mirror of
https://github.com/pgvector/pgvector.git
synced 2026-07-22 12:07:34 +08:00
Compare commits
1 Commits
concat-v2
...
hnsw-datum
| Author | SHA1 | Date | |
|---|---|---|---|
|
c8e4bf1c3c |
46
.github/workflows/build.yml
vendored
46
.github/workflows/build.yml
vendored
@@ -28,7 +28,7 @@ jobs:
|
||||
dev-files: true
|
||||
- run: make
|
||||
env:
|
||||
PG_CFLAGS: -DUSE_ASSERT_CHECKING -Wall -Wextra -Werror -Wno-unused-parameter -Wno-sign-compare
|
||||
PG_CFLAGS: -Wall -Wextra -Werror -Wno-unused-parameter -Wno-sign-compare
|
||||
- run: |
|
||||
export PG_CONFIG=`which pg_config`
|
||||
sudo --preserve-env=PG_CONFIG make install
|
||||
@@ -40,43 +40,27 @@ jobs:
|
||||
sudo apt-get install libipc-run-perl
|
||||
- run: make prove_installcheck
|
||||
mac:
|
||||
runs-on: ${{ matrix.os }}
|
||||
runs-on: macos-latest
|
||||
if: ${{ !startsWith(github.ref_name, 'windows') }}
|
||||
strategy:
|
||||
fail-fast: false
|
||||
matrix:
|
||||
include:
|
||||
- postgres: 16
|
||||
os: macos-14
|
||||
- postgres: 14
|
||||
os: macos-12
|
||||
steps:
|
||||
- uses: actions/checkout@v4
|
||||
- uses: ankane/setup-postgres@v1
|
||||
with:
|
||||
postgres-version: ${{ matrix.postgres }}
|
||||
postgres-version: 14
|
||||
- run: make
|
||||
env:
|
||||
PG_CFLAGS: -DUSE_ASSERT_CHECKING -Wall -Wextra -Werror -Wno-unused-parameter
|
||||
PG_CFLAGS: -Wall -Wextra -Werror -Wno-unused-parameter
|
||||
- run: make install
|
||||
- run: make installcheck
|
||||
- if: ${{ failure() }}
|
||||
run: cat regression.diffs
|
||||
# Homebrew Postgres does not enable TAP tests, so need to download
|
||||
- run: |
|
||||
brew install cpanm
|
||||
cpanm --notest IPC::Run
|
||||
wget -q https://github.com/postgres/postgres/archive/refs/tags/$TAG.tar.gz
|
||||
tar xf $TAG.tar.gz
|
||||
mv postgres-$TAG postgres
|
||||
env:
|
||||
TAG: ${{ matrix.postgres == 16 && 'REL_16_2' || 'REL_14_11' }}
|
||||
- run: make prove_installcheck PROVE_FLAGS="-I ./postgres/src/test/perl -I ./test/perl"
|
||||
env:
|
||||
PERL5LIB: /Users/runner/perl5/lib/perl5
|
||||
- run: make clean && $(brew --prefix llvm@15)/bin/scan-build --status-bugs make
|
||||
env:
|
||||
PG_CFLAGS: -DUSE_ASSERT_CHECKING
|
||||
wget -q https://github.com/postgres/postgres/archive/refs/tags/REL_14_10.tar.gz
|
||||
tar xf REL_14_10.tar.gz
|
||||
- run: make prove_installcheck PROVE_FLAGS="-I ./postgres-REL_14_10/src/test/perl" PERL5LIB="/Users/runner/perl5/lib/perl5"
|
||||
- run: make clean && /usr/local/opt/llvm@15/bin/scan-build --status-bugs make PG_CFLAGS="-DUSE_ASSERT_CHECKING"
|
||||
windows:
|
||||
runs-on: windows-latest
|
||||
if: ${{ !startsWith(github.ref_name, 'mac') }}
|
||||
@@ -87,7 +71,6 @@ jobs:
|
||||
postgres-version: 14
|
||||
- run: |
|
||||
call "C:\Program Files\Microsoft Visual Studio\2022\Enterprise\VC\Auxiliary\Build\vcvars64.bat" && ^
|
||||
cd %TEMP% && ^
|
||||
nmake /NOLOGO /F Makefile.win && ^
|
||||
nmake /NOLOGO /F Makefile.win install && ^
|
||||
nmake /NOLOGO /F Makefile.win installcheck && ^
|
||||
@@ -114,15 +97,4 @@ jobs:
|
||||
sudo -u postgres make installcheck
|
||||
sudo -u postgres make prove_installcheck
|
||||
env:
|
||||
PG_CFLAGS: -DUSE_ASSERT_CHECKING -Wall -Wextra -Werror -Wno-unused-parameter -Wno-sign-compare
|
||||
valgrind:
|
||||
if: ${{ !startsWith(github.ref_name, 'mac') && !startsWith(github.ref_name, 'windows') }}
|
||||
runs-on: ubuntu-latest
|
||||
steps:
|
||||
- uses: actions/checkout@v4
|
||||
- uses: ankane/setup-postgres-valgrind@v1
|
||||
with:
|
||||
postgres-version: 16
|
||||
- run: make
|
||||
- run: sudo --preserve-env=PG_CONFIG make install
|
||||
- run: make installcheck
|
||||
PG_CFLAGS: -Wall -Wextra -Werror -Wno-unused-parameter -Wno-sign-compare
|
||||
|
||||
18
CHANGELOG.md
18
CHANGELOG.md
@@ -1,22 +1,6 @@
|
||||
## 0.7.0 (unreleased)
|
||||
|
||||
- Added support for bit vectors to HNSW
|
||||
- Added `hamming_distance` function
|
||||
- Added `jaccard_distance` function
|
||||
- Added `quantize_binary` function
|
||||
- Added `subvector` function
|
||||
- Added concatenate operator for vectors
|
||||
- Updated comparison operators to support vectors with different dimensions
|
||||
|
||||
## 0.6.2 (2024-03-18)
|
||||
|
||||
- Reduced lock contention with parallel HNSW index builds
|
||||
|
||||
## 0.6.1 (2024-03-04)
|
||||
## 0.6.1 (unreleased)
|
||||
|
||||
- Fixed error with `ANALYZE` and vectors with different dimensions
|
||||
- Fixed segmentation fault with `shared_preload_libraries`
|
||||
- Fixed vector subtraction being marked as commutative
|
||||
|
||||
## 0.6.0 (2024-01-29)
|
||||
|
||||
|
||||
2
LICENSE
2
LICENSE
@@ -1,4 +1,4 @@
|
||||
Portions Copyright (c) 1996-2024, PostgreSQL Global Development Group
|
||||
Portions Copyright (c) 1996-2023, PostgreSQL Global Development Group
|
||||
|
||||
Portions Copyright (c) 1994, The Regents of the University of California
|
||||
|
||||
|
||||
@@ -2,7 +2,7 @@
|
||||
"name": "vector",
|
||||
"abstract": "Open-source vector similarity search for Postgres",
|
||||
"description": "Supports L2 distance, inner product, and cosine distance",
|
||||
"version": "0.6.2",
|
||||
"version": "0.6.0",
|
||||
"maintainer": [
|
||||
"Andrew Kane <andrew@ankane.org>"
|
||||
],
|
||||
@@ -20,7 +20,7 @@
|
||||
"vector": {
|
||||
"file": "sql/vector.sql",
|
||||
"docfile": "README.md",
|
||||
"version": "0.6.2",
|
||||
"version": "0.6.0",
|
||||
"abstract": "Open-source vector similarity search for Postgres"
|
||||
}
|
||||
},
|
||||
|
||||
12
Makefile
12
Makefile
@@ -1,9 +1,9 @@
|
||||
EXTENSION = vector
|
||||
EXTVERSION = 0.6.2
|
||||
EXTVERSION = 0.6.0
|
||||
|
||||
MODULE_big = vector
|
||||
DATA = $(wildcard sql/*--*.sql)
|
||||
OBJS = src/bitvector.o src/hnsw.o src/hnswbuild.o src/hnswinsert.o src/hnswscan.o src/hnswutils.o src/hnswvacuum.o src/ivfbuild.o src/ivfflat.o src/ivfinsert.o src/ivfkmeans.o src/ivfscan.o src/ivfutils.o src/ivfvacuum.o src/vector.o
|
||||
OBJS = src/hnsw.o src/hnswbuild.o src/hnswinsert.o src/hnswscan.o src/hnswutils.o src/hnswvacuum.o src/ivfbuild.o src/ivfflat.o src/ivfinsert.o src/ivfkmeans.o src/ivfscan.o src/ivfutils.o src/ivfvacuum.o src/vector.o
|
||||
HEADERS = src/vector.h
|
||||
|
||||
TESTS = $(wildcard test/sql/*.sql)
|
||||
@@ -12,7 +12,7 @@ REGRESS_OPTS = --inputdir=test --load-extension=$(EXTENSION)
|
||||
|
||||
OPTFLAGS = -march=native
|
||||
|
||||
# Mac ARM doesn't always support -march=native
|
||||
# Mac ARM doesn't support -march=native
|
||||
ifeq ($(shell uname -s), Darwin)
|
||||
ifeq ($(shell uname -p), arm)
|
||||
# no difference with -march=armv8.5-a
|
||||
@@ -71,9 +71,11 @@ PG_MAJOR ?= 16
|
||||
.PHONY: docker
|
||||
|
||||
docker:
|
||||
docker build --pull --no-cache --build-arg PG_MAJOR=$(PG_MAJOR) -t pgvector/pgvector:pg$(PG_MAJOR) -t pgvector/pgvector:$(EXTVERSION)-pg$(PG_MAJOR) .
|
||||
docker build --pull --no-cache --build-arg PG_MAJOR=$(PG_MAJOR) -t pgvector/pgvector:pg$(PG_MAJOR) .
|
||||
docker build --build-arg PG_MAJOR=$(PG_MAJOR) -t pgvector/pgvector:$(EXTVERSION)-pg$(PG_MAJOR) .
|
||||
|
||||
.PHONY: docker-release
|
||||
|
||||
docker-release:
|
||||
docker buildx build --push --pull --no-cache --platform linux/amd64,linux/arm64 --build-arg PG_MAJOR=$(PG_MAJOR) -t pgvector/pgvector:pg$(PG_MAJOR) -t pgvector/pgvector:$(EXTVERSION)-pg$(PG_MAJOR) .
|
||||
docker buildx build --push --pull --no-cache --platform linux/amd64,linux/arm64 --build-arg PG_MAJOR=$(PG_MAJOR) -t pgvector/pgvector:pg$(PG_MAJOR) .
|
||||
docker buildx build --push --platform linux/amd64,linux/arm64 --build-arg PG_MAJOR=$(PG_MAJOR) -t pgvector/pgvector:$(EXTVERSION)-pg$(PG_MAJOR) .
|
||||
|
||||
@@ -1,10 +1,10 @@
|
||||
EXTENSION = vector
|
||||
EXTVERSION = 0.6.2
|
||||
EXTVERSION = 0.6.0
|
||||
|
||||
OBJS = src\bitvector.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\vector.obj
|
||||
OBJS = src\hnsw.obj src\hnswbuild.obj src\hnswinsert.obj src\hnswscan.obj src\hnswutils.obj src\hnswvacuum.obj src\ivfbuild.obj src\ivfflat.obj src\ivfinsert.obj src\ivfkmeans.obj src\ivfscan.obj src\ivfutils.obj src\ivfvacuum.obj src\vector.obj
|
||||
HEADERS = src\vector.h
|
||||
|
||||
REGRESS = bit_functions btree cast copy hnsw_cosine hnsw_hamming hnsw_ip hnsw_jaccard hnsw_l2 hnsw_options hnsw_unlogged ivfflat_cosine ivfflat_ip ivfflat_l2 ivfflat_options ivfflat_unlogged vector_functions vector_input
|
||||
REGRESS = btree cast copy functions input ivfflat_cosine ivfflat_ip ivfflat_l2 ivfflat_options ivfflat_unlogged
|
||||
REGRESS_OPTS = --inputdir=test --load-extension=$(EXTENSION)
|
||||
|
||||
# For /arch flags
|
||||
|
||||
225
README.md
225
README.md
@@ -20,15 +20,15 @@ Compile and install the extension (supports Postgres 12+)
|
||||
|
||||
```sh
|
||||
cd /tmp
|
||||
git clone --branch v0.6.2 https://github.com/pgvector/pgvector.git
|
||||
git clone --branch v0.6.0 https://github.com/pgvector/pgvector.git
|
||||
cd pgvector
|
||||
make
|
||||
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) if you run into issues
|
||||
|
||||
You can also install it with [Docker](#docker), [Homebrew](#homebrew), [PGXN](#pgxn), [APT](#apt), [Yum](#yum), [pkg](#pkg), or [conda-forge](#conda-forge), and it comes preinstalled with [Postgres.app](#postgresapp) and many [hosted providers](#hosted-postgres). There are also instructions for [GitHub Actions](https://github.com/pgvector/setup-pgvector).
|
||||
You can also install it with [Docker](#docker), [Homebrew](#homebrew), [PGXN](#pgxn), [APT](#apt), [Yum](#yum), or [conda-forge](#conda-forge), and it comes preinstalled with [Postgres.app](#postgresapp) and many [hosted providers](#hosted-postgres). There are also instructions for [GitHub Actions](https://github.com/pgvector/setup-pgvector).
|
||||
|
||||
### Windows
|
||||
|
||||
@@ -44,15 +44,12 @@ Then use `nmake` to build:
|
||||
|
||||
```cmd
|
||||
set "PGROOT=C:\Program Files\PostgreSQL\16"
|
||||
cd %TEMP%
|
||||
git clone --branch v0.6.2 https://github.com/pgvector/pgvector.git
|
||||
git clone --branch v0.6.0 https://github.com/pgvector/pgvector.git
|
||||
cd pgvector
|
||||
nmake /F Makefile.win
|
||||
nmake /F Makefile.win install
|
||||
```
|
||||
|
||||
See the [installation notes](#installation-notes---windows) if you run into issues
|
||||
|
||||
You can also install it with [Docker](#docker) or [conda-forge](#conda-forge).
|
||||
|
||||
## Getting Started
|
||||
@@ -105,12 +102,6 @@ Insert vectors
|
||||
INSERT INTO items (embedding) VALUES ('[1,2,3]'), ('[4,5,6]');
|
||||
```
|
||||
|
||||
Or load vectors in bulk using `COPY` ([example](https://github.com/pgvector/pgvector-python/blob/master/examples/bulk_loading.py))
|
||||
|
||||
```sql
|
||||
COPY items (embedding) FROM STDIN WITH (FORMAT BINARY);
|
||||
```
|
||||
|
||||
Upsert vectors
|
||||
|
||||
```sql
|
||||
@@ -221,19 +212,7 @@ Cosine distance
|
||||
CREATE INDEX ON items USING hnsw (embedding vector_cosine_ops);
|
||||
```
|
||||
|
||||
Hamming distance - unreleased
|
||||
|
||||
```sql
|
||||
CREATE INDEX ON items USING hnsw (embedding bit_hamming_ops);
|
||||
```
|
||||
|
||||
Jaccard distance - unreleased
|
||||
|
||||
```sql
|
||||
CREATE INDEX ON items USING hnsw (embedding bit_jaccard_ops);
|
||||
```
|
||||
|
||||
Vectors with up to 2,000 dimensions can be indexed, or bit vectors with up to 64,000 dimensions.
