and run:
```sh ```sh
sudo yum install pgvector_18 sudo yum install pgvector_17
# or # or
sudo dnf install pgvector_18 sudo dnf install pgvector_17
``` ```
Note: Replace `18` with your Postgres server version Note: Replace `17` with your Postgres server version
### pkg ### pkg
Install the FreeBSD package with: Install the FreeBSD package with:
```sh ```sh
pkg install postgresql17-pgvector pkg install postgresql16-pgvector
``` ```
or the port with: or the port with:
@@ -1234,14 +1220,6 @@ 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

@@ -1,2 +0,0 @@
-- complain if script is sourced in psql, rather than via CREATE EXTENSION
\echo Use "ALTER EXTENSION vector UPDATE TO '0.8.1'" to load this file. \quit

View File

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

View File

@@ -8,6 +8,9 @@
#error "Requires PostgreSQL 13+" #error "Requires PostgreSQL 13+"
#endif #endif
/* Check architecture in first header */
StaticAssertDecl(sizeof(Datum) == SIZEOF_DATUM, "Architecture mismatch");
extern uint64 (*BitHammingDistance) (uint32 bytes, unsigned char *ax, unsigned char *bx, uint64 distance); extern uint64 (*BitHammingDistance) (uint32 bytes, unsigned char *ax, unsigned char *bx, uint64 distance);
extern double (*BitJaccardDistance) (uint32 bytes, unsigned char *ax, unsigned char *bx, uint64 ab, uint64 aa, uint64 bb); extern double (*BitJaccardDistance) (uint32 bytes, unsigned char *ax, unsigned char *bx, uint64 ab, uint64 aa, uint64 bb);

View File

@@ -2,7 +2,6 @@
#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

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

View File

@@ -13,20 +13,12 @@
#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/varbit.h" #include "utils/numeric.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))
@@ -137,9 +129,9 @@ InitHalfVector(int dim)
return result; return result;
} }
#if PG_VERSION_NUM >= 170000 /*
#define halfvec_isspace(ch) scanner_isspace(ch) * Check for whitespace, since array_isspace() is static
#else */
static inline bool static inline bool
halfvec_isspace(char ch) halfvec_isspace(char ch)
{ {
@@ -152,7 +144,6 @@ halfvec_isspace(char ch)
return true; return true;
return false; return false;
} }
#endif
/* /*
* Check state array * Check state array
@@ -907,21 +898,8 @@ halfvec_binary_quantize(PG_FUNCTION_ARGS)
half *ax = a->x; half *ax = a->x;
VarBit *result = InitBitVector(a->dim); VarBit *result = InitBitVector(a->dim);
unsigned char *rx = VARBITS(result); unsigned char *rx = VARBITS(result);
int i = 0;
int count = (a->dim / 8) * 8;
/* Auto-vectorized on aarch64 */ for (int i = 0; i < a->dim; i++)
for (; i < count; i += 8)
{
unsigned char result_byte = 0;
for (int j = 0; j < 8; j++)
result_byte |= (HalfToFloat4(ax[i + j]) > 0) << (7 - j);
rx[i / 8] = result_byte;
}
for (; i < a->dim; i++)
rx[i / 8] |= (HalfToFloat4(ax[i]) > 0) << (7 - (i % 8)); rx[i / 8] |= (HalfToFloat4(ax[i]) > 0) << (7 - (i % 8));
PG_RETURN_VARBIT_P(result); PG_RETURN_VARBIT_P(result);