|
||||
Vectors with up to 2,000 dimensions can be indexed.
|
||||
|
||||
### Index Options
|
||||
|
||||
@@ -285,8 +264,6 @@ HINT: Increase maintenance_work_mem to speed up builds.
|
||||
|
||||
Note: Do not set `maintenance_work_mem` so high that it exhausts the memory on the server
|
||||
|
||||
Like other index types, it’s faster to create an index after loading your initial data
|
||||
|
||||
Starting with 0.6.0, you can also speed up index creation by increasing the number of parallel workers (2 by default)
|
||||
|
||||
```sql
|
||||
@@ -425,51 +402,13 @@ You can use [Reciprocal Rank Fusion](https://github.com/pgvector/pgvector-python
|
||||
|
||||
## Performance
|
||||
|
||||
### Tuning
|
||||
|
||||
Use a tool like [PgTune](https://pgtune.leopard.in.ua/) to set initial values for Postgres server parameters. For instance, `shared_buffers` should typically be 25% of the server’s memory. You can find the config file with:
|
||||
|
||||
```sql
|
||||
SHOW config_file;
|
||||
```
|
||||
|
||||
And check individual settings with:
|
||||
|
||||
```sql
|
||||
SHOW shared_buffers;
|
||||
```
|
||||
|
||||
Be sure to restart Postgres for changes to take effect.
|
||||
|
||||
### Loading
|
||||
|
||||
Use `COPY` for bulk loading data ([example](https://github.com/pgvector/pgvector-python/blob/master/examples/bulk_loading.py)).
|
||||
|
||||
```sql
|
||||
COPY items (embedding) FROM STDIN WITH (FORMAT BINARY);
|
||||
```
|
||||
|
||||
Add any indexes *after* loading the initial data for best performance.
|
||||
|
||||
### Indexing
|
||||
|
||||
See index build time for [HNSW](#index-build-time) and [IVFFlat](#index-build-time-1).
|
||||
|
||||
In production environments, create indexes concurrently to avoid blocking writes.
|
||||
|
||||
```sql
|
||||
CREATE INDEX CONCURRENTLY ...
|
||||
```
|
||||
|
||||
### Querying
|
||||
|
||||
Use `EXPLAIN ANALYZE` to debug performance.
|
||||
|
||||
```sql
|
||||
EXPLAIN ANALYZE SELECT * FROM items ORDER BY embedding <-> '[3,1,2]' LIMIT 5;
|
||||
```
|
||||
|
||||
#### Exact Search
|
||||
### Exact Search
|
||||
|
||||
To speed up queries without an index, increase `max_parallel_workers_per_gather`.
|
||||
|
||||
@@ -483,7 +422,7 @@ If vectors are normalized to length 1 (like [OpenAI embeddings](https://platform
|
||||
SELECT * FROM items ORDER BY embedding <#> '[3,1,2]' LIMIT 5;
|
||||
```
|
||||
|
||||
#### Approximate Search
|
||||
### Approximate Search
|
||||
|
||||
To speed up queries with an IVFFlat index, increase the number of inverted lists (at the expense of recall).
|
||||
|
||||
@@ -491,7 +430,7 @@ To speed up queries with an IVFFlat index, increase the number of inverted lists
|
||||
CREATE INDEX ON items USING ivfflat (embedding vector_l2_ops) WITH (lists = 1000);
|
||||
```
|
||||
|
||||
### Vacuuming
|
||||
## Vacuuming
|
||||
|
||||
Vacuuming can take a while for HNSW indexes. Speed it up by reindexing first.
|
||||
|
||||
@@ -500,41 +439,6 @@ REINDEX INDEX CONCURRENTLY index_name;
|
||||
VACUUM table_name;
|
||||
```
|
||||
|
||||
## Monitoring
|
||||
|
||||
Monitor performance with [pg_stat_statements](https://www.postgresql.org/docs/current/pgstatstatements.html) (be sure to add it to `shared_preload_libraries`).
|
||||
|
||||
```sql
|
||||
CREATE EXTENSION pg_stat_statements;
|
||||
```
|
||||
|
||||
Get the most time-consuming queries with:
|
||||
|
||||
```sql
|
||||
SELECT query, calls, ROUND((total_plan_time + total_exec_time) / calls) AS avg_time_ms,
|
||||
ROUND((total_plan_time + total_exec_time) / 60000) AS total_time_min
|
||||
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.
|
||||
|
||||
```sql
|
||||
BEGIN;
|
||||
SET LOCAL enable_indexscan = off; -- use exact search
|
||||
SELECT ...
|
||||
COMMIT;
|
||||
```
|
||||
|
||||
## Scaling
|
||||
|
||||
Scale pgvector the same way you scale Postgres.
|
||||
|
||||
Scale vertically by increasing memory, CPU, and storage on a single instance. Use existing tools to [tune parameters](#tuning) and [monitor performance](#monitoring).
|
||||
|
||||
Scale horizontally with [replicas](https://www.postgresql.org/docs/current/hot-standby.html), or use [Citus](https://github.com/citusdata/citus) or another approach for sharding ([example](https://github.com/pgvector/pgvector-python/blob/master/examples/citus.py)).
|
||||
|
||||
## Languages
|
||||
|
||||
Use pgvector from any language with a Postgres client. You can even generate and store vectors in one language and query them in another.
|
||||
@@ -628,18 +532,6 @@ and query with:
|
||||
SELECT * FROM items ORDER BY embedding::vector(3) <-> '[3,1,2]' LIMIT 5;
|
||||
```
|
||||
|
||||
#### Are binary vectors supported?
|
||||
|
||||
You can store binary vectors and perform exact nearest neighbor search by Hamming distance in Postgres without an extension ([example](https://github.com/pgvector/pgvector-python/blob/master/examples/hash_image_search.py)).
|
||||
|
||||
```tsql
|
||||
CREATE TABLE items (id bigserial PRIMARY KEY, embedding bit(3));
|
||||
INSERT INTO items (embedding) VALUES (B'000'), (B'111');
|
||||
SELECT * FROM items ORDER BY bit_count(embedding # B'101') LIMIT 5;
|
||||
```
|
||||
|
||||
Indexing is not currently supported.
|
||||
|
||||
#### Do indexes need to fit into memory?
|
||||
|
||||
No, but like other index types, you’ll likely see better performance if they do. You can get the size of an index with:
|
||||
@@ -652,17 +544,7 @@ SELECT pg_size_pretty(pg_relation_size('index_name'));
|
||||
|
||||
#### Why isn’t a query using an index?
|
||||
|
||||
The query needs to have an `ORDER BY` and `LIMIT`, and the `ORDER BY` must be the result of a distance operator, not an expression.
|
||||
|
||||
```sql
|
||||
-- index
|
||||
ORDER BY embedding <=> '[3,1,2]' LIMIT 5;
|
||||
|
||||
-- no index
|
||||
ORDER BY 1 - (embedding <=> '[3,1,2]') DESC LIMIT 5;
|
||||
```
|
||||
|
||||
You can encourage the planner to use an index for a query with:
|
||||
The cost estimation in pgvector < 0.4.3 does not always work well with the planner. You can encourage the planner to use an index for a query with:
|
||||
|
||||
```sql
|
||||
BEGIN;
|
||||
@@ -691,12 +573,6 @@ or choose to store vectors inline:
|
||||
ALTER TABLE items ALTER COLUMN embedding SET STORAGE PLAIN;
|
||||
```
|
||||
|
||||
#### Why are there less results for a query after adding an HNSW index?
|
||||
|
||||
Results are limited by the size of the dynamic candidate list (`hnsw.ef_search`). There may be even less results due to dead tuples or filtering conditions in the query. We recommend setting `hnsw.ef_search` to at least twice the `LIMIT` of the query. If you need more than 500 results, use an IVFFlat index instead.
|
||||
|
||||
Also, note that `NULL` vectors are not indexed (as well as zero vectors for cosine distance).
|
||||
|
||||
#### Why are there less results for a query after adding an IVFFlat index?
|
||||
|
||||
The index was likely created with too little data for the number of lists. Drop the index until the table has more data.
|
||||
@@ -705,18 +581,11 @@ The index was likely created with too little data for the number of lists. Drop
|
||||
DROP INDEX index_name;
|
||||
```
|
||||
|
||||
Results can also be limited by the number of probes (`ivfflat.probes`).
|
||||
|
||||
Also, note that `NULL` vectors are not indexed (as well as zero vectors for cosine distance).
|
||||
|
||||
## Reference
|
||||
|
||||
- [Vector](#vector-type)
|
||||
- [Bit](#bit-type)
|
||||
|
||||
### Vector Type
|
||||
|
||||
Each vector takes `4 * dimensions + 8` bytes of storage. Each element is a single-precision floating-point number (like the `real` type in Postgres), and all elements must be finite (no `NaN`, `Infinity` or `-Infinity`). Vectors can have up to 16,000 dimensions.
|
||||
Each vector takes `4 * dimensions + 8` bytes of storage. Each element is a single precision floating-point number (like the `real` type in Postgres), and all elements must be finite (no `NaN`, `Infinity` or `-Infinity`). Vectors can have up to 16,000 dimensions.
|
||||
|
||||
### Vector Operators
|
||||
|
||||
@@ -725,7 +594,6 @@ Operator | Description | Added
|
||||
\+ | element-wise addition |
|
||||
\- | element-wise subtraction |
|
||||
\* | element-wise multiplication | 0.5.0
|
||||
\|\| | concatenate | unreleased
|
||||
<-> | Euclidean distance |
|
||||
<#> | negative inner product |
|
||||
<=> | cosine distance |
|
||||
@@ -738,37 +606,17 @@ cosine_distance(vector, vector) → double precision | cosine distance |
|
||||
inner_product(vector, vector) → double precision | inner product |
|
||||
l2_distance(vector, vector) → double precision | Euclidean distance |
|
||||
l1_distance(vector, vector) → double precision | taxicab distance | 0.5.0
|
||||
quantize_binary(vector) → bit | quantize | unreleased
|
||||
subvector(vector, integer, integer) → vector | subvector | unreleased
|
||||
vector_dims(vector) → integer | number of dimensions |
|
||||
vector_norm(vector) → double precision | Euclidean norm |
|
||||
|
||||
### Vector Aggregate Functions
|
||||
### Aggregate Functions
|
||||
|
||||
Function | Description | Added
|
||||
--- | --- | ---
|
||||
avg(vector) → vector | average |
|
||||
sum(vector) → vector | sum | 0.5.0
|
||||
|
||||
### Bit Type
|
||||
|
||||
Each bit vector takes `dimensions / 8 + (5 or 8)` bytes of storage. See the [Postgres docs](https://www.postgresql.org/docs/current/datatype-bit.html) for more info.
|
||||
|
||||
### Bit Operators
|
||||
|
||||
Operator | Description | Added
|
||||
--- | --- | ---
|
||||
<~> | Hamming distance | unreleased
|
||||
<%> | Jaccard distance | unreleased
|
||||
|
||||
### Bit Functions
|
||||
|
||||
Function | Description | Added
|
||||
--- | --- | ---
|
||||
hamming_distance(bit, bit) → double precision | Hamming distance | unreleased
|
||||
jaccard_distance(bit, bit) → double precision | Jaccard distance | unreleased
|
||||
|
||||
## Installation Notes - Linux and Mac
|
||||
## Installation Notes
|
||||
|
||||
### Postgres Location
|
||||
|
||||
@@ -808,26 +656,6 @@ Note: Replace `16` with your Postgres server version
|
||||
|
||||
If compilation fails and the output includes `warning: no such sysroot directory` on Mac, reinstall Xcode Command Line Tools.
|
||||
|
||||
### Portability
|
||||
|
||||
By default, pgvector compiles with `-march=native` on some platforms for best performance. However, this can lead to `Illegal instruction` errors if trying to run the compiled extension on a different machine.
|
||||
|
||||
To compile for portability, use:
|
||||
|
||||
```sh
|
||||
make OPTFLAGS=""
|
||||
```
|
||||
|
||||
## Installation Notes - Windows
|
||||
|
||||
### Missing Header
|
||||
|
||||
If compilation fails with `Cannot open include file: 'postgres.h': No such file or directory`, make sure `PGROOT` is correct.
|
||||
|
||||
### Permissions
|
||||
|
||||
If installation fails with `Access is denied`, re-run the installation instructions as an administrator.
|
||||
|
||||
## Additional Installation Methods
|
||||
|
||||
### Docker
|
||||
@@ -843,7 +671,7 @@ This adds pgvector to the [Postgres image](https://hub.docker.com/_/postgres) (r
|
||||
You can also build the image manually:
|
||||
|
||||
```sh
|
||||
git clone --branch v0.6.2 https://github.com/pgvector/pgvector.git
|
||||
git clone --branch v0.6.0 https://github.com/pgvector/pgvector.git
|
||||
cd pgvector
|
||||
docker build --build-arg PG_MAJOR=16 -t myuser/pgvector .
|
||||
```
|
||||
@@ -888,21 +716,6 @@ sudo dnf install pgvector_16
|
||||
|
||||
Note: Replace `16` with your Postgres server version
|
||||
|
||||
### pkg
|
||||
|
||||
Install the FreeBSD package with:
|
||||
|
||||
```sh
|
||||
pkg install postgresql15-pg_vector
|
||||
```
|
||||
|
||||
or the port with:
|
||||
|
||||
```sh
|
||||
cd /usr/ports/databases/pgvector
|
||||
make install
|
||||
```
|
||||
|
||||
### conda-forge
|
||||
|
||||
With Conda Postgres, install from [conda-forge](https://anaconda.org/conda-forge/pgvector) with:
|
||||
@@ -1012,12 +825,6 @@ make installcheck REGRESS=functions # regression test
|
||||
make prove_installcheck PROVE_TESTS=test/t/001_ivfflat_wal.pl # TAP test
|
||||
```
|
||||
|
||||
To enable assertions:
|
||||
|
||||
```sh
|
||||
make clean && PG_CFLAGS="-DUSE_ASSERT_CHECKING" make && make install
|
||||
```
|
||||
|
||||
To enable benchmarking:
|
||||
|
||||
```sh
|
||||
@@ -1030,6 +837,12 @@ To show memory usage:
|
||||
make clean && PG_CFLAGS="-DHNSW_MEMORY -DIVFFLAT_MEMORY" make && make install
|
||||
```
|
||||
|
||||
To enable assertions:
|
||||
|
||||
```sh
|
||||
make clean && PG_CFLAGS="-DUSE_ASSERT_CHECKING" make && make install
|
||||
```
|
||||
|
||||
To get k-means metrics:
|
||||
|
||||
```sh
|
||||
|
||||
@@ -1,16 +0,0 @@
|
||||
-- complain if script is sourced in psql, rather than via CREATE EXTENSION
|
||||
\echo Use "ALTER EXTENSION vector UPDATE TO '0.6.1'" to load this file. \quit
|
||||
|
||||
DROP OPERATOR - (vector, vector);
|
||||
|
||||
CREATE OPERATOR - (
|
||||
LEFTARG = vector, RIGHTARG = vector, PROCEDURE = vector_sub
|
||||
);
|
||||
|
||||
ALTER OPERATOR <= (vector, vector) SET (
|
||||
RESTRICT = scalarlesel, JOIN = scalarlejoinsel
|
||||
);
|
||||
|
||||
ALTER OPERATOR >= (vector, vector) SET (
|
||||
RESTRICT = scalargesel, JOIN = scalargejoinsel
|
||||
);
|
||||
@@ -1,2 +0,0 @@
|
||||
-- complain if script is sourced in psql, rather than via CREATE EXTENSION
|
||||
\echo Use "ALTER EXTENSION vector UPDATE TO '0.6.2'" to load this file. \quit
|
||||
@@ -1,41 +0,0 @@
|
||||
-- complain if script is sourced in psql, rather than via CREATE EXTENSION
|
||||
\echo Use "ALTER EXTENSION vector UPDATE TO '0.7.0'" to load this file. \quit
|
||||
|
||||
CREATE FUNCTION quantize_binary(vector) RETURNS bit
|
||||
AS 'MODULE_PATHNAME' LANGUAGE C IMMUTABLE STRICT PARALLEL SAFE;
|
||||
|
||||
CREATE FUNCTION subvector(vector, int, int) RETURNS vector
|
||||
AS 'MODULE_PATHNAME' LANGUAGE C IMMUTABLE STRICT PARALLEL SAFE;
|
||||
|
||||
CREATE FUNCTION vector_concat(vector, vector) RETURNS vector
|
||||
AS 'MODULE_PATHNAME' LANGUAGE C IMMUTABLE STRICT PARALLEL SAFE;
|
||||
|
||||
CREATE OPERATOR || (
|
||||
LEFTARG = vector, RIGHTARG = vector, PROCEDURE = vector_concat
|
||||
);
|
||||
|
||||
CREATE FUNCTION hamming_distance(bit, bit) RETURNS float8
|
||||
AS 'MODULE_PATHNAME' LANGUAGE C IMMUTABLE STRICT PARALLEL SAFE;
|
||||
|
||||
CREATE FUNCTION jaccard_distance(bit, bit) RETURNS float8
|
||||
AS 'MODULE_PATHNAME' LANGUAGE C IMMUTABLE STRICT PARALLEL SAFE;
|
||||
|
||||
CREATE OPERATOR <~> (
|
||||
LEFTARG = bit, RIGHTARG = bit, PROCEDURE = hamming_distance,
|
||||
COMMUTATOR = '<~>'
|
||||
);
|
||||
|
||||
CREATE OPERATOR <%> (
|
||||
LEFTARG = bit, RIGHTARG = bit, PROCEDURE = jaccard_distance,
|
||||
COMMUTATOR = '<%>'
|
||||
);
|
||||
|
||||
CREATE OPERATOR CLASS bit_hamming_ops
|
||||
FOR TYPE bit USING hnsw AS
|
||||
OPERATOR 1 <~> (bit, bit) FOR ORDER BY float_ops,
|
||||
FUNCTION 1 hamming_distance(bit, bit);
|
||||
|
||||
CREATE OPERATOR CLASS bit_jaccard_ops
|
||||
FOR TYPE bit USING hnsw AS
|
||||
OPERATOR 1 <%> (bit, bit) FOR ORDER BY float_ops,
|
||||
FUNCTION 1 jaccard_distance(bit, bit);
|
||||
@@ -1,7 +1,7 @@
|
||||
-- complain if script is sourced in psql, rather than via CREATE EXTENSION
|
||||
\echo Use "CREATE EXTENSION vector" to load this file. \quit
|
||||
|
||||
-- vector type
|
||||
-- type
|
||||
|
||||
CREATE TYPE vector;
|
||||
|
||||
@@ -29,7 +29,7 @@ CREATE TYPE vector (
|
||||
STORAGE = external
|
||||
);
|
||||
|
||||
-- vector functions
|
||||
-- functions
|
||||
|
||||
CREATE FUNCTION l2_distance(vector, vector) RETURNS float8
|
||||
AS 'MODULE_PATHNAME' LANGUAGE C IMMUTABLE STRICT PARALLEL SAFE;
|
||||
@@ -58,13 +58,7 @@ CREATE FUNCTION vector_sub(vector, vector) RETURNS vector
|
||||
CREATE FUNCTION vector_mul(vector, vector) RETURNS vector
|
||||
AS 'MODULE_PATHNAME' LANGUAGE C IMMUTABLE STRICT PARALLEL SAFE;
|
||||
|
||||
CREATE FUNCTION quantize_binary(vector) RETURNS bit
|
||||
AS 'MODULE_PATHNAME' LANGUAGE C IMMUTABLE STRICT PARALLEL SAFE;
|
||||
|
||||
CREATE FUNCTION subvector(vector, int, int) RETURNS vector
|
||||
AS 'MODULE_PATHNAME' LANGUAGE C IMMUTABLE STRICT PARALLEL SAFE;
|
||||
|
||||
-- vector private functions
|
||||
-- private functions
|
||||
|
||||
CREATE FUNCTION vector_lt(vector, vector) RETURNS bool
|
||||
AS 'MODULE_PATHNAME' LANGUAGE C IMMUTABLE STRICT PARALLEL SAFE;
|
||||
@@ -105,10 +99,7 @@ CREATE FUNCTION vector_avg(double precision[]) RETURNS vector
|
||||
CREATE FUNCTION vector_combine(double precision[], double precision[]) RETURNS double precision[]
|
||||
AS 'MODULE_PATHNAME' LANGUAGE C IMMUTABLE STRICT PARALLEL SAFE;
|
||||
|
||||
CREATE FUNCTION vector_concat(vector, vector) RETURNS vector
|
||||
AS 'MODULE_PATHNAME' LANGUAGE C IMMUTABLE STRICT PARALLEL SAFE;
|
||||
|
||||
-- vector aggregates
|
||||
-- aggregates
|
||||
|
||||
CREATE AGGREGATE avg(vector) (
|
||||
SFUNC = vector_accum,
|
||||
@@ -126,7 +117,7 @@ CREATE AGGREGATE sum(vector) (
|
||||
PARALLEL = SAFE
|
||||
);
|
||||
|
||||
-- vector cast functions
|
||||
-- cast functions
|
||||
|
||||
CREATE FUNCTION vector(vector, integer, boolean) RETURNS vector
|
||||
AS 'MODULE_PATHNAME' LANGUAGE C IMMUTABLE STRICT PARALLEL SAFE;
|
||||
@@ -146,7 +137,7 @@ CREATE FUNCTION array_to_vector(numeric[], integer, boolean) RETURNS vector
|
||||
CREATE FUNCTION vector_to_float4(vector, integer, boolean) RETURNS real[]
|
||||
AS 'MODULE_PATHNAME' LANGUAGE C IMMUTABLE STRICT PARALLEL SAFE;
|
||||
|
||||
-- vector casts
|
||||
-- casts
|
||||
|
||||
CREATE CAST (vector AS vector)
|
||||
WITH FUNCTION vector(vector, integer, boolean) AS IMPLICIT;
|
||||
@@ -166,7 +157,7 @@ CREATE CAST (double precision[] AS vector)
|
||||
CREATE CAST (numeric[] AS vector)
|
||||
WITH FUNCTION array_to_vector(numeric[], integer, boolean) AS ASSIGNMENT;
|
||||
|
||||
-- vector operators
|
||||
-- operators
|
||||
|
||||
CREATE OPERATOR <-> (
|
||||
LEFTARG = vector, RIGHTARG = vector, PROCEDURE = l2_distance,
|
||||
@@ -189,7 +180,8 @@ CREATE OPERATOR + (
|
||||
);
|
||||
|
||||
CREATE OPERATOR - (
|
||||
LEFTARG = vector, RIGHTARG = vector, PROCEDURE = vector_sub
|
||||
LEFTARG = vector, RIGHTARG = vector, PROCEDURE = vector_sub,
|
||||
COMMUTATOR = -
|
||||
);
|
||||
|
||||
CREATE OPERATOR * (
|
||||
@@ -197,20 +189,17 @@ CREATE OPERATOR * (
|
||||
COMMUTATOR = *
|
||||
);
|
||||
|
||||
CREATE OPERATOR || (
|
||||
LEFTARG = vector, RIGHTARG = vector, PROCEDURE = vector_concat
|
||||
);
|
||||
|
||||
CREATE OPERATOR < (
|
||||
LEFTARG = vector, RIGHTARG = vector, PROCEDURE = vector_lt,
|
||||
COMMUTATOR = > , NEGATOR = >= ,
|
||||
RESTRICT = scalarltsel, JOIN = scalarltjoinsel
|
||||
);
|
||||
|
||||
-- should use scalarlesel and scalarlejoinsel, but not supported in Postgres < 11
|
||||
CREATE OPERATOR <= (
|
||||
LEFTARG = vector, RIGHTARG = vector, PROCEDURE = vector_le,
|
||||
COMMUTATOR = >= , NEGATOR = > ,
|
||||
RESTRICT = scalarlesel, JOIN = scalarlejoinsel
|
||||
RESTRICT = scalarltsel, JOIN = scalarltjoinsel
|
||||
);
|
||||
|
||||
CREATE OPERATOR = (
|
||||
@@ -225,10 +214,11 @@ CREATE OPERATOR <> (
|
||||
RESTRICT = eqsel, JOIN = eqjoinsel
|
||||
);
|
||||
|
||||
-- should use scalargesel and scalargejoinsel, but not supported in Postgres < 11
|
||||
CREATE OPERATOR >= (
|
||||
LEFTARG = vector, RIGHTARG = vector, PROCEDURE = vector_ge,
|
||||
COMMUTATOR = <= , NEGATOR = < ,
|
||||
RESTRICT = scalargesel, JOIN = scalargejoinsel
|
||||
RESTRICT = scalargtsel, JOIN = scalargtjoinsel
|
||||
);
|
||||
|
||||
CREATE OPERATOR > (
|
||||
@@ -253,7 +243,7 @@ CREATE ACCESS METHOD hnsw TYPE INDEX HANDLER hnswhandler;
|
||||
|
||||
COMMENT ON ACCESS METHOD hnsw IS 'hnsw index access method';
|
||||
|
||||
-- vector opclasses
|
||||
-- opclasses
|
||||
|
||||
CREATE OPERATOR CLASS vector_ops
|
||||
DEFAULT FOR TYPE vector USING btree AS
|
||||
@@ -300,31 +290,3 @@ CREATE OPERATOR CLASS vector_cosine_ops
|
||||
OPERATOR 1 <=> (vector, vector) FOR ORDER BY float_ops,
|
||||
FUNCTION 1 vector_negative_inner_product(vector, vector),
|
||||
FUNCTION 2 vector_norm(vector);
|
||||
|
||||
-- bit functions
|
||||
|
||||
CREATE FUNCTION hamming_distance(bit, bit) RETURNS float8
|
||||
AS 'MODULE_PATHNAME' LANGUAGE C IMMUTABLE STRICT PARALLEL SAFE;
|
||||
|
||||
CREATE FUNCTION jaccard_distance(bit, bit) RETURNS float8
|
||||
AS 'MODULE_PATHNAME' LANGUAGE C IMMUTABLE STRICT PARALLEL SAFE;
|
||||
|
||||
CREATE OPERATOR <~> (
|
||||
LEFTARG = bit, RIGHTARG = bit, PROCEDURE = hamming_distance,
|
||||
COMMUTATOR = '<~>'
|
||||
);
|
||||
|
||||
CREATE OPERATOR <%> (
|
||||
LEFTARG = bit, RIGHTARG = bit, PROCEDURE = jaccard_distance,
|
||||
COMMUTATOR = '<%>'
|
||||
);
|
||||
|
||||
CREATE OPERATOR CLASS bit_hamming_ops
|
||||
FOR TYPE bit USING hnsw AS
|
||||
OPERATOR 1 <~> (bit, bit) FOR ORDER BY float_ops,
|
||||
FUNCTION 1 hamming_distance(bit, bit);
|
||||
|
||||
CREATE OPERATOR CLASS bit_jaccard_ops
|
||||
FOR TYPE bit USING hnsw AS
|
||||
OPERATOR 1 <%> (bit, bit) FOR ORDER BY float_ops,
|
||||
FUNCTION 1 jaccard_distance(bit, bit);
|
||||
|
||||
@@ -1,90 +0,0 @@
|
||||
#include "postgres.h"
|
||||
|
||||
#include "bitvector.h"
|
||||
#include "port/pg_bitutils.h"
|
||||
#include "utils/varbit.h"
|
||||
|
||||
#if PG_VERSION_NUM >= 160000
|
||||
#include "varatt.h"
|
||||
#endif
|
||||
|
||||
/*
|
||||
* Allocate and initialize a new bit vector
|
||||
*/
|
||||
VarBit *
|
||||
InitBitVector(int dim)
|
||||
{
|
||||
VarBit *result;
|
||||
int size;
|
||||
|
||||
size = VARBITTOTALLEN(dim);
|
||||
result = (VarBit *) palloc0(size);
|
||||
SET_VARSIZE(result, size);
|
||||
VARBITLEN(result) = dim;
|
||||
|
||||
return result;
|
||||
}
|
||||
|
||||
/*
|
||||
* Ensure same dimensions
|
||||
*/
|
||||
static inline void
|
||||
CheckDims(VarBit *a, VarBit *b)
|
||||
{
|
||||
if (VARBITLEN(a) != VARBITLEN(b))
|
||||
ereport(ERROR,
|
||||
(errcode(ERRCODE_DATA_EXCEPTION),
|
||||
errmsg("different bit lengths %u and %u", VARBITLEN(a), VARBITLEN(b))));
|
||||
}
|
||||
|
||||
/*
|
||||
* Get the Hamming distance between two bit vectors
|
||||
*/
|
||||
PGDLLEXPORT PG_FUNCTION_INFO_V1(hamming_distance);
|
||||
Datum
|
||||
hamming_distance(PG_FUNCTION_ARGS)
|
||||
{
|
||||
VarBit *a = PG_GETARG_VARBIT_P(0);
|
||||
VarBit *b = PG_GETARG_VARBIT_P(1);
|
||||
unsigned char *ax = VARBITS(a);
|
||||
unsigned char *bx = VARBITS(b);
|
||||
uint64 distance = 0;
|
||||
|
||||
CheckDims(a, b);
|
||||
|
||||
/* TODO Improve performance */
|
||||
for (uint32 i = 0; i < VARBITBYTES(a); i++)
|
||||
distance += pg_number_of_ones[ax[i] ^ bx[i]];
|
||||
|
||||
PG_RETURN_FLOAT8((double) distance);
|
||||
}
|
||||
|
||||
/*
|
||||
* Get the Jaccard distance between two bit vectors
|
||||
*/
|
||||
PGDLLEXPORT PG_FUNCTION_INFO_V1(jaccard_distance);
|
||||
Datum
|
||||
jaccard_distance(PG_FUNCTION_ARGS)
|
||||
{
|
||||
VarBit *a = PG_GETARG_VARBIT_P(0);
|
||||
VarBit *b = PG_GETARG_VARBIT_P(1);
|
||||
unsigned char *ax = VARBITS(a);
|
||||
unsigned char *bx = VARBITS(b);
|
||||
uint64 ab = 0;
|
||||
uint64 aa;
|
||||
uint64 bb;
|
||||
|
||||
CheckDims(a, b);
|
||||
|
||||
/* TODO Improve performance */
|
||||
for (uint32 i = 0; i < VARBITBYTES(a); i++)
|
||||
ab += pg_number_of_ones[ax[i] & bx[i]];
|
||||
|
||||
if (ab == 0)
|
||||
PG_RETURN_FLOAT8(1);
|
||||
|
||||
aa = pg_popcount((char *) ax, VARBITBYTES(a));
|
||||
bb = pg_popcount((char *) bx, VARBITBYTES(b));
|
||||
|
||||
PG_RETURN_FLOAT8(1 - (ab / ((double) (aa + bb - ab))));
|
||||
}
|
||||
@@ -1,8 +0,0 @@
|
||||
#ifndef BITVECTOR_H
|
||||
#define BITVECTOR_H
|
||||
|
||||
#include "utils/varbit.h"
|
||||
|
||||
VarBit *InitBitVector(int dim);
|
||||
|
||||
#endif
|
||||
@@ -8,7 +8,6 @@
|
||||
#include "commands/progress.h"
|
||||
#include "commands/vacuum.h"
|
||||
#include "hnsw.h"
|
||||
#include "miscadmin.h"
|
||||
#include "utils/guc.h"
|
||||
#include "utils/selfuncs.h"
|
||||
|
||||
@@ -29,7 +28,7 @@ static relopt_kind hnsw_relopt_kind;
|
||||
* this grows bigger, we should use a shmem_request_hook and
|
||||
* RequestAddinShmemSpace() to pre-reserve space for this.