View File

@@ -1,24 +1,18 @@
#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)
@@ -58,20 +52,12 @@ HnswInitLockTranche(void)
sizeof(int) * 1, sizeof(int) * 1,
&found); &found);
if (!found) if (!found)
{
#if PG_VERSION_NUM >= 190000
tranche_ids[0] = LWLockNewTrancheId("HnswBuild");
#else
tranche_ids[0] = LWLockNewTrancheId(); tranche_ids[0] = LWLockNewTrancheId();
#endif
}
hnsw_lock_tranche_id = tranche_ids[0]; hnsw_lock_tranche_id = tranche_ids[0];
LWLockRelease(AddinShmemInitLock); LWLockRelease(AddinShmemInitLock);
#if PG_VERSION_NUM < 190000
/* Per-backend registration of the tranche ID */ /* Per-backend registration of the tranche ID */
LWLockRegisterTranche(hnsw_lock_tranche_id, "HnswBuild"); LWLockRegisterTranche(hnsw_lock_tranche_id, "HnswBuild");
#endif
} }
/* /*
@@ -144,7 +130,7 @@ hnswcostestimate(PlannerInfo *root, IndexPath *path, double loop_count,
Relation index; Relation index;
/* Never use index without order */ /* Never use index without order */
if (path->indexorderbys == NIL) if (path->indexorderbys == NULL)
{ {
*indexStartupCost = get_float8_infinity(); *indexStartupCost = get_float8_infinity();
*indexTotalCost = get_float8_infinity(); *indexTotalCost = get_float8_infinity();
@@ -266,64 +252,6 @@ 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;
@@ -331,11 +259,6 @@ hnswhandler(PG_FUNCTION_ARGS)
amroutine->amoptsprocnum = 0; amroutine->amoptsprocnum = 0;
amroutine->amcanorder = false; amroutine->amcanorder = false;
amroutine->amcanorderbyop = true; amroutine->amcanorderbyop = true;
#if PG_VERSION_NUM >= 180000
amroutine->amcanhash = false;
amroutine->amconsistentequality = false;
amroutine->amconsistentordering = false;
#endif
amroutine->amcanbackward = false; /* can change direction mid-scan */ amroutine->amcanbackward = false; /* can change direction mid-scan */
amroutine->amcanunique = false; amroutine->amcanunique = false;
amroutine->amcanmulticol = false; amroutine->amcanmulticol = false;
@@ -368,9 +291,6 @@ hnswhandler(PG_FUNCTION_ARGS)
amroutine->amvacuumcleanup = hnswvacuumcleanup; amroutine->amvacuumcleanup = hnswvacuumcleanup;
amroutine->amcanreturn = NULL; amroutine->amcanreturn = NULL;
amroutine->amcostestimate = hnswcostestimate; amroutine->amcostestimate = hnswcostestimate;
#if PG_VERSION_NUM >= 180000
amroutine->amgettreeheight = NULL;
#endif
amroutine->amoptions = hnswoptions; amroutine->amoptions = hnswoptions;
amroutine->amproperty = NULL; /* TODO AMPROP_DISTANCE_ORDERABLE */ amroutine->amproperty = NULL; /* TODO AMPROP_DISTANCE_ORDERABLE */
amroutine->ambuildphasename = hnswbuildphasename; amroutine->ambuildphasename = hnswbuildphasename;
@@ -391,11 +311,5 @@ hnswhandler(PG_FUNCTION_ARGS)
amroutine->aminitparallelscan = NULL; amroutine->aminitparallelscan = NULL;
amroutine->amparallelrescan = NULL; amroutine->amparallelrescan = NULL;
#if PG_VERSION_NUM >= 180000
amroutine->amtranslatestrategy = NULL;
amroutine->amtranslatecmptype = NULL;
#endif
PG_RETURN_POINTER(amroutine); PG_RETURN_POINTER(amroutine);
#endif
} }

View File

@@ -3,8 +3,6 @@
#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"
@@ -14,10 +12,6 @@
#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
@@ -430,7 +424,7 @@ 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); void HnswFindElementNeighbors(char *base, HnswElement element, HnswElement entryPoint, Relation index, HnswSupport * support, int m, int efConstruction, bool existing);
HnswSearchCandidate *HnswEntryCandidate(char *base, HnswElement entryPoint, HnswQuery * q, Relation index, HnswSupport * support, bool loadVec); HnswSearchCandidate *HnswEntryCandidate(char *base, HnswElement em, HnswQuery * q, Relation rel, 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