|
||||
*/
|
||||
void
|
||||
static void
|
||||
HnswInitLockTranche(void)
|
||||
{
|
||||
int *tranche_ids;
|
||||
@@ -54,8 +53,7 @@ HnswInitLockTranche(void)
|
||||
void
|
||||
HnswInit(void)
|
||||
{
|
||||
if (!process_shared_preload_libraries_in_progress)
|
||||
HnswInitLockTranche();
|
||||
HnswInitLockTranche();
|
||||
|
||||
hnsw_relopt_kind = add_reloption_kind();
|
||||
add_int_reloption(hnsw_relopt_kind, "m", "Max number of connections",
|
||||
|
||||
23
src/hnsw.h
23
src/hnsw.h
@@ -55,12 +55,6 @@
|
||||
#define HNSW_UPDATE_ENTRY_GREATER 1
|
||||
#define HNSW_UPDATE_ENTRY_ALWAYS 2
|
||||
|
||||
typedef enum HnswType
|
||||
{
|
||||
HNSW_TYPE_VECTOR,
|
||||
HNSW_TYPE_BIT
|
||||
} HnswType;
|
||||
|
||||
/* Build phases */
|
||||
/* PROGRESS_CREATEIDX_SUBPHASE_INITIALIZE is 1 */
|
||||
#define PROGRESS_HNSW_PHASE_LOAD 2
|
||||
@@ -86,7 +80,7 @@ typedef enum HnswType
|
||||
|
||||
#if PG_VERSION_NUM < 130000
|
||||
#define list_delete_last(list) list_truncate(list, list_length(list) - 1)
|
||||
#define list_sort(list, cmp) ((list) = list_qsort(list, cmp))
|
||||
#define list_sort(list, cmp) list_qsort(list, cmp)
|
||||
#endif
|
||||
|
||||
#define HnswIsElementTuple(tup) ((tup)->type == HNSW_ELEMENT_TUPLE_TYPE)
|
||||
@@ -135,7 +129,7 @@ HnswPtrDeclare(HnswNeighborArray, HnswNeighborArrayRelptr, HnswNeighborArrayPtr)
|
||||
HnswPtrDeclare(HnswNeighborArrayPtr, HnswNeighborsRelptr, HnswNeighborsPtr);
|
||||
HnswPtrDeclare(char, DatumRelptr, DatumPtr);
|
||||
|
||||
struct HnswElementData
|
||||
typedef struct HnswElementData
|
||||
{
|
||||
HnswElementPtr next;
|
||||
ItemPointerData heaptids[HNSW_HEAPTIDS];
|
||||
@@ -150,7 +144,7 @@ struct HnswElementData
|
||||
BlockNumber neighborPage;
|
||||
DatumPtr value;
|
||||
LWLock lock;
|
||||
};
|
||||
} HnswElementData;
|
||||
|
||||
typedef HnswElementData * HnswElement;
|
||||
|
||||
@@ -161,12 +155,12 @@ typedef struct HnswCandidate
|
||||
bool closer;
|
||||
} HnswCandidate;
|
||||
|
||||
struct HnswNeighborArray
|
||||
typedef struct HnswNeighborArray
|
||||
{
|
||||
int length;
|
||||
bool closerSet;
|
||||
HnswCandidate items[FLEXIBLE_ARRAY_MEMBER];
|
||||
};
|
||||
} HnswNeighborArray;
|
||||
|
||||
typedef struct HnswPairingHeapNode
|
||||
{
|
||||
@@ -191,7 +185,6 @@ typedef struct HnswGraph
|
||||
|
||||
/* Entry state */
|
||||
LWLock entryLock;
|
||||
LWLock entryWaitLock;
|
||||
HnswElementPtr entryPoint;
|
||||
|
||||
/* Allocations state */
|
||||
@@ -248,7 +241,6 @@ typedef struct HnswBuildState
|
||||
Relation index;
|
||||
IndexInfo *indexInfo;
|
||||
ForkNumber forkNum;
|
||||
HnswType type;
|
||||
|
||||
/* Settings */
|
||||
int dimensions;
|
||||
@@ -269,6 +261,7 @@ typedef struct HnswBuildState
|
||||
HnswGraph *graph;
|
||||
double ml;
|
||||
int maxLevel;
|
||||
Vector *normvec;
|
||||
|
||||
/* Memory */
|
||||
MemoryContext graphCtx;
|
||||
@@ -373,8 +366,7 @@ typedef struct HnswVacuumState
|
||||
int HnswGetM(Relation index);
|
||||
int HnswGetEfConstruction(Relation index);
|
||||
FmgrInfo *HnswOptionalProcInfo(Relation index, uint16 procnum);
|
||||
HnswType HnswGetType(Relation index);
|
||||
bool HnswNormValue(FmgrInfo *procinfo, Oid collation, Datum *value, HnswType type);
|
||||
bool HnswNormValue(FmgrInfo *procinfo, Oid collation, Datum *value, Vector * result);
|
||||
Buffer HnswNewBuffer(Relation index, ForkNumber forkNum);
|
||||
void HnswInitPage(Buffer buf, Page page);
|
||||
void HnswInit(void);
|
||||
@@ -397,7 +389,6 @@ void HnswLoadElement(HnswElement element, float *distance, Datum *q, Relation i
|
||||
void HnswSetElementTuple(char *base, HnswElementTuple etup, HnswElement element);
|
||||
void HnswUpdateConnection(char *base, HnswElement element, HnswCandidate * hc, int lm, int lc, int *updateIdx, Relation index, FmgrInfo *procinfo, Oid collation);
|
||||
void HnswLoadNeighbors(HnswElement element, Relation index, int m);
|
||||
void HnswInitLockTranche(void);
|
||||
PGDLLEXPORT void HnswParallelBuildMain(dsm_segment *seg, shm_toc *toc);
|
||||
|
||||
/* Index access methods */
|
||||
|
||||
@@ -44,7 +44,6 @@
|
||||
#include "access/xact.h"
|
||||
#include "access/xloginsert.h"
|
||||
#include "catalog/index.h"
|
||||
#include "catalog/pg_type_d.h"
|
||||
#include "commands/progress.h"
|
||||
#include "hnsw.h"
|
||||
#include "miscadmin.h"
|
||||
@@ -177,7 +176,7 @@ CreateGraphPages(HnswBuildState * buildstate)
|
||||
Size etupSize;
|
||||
Size ntupSize;
|
||||
Size combinedSize;
|
||||
Pointer valuePtr = HnswPtrAccess(base, element->value);
|
||||
void *valuePtr = HnswPtrAccess(base, element->value);
|
||||
|
||||
/* Update iterator */
|
||||
iter = element->next;
|
||||
@@ -432,15 +431,10 @@ InsertTupleInMemory(HnswBuildState * buildstate, HnswElement element)
|
||||
HnswGraph *graph = buildstate->graph;
|
||||
HnswElement entryPoint;
|
||||
LWLock *entryLock = &graph->entryLock;
|
||||
LWLock *entryWaitLock = &graph->entryWaitLock;
|
||||
int efConstruction = buildstate->efConstruction;
|
||||
int m = buildstate->m;
|
||||
char *base = buildstate->hnswarea;
|
||||
|
||||
/* Wait if another process needs exclusive lock on entry lock */
|
||||
LWLockAcquire(entryWaitLock, LW_EXCLUSIVE);
|
||||
LWLockRelease(entryWaitLock);
|
||||
|
||||
/* Get entry point */
|
||||
LWLockAcquire(entryLock, LW_SHARED);
|
||||
entryPoint = HnswPtrAccess(base, graph->entryPoint);
|
||||
@@ -451,10 +445,8 @@ InsertTupleInMemory(HnswBuildState * buildstate, HnswElement element)
|
||||
/* Release shared lock */
|
||||
LWLockRelease(entryLock);
|
||||
|
||||
/* Tell other processes to wait and get exclusive lock */
|
||||
LWLockAcquire(entryWaitLock, LW_EXCLUSIVE);
|
||||
/* Get exclusive lock */
|
||||
LWLockAcquire(entryLock, LW_EXCLUSIVE);
|
||||
LWLockRelease(entryWaitLock);
|
||||
|
||||
/* Get latest entry point after lock is acquired */
|
||||
entryPoint = HnswPtrAccess(base, graph->entryPoint);
|
||||
@@ -490,7 +482,7 @@ InsertTuple(Relation index, Datum *values, bool *isnull, ItemPointer heaptid, Hn
|
||||
/* Normalize if needed */
|
||||
if (buildstate->normprocinfo != NULL)
|
||||
{
|
||||
if (!HnswNormValue(buildstate->normprocinfo, buildstate->collation, &value, buildstate->type))
|
||||
if (!HnswNormValue(buildstate->normprocinfo, buildstate->collation, &value, buildstate->normvec))
|
||||
return false;
|
||||
}
|
||||
|
||||
@@ -609,9 +601,6 @@ BuildCallback(Relation index, CALLBACK_ITEM_POINTER, Datum *values,
|
||||
static void
|
||||
InitGraph(HnswGraph * graph, char *base, long memoryTotal)
|
||||
{
|
||||
/* Initialize the lock tranche if needed */
|
||||
HnswInitLockTranche();
|
||||
|
||||
HnswPtrStore(base, graph->head, (HnswElement) NULL);
|
||||
HnswPtrStore(base, graph->entryPoint, (HnswElement) NULL);
|
||||
graph->memoryUsed = 0;
|
||||
@@ -620,7 +609,6 @@ InitGraph(HnswGraph * graph, char *base, long memoryTotal)
|
||||
graph->indtuples = 0;
|
||||
SpinLockInit(&graph->lock);
|
||||
LWLockInitialize(&graph->entryLock, hnsw_lock_tranche_id);
|
||||
LWLockInitialize(&graph->entryWaitLock, hnsw_lock_tranche_id);
|
||||
LWLockInitialize(&graph->allocatorLock, hnsw_lock_tranche_id);
|
||||
LWLockInitialize(&graph->flushLock, hnsw_lock_tranche_id);
|
||||
}
|
||||
@@ -666,46 +654,23 @@ HnswSharedMemoryAlloc(Size size, void *state)
|
||||
return chunk;
|
||||
}
|
||||
|
||||
/*
|
||||
* Get max dimensions
|
||||
*/
|
||||
static int
|
||||
GetMaxDimensions(HnswType type)
|
||||
{
|
||||
int maxDimensions = HNSW_MAX_DIM;
|
||||
|
||||
if (type == HNSW_TYPE_BIT)
|
||||
maxDimensions *= 32;
|
||||
|
||||
return maxDimensions;
|
||||
}
|
||||
|
||||
/*
|
||||
* Initialize the build state
|
||||
*/
|
||||
static void
|
||||
InitBuildState(HnswBuildState * buildstate, Relation heap, Relation index, IndexInfo *indexInfo, ForkNumber forkNum)
|
||||
{
|
||||
int maxDimensions;
|
||||
|
||||
buildstate->heap = heap;
|
||||
buildstate->index = index;
|
||||
buildstate->indexInfo = indexInfo;
|
||||
buildstate->forkNum = forkNum;
|
||||
buildstate->type = HnswGetType(index);
|
||||
|
||||
buildstate->m = HnswGetM(index);
|
||||
buildstate->efConstruction = HnswGetEfConstruction(index);
|
||||
buildstate->dimensions = TupleDescAttr(index->rd_att, 0)->atttypmod;
|
||||
|
||||
maxDimensions = GetMaxDimensions(buildstate->type);
|
||||
|
||||
/* Require column to have dimensions to be indexed */
|
||||
if (buildstate->dimensions < 0)
|
||||
elog(ERROR, "column does not have dimensions");
|
||||
|
||||
if (buildstate->dimensions > maxDimensions)
|
||||
elog(ERROR, "column cannot have more than %d dimensions for hnsw index", maxDimensions);
|
||||
if (buildstate->dimensions > HNSW_MAX_DIM)
|
||||
elog(ERROR, "column cannot have more than %d dimensions for hnsw index", HNSW_MAX_DIM);
|
||||
|
||||
if (buildstate->efConstruction < 2 * buildstate->m)
|
||||
elog(ERROR, "ef_construction must be greater than or equal to 2 * m");
|
||||
@@ -723,6 +688,9 @@ InitBuildState(HnswBuildState * buildstate, Relation heap, Relation index, Index
|
||||
buildstate->ml = HnswGetMl(buildstate->m);
|
||||
buildstate->maxLevel = HnswGetMaxLevel(buildstate->m);
|
||||
|
||||
/* Reuse for each tuple */
|
||||
buildstate->normvec = InitVector(buildstate->dimensions);
|
||||
|
||||
buildstate->graphCtx = GenerationContextCreate(CurrentMemoryContext,
|
||||
"Hnsw build graph context",
|
||||
#if PG_VERSION_NUM >= 150000
|
||||
@@ -746,6 +714,7 @@ InitBuildState(HnswBuildState * buildstate, Relation heap, Relation index, Index
|
||||
static void
|
||||
FreeBuildState(HnswBuildState * buildstate)
|
||||
{
|
||||
pfree(buildstate->normvec);
|
||||
MemoryContextDelete(buildstate->graphCtx);
|
||||
MemoryContextDelete(buildstate->tmpCtx);
|
||||
}
|
||||
@@ -1001,14 +970,6 @@ HnswBeginParallel(HnswBuildState * buildstate, bool isconcurrent, int request)
|
||||
/* Report less than allocated so never fails */
|
||||
InitGraph(&hnswshared->graphData, hnswarea, esthnswarea - 1024 * 1024);
|
||||
|
||||
/*
|
||||
* Avoid base address for relptr for Postgres < 14.5
|
||||
* https://github.com/postgres/postgres/commit/7201cd18627afc64850537806da7f22150d1a83b
|
||||
*/
|
||||
#if PG_VERSION_NUM < 140005
|
||||
hnswshared->graphData.memoryUsed += MAXALIGN(1);
|
||||
#endif
|
||||
|
||||
shm_toc_insert(pcxt->toc, PARALLEL_KEY_HNSW_SHARED, hnswshared);