@@ -36,14 +36,11 @@
*/ */
#include "postgres.h" #include "postgres.h"
#include <limits.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"
@@ -51,18 +48,11 @@
#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
#include "varatt.h"
#endif
#if PG_VERSION_NUM >= 140000 #if PG_VERSION_NUM >= 140000
#include "utils/backend_progress.h" #include "utils/backend_progress.h"
@@ -79,8 +69,6 @@
#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
*/ */
@@ -410,7 +398,7 @@ UpdateNeighborsInMemory(char *base, HnswSupport * support, HnswElement e, int m)
* Update graph in memory * Update graph in memory
*/ */
static void static void
UpdateGraphInMemory(HnswSupport * support, HnswElement element, int m, HnswElement entryPoint, HnswBuildState * buildstate) UpdateGraphInMemory(HnswSupport * support, HnswElement element, int m, int efConstruction, HnswElement entryPoint, HnswBuildState * buildstate)
{ {
HnswGraph *graph = buildstate->graph; HnswGraph *graph = buildstate->graph;
char *base = buildstate->hnswarea; char *base = buildstate->hnswarea;
@@ -472,7 +460,7 @@ InsertTupleInMemory(HnswBuildState * buildstate, HnswElement element)
HnswFindElementNeighbors(base, element, entryPoint, NULL, support, m, efConstruction, false); 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, efConstruction, entryPoint, buildstate);
/* Release entry lock */ /* Release entry lock */
LWLockRelease(entryLock); LWLockRelease(entryLock);
@@ -493,7 +481,6 @@ 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))
@@ -502,9 +489,6 @@ 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);
@@ -526,7 +510,7 @@ 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 + memoryMargin >= graph->memoryTotal) if (graph->memoryUsed >= graph->memoryTotal)
{ {
LWLockRelease(&graph->allocatorLock); LWLockRelease(&graph->allocatorLock);
@@ -561,7 +545,7 @@ 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, (char *) valuePtr); HnswPtrStore(base, element->value, 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);
@@ -619,7 +603,7 @@ 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 = Min(memoryTotal, HNSW_MAX_GRAPH_MEMORY); graph->memoryTotal = memoryTotal;
graph->flushed = false; graph->flushed = false;
graph->indtuples = 0; graph->indtuples = 0;
SpinLockInit(&graph->lock); SpinLockInit(&graph->lock);
@@ -660,17 +644,9 @@ static void *
HnswSharedMemoryAlloc(Size size, void *state) HnswSharedMemoryAlloc(Size size, void *state)
{ {
HnswBuildState *buildstate = (HnswBuildState *) state; HnswBuildState *buildstate = (HnswBuildState *) state;
Size alignedSize = MAXALIGN(size); void *chunk = buildstate->hnswarea + buildstate->graph->memoryUsed;
void *chunk;
if (alignedSize > 1024 * 1024) buildstate->graph->memoryUsed += MAXALIGN(size);
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;
} }
@@ -956,8 +932,6 @@ 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);
@@ -1000,7 +974,8 @@ 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);
InitGraph(&hnswshared->graphData, hnswarea, esthnswarea); /* Report less than allocated so never fails */
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
@@ -1079,7 +1054,7 @@ ComputeParallelWorkers(Relation heap, Relation index)
* Build graph * Build graph
*/ */
static void static void
BuildGraph(HnswBuildState * buildstate) BuildGraph(HnswBuildState * buildstate, ForkNumber forkNum)
{ {
int parallel_workers = 0; int parallel_workers = 0;
@@ -1127,7 +1102,7 @@ BuildIndex(Relation heap, Relation index, IndexInfo *indexInfo,
InitBuildState(buildstate, heap, index, indexInfo, forkNum); InitBuildState(buildstate, heap, index, indexInfo, forkNum);
BuildGraph(buildstate); BuildGraph(buildstate, forkNum);
if (RelationNeedsWAL(index) || forkNum == INIT_FORKNUM) if (RelationNeedsWAL(index) || forkNum == INIT_FORKNUM)
log_newpage_range(index, forkNum, 0, RelationGetNumberOfBlocksInFork(index, forkNum), true); log_newpage_range(index, forkNum, 0, RelationGetNumberOfBlocksInFork(index, forkNum), true);

View File

@@ -1,18 +1,13 @@
#include "postgres.h" #include "postgres.h"
#include "access/genam.h" #include <math.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
#include "varatt.h"
#endif
/* /*
* Get the insert page * Get the insert page
@@ -665,7 +660,7 @@ FindDuplicateOnDisk(Relation index, HnswElement element, bool building)
* Update graph on disk * Update graph on disk
*/ */
static void static void
UpdateGraphOnDisk(Relation index, HnswSupport * support, HnswElement element, int m, HnswElement entryPoint, bool building) UpdateGraphOnDisk(Relation index, HnswSupport * support, HnswElement element, int m, int efConstruction, HnswElement entryPoint, bool building)
{ {
BlockNumber newInsertPage = InvalidBlockNumber; BlockNumber newInsertPage = InvalidBlockNumber;
@@ -713,7 +708,7 @@ 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, (char *) DatumGetPointer(value)); HnswPtrStore(base, element->value, 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)
@@ -733,7 +728,7 @@ HnswInsertTupleOnDisk(Relation index, HnswSupport * support, Datum value, ItemPo
HnswFindElementNeighbors(base, element, entryPoint, index, support, m, efConstruction, false); 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, efConstruction, entryPoint, building);
/* Release lock */ /* Release lock */
UnlockPage(index, HNSW_UPDATE_LOCK, lockmode); UnlockPage(index, HNSW_UPDATE_LOCK, lockmode);

View File

@@ -1,23 +1,12 @@
#include "postgres.h" #include "postgres.h"
#include <limits.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
@@ -153,7 +142,7 @@ hnswbeginscan(Relation index, int nkeys, int norderbys)
/* Calculate max memory */ /* Calculate max memory */
/* Add 256 extra bytes to fill last block when close */ /* Add 256 extra bytes to fill last block when close */
maxMemory = (double) work_mem * hnsw_scan_mem_multiplier * 1024.0 + 256; maxMemory = (double) work_mem * hnsw_scan_mem_multiplier * 1024.0 + 256;
so->maxMemory = Min(maxMemory, (double) (SIZE_MAX / 2)); so->maxMemory = Min(maxMemory, (double) SIZE_MAX);
scan->opaque = so; scan->opaque = so;
@@ -204,10 +193,6 @@ hnswgettuple(IndexScanDesc scan, ScanDirection dir)
/* Count index scan for stats */ /* Count index scan for stats */
pgstat_count_index_scan(scan->indexRelation); pgstat_count_index_scan(scan->indexRelation);
#if PG_VERSION_NUM >= 180000
if (scan->instrument)
scan->instrument->nsearches++;
#endif
/* Safety check */ /* Safety check */
if (scan->orderByData == NULL) if (scan->orderByData == NULL)