|
||||
shm_toc_insert(pcxt->toc, PARALLEL_KEY_HNSW_AREA, hnswarea);
|
||||
|
||||
|
||||
@@ -622,7 +622,7 @@ HnswInsertTuple(Relation index, Datum *values, bool *isnull, ItemPointer heap_ti
|
||||
normprocinfo = HnswOptionalProcInfo(index, HNSW_NORM_PROC);
|
||||
if (normprocinfo != NULL)
|
||||
{
|
||||
if (!HnswNormValue(normprocinfo, collation, &value, HnswGetType(index)))
|
||||
if (!HnswNormValue(normprocinfo, collation, &value, NULL))
|
||||
return;
|
||||
}
|
||||
|
||||
|
||||
@@ -1,8 +1,6 @@
|
||||
#include "postgres.h"
|
||||
|
||||
#include "access/relscan.h"
|
||||
#include "bitvector.h"
|
||||
#include "catalog/pg_type_d.h"
|
||||
#include "hnsw.h"
|
||||
#include "pgstat.h"
|
||||
#include "storage/bufmgr.h"
|
||||
@@ -42,6 +40,29 @@ GetScanItems(IndexScanDesc scan, Datum q)
|
||||
return HnswSearchLayer(base, q, ep, hnsw_ef_search, 0, index, procinfo, collation, m, false, NULL);
|
||||
}
|
||||
|
||||
/*
|
||||
* Get dimensions from metapage
|
||||
*/
|
||||
static int
|
||||
GetDimensions(Relation index)
|
||||
{
|
||||
Buffer buf;
|
||||
Page page;
|
||||
HnswMetaPage metap;
|
||||
int dimensions;
|
||||
|
||||
buf = ReadBuffer(index, HNSW_METAPAGE_BLKNO);
|
||||
LockBuffer(buf, BUFFER_LOCK_SHARE);
|
||||
page = BufferGetPage(buf);
|
||||
metap = HnswPageGetMeta(page);
|
||||
|
||||
dimensions = metap->dimensions;
|
||||
|
||||
UnlockReleaseBuffer(buf);
|
||||
|
||||
return dimensions;
|
||||
}
|
||||
|
||||
/*
|
||||
* Get scan value
|
||||
*/
|
||||
@@ -52,7 +73,7 @@ GetScanValue(IndexScanDesc scan)
|
||||
Datum value;
|
||||
|
||||
if (scan->orderByData->sk_flags & SK_ISNULL)
|
||||
value = PointerGetDatum(NULL);
|
||||
value = PointerGetDatum(InitVector(GetDimensions(scan->indexRelation)));
|
||||
else
|
||||
{
|
||||
value = scan->orderByData->sk_argument;
|
||||
@@ -63,7 +84,7 @@ GetScanValue(IndexScanDesc scan)
|
||||
|
||||
/* Fine if normalization fails */
|
||||
if (so->normprocinfo != NULL)
|
||||
HnswNormValue(so->normprocinfo, so->collation, &value, HnswGetType(scan->indexRelation));
|
||||
HnswNormValue(so->normprocinfo, so->collation, &value, NULL);
|
||||
}
|
||||
|
||||
return value;
|
||||
|
||||
@@ -3,7 +3,6 @@
|
||||
#include <math.h>
|
||||
|
||||
#include "access/generic_xlog.h"
|
||||
#include "catalog/pg_type_d.h"
|
||||
#include "hnsw.h"
|
||||
#include "lib/pairingheap.h"
|
||||
#include "storage/bufmgr.h"
|
||||
@@ -150,20 +149,6 @@ HnswOptionalProcInfo(Relation index, uint16 procnum)
|
||||
return index_getprocinfo(index, 1, procnum);
|
||||
}
|
||||
|
||||
/*
|
||||
* Get type
|
||||
*/
|
||||
HnswType
|
||||
HnswGetType(Relation index)
|
||||
{
|
||||
Oid typid = TupleDescAttr(index->rd_att, 0)->atttypid;
|
||||
|
||||
if (typid == BITOID || typid == VARBITOID)
|
||||
return HNSW_TYPE_BIT;
|
||||
|
||||
return HNSW_TYPE_VECTOR;
|
||||
}
|
||||
|
||||
/*
|
||||
* Divide by the norm
|
||||
*
|
||||
@@ -173,25 +158,21 @@ HnswGetType(Relation index)
|
||||
* if it's different than the original value
|
||||
*/
|
||||
bool
|
||||
HnswNormValue(FmgrInfo *procinfo, Oid collation, Datum *value, HnswType type)
|
||||
HnswNormValue(FmgrInfo *procinfo, Oid collation, Datum *value, Vector * result)
|
||||
{
|
||||
double norm = DatumGetFloat8(FunctionCall1Coll(procinfo, collation, *value));
|
||||
|
||||
if (norm > 0)
|
||||
{
|
||||
/* TODO Remove vector-specific code */
|
||||
if (type == HNSW_TYPE_VECTOR)
|
||||
{
|
||||
Vector *v = DatumGetVector(*value);
|
||||
Vector *result = InitVector(v->dim);
|
||||
Vector *v = DatumGetVector(*value);
|
||||
|
||||
for (int i = 0; i < v->dim; i++)
|
||||
result->x[i] = v->x[i] / norm;
|
||||
if (result == NULL)
|
||||
result = InitVector(v->dim);
|
||||
|
||||
*value = PointerGetDatum(result);
|
||||
}
|
||||
else
|
||||
elog(ERROR, "Unsupported type");
|
||||
for (int i = 0; i < v->dim; i++)
|
||||
result->x[i] = v->x[i] / norm;
|
||||
|
||||
*value = PointerGetDatum(result);
|
||||
|
||||
return true;
|
||||
}
|
||||
@@ -594,12 +575,7 @@ HnswLoadElement(HnswElement element, float *distance, Datum *q, Relation index,
|
||||
|
||||
/* Calculate distance */
|
||||
if (distance != NULL)
|
||||
{
|
||||
if (DatumGetPointer(*q) == NULL)
|
||||
*distance = 0;
|
||||
else
|
||||
*distance = (float) DatumGetFloat8(FunctionCall2Coll(procinfo, collation, *q, PointerGetDatum(&etup->data)));
|
||||
}
|
||||
*distance = (float) DatumGetFloat8(FunctionCall2Coll(procinfo, collation, *q, PointerGetDatum(&etup->data)));
|
||||
|
||||
UnlockReleaseBuffer(buf);
|
||||
}
|
||||
@@ -884,15 +860,12 @@ HnswSearchLayer(char *base, Datum q, List *ep, int ef, int lc, Relation index, F
|
||||
static int
|
||||
#if PG_VERSION_NUM >= 130000
|
||||
CompareCandidateDistances(const ListCell *a, const ListCell *b)
|
||||
{
|
||||
HnswCandidate *hca = lfirst(a);
|
||||
HnswCandidate *hcb = lfirst(b);
|
||||
#else
|
||||
CompareCandidateDistances(const void *a, const void *b)
|
||||
{
|
||||
HnswCandidate *hca = lfirst(*(ListCell **) a);
|
||||
HnswCandidate *hcb = lfirst(*(ListCell **) b);
|
||||
#endif
|
||||
{
|
||||
HnswCandidate *hca = lfirst((ListCell *) a);
|
||||
HnswCandidate *hcb = lfirst((ListCell *) b);
|
||||
|
||||
if (hca->distance < hcb->distance)
|
||||
return 1;
|
||||
@@ -915,15 +888,12 @@ CompareCandidateDistances(const void *a, const void *b)
|
||||
static int
|
||||
#if PG_VERSION_NUM >= 130000
|
||||
CompareCandidateDistancesOffset(const ListCell *a, const ListCell *b)
|
||||
{
|
||||
HnswCandidate *hca = lfirst(a);
|
||||
HnswCandidate *hcb = lfirst(b);
|
||||
#else
|
||||
CompareCandidateDistancesOffset(const void *a, const void *b)
|
||||
{
|
||||
HnswCandidate *hca = lfirst(*(ListCell **) a);
|
||||
HnswCandidate *hcb = lfirst(*(ListCell **) b);
|
||||
#endif
|
||||
{
|
||||
HnswCandidate *hca = lfirst((ListCell *) a);
|
||||
HnswCandidate *hcb = lfirst((ListCell *) b);
|
||||
|
||||
if (hca->distance < hcb->distance)
|
||||
return 1;
|
||||
@@ -982,9 +952,7 @@ SelectNeighbors(char *base, List *c, int lm, int lc, FmgrInfo *procinfo, Oid col
|
||||
{
|
||||
List *r = NIL;
|
||||
List *w = list_copy(c);
|
||||
HnswCandidate **wd;
|
||||
int wdlen = 0;
|
||||
int wdoff = 0;
|
||||
pairingheap *wd;
|
||||
HnswNeighborArray *neighbors = HnswGetNeighbors(base, e2, lc);
|
||||
bool mustCalculate = !neighbors->closerSet;
|
||||
List *added = NIL;
|
||||
@@ -993,7 +961,7 @@ SelectNeighbors(char *base, List *c, int lm, int lc, FmgrInfo *procinfo, Oid col
|
||||
if (list_length(w) <= lm)
|
||||
return w;
|
||||
|
||||
wd = palloc(sizeof(HnswCandidate *) * list_length(w));
|
||||
wd = pairingheap_allocate(CompareNearestCandidates, NULL);
|
||||
|
||||
/* Ensure order of candidates is deterministic for closer caching */
|
||||
if (sortCandidates)
|
||||
@@ -1059,21 +1027,21 @@ SelectNeighbors(char *base, List *c, int lm, int lc, FmgrInfo *procinfo, Oid col
|
||||
if (e->closer)
|
||||
r = lappend(r, e);
|
||||
else
|
||||
wd[wdlen++] = e;
|
||||
pairingheap_add(wd, &(CreatePairingHeapNode(e)->ph_node));
|
||||
}
|
||||
|
||||
/* Cached value can only be used in future if sorted deterministically */
|
||||
neighbors->closerSet = sortCandidates;
|
||||
|
||||
/* Keep pruned connections */
|
||||
while (wdoff < wdlen && list_length(r) < lm)
|
||||
r = lappend(r, wd[wdoff++]);
|
||||
while (!pairingheap_is_empty(wd) && list_length(r) < lm)
|
||||
r = lappend(r, ((HnswPairingHeapNode *) pairingheap_remove_first(wd))->inner);
|
||||
|
||||
/* Return pruned for update connections */
|
||||
if (pruned != NULL)
|
||||
{
|
||||
if (wdoff < wdlen)
|
||||
*pruned = wd[wdoff];
|
||||
if (!pairingheap_is_empty(wd))
|
||||
*pruned = ((HnswPairingHeapNode *) pairingheap_first(wd))->inner;
|
||||
else
|
||||
*pruned = linitial(w);
|
||||
}
|
||||
|
||||
@@ -57,7 +57,7 @@ AddSample(Datum *values, IvfflatBuildState * buildstate)
|
||||
*/
|
||||
if (buildstate->kmeansnormprocinfo != NULL)
|
||||
{
|
||||
if (!IvfflatNormValue(buildstate->kmeansnormprocinfo, buildstate->collation, &value))
|
||||
if (!IvfflatNormValue(buildstate->kmeansnormprocinfo, buildstate->collation, &value, buildstate->normvec))
|
||||
return;
|
||||
}
|
||||
|
||||
@@ -105,7 +105,7 @@ SampleCallback(Relation index, CALLBACK_ITEM_POINTER, Datum *values,
|
||||
oldCtx = MemoryContextSwitchTo(buildstate->tmpCtx);
|
||||
|
||||
/* Add sample */
|
||||
AddSample(values, buildstate);
|
||||
AddSample(values, state);
|
||||
|
||||
/* Reset memory context */
|
||||
MemoryContextSwitchTo(oldCtx);
|
||||
@@ -153,7 +153,7 @@ AddTupleToSort(Relation index, ItemPointer tid, Datum *values, IvfflatBuildState
|
||||
/* Normalize if needed */
|
||||
if (buildstate->normprocinfo != NULL)
|
||||
{
|
||||
if (!IvfflatNormValue(buildstate->normprocinfo, buildstate->collation, &value))
|
||||
if (!IvfflatNormValue(buildstate->normprocinfo, buildstate->collation, &value, buildstate->normvec))
|
||||
return;
|
||||
}
|
||||
|
||||
@@ -356,6 +356,9 @@ InitBuildState(IvfflatBuildState * buildstate, Relation heap, Relation index, In
|
||||
buildstate->centers = VectorArrayInit(buildstate->lists, buildstate->dimensions);
|
||||
buildstate->listInfo = palloc(sizeof(ListInfo) * buildstate->lists);
|
||||
|
||||
/* Reuse for each tuple */
|
||||
buildstate->normvec = InitVector(buildstate->dimensions);
|
||||
|
||||
buildstate->tmpCtx = AllocSetContextCreate(CurrentMemoryContext,
|
||||
"Ivfflat build temporary context",
|
||||
ALLOCSET_DEFAULT_SIZES);
|
||||
@@ -377,6 +380,7 @@ FreeBuildState(IvfflatBuildState * buildstate)
|
||||
{
|
||||
VectorArrayFree(buildstate->centers);
|
||||
pfree(buildstate->listInfo);
|
||||
pfree(buildstate->normvec);
|
||||
|
||||
#ifdef IVFFLAT_KMEANS_DEBUG
|
||||
pfree(buildstate->listSums);
|
||||
|
||||
@@ -172,6 +172,7 @@ typedef struct IvfflatBuildState
|
||||
VectorArray samples;
|
||||
VectorArray centers;
|
||||
ListInfo *listInfo;
|
||||
Vector *normvec;
|
||||
|
||||
#ifdef IVFFLAT_KMEANS_DEBUG
|
||||
double inertia;
|
||||
@@ -266,7 +267,7 @@ void VectorArrayFree(VectorArray arr);
|
||||
void PrintVectorArray(char *msg, VectorArray arr);
|
||||
void IvfflatKmeans(Relation index, VectorArray samples, VectorArray centers);
|
||||
FmgrInfo *IvfflatOptionalProcInfo(Relation index, uint16 procnum);
|
||||
bool IvfflatNormValue(FmgrInfo *procinfo, Oid collation, Datum *value);
|
||||
bool IvfflatNormValue(FmgrInfo *procinfo, Oid collation, Datum *value, Vector * result);
|
||||
int IvfflatGetLists(Relation index);
|
||||
void IvfflatGetMetaPageInfo(Relation index, int *lists, int *dimensions);
|
||||
void IvfflatUpdateList(Relation index, ListInfo listInfo, BlockNumber insertPage, BlockNumber originalInsertPage, BlockNumber startPage, ForkNumber forkNum);
|
||||
|
||||
@@ -85,7 +85,7 @@ InsertTuple(Relation index, Datum *values, bool *isnull, ItemPointer heap_tid, R
|
||||
normprocinfo = IvfflatOptionalProcInfo(index, IVFFLAT_NORM_PROC);
|
||||
if (normprocinfo != NULL)
|
||||
{
|
||||
if (!IvfflatNormValue(normprocinfo, index->rd_indcollation[0], &value))
|
||||
if (!IvfflatNormValue(normprocinfo, index->rd_indcollation[0], &value, NULL))
|
||||
return;
|
||||
}
|
||||
|
||||
|
||||
@@ -293,7 +293,7 @@ ivfflatgettuple(IndexScanDesc scan, ScanDirection dir)
|
||||
|
||||
/* Fine if normalization fails */
|
||||
if (so->normprocinfo != NULL)
|
||||
IvfflatNormValue(so->normprocinfo, so->collation, &value);
|
||||
IvfflatNormValue(so->normprocinfo, so->collation, &value, NULL);
|
||||
}
|
||||
|
||||
IvfflatBench("GetScanLists", GetScanLists(scan, value));
|
||||
|
||||
@@ -75,14 +75,16 @@ IvfflatOptionalProcInfo(Relation index, uint16 procnum)
|
||||
* if it's different than the original value
|
||||
*/
|
||||
bool
|
||||
IvfflatNormValue(FmgrInfo *procinfo, Oid collation, Datum *value)
|
||||
IvfflatNormValue(FmgrInfo *procinfo, Oid collation, Datum *value, Vector * result)
|
||||
{
|
||||
double norm = DatumGetFloat8(FunctionCall1Coll(procinfo, collation, *value));
|
||||
|
||||
if (norm > 0)
|
||||
{
|
||||
Vector *v = DatumGetVector(*value);
|
||||
Vector *result = InitVector(v->dim);
|
||||
|
||||
if (result == NULL)
|
||||
result = InitVector(v->dim);
|
||||
|
||||
for (int i = 0; i < v->dim; i++)
|
||||
result->x[i] = v->x[i] / norm;
|
||||
|
||||
113
src/vector.c
113
src/vector.c
@@ -2,7 +2,6 @@
|
||||
|
||||
#include <math.h>
|
||||
|
||||
#include "bitvector.h"
|
||||
#include "catalog/pg_type.h"
|
||||
#include "common/shortest_dec.h"
|
||||
#include "fmgr.h"
|
||||
@@ -276,9 +275,6 @@ vector_in(PG_FUNCTION_ARGS)
|
||||
PG_RETURN_POINTER(result);
|
||||
}
|
||||
|
||||
#define AppendChar(ptr, c) (*(ptr)++ = (c))
|
||||
#define AppendFloat(ptr, f) ((ptr) += float_to_shortest_decimal_bufn((f), (ptr)))
|
||||
|
||||
/*
|
||||
* Convert internal representation to textual representation
|
||||
*/
|
||||
@@ -290,6 +286,7 @@ vector_out(PG_FUNCTION_ARGS)
|
||||
int dim = vector->dim;
|
||||
char *buf;
|
||||
char *ptr;
|
||||
int n;
|
||||
|
||||
/*
|
||||
* Need:
|
||||
@@ -304,17 +301,21 @@ vector_out(PG_FUNCTION_ARGS)
|
||||
buf = (char *) palloc(FLOAT_SHORTEST_DECIMAL_LEN * dim + 2);
|
||||
ptr = buf;
|
||||
|
||||
AppendChar(ptr, '[');
|
||||
|
||||
*ptr = '[';
|
||||
ptr++;
|
||||
for (int i = 0; i < dim; i++)
|
||||
{
|
||||
if (i > 0)
|
||||
AppendChar(ptr, ',');
|
||||
{
|
||||
*ptr = ',';
|
||||
ptr++;
|
||||
}
|
||||
|
||||
AppendFloat(ptr, vector->x[i]);
|
||||
n = float_to_shortest_decimal_bufn(vector->x[i], ptr);
|
||||
ptr += n;
|
||||
}
|
||||
|
||||
AppendChar(ptr, ']');
|
||||
*ptr = ']';
|
||||
ptr++;
|
||||
*ptr = '\0';
|
||||
|
||||
PG_FREE_IF_COPY(vector, 0);
|
||||
@@ -859,80 +860,6 @@ vector_mul(PG_FUNCTION_ARGS)
|
||||
PG_RETURN_POINTER(result);
|
||||
}
|
||||
|
||||
/*
|
||||
* Quantize a vector
|
||||
*/
|
||||
PGDLLEXPORT PG_FUNCTION_INFO_V1(quantize_binary);
|
||||
Datum
|
||||
quantize_binary(PG_FUNCTION_ARGS)
|
||||
{
|
||||
Vector *a = PG_GETARG_VECTOR_P(0);
|
||||
float *ax = a->x;
|
||||
VarBit *result = InitBitVector(a->dim);
|
||||
unsigned char *rx = VARBITS(result);
|
||||
|
||||
for (int i = 0; i < a->dim; i++)
|
||||
rx[i / 8] |= (ax[i] > 0) << (7 - (i % 8));
|
||||
|
||||
PG_RETURN_VARBIT_P(result);
|
||||
}
|
||||
|
||||
/*
|
||||
* Get a subvector
|
||||
*/
|
||||
PGDLLEXPORT PG_FUNCTION_INFO_V1(subvector);
|
||||
Datum
|
||||
subvector(PG_FUNCTION_ARGS)
|
||||
{
|
||||
Vector *a = PG_GETARG_VECTOR_P(0);
|
||||
int32 start = PG_GETARG_INT32(1);
|
||||
int32 count = PG_GETARG_INT32(2);
|
||||
int32 end = start + count;
|
||||
float *ax = a->x;
|
||||
Vector *result;
|
||||
int dim;
|
||||
|
||||
/* Indexing starts at 1, like substring */
|
||||
if (start < 1)
|
||||
start = 1;
|
||||
|
||||
if (end > a->dim)
|
||||
end = a->dim + 1;
|
||||
|
||||
dim = end - start;
|
||||
CheckDim(dim);
|
||||
result = InitVector(dim);
|
||||
|
||||
for (int i = 0; i < dim; i++)
|
||||
result->x[i] = ax[start - 1 + i];
|
||||
|
||||
PG_RETURN_POINTER(result);
|
||||
}
|
||||
|
||||
/*
|
||||
* Concatenate vectors
|
||||
*/
|
||||
PGDLLEXPORT PG_FUNCTION_INFO_V1(vector_concat);
|
||||
Datum
|
||||
vector_concat(PG_FUNCTION_ARGS)
|
||||
{
|
||||
Vector *a = PG_GETARG_VECTOR_P(0);
|
||||
Vector *b = PG_GETARG_VECTOR_P(1);
|
||||
Vector *result;
|
||||
int dim = a->dim + b->dim;
|
||||
|
||||
CheckDim(dim);
|
||||
result = InitVector(dim);
|
||||
|
||||
for (int i = 0; i < a->dim; i++)
|
||||
result->x[i] = a->x[i];
|
||||
|
||||
for (int i = 0; i < b->dim; i++)
|
||||
result->x[i + a->dim] = b->x[i];
|
||||
|
||||
PG_RETURN_POINTER(result);
|
||||
}
|
||||
|
||||
/*
|
||||
* Internal helper to compare vectors
|
||||
*/
|
||||
@@ -970,6 +897,9 @@ vector_lt(PG_FUNCTION_ARGS)
|
||||
Vector *a = PG_GETARG_VECTOR_P(0);
|
||||
Vector *b = PG_GETARG_VECTOR_P(1);
|
||||
|
||||
/* TODO Remove in 0.7.0 */
|
||||
CheckDims(a, b);
|
||||
|
||||
PG_RETURN_BOOL(vector_cmp_internal(a, b) < 0);
|
||||
}
|
||||
|
||||
@@ -983,6 +913,9 @@ vector_le(PG_FUNCTION_ARGS)
|
||||
Vector *a = PG_GETARG_VECTOR_P(0);
|
||||
Vector *b = PG_GETARG_VECTOR_P(1);
|
||||
|
||||
/* TODO Remove in 0.7.0 */
|
||||
CheckDims(a, b);
|
||||
|
||||
PG_RETURN_BOOL(vector_cmp_internal(a, b) <= 0);
|
||||
}
|
||||
|
||||
@@ -996,6 +929,9 @@ vector_eq(PG_FUNCTION_ARGS)
|
||||
Vector *a = PG_GETARG_VECTOR_P(0);
|
||||
Vector *b = PG_GETARG_VECTOR_P(1);
|
||||
|
||||
/* TODO Remove in 0.7.0 */
|
||||
CheckDims(a, b);
|
||||
|
||||
PG_RETURN_BOOL(vector_cmp_internal(a, b) == 0);
|
||||
}
|
||||
|
||||
@@ -1009,6 +945,9 @@ vector_ne(PG_FUNCTION_ARGS)
|
||||
Vector *a = PG_GETARG_VECTOR_P(0);
|
||||
Vector *b = PG_GETARG_VECTOR_P(1);
|
||||
|
||||
/* TODO Remove in 0.7.0 */
|
||||
CheckDims(a, b);
|
||||
|
||||
PG_RETURN_BOOL(vector_cmp_internal(a, b) != 0);
|
||||
}
|
||||
|
||||
@@ -1022,6 +961,9 @@ vector_ge(PG_FUNCTION_ARGS)
|
||||
Vector *a = PG_GETARG_VECTOR_P(0);
|
||||
Vector *b = PG_GETARG_VECTOR_P(1);
|
||||
|
||||
/* TODO Remove in 0.7.0 */
|
||||
CheckDims(a, b);
|
||||
|
||||
PG_RETURN_BOOL(vector_cmp_internal(a, b) >= 0);
|
||||
}
|
||||
|
||||
@@ -1035,6 +977,9 @@ vector_gt(PG_FUNCTION_ARGS)
|
||||
Vector *a = PG_GETARG_VECTOR_P(0);
|
||||
Vector *b = PG_GETARG_VECTOR_P(1);
|
||||
|
||||
/* TODO Remove in 0.7.0 */
|
||||
CheckDims(a, b);
|
||||
|
||||
PG_RETURN_BOOL(vector_cmp_internal(a, b) > 0);
|
||||
}
|
||||
|
||||
|
||||
@@ -1,64 +0,0 @@
|
||||
SELECT hamming_distance(B'111', B'111');
|
||||
hamming_distance
|
||||
------------------
|
||||
0
|
||||
(1 row)
|
||||
|
||||
SELECT hamming_distance(B'111', B'110');
|
||||
hamming_distance
|
||||
------------------
|
||||
1
|
||||
(1 row)
|
||||
|
||||
SELECT hamming_distance(B'111', B'100');
|
||||
hamming_distance
|
||||
------------------
|
||||
2
|
||||
(1 row)
|
||||
|
||||
SELECT hamming_distance(B'111', B'000');
|
||||
hamming_distance
|
||||
------------------
|
||||
3
|
||||
(1 row)
|
||||
|
||||
SELECT hamming_distance(B'111', B'00');
|
||||
ERROR: different bit lengths 3 and 2
|
||||
SELECT jaccard_distance(B'1111', B'1111');
|
||||
jaccard_distance
|
||||
------------------
|
||||
0
|
||||
(1 row)
|
||||
|
||||
SELECT jaccard_distance(B'1111', B'1110');
|
||||
jaccard_distance
|
||||
------------------
|
||||
0.25
|
||||
(1 row)
|
||||
|
||||
SELECT jaccard_distance(B'1111', B'1100');
|
||||
jaccard_distance
|
||||
------------------
|
||||
0.5
|
||||
(1 row)
|
||||
|
||||
SELECT jaccard_distance(B'1111', B'1000');
|
||||
jaccard_distance
|
||||
------------------
|
||||
0.75
|
||||
(1 row)
|
||||
|
||||
SELECT jaccard_distance(B'1111', B'0000');
|
||||
jaccard_distance
|
||||
------------------
|
||||
1
|
||||
(1 row)
|
||||
|
||||
SELECT jaccard_distance(B'1100', B'1000');
|
||||
jaccard_distance
|
||||
------------------
|
||||
0.5
|
||||
(1 row)
|
||||
|
||||
SELECT jaccard_distance(B'1111', B'000');
|
||||
ERROR: different bit lengths 4 and 3
|
||||
@@ -24,38 +24,6 @@ SELECT '[1e37]'::vector * '[1e37]';
|
||||
ERROR: value out of range: overflow
|
||||
SELECT '[1e-37]'::vector * '[1e-37]';
|
||||
ERROR: value out of range: underflow
|
||||
SELECT '[1,2,3]'::vector || '[4,5]'::vector;
|
||||
?column?
|
||||
-------------
|
||||
[1,2,3,4,5]
|
||||
(1 row)
|
||||
|
||||
SELECT array_fill(0, ARRAY[16000])::vector || '[1]'::vector;
|
||||
ERROR: vector cannot have more than 16000 dimensions
|
||||
SELECT '[1,2,3]'::vector < '[1,2,3]';
|
||||
?column?
|
||||
----------
|
||||
f
|
||||
(1 row)
|
||||
|
||||
SELECT '[1,2,3]'::vector < '[1,2]';
|
||||
?column?
|
||||
----------
|
||||
f
|
||||
(1 row)
|
||||
|
||||
SELECT '[1,2,3]'::vector <= '[1,2,3]';
|
||||
?column?
|
||||
----------
|
||||
t
|
||||
(1 row)
|
||||
|
||||
SELECT '[1,2,3]'::vector <= '[1,2]';
|
||||
?column?
|
||||
----------
|
||||
f
|
||||
(1 row)
|
||||
|
||||
SELECT '[1,2,3]'::vector = '[1,2,3]';
|
||||
?column?
|
||||
----------
|
||||
@@ -63,47 +31,7 @@ SELECT '[1,2,3]'::vector = '[1,2,3]';
|
||||
(1 row)
|
||||
|
||||
SELECT '[1,2,3]'::vector = '[1,2]';
|
||||
?column?
|
||||
----------
|
||||
f
|
||||
(1 row)
|
||||
|
||||
SELECT '[1,2,3]'::vector != '[1,2,3]';
|
||||
?column?
|
||||
----------
|
||||
f
|
||||
(1 row)
|
||||
|
||||
SELECT '[1,2,3]'::vector != '[1,2]';
|
||||
?column?
|
||||
----------
|
||||
t
|
||||
(1 row)
|
||||
|
||||
SELECT '[1,2,3]'::vector >= '[1,2,3]';
|
||||
?column?
|
||||
----------
|
||||
t
|
||||
(1 row)
|
||||
|
||||
SELECT '[1,2,3]'::vector >= '[1,2]';
|
||||
?column?
|
||||
----------
|
||||
t
|
||||
(1 row)
|
||||
|
||||
SELECT '[1,2,3]'::vector > '[1,2,3]';
|
||||
?column?
|
||||
----------
|
||||
f
|
||||
(1 row)
|
||||
|
||||
SELECT '[1,2,3]'::vector > '[1,2]';
|
||||
?column?