View File

@@ -2,24 +2,18 @@
#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
#include "varatt.h"
#endif
#if PG_VERSION_NUM < 170000 #if PG_VERSION_NUM < 170000
static inline uint64 static inline uint64
@@ -262,7 +256,7 @@ 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, (char *) NULL); HnswPtrStore(base, element->value, (Pointer) NULL);
return element; return element;
} }
@@ -288,7 +282,7 @@ 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, (char *) NULL); HnswPtrStore(base, element->value, (Pointer) NULL);
return element; return element;
} }
@@ -514,7 +508,7 @@ 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, (char *) DatumGetPointer(value)); HnswPtrStore(base, element->value, DatumGetPointer(value));
} }
} }
@@ -928,7 +922,7 @@ HnswSearchLayer(char *base, HnswQuery * q, List *ep, int ef, int lc, Relation in
continue; continue;
} }
if (!(eDistance < f->distance || alwaysAdd)) if (eElement == NULL || !(eDistance < f->distance || alwaysAdd))
{ {
if (discarded != NULL) if (discarded != NULL)
{ {

View File

@@ -1,18 +1,13 @@
#include "postgres.h" #include "postgres.h"
#include "access/genam.h" #include <math.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
#include "varatt.h"
#endif
#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)
@@ -179,12 +174,7 @@ 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);
@@ -534,9 +524,8 @@ 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

@@ -2,36 +2,23 @@
#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 "access/xloginsert.h" #include "bitvec.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 "fmgr.h" #include "halfvec.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 "utils/rel.h" #include "vector.h"
#include "utils/sampling.h"
#include "utils/snapmgr.h"
#include "utils/tuplesort.h"
#if PG_VERSION_NUM >= 160000
#include "varatt.h"
#endif
#if PG_VERSION_NUM >= 140000 #if PG_VERSION_NUM >= 140000
#include "utils/backend_progress.h" #include "utils/backend_progress.h"
@@ -151,7 +138,7 @@ SampleRows(IvfflatBuildState * buildstate)
* Add tuple to sort * Add tuple to sort
*/ */
static void static void
AddTupleToSort(ItemPointer tid, Datum *values, IvfflatBuildState * buildstate) AddTupleToSort(Relation index, ItemPointer tid, Datum *values, IvfflatBuildState * buildstate)
{ {
double distance; double distance;
double minDistance = DBL_MAX; double minDistance = DBL_MAX;
@@ -228,7 +215,7 @@ BuildCallback(Relation index, ItemPointer tid, Datum *values,
oldCtx = MemoryContextSwitchTo(buildstate->tmpCtx); oldCtx = MemoryContextSwitchTo(buildstate->tmpCtx);
/* Add tuple to sort */ /* Add tuple to sort */
AddTupleToSort(tid, values, buildstate); AddTupleToSort(index, tid, values, buildstate);
/* Reset memory context */ /* Reset memory context */
MemoryContextSwitchTo(oldCtx); MemoryContextSwitchTo(oldCtx);
@@ -483,8 +470,8 @@ CreateMetaPage(Relation index, int dimensions, int lists, ForkNumber forkNum)
* Create list pages * Create list pages
*/ */
static void static void
CreateListPages(Relation index, VectorArray centers, int lists, CreateListPages(Relation index, VectorArray centers, int dimensions,
ForkNumber forkNum, ListInfo * *listInfo) int lists, ForkNumber forkNum, ListInfo * *listInfo)
{ {
Buffer buf; Buffer buf;
Page page; Page page;
@@ -1017,7 +1004,7 @@ BuildIndex(Relation heap, Relation index, IndexInfo *indexInfo,
/* Create pages */ /* Create pages */
CreateMetaPage(index, buildstate->dimensions, buildstate->lists, forkNum); CreateMetaPage(index, buildstate->dimensions, buildstate->lists, forkNum);
CreateListPages(index, buildstate->centers, buildstate->lists, forkNum, &buildstate->listInfo); CreateListPages(index, buildstate->centers, buildstate->dimensions, buildstate->lists, forkNum, &buildstate->listInfo);
CreateEntryPages(buildstate, forkNum); CreateEntryPages(buildstate, forkNum);
/* Write WAL for initialization fork since GenericXLog functions do not */ /* Write WAL for initialization fork since GenericXLog functions do not */
@@ -1036,10 +1023,6 @@ ivfflatbuild(Relation heap, Relation index, IndexInfo *indexInfo)
IndexBuildResult *result; IndexBuildResult *result;
IvfflatBuildState buildstate; IvfflatBuildState buildstate;
#ifdef IVFFLAT_BENCH
SeedRandom(42);
#endif
BuildIndex(heap, index, indexInfo, &buildstate, MAIN_FORKNUM); BuildIndex(heap, index, indexInfo, &buildstate, MAIN_FORKNUM);
result = (IndexBuildResult *) palloc(sizeof(IndexBuildResult)); result = (IndexBuildResult *) palloc(sizeof(IndexBuildResult));