|
||||
----------
|
||||
t
|
||||
(1 row)
|
||||
|
||||
ERROR: different vector dimensions 3 and 2
|
||||
SELECT vector_cmp('[1,2,3]', '[1,2,3]');
|
||||
vector_cmp
|
||||
------------
|
||||
@@ -176,152 +104,110 @@ SELECT vector_norm('[3e37,4e37]')::real;
|
||||
5e+37
|
||||
(1 row)
|
||||
|
||||
SELECT l2_distance('[0,0]'::vector, '[3,4]');
|
||||
SELECT l2_distance('[0,0]', '[3,4]');
|
||||
l2_distance
|
||||
-------------
|
||||
5
|
||||
(1 row)
|
||||
|
||||
SELECT l2_distance('[0,0]'::vector, '[0,1]');
|
||||
SELECT l2_distance('[0,0]', '[0,1]');
|
||||
l2_distance
|
||||
-------------
|
||||
1
|
||||
(1 row)
|
||||
|
||||
SELECT l2_distance('[1,2]'::vector, '[3]');
|
||||
SELECT l2_distance('[1,2]', '[3]');
|
||||
ERROR: different vector dimensions 2 and 1
|
||||
SELECT l2_distance('[3e38]'::vector, '[-3e38]');
|
||||
SELECT l2_distance('[3e38]', '[-3e38]');
|
||||
l2_distance
|
||||
-------------
|
||||
Infinity
|
||||
(1 row)
|
||||
|
||||
SELECT inner_product('[1,2]'::vector, '[3,4]');
|
||||
SELECT inner_product('[1,2]', '[3,4]');
|
||||
inner_product
|
||||
---------------
|
||||
11
|
||||
(1 row)
|
||||
|
||||
SELECT inner_product('[1,2]'::vector, '[3]');
|
||||
SELECT inner_product('[1,2]', '[3]');
|
||||
ERROR: different vector dimensions 2 and 1
|
||||
SELECT inner_product('[3e38]'::vector, '[3e38]');
|
||||
SELECT inner_product('[3e38]', '[3e38]');
|
||||
inner_product
|
||||
---------------
|
||||
Infinity
|
||||
(1 row)
|
||||
|
||||
SELECT cosine_distance('[1,2]'::vector, '[2,4]');
|
||||
SELECT cosine_distance('[1,2]', '[2,4]');
|
||||
cosine_distance
|
||||
-----------------
|
||||
0
|
||||
(1 row)
|
||||
|
||||
SELECT cosine_distance('[1,2]'::vector, '[0,0]');
|
||||
SELECT cosine_distance('[1,2]', '[0,0]');
|
||||
cosine_distance
|
||||
-----------------
|
||||
NaN
|
||||
(1 row)
|
||||
|
||||
SELECT cosine_distance('[1,1]'::vector, '[1,1]');
|
||||
SELECT cosine_distance('[1,1]', '[1,1]');
|
||||
cosine_distance
|
||||
-----------------
|
||||
0
|
||||
(1 row)
|
||||
|
||||
SELECT cosine_distance('[1,0]'::vector, '[0,2]');
|
||||
SELECT cosine_distance('[1,0]', '[0,2]');
|
||||
cosine_distance
|
||||
-----------------
|
||||
1
|
||||
(1 row)
|
||||
|
||||
SELECT cosine_distance('[1,1]'::vector, '[-1,-1]');
|
||||
SELECT cosine_distance('[1,1]', '[-1,-1]');
|
||||
cosine_distance
|
||||
-----------------
|
||||
2
|
||||
(1 row)
|
||||
|
||||
SELECT cosine_distance('[1,2]'::vector, '[3]');
|
||||
SELECT cosine_distance('[1,2]', '[3]');
|
||||
ERROR: different vector dimensions 2 and 1
|
||||
SELECT cosine_distance('[1,1]'::vector, '[1.1,1.1]');
|
||||
SELECT cosine_distance('[1,1]', '[1.1,1.1]');
|
||||
cosine_distance
|
||||
-----------------
|
||||
0
|
||||
(1 row)
|
||||
|
||||
SELECT cosine_distance('[1,1]'::vector, '[-1.1,-1.1]');
|
||||
SELECT cosine_distance('[1,1]', '[-1.1,-1.1]');
|
||||
cosine_distance
|
||||
-----------------
|
||||
2
|
||||
(1 row)
|
||||
|
||||
SELECT cosine_distance('[3e38]'::vector, '[3e38]');
|
||||
SELECT cosine_distance('[3e38]', '[3e38]');
|
||||
cosine_distance
|
||||
-----------------
|
||||
NaN
|
||||
(1 row)
|
||||
|
||||
SELECT l1_distance('[0,0]'::vector, '[3,4]');
|
||||
SELECT l1_distance('[0,0]', '[3,4]');
|
||||
l1_distance
|
||||
-------------
|
||||
7
|
||||
(1 row)
|
||||
|
||||
SELECT l1_distance('[0,0]'::vector, '[0,1]');
|
||||
SELECT l1_distance('[0,0]', '[0,1]');
|
||||
l1_distance
|
||||
-------------
|
||||
1
|
||||
(1 row)
|
||||
|
||||
SELECT l1_distance('[1,2]'::vector, '[3]');
|
||||
SELECT l1_distance('[1,2]', '[3]');
|
||||
ERROR: different vector dimensions 2 and 1
|
||||
SELECT l1_distance('[3e38]'::vector, '[-3e38]');
|
||||
SELECT l1_distance('[3e38]', '[-3e38]');
|
||||
l1_distance
|
||||
-------------
|
||||
Infinity
|
||||
(1 row)
|
||||
|
||||
SELECT quantize_binary('[1,0,-1]');
|
||||
quantize_binary
|
||||
-----------------
|
||||
100
|
||||
(1 row)
|
||||
|
||||
SELECT quantize_binary('[0,0.1,-0.2,-0.3,0.4,0.5,0.6,-0.7,0.8,-0.9,1]');
|
||||
quantize_binary
|
||||
-----------------
|
||||
01001110101
|
||||
(1 row)
|
||||
|
||||
SELECT subvector('[1,2,3,4,5]', 1, 3);
|
||||
subvector
|
||||
-----------
|
||||
[1,2,3]
|
||||
(1 row)
|
||||
|
||||
SELECT subvector('[1,2,3,4,5]', 3, 2);
|
||||
subvector
|
||||
-----------
|
||||
[3,4]
|
||||
(1 row)
|
||||
|
||||
SELECT subvector('[1,2,3,4,5]', -1, 3);
|
||||
subvector
|
||||
-----------
|
||||
[1]
|
||||
(1 row)
|
||||
|
||||
SELECT subvector('[1,2,3,4,5]', 3, 9);
|
||||
subvector
|
||||
-----------
|
||||
[3,4,5]
|
||||
(1 row)
|
||||
|
||||
SELECT subvector('[1,2,3,4,5]', 1, 0);
|
||||
ERROR: vector must have at least 1 dimension
|
||||
SELECT subvector('[1,2,3,4,5]', 3, -1);
|
||||
ERROR: vector must have at least 1 dimension
|
||||
SELECT subvector('[1,2,3,4,5]', -1, 2);
|
||||
ERROR: vector must have at least 1 dimension
|
||||
SELECT avg(v) FROM unnest(ARRAY['[1,2,3]'::vector, '[3,5,7]']) v;
|
||||
avg
|
||||
-----------
|
||||
@@ -1,21 +0,0 @@
|
||||
SET enable_seqscan = off;
|
||||
CREATE TABLE t (val bit(3));
|
||||
INSERT INTO t (val) VALUES (B'000'), (B'100'), (B'111'), (NULL);
|
||||
CREATE INDEX ON t USING hnsw (val bit_hamming_ops);
|
||||
INSERT INTO t (val) VALUES (B'110');
|
||||
SELECT * FROM t ORDER BY val <~> B'111';
|
||||
val
|
||||
-----
|
||||
111
|
||||
110
|
||||
100
|
||||
000
|
||||
(4 rows)
|
||||
|
||||
SELECT COUNT(*) FROM (SELECT * FROM t ORDER BY val <~> (SELECT NULL::bit)) t2;
|
||||
count
|
||||
-------
|
||||
4
|
||||
(1 row)
|
||||
|
||||
DROP TABLE t;
|
||||
@@ -1,21 +0,0 @@
|
||||
SET enable_seqscan = off;
|
||||
CREATE TABLE t (val bit(4));
|
||||
INSERT INTO t (val) VALUES (B'0000'), (B'1100'), (B'1111'), (NULL);
|
||||
CREATE INDEX ON t USING hnsw (val bit_jaccard_ops);
|
||||
INSERT INTO t (val) VALUES (B'1110');
|
||||
SELECT * FROM t ORDER BY val <%> B'1111';
|
||||
val
|
||||
------
|
||||
1111
|
||||
1110
|
||||
1100
|
||||
0000
|
||||
(4 rows)
|
||||
|
||||
SELECT COUNT(*) FROM (SELECT * FROM t ORDER BY val <%> (SELECT NULL::bit)) t2;
|
||||
count
|
||||
-------
|
||||
4
|
||||
(1 row)
|
||||
|
||||
DROP TABLE t;
|
||||
@@ -12,11 +12,14 @@ SELECT * FROM t ORDER BY val <-> '[3,3,3]';
|
||||
[0,0,0]
|
||||
(4 rows)
|
||||
|
||||
SELECT COUNT(*) FROM (SELECT * FROM t ORDER BY val <-> (SELECT NULL::vector)) t2;
|
||||
count
|
||||
-------
|
||||
4
|
||||
(1 row)
|
||||
SELECT * FROM t ORDER BY val <-> (SELECT NULL::vector);
|
||||
val
|
||||
---------
|
||||
[0,0,0]
|
||||
[1,1,1]
|
||||
[1,2,3]
|
||||
[1,2,4]
|
||||
(4 rows)
|
||||
|
||||
SELECT COUNT(*) FROM t;
|
||||
count
|
||||
|
||||
@@ -66,22 +66,6 @@ SELECT '[4e38,1]'::vector;
|
||||
ERROR: infinite value not allowed in vector
|
||||
LINE 1: SELECT '[4e38,1]'::vector;
|
||||
^
|
||||
SELECT '[-4e38,1]'::vector;
|
||||
ERROR: infinite value not allowed in vector
|
||||
LINE 1: SELECT '[-4e38,1]'::vector;
|
||||
^
|
||||
SELECT '[1e-46,1]'::vector;
|
||||
vector
|
||||
--------
|
||||
[0,1]
|
||||
(1 row)
|
||||
|
||||
SELECT '[-1e-46,1]'::vector;
|
||||
vector
|
||||
--------
|
||||
[-0,1]
|
||||
(1 row)
|
||||
|
||||
SELECT '[1,2,3'::vector;
|
||||
ERROR: malformed vector literal: "[1,2,3"
|
||||
LINE 1: SELECT '[1,2,3'::vector;
|
||||
@@ -132,30 +116,8 @@ SELECT '[1, ,3]'::vector;
|
||||
ERROR: invalid input syntax for type vector: "[1, ,3]"
|
||||
LINE 1: SELECT '[1, ,3]'::vector;
|
||||
^
|
||||
SELECT '[1,2,3]'::vector(3);
|
||||
vector
|
||||
---------
|
||||
[1,2,3]
|
||||
(1 row)
|
||||
|
||||
SELECT '[1,2,3]'::vector(2);
|
||||
ERROR: expected 2 dimensions, not 3
|
||||
SELECT '[1,2,3]'::vector(3, 2);
|
||||
ERROR: invalid type modifier
|
||||
LINE 1: SELECT '[1,2,3]'::vector(3, 2);
|
||||
^
|
||||
SELECT '[1,2,3]'::vector('a');
|
||||
ERROR: invalid input syntax for type integer: "a"
|
||||
LINE 1: SELECT '[1,2,3]'::vector('a');
|
||||
^
|
||||
SELECT '[1,2,3]'::vector(0);
|
||||
ERROR: dimensions for type vector must be at least 1
|
||||
LINE 1: SELECT '[1,2,3]'::vector(0);
|
||||
^
|
||||
SELECT '[1,2,3]'::vector(16001);
|
||||
ERROR: dimensions for type vector cannot exceed 16000
|
||||
LINE 1: SELECT '[1,2,3]'::vector(16001);
|
||||
^
|
||||
SELECT unnest('{"[1,2,3]", "[4,5,6]"}'::vector[]);
|
||||
unnest
|
||||
---------
|
||||
@@ -1,13 +0,0 @@
|
||||
SELECT hamming_distance(B'111', B'111');
|
||||
SELECT hamming_distance(B'111', B'110');
|
||||
SELECT hamming_distance(B'111', B'100');
|
||||
SELECT hamming_distance(B'111', B'000');
|
||||
SELECT hamming_distance(B'111', B'00');
|
||||
|
||||
SELECT jaccard_distance(B'1111', B'1111');
|
||||
SELECT jaccard_distance(B'1111', B'1110');
|
||||
SELECT jaccard_distance(B'1111', B'1100');
|
||||
SELECT jaccard_distance(B'1111', B'1000');
|
||||
SELECT jaccard_distance(B'1111', B'0000');
|
||||
SELECT jaccard_distance(B'1100', B'1000');
|
||||
SELECT jaccard_distance(B'1111', B'000');
|
||||
62
test/sql/functions.sql
Normal file
62
test/sql/functions.sql
Normal file
@@ -0,0 +1,62 @@
|
||||
SELECT '[1,2,3]'::vector + '[4,5,6]';
|
||||
SELECT '[3e38]'::vector + '[3e38]';
|
||||
SELECT '[1,2,3]'::vector - '[4,5,6]';
|
||||
SELECT '[-3e38]'::vector - '[3e38]';
|
||||
SELECT '[1,2,3]'::vector * '[4,5,6]';
|
||||
SELECT '[1e37]'::vector * '[1e37]';
|
||||
SELECT '[1e-37]'::vector * '[1e-37]';
|
||||
|
||||
SELECT '[1,2,3]'::vector = '[1,2,3]';
|
||||
SELECT '[1,2,3]'::vector = '[1,2]';
|
||||
|
||||
SELECT vector_cmp('[1,2,3]', '[1,2,3]');
|
||||
SELECT vector_cmp('[1,2,3]', '[0,0,0]');
|
||||
SELECT vector_cmp('[0,0,0]', '[1,2,3]');
|
||||
SELECT vector_cmp('[1,2]', '[1,2,3]');
|
||||
SELECT vector_cmp('[1,2,3]', '[1,2]');
|
||||
SELECT vector_cmp('[1,2]', '[2,3,4]');
|
||||
SELECT vector_cmp('[2,3]', '[1,2,3]');
|
||||
|
||||
SELECT vector_dims('[1,2,3]');
|
||||
|
||||
SELECT round(vector_norm('[1,1]')::numeric, 5);
|
||||
SELECT vector_norm('[3,4]');
|
||||
SELECT vector_norm('[0,1]');
|
||||
SELECT vector_norm('[3e37,4e37]')::real;
|
||||
|
||||
SELECT l2_distance('[0,0]', '[3,4]');
|
||||
SELECT l2_distance('[0,0]', '[0,1]');
|
||||
SELECT l2_distance('[1,2]', '[3]');