View File

@@ -3,19 +3,14 @@
#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)
@@ -97,7 +92,7 @@ ivfflatcostestimate(PlannerInfo *root, IndexPath *path, double loop_count,
Relation index; Relation index;
/* Never use index without order */ /* Never use index without order */
if (path->indexorderbys == NIL) if (path->indexorderbys == NULL)
{ {
*indexStartupCost = get_float8_infinity(); *indexStartupCost = get_float8_infinity();
*indexTotalCost = get_float8_infinity(); *indexTotalCost = get_float8_infinity();
@@ -184,64 +179,6 @@ 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;
@@ -249,11 +186,6 @@ ivfflathandler(PG_FUNCTION_ARGS)
amroutine->amoptsprocnum = 0; amroutine->amoptsprocnum = 0;
amroutine->amcanorder = false; amroutine->amcanorder = false;
amroutine->amcanorderbyop = true; amroutine->amcanorderbyop = true;
#if PG_VERSION_NUM >= 180000
amroutine->amcanhash = false;
amroutine->amconsistentequality = false;
amroutine->amconsistentordering = false;
#endif
amroutine->amcanbackward = false; /* can change direction mid-scan */ amroutine->amcanbackward = false; /* can change direction mid-scan */
amroutine->amcanunique = false; amroutine->amcanunique = false;
amroutine->amcanmulticol = false; amroutine->amcanmulticol = false;
@@ -286,9 +218,6 @@ ivfflathandler(PG_FUNCTION_ARGS)
amroutine->amvacuumcleanup = ivfflatvacuumcleanup; amroutine->amvacuumcleanup = ivfflatvacuumcleanup;
amroutine->amcanreturn = NULL; /* tuple not included in heapsort */ amroutine->amcanreturn = NULL; /* tuple not included in heapsort */
amroutine->amcostestimate = ivfflatcostestimate; amroutine->amcostestimate = ivfflatcostestimate;
#if PG_VERSION_NUM >= 180000
amroutine->amgettreeheight = NULL;
#endif
amroutine->amoptions = ivfflatoptions; amroutine->amoptions = ivfflatoptions;
amroutine->amproperty = NULL; /* TODO AMPROP_DISTANCE_ORDERABLE */ amroutine->amproperty = NULL; /* TODO AMPROP_DISTANCE_ORDERABLE */
amroutine->ambuildphasename = ivfflatbuildphasename; amroutine->ambuildphasename = ivfflatbuildphasename;
@@ -309,11 +238,5 @@ ivfflathandler(PG_FUNCTION_ARGS)
amroutine->aminitparallelscan = NULL; amroutine->aminitparallelscan = NULL;
amroutine->amparallelrescan = NULL; amroutine->amparallelrescan = NULL;
#if PG_VERSION_NUM >= 180000
amroutine->amtranslatestrategy = NULL;
amroutine->amtranslatecmptype = NULL;
#endif
PG_RETURN_POINTER(amroutine); PG_RETURN_POINTER(amroutine);
#endif
} }

View File

@@ -13,10 +13,6 @@
#include "utils/tuplesort.h" #include "utils/tuplesort.h"
#include "vector.h" #include "vector.h"
#if PG_VERSION_NUM >= 160000
#include "varatt.h"
#endif
#if PG_VERSION_NUM >= 150000 #if PG_VERSION_NUM >= 150000
#include "common/pg_prng.h" #include "common/pg_prng.h"
#endif #endif
@@ -25,10 +21,6 @@
#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 */
@@ -81,11 +73,9 @@ typedef Pointer Item;
#if PG_VERSION_NUM >= 150000 #if PG_VERSION_NUM >= 150000
#define RandomDouble() pg_prng_double(&pg_global_prng_state) #define RandomDouble() pg_prng_double(&pg_global_prng_state)
#define RandomInt() pg_prng_uint32(&pg_global_prng_state) #define RandomInt() pg_prng_uint32(&pg_global_prng_state)
#define SeedRandom(seed) pg_prng_seed(&pg_global_prng_state, seed)
#else #else
#define RandomDouble() (((double) random()) / MAX_RANDOM_VALUE) #define RandomDouble() (((double) random()) / MAX_RANDOM_VALUE)
#define RandomInt() random() #define RandomInt() random()
#define SeedRandom(seed) srandom(seed)
#endif #endif
/* Variables */ /* Variables */

View File

@@ -2,16 +2,11 @@
#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
@@ -70,7 +65,7 @@ FindInsertPage(Relation index, Datum *values, BlockNumber *insertPage, ListInfo
* Insert a tuple into the index * Insert a tuple into the index
*/ */
static void static void
InsertTuple(Relation index, Datum *values, bool *isnull, ItemPointer heap_tid) InsertTuple(Relation index, Datum *values, bool *isnull, ItemPointer heap_tid, Relation heapRel)
{ {
const IvfflatTypeInfo *typeInfo = IvfflatGetTypeInfo(index); const IvfflatTypeInfo *typeInfo = IvfflatGetTypeInfo(index);
IndexTuple itup; IndexTuple itup;
@@ -209,7 +204,7 @@ ivfflatinsert(Relation index, Datum *values, bool *isnull, ItemPointer heap_tid,
oldCtx = MemoryContextSwitchTo(insertCtx); oldCtx = MemoryContextSwitchTo(insertCtx);
/* Insert tuple */ /* Insert tuple */
InsertTuple(index, values, isnull, heap_tid); InsertTuple(index, values, isnull, heap_tid, heap);
/* Delete memory context */ /* Delete memory context */
MemoryContextSwitchTo(oldCtx); MemoryContextSwitchTo(oldCtx);

View File

@@ -1,19 +1,17 @@
#include "postgres.h" #include "postgres.h"
#include <float.h> #include <float.h>
#include <limits.h>
#include <math.h> #include <math.h>
#include "access/genam.h" #include "bitvec.h"
#include "fmgr.h" #include "halfutils.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 "utils/relcache.h" #include "vector.h"
#if PG_VERSION_NUM >= 160000
#include "varatt.h"
#endif
/* /*
* Initialize with kmeans++ * Initialize with kmeans++

View File

@@ -2,26 +2,15 @@
#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)
@@ -366,10 +355,6 @@ ivfflatgettuple(IndexScanDesc scan, ScanDirection dir)
/* Count index scan for stats */ /* Count index scan for stats */
pgstat_count_index_scan(scan->indexRelation); pgstat_count_index_scan(scan->indexRelation);
#if PG_VERSION_NUM >= 180000
if (scan->instrument)
scan->instrument->nsearches++;
#endif
/* Safety check */ /* Safety check */
if (scan->orderByData == NULL) if (scan->orderByData == NULL)

View File

@@ -1,19 +1,13 @@
#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
@@ -265,8 +259,8 @@ VectorUpdateCenter(Pointer v, int dimensions, float *x)
SET_VARSIZE(vec, VECTOR_SIZE(dimensions)); SET_VARSIZE(vec, VECTOR_SIZE(dimensions));
vec->dim = dimensions; vec->dim = dimensions;
for (int i = 0; i < dimensions; i++) for (int k = 0; k < dimensions; k++)
vec->x[i] = x[i]; vec->x[k] = x[k];
} }
static void static void
@@ -277,8 +271,8 @@ HalfvecUpdateCenter(Pointer v, int dimensions, float *x)
SET_VARSIZE(vec, HALFVEC_SIZE(dimensions)); SET_VARSIZE(vec, HALFVEC_SIZE(dimensions));
vec->dim = dimensions; vec->dim = dimensions;
for (int i = 0; i < dimensions; i++) for (int k = 0; k < dimensions; k++)
vec->x[i] = Float4ToHalfUnchecked(x[i]); vec->x[k] = Float4ToHalfUnchecked(x[k]);
} }
static void static void
@@ -290,33 +284,29 @@ BitUpdateCenter(Pointer v, int dimensions, float *x)
SET_VARSIZE(vec, VARBITTOTALLEN(dimensions)); SET_VARSIZE(vec, VARBITTOTALLEN(dimensions));
VARBITLEN(vec) = dimensions; VARBITLEN(vec) = dimensions;
for (uint32 i = 0; i < VARBITBYTES(vec); i++) for (uint32 k = 0; k < VARBITBYTES(vec); k++)
nx[i] = 0; nx[k] = 0;
for (int i = 0; i < dimensions; i++) for (int k = 0; k < dimensions; k++)
nx[i / 8] |= (x[i] > 0.5 ? 1 : 0) << (7 - (i % 8)); nx[k / 8] |= (x[k] > 0.5 ? 1 : 0) << (7 - (k % 8));
} }
static void static void
VectorSumCenter(Pointer v, float *x) VectorSumCenter(Pointer v, float *x)
{ {
Vector *vec = (Vector *) v; Vector *vec = (Vector *) v;
int dim = vec->dim;
/* Auto-vectorized */ for (int k = 0; k < vec->dim; k++)
for (int i = 0; i < dim; i++) x[k] += vec->x[k];
x[i] += vec->x[i];
} }
static void static void
HalfvecSumCenter(Pointer v, float *x) HalfvecSumCenter(Pointer v, float *x)
{ {
HalfVector *vec = (HalfVector *) v; HalfVector *vec = (HalfVector *) v;
int dim = vec->dim;
/* Auto-vectorized on aarch64 */ for (int k = 0; k < vec->dim; k++)
for (int i = 0; i < dim; i++) x[k] += HalfToFloat4(vec->x[k]);
x[i] += HalfToFloat4(vec->x[i]);
} }
static void static void
@@ -324,8 +314,8 @@ BitSumCenter(Pointer v, float *x)
{ {
VarBit *vec = (VarBit *) v; VarBit *vec = (VarBit *) v;
for (int i = 0; i < VARBITLEN(vec); i++) for (int k = 0; k < VARBITLEN(vec); k++)
x[i] += (float) (((VARBITS(vec)[i / 8]) >> (7 - (i % 8))) & 0x01); x[k] += (float) (((VARBITS(vec)[k / 8]) >> (7 - (k % 8))) & 0x01);
} }
/* /*

View File

@@ -1,12 +1,9 @@
#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

@@ -5,27 +5,18 @@
#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;
@@ -159,9 +150,9 @@ InitSparseVector(int dim, int nnz)
return result; return result;
} }
#if PG_VERSION_NUM >= 170000 /*
#define sparsevec_isspace(ch) scanner_isspace(ch) * Check for whitespace, since array_isspace() is static
#else */
static inline bool static inline bool
sparsevec_isspace(char ch) sparsevec_isspace(char ch)
{ {
@@ -174,7 +165,6 @@ sparsevec_isspace(char ch)
return true; return true;
return false; return false;
} }
#endif
/* /*
* Compare indices * Compare indices
@@ -895,24 +885,28 @@ SparsevecInnerProduct(SparseVector * a, SparseVector * b)
float *ax = SPARSEVEC_VALUES(a); float *ax = SPARSEVEC_VALUES(a);
float *bx = SPARSEVEC_VALUES(b); float *bx = SPARSEVEC_VALUES(b);
float distance = 0.0; float distance = 0.0;
int i = 0; int bpos = 0;
int j = 0;
while (i < a->nnz && j < b->nnz) for (int i = 0; i < a->nnz; i++)
{ {
int ai = a->indices[i]; int ai = a->indices[i];
for (int j = bpos; j < b->nnz; j++)
{
int bi = b->indices[j]; int bi = b->indices[j];
/* Only update when the same index */
if (ai == bi) if (ai == bi)
{
distance += ax[i] * bx[j]; distance += ax[i] * bx[j];
i++;
j++; /* Update start for next iteration */
if (ai >= bi)
bpos = j + 1;
/* Found or passed it */
if (bi >= ai)
break;
} }
else if (ai < bi)
i++;
else
j++;
} }
return distance; return distance;

View File

@@ -16,20 +16,16 @@
#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/varbit.h" #include "utils/numeric.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))
@@ -39,11 +35,7 @@
#define VECTOR_TARGET_CLONES #define VECTOR_TARGET_CLONES
#endif #endif
#if PG_VERSION_NUM >= 180000
PG_MODULE_MAGIC_EXT(.name = "vector",.version = "0.8.2");
#else
PG_MODULE_MAGIC; PG_MODULE_MAGIC;
#endif
/* /*
* Initialize index options and variables * Initialize index options and variables
@@ -133,9 +125,9 @@ InitVector(int dim)
return result; return result;
} }
#if PG_VERSION_NUM >= 170000 /*
#define vector_isspace(ch) scanner_isspace(ch) * Check for whitespace, since array_isspace() is static
#else */
static inline bool static inline bool
vector_isspace(char ch) vector_isspace(char ch)
{ {
@@ -148,7 +140,6 @@ vector_isspace(char ch)
return true; return true;
return false; return false;
} }
#endif
/* /*
* Check state array * Check state array
@@ -929,13 +920,11 @@ vector_concat(PG_FUNCTION_ARGS)
CheckDim(dim); CheckDim(dim);
result = InitVector(dim); result = InitVector(dim);
/* Auto-vectorized */ for (int i = 0; i < a->dim; i++)
for (int i = 0, imax = a->dim; i < imax; i++)
result->x[i] = a->x[i]; result->x[i] = a->x[i];
/* Auto-vectorized */ for (int i = 0; i < b->dim; i++)
for (int i = 0, imax = b->dim, start = a->dim; i < imax; i++) result->x[i + a->dim] = b->x[i];
result->x[i + start] = b->x[i];
PG_RETURN_POINTER(result); PG_RETURN_POINTER(result);
} }
@@ -951,21 +940,8 @@ binary_quantize(PG_FUNCTION_ARGS)
float *ax = a->x; float *ax = a->x;
VarBit *result = InitBitVector(a->dim); VarBit *result = InitBitVector(a->dim);
unsigned char *rx = VARBITS(result); unsigned char *rx = VARBITS(result);
int i = 0;
int count = (a->dim / 8) * 8;
/* Auto-vectorized */ for (int i = 0; i < a->dim; i++)
for (; i < count; i += 8)
{
unsigned char result_byte = 0;
for (int j = 0; j < 8; j++)
result_byte |= (ax[i + j] > 0) << (7 - j);
rx[i / 8] = result_byte;
}
for (; i < a->dim; i++)
rx[i / 8] |= (ax[i] > 0) << (7 - (i % 8)); rx[i / 8] |= (ax[i] > 0) << (7 - (i % 8));
PG_RETURN_VARBIT_P(result); PG_RETURN_VARBIT_P(result);

View File

@@ -540,12 +540,6 @@ SELECT binary_quantize('[0,0.1,-0.2,-0.3,0.4,0.5,0.6,-0.7,0.8,-0.9,1]'::halfvec)
01001110101 01001110101
(1 row) (1 row)
SELECT binary_quantize('[1,2,3,-4,5,6,-7,8,1,-2,-3,4,5,-6,7,8,-1,2,3]'::halfvec);
binary_quantize
---------------------
1110110110011011011
(1 row)
SELECT subvector('[1,2,3,4,5]'::halfvec, 1, 3); SELECT subvector('[1,2,3,4,5]'::halfvec, 1, 3);
subvector subvector
----------- -----------

View File

@@ -576,12 +576,6 @@ SELECT binary_quantize('[0,0.1,-0.2,-0.3,0.4,0.5,0.6,-0.7,0.8,-0.9,1]'::vector);
01001110101 01001110101
(1 row) (1 row)
SELECT binary_quantize('[1,2,3,-4,5,6,-7,8,1,-2,-3,4,5,-6,7,8,-1,2,3]'::vector);
binary_quantize
---------------------
1110110110011011011
(1 row)
SELECT subvector('[1,2,3,4,5]'::vector, 1, 3); SELECT subvector('[1,2,3,4,5]'::vector, 1, 3);
subvector subvector
----------- -----------

View File

@@ -121,7 +121,6 @@ SELECT l2_normalize('[65504]'::halfvec);
SELECT binary_quantize('[1,0,-1]'::halfvec); SELECT binary_quantize('[1,0,-1]'::halfvec);
SELECT binary_quantize('[0,0.1,-0.2,-0.3,0.4,0.5,0.6,-0.7,0.8,-0.9,1]'::halfvec); SELECT binary_quantize('[0,0.1,-0.2,-0.3,0.4,0.5,0.6,-0.7,0.8,-0.9,1]'::halfvec);
SELECT binary_quantize('[1,2,3,-4,5,6,-7,8,1,-2,-3,4,5,-6,7,8,-1,2,3]'::halfvec);
SELECT subvector('[1,2,3,4,5]'::halfvec, 1, 3); SELECT subvector('[1,2,3,4,5]'::halfvec, 1, 3);
SELECT subvector('[1,2,3,4,5]'::halfvec, 3, 2); SELECT subvector('[1,2,3,4,5]'::halfvec, 3, 2);

View File

@@ -128,7 +128,6 @@ SELECT l2_normalize('[3e38]'::vector);
SELECT binary_quantize('[1,0,-1]'::vector); SELECT binary_quantize('[1,0,-1]'::vector);
SELECT binary_quantize('[0,0.1,-0.2,-0.3,0.4,0.5,0.6,-0.7,0.8,-0.9,1]'::vector); SELECT binary_quantize('[0,0.1,-0.2,-0.3,0.4,0.5,0.6,-0.7,0.8,-0.9,1]'::vector);
SELECT binary_quantize('[1,2,3,-4,5,6,-7,8,1,-2,-3,4,5,-6,7,8,-1,2,3]'::vector);
SELECT subvector('[1,2,3,4,5]'::vector, 1, 3); SELECT subvector('[1,2,3,4,5]'::vector, 1, 3);
SELECT subvector('[1,2,3,4,5]'::vector, 3, 2); SELECT subvector('[1,2,3,4,5]'::vector, 3, 2);

View File

@@ -16,8 +16,17 @@ $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
my @operators = ("<->", "<#>", "<=>");
my @opclasses = ("vector_l2_ops", "vector_ip_ops", "vector_cosine_ops");
for my $i (0 .. $#operators)
{
my $operator = $operators[$i];
my $opclass = $opclasses[$i];
# Add index # Add index
$node->safe_psql("postgres", "CREATE INDEX ON tst USING ivfflat (v vector_l2_ops);"); $node->safe_psql("postgres", "CREATE INDEX ON tst USING ivfflat (v $opclass);");
# Test 100% recall # Test 100% recall
for (1 .. 20) for (1 .. 20)
@@ -30,5 +39,6 @@ for (1 .. 20)
)); ));
is($res, $query); is($res, $query);
} }
}
done_testing(); done_testing();

View File

@@ -1,29 +0,0 @@
use strict;
use warnings FATAL => 'all';
use PostgreSQL::Test::Cluster;
use PostgreSQL::Test::Utils;
use Test::More;
# Initialize node
my $node = PostgreSQL::Test::Cluster->new('node');
$node->init;
$node->start;
# Create table
$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 = '0.8.2' default_version = '0.8.0'
module_pathname = '$libdir/vector' module_pathname = '$libdir/vector'
relocatable = true relocatable = true