|
||||
SELECT l2_distance('[3e38]', '[-3e38]');
|
||||
|
||||
SELECT inner_product('[1,2]', '[3,4]');
|
||||
SELECT inner_product('[1,2]', '[3]');
|
||||
SELECT inner_product('[3e38]', '[3e38]');
|
||||
|
||||
SELECT cosine_distance('[1,2]', '[2,4]');
|
||||
SELECT cosine_distance('[1,2]', '[0,0]');
|
||||
SELECT cosine_distance('[1,1]', '[1,1]');
|
||||
SELECT cosine_distance('[1,0]', '[0,2]');
|
||||
SELECT cosine_distance('[1,1]', '[-1,-1]');
|
||||
SELECT cosine_distance('[1,2]', '[3]');
|
||||
SELECT cosine_distance('[1,1]', '[1.1,1.1]');
|
||||
SELECT cosine_distance('[1,1]', '[-1.1,-1.1]');
|
||||
SELECT cosine_distance('[3e38]', '[3e38]');
|
||||
|
||||
SELECT l1_distance('[0,0]', '[3,4]');
|
||||
SELECT l1_distance('[0,0]', '[0,1]');
|
||||
SELECT l1_distance('[1,2]', '[3]');
|
||||
SELECT l1_distance('[3e38]', '[-3e38]');
|
||||
|
||||
SELECT avg(v) FROM unnest(ARRAY['[1,2,3]'::vector, '[3,5,7]']) v;
|
||||
SELECT avg(v) FROM unnest(ARRAY['[1,2,3]'::vector, '[3,5,7]', NULL]) v;
|
||||
SELECT avg(v) FROM unnest(ARRAY[]::vector[]) v;
|
||||
SELECT avg(v) FROM unnest(ARRAY['[1,2]'::vector, '[3]']) v;
|
||||
SELECT avg(v) FROM unnest(ARRAY['[3e38]'::vector, '[3e38]']) v;
|
||||
SELECT vector_avg(array_agg(n)) FROM generate_series(1, 16002) n;
|
||||
|
||||
SELECT sum(v) FROM unnest(ARRAY['[1,2,3]'::vector, '[3,5,7]']) v;
|
||||
SELECT sum(v) FROM unnest(ARRAY['[1,2,3]'::vector, '[3,5,7]', NULL]) v;
|
||||
SELECT sum(v) FROM unnest(ARRAY[]::vector[]) v;
|
||||
SELECT sum(v) FROM unnest(ARRAY['[1,2]'::vector, '[3]']) v;
|
||||
SELECT sum(v) FROM unnest(ARRAY['[3e38]'::vector, '[3e38]']) v;
|
||||
@@ -1,12 +0,0 @@
|
||||
SET enable_seqscan = off;
|
||||
|
||||
CREATE TABLE t (val bit(3));
|
||||
INSERT INTO t (val) VALUES (B'000'), (B'100'), (B'111'), (NULL);
|
||||
CREATE INDEX ON t USING hnsw (val bit_hamming_ops);
|
||||
|
||||
INSERT INTO t (val) VALUES (B'110');
|
||||
|
||||
SELECT * FROM t ORDER BY val <~> B'111';
|
||||
SELECT COUNT(*) FROM (SELECT * FROM t ORDER BY val <~> (SELECT NULL::bit)) t2;
|
||||
|
||||
DROP TABLE t;
|
||||
@@ -1,12 +0,0 @@
|
||||
SET enable_seqscan = off;
|
||||
|
||||
CREATE TABLE t (val bit(4));
|
||||
INSERT INTO t (val) VALUES (B'0000'), (B'1100'), (B'1111'), (NULL);
|
||||
CREATE INDEX ON t USING hnsw (val bit_jaccard_ops);
|
||||
|
||||
INSERT INTO t (val) VALUES (B'1110');
|
||||
|
||||
SELECT * FROM t ORDER BY val <%> B'1111';
|
||||
SELECT COUNT(*) FROM (SELECT * FROM t ORDER BY val <%> (SELECT NULL::bit)) t2;
|
||||
|
||||
DROP TABLE t;
|
||||
@@ -7,7 +7,7 @@ CREATE INDEX ON t USING hnsw (val vector_l2_ops);
|
||||
INSERT INTO t (val) VALUES ('[1,2,4]');
|
||||
|
||||
SELECT * FROM t ORDER BY val <-> '[3,3,3]';
|
||||
SELECT COUNT(*) FROM (SELECT * FROM t ORDER BY val <-> (SELECT NULL::vector)) t2;
|
||||
SELECT * FROM t ORDER BY val <-> (SELECT NULL::vector);
|
||||
SELECT COUNT(*) FROM t;
|
||||
|
||||
TRUNCATE t;
|
||||
|
||||
@@ -11,9 +11,6 @@ SELECT '[1.5e38,-1.5e38]'::vector;
|
||||
SELECT '[1.5e+38,-1.5e+38]'::vector;
|
||||
SELECT '[1.5e-38,-1.5e-38]'::vector;
|
||||
SELECT '[4e38,1]'::vector;
|
||||
SELECT '[-4e38,1]'::vector;
|
||||
SELECT '[1e-46,1]'::vector;
|
||||
SELECT '[-1e-46,1]'::vector;
|
||||
SELECT '[1,2,3'::vector;
|
||||
SELECT '[1,2,3]9'::vector;
|
||||
SELECT '1,2,3'::vector;
|
||||
@@ -25,13 +22,7 @@ SELECT '[1,]'::vector;
|
||||
SELECT '[1a]'::vector;
|
||||
SELECT '[1,,3]'::vector;
|
||||
SELECT '[1, ,3]'::vector;
|
||||
|
||||
SELECT '[1,2,3]'::vector(3);
|
||||
SELECT '[1,2,3]'::vector(2);
|
||||
SELECT '[1,2,3]'::vector(3, 2);
|
||||
SELECT '[1,2,3]'::vector('a');
|
||||
SELECT '[1,2,3]'::vector(0);
|
||||
SELECT '[1,2,3]'::vector(16001);
|
||||
|
||||
SELECT unnest('{"[1,2,3]", "[4,5,6]"}'::vector[]);
|
||||
SELECT '{"[1,2,3]"}'::vector(2)[];
|
||||
@@ -1,86 +0,0 @@
|
||||
SELECT '[1,2,3]'::vector + '[4,5,6]';
|
||||
SELECT '[3e38]'::vector + '[3e38]';
|
||||
SELECT '[1,2,3]'::vector - '[4,5,6]';
|
||||
SELECT '[-3e38]'::vector - '[3e38]';
|
||||
SELECT '[1,2,3]'::vector * '[4,5,6]';
|
||||
SELECT '[1e37]'::vector * '[1e37]';
|
||||
SELECT '[1e-37]'::vector * '[1e-37]';
|
||||
|
||||
SELECT '[1,2,3]'::vector || '[4,5]'::vector;
|
||||
SELECT array_fill(0, ARRAY[16000])::vector || '[1]'::vector;
|
||||
|
||||
SELECT '[1,2,3]'::vector < '[1,2,3]';
|
||||
SELECT '[1,2,3]'::vector < '[1,2]';
|
||||
SELECT '[1,2,3]'::vector <= '[1,2,3]';
|
||||
SELECT '[1,2,3]'::vector <= '[1,2]';
|
||||
SELECT '[1,2,3]'::vector = '[1,2,3]';
|
||||
SELECT '[1,2,3]'::vector = '[1,2]';
|
||||
SELECT '[1,2,3]'::vector != '[1,2,3]';
|
||||
SELECT '[1,2,3]'::vector != '[1,2]';
|
||||
SELECT '[1,2,3]'::vector >= '[1,2,3]';
|
||||
SELECT '[1,2,3]'::vector >= '[1,2]';
|
||||
SELECT '[1,2,3]'::vector > '[1,2,3]';
|
||||
SELECT '[1,2,3]'::vector > '[1,2]';
|
||||
|
||||
SELECT vector_cmp('[1,2,3]', '[1,2,3]');
|
||||
SELECT vector_cmp('[1,2,3]', '[0,0,0]');
|
||||
SELECT vector_cmp('[0,0,0]', '[1,2,3]');
|
||||
SELECT vector_cmp('[1,2]', '[1,2,3]');
|
||||
SELECT vector_cmp('[1,2,3]', '[1,2]');
|
||||
SELECT vector_cmp('[1,2]', '[2,3,4]');
|
||||
SELECT vector_cmp('[2,3]', '[1,2,3]');
|
||||
|
||||
SELECT vector_dims('[1,2,3]');
|
||||
|
||||
SELECT round(vector_norm('[1,1]')::numeric, 5);
|
||||
SELECT vector_norm('[3,4]');
|
||||
SELECT vector_norm('[0,1]');
|
||||
SELECT vector_norm('[3e37,4e37]')::real;
|
||||
|
||||
SELECT l2_distance('[0,0]'::vector, '[3,4]');
|
||||
SELECT l2_distance('[0,0]'::vector, '[0,1]');
|
||||
SELECT l2_distance('[1,2]'::vector, '[3]');
|
||||
SELECT l2_distance('[3e38]'::vector, '[-3e38]');
|
||||
|
||||
SELECT inner_product('[1,2]'::vector, '[3,4]');
|
||||
SELECT inner_product('[1,2]'::vector, '[3]');
|
||||
SELECT inner_product('[3e38]'::vector, '[3e38]');
|
||||
|
||||
SELECT cosine_distance('[1,2]'::vector, '[2,4]');
|
||||
SELECT cosine_distance('[1,2]'::vector, '[0,0]');
|
||||
SELECT cosine_distance('[1,1]'::vector, '[1,1]');
|
||||
SELECT cosine_distance('[1,0]'::vector, '[0,2]');
|
||||
SELECT cosine_distance('[1,1]'::vector, '[-1,-1]');
|
||||
SELECT cosine_distance('[1,2]'::vector, '[3]');
|
||||
SELECT cosine_distance('[1,1]'::vector, '[1.1,1.1]');
|
||||
SELECT cosine_distance('[1,1]'::vector, '[-1.1,-1.1]');
|
||||
SELECT cosine_distance('[3e38]'::vector, '[3e38]');
|
||||
|
||||
SELECT l1_distance('[0,0]'::vector, '[3,4]');
|
||||
SELECT l1_distance('[0,0]'::vector, '[0,1]');
|
||||
SELECT l1_distance('[1,2]'::vector, '[3]');
|
||||
SELECT l1_distance('[3e38]'::vector, '[-3e38]');
|
||||
|
||||
SELECT quantize_binary('[1,0,-1]');
|
||||
SELECT quantize_binary('[0,0.1,-0.2,-0.3,0.4,0.5,0.6,-0.7,0.8,-0.9,1]');
|
||||
|
||||
SELECT subvector('[1,2,3,4,5]', 1, 3);
|
||||
SELECT subvector('[1,2,3,4,5]', 3, 2);
|
||||
SELECT subvector('[1,2,3,4,5]', -1, 3);
|
||||
SELECT subvector('[1,2,3,4,5]', 3, 9);
|
||||
SELECT subvector('[1,2,3,4,5]', 1, 0);
|
||||
SELECT subvector('[1,2,3,4,5]', 3, -1);
|
||||
SELECT subvector('[1,2,3,4,5]', -1, 2);
|
||||
|
||||
SELECT avg(v) FROM unnest(ARRAY['[1,2,3]'::vector, '[3,5,7]']) v;
|
||||
SELECT avg(v) FROM unnest(ARRAY['[1,2,3]'::vector, '[3,5,7]', NULL]) v;
|
||||
SELECT avg(v) FROM unnest(ARRAY[]::vector[]) v;
|
||||
SELECT avg(v) FROM unnest(ARRAY['[1,2]'::vector, '[3]']) v;
|
||||
SELECT avg(v) FROM unnest(ARRAY['[3e38]'::vector, '[3e38]']) v;
|
||||
SELECT vector_avg(array_agg(n)) FROM generate_series(1, 16002) n;
|
||||
|
||||
SELECT sum(v) FROM unnest(ARRAY['[1,2,3]'::vector, '[3,5,7]']) v;
|
||||
SELECT sum(v) FROM unnest(ARRAY['[1,2,3]'::vector, '[3,5,7]', NULL]) v;
|
||||
SELECT sum(v) FROM unnest(ARRAY[]::vector[]) v;
|
||||
SELECT sum(v) FROM unnest(ARRAY['[1,2]'::vector, '[3]']) v;
|
||||
SELECT sum(v) FROM unnest(ARRAY['[3e38]'::vector, '[3e38]']) v;
|
||||
@@ -86,7 +86,7 @@ foreach (@queries)
|
||||
push(@expected, $res);
|
||||
}
|
||||
|
||||
test_recall(0.19, $limit, "before vacuum");
|
||||
test_recall(0.20, $limit, "before vacuum");
|
||||
test_recall(0.95, 100, "before vacuum");
|
||||
|
||||
# TODO Test concurrent inserts with vacuum
|
||||
|
||||
113
test/t/020_hnsw_array.pl
Normal file
113
test/t/020_hnsw_array.pl
Normal file
@@ -0,0 +1,113 @@
|
||||
use strict;
|
||||
use warnings;
|
||||
use PostgresNode;
|
||||
use TestLib;
|
||||
use Test::More;
|
||||
|
||||
my $node;
|
||||
my @queries = ();
|
||||
my @expected;
|
||||
my $limit = 20;
|
||||
|
||||
sub test_recall
|
||||
{
|
||||
my ($min, $operator) = @_;
|
||||
my $correct = 0;
|
||||
my $total = 0;
|
||||
|
||||
my $explain = $node->safe_psql("postgres", qq(
|
||||
SET enable_seqscan = off;
|
||||
EXPLAIN ANALYZE SELECT i FROM tst ORDER BY v $operator '$queries[0]' LIMIT $limit;
|
||||
));
|
||||
like($explain, qr/Index Scan/);
|
||||
|
||||
for my $i (0 .. $#queries)
|
||||
{
|
||||
my $actual = $node->safe_psql("postgres", qq(
|
||||
SET enable_seqscan = off;
|
||||
SELECT i FROM tst ORDER BY v $operator '$queries[$i]' LIMIT $limit;
|
||||
));
|
||||
my @actual_ids = split("\n", $actual);
|
||||
my %actual_set = map { $_ => 1 } @actual_ids;
|
||||
|
||||
my @expected_ids = split("\n", $expected[$i]);
|
||||
|
||||
foreach (@expected_ids)
|
||||
{
|
||||
if (exists($actual_set{$_}))
|
||||
{
|
||||
$correct++;
|
||||
}
|
||||
$total++;
|
||||
}
|
||||
}
|
||||
|
||||
cmp_ok($correct / $total, ">=", $min, $operator);
|
||||
}
|
||||
|
||||
# Initialize node
|
||||
$node = get_new_node('node');
|
||||
$node->init;
|
||||
$node->start;
|
||||
|
||||
# Create table
|
||||
$node->safe_psql("postgres", "CREATE EXTENSION vector;");
|
||||
$node->safe_psql("postgres", "CREATE TABLE tst (i int4, v float4[3]);");
|
||||
$node->safe_psql("postgres",
|
||||
"INSERT INTO tst SELECT i, ARRAY[random(), random(), random()] FROM generate_series(1, 10000) i;"
|
||||
);
|
||||
|
||||
$node->safe_psql("postgres", qq(
|
||||
CREATE FUNCTION float4_l2_distance(float4[], float4[]) RETURNS float8
|
||||
AS 'BEGIN RETURN l2_distance(\$1::vector, \$2::vector); END;'
|
||||
LANGUAGE plpgsql IMMUTABLE STRICT PARALLEL SAFE;
|
||||
|
||||
CREATE FUNCTION float4_l2_squared_distance(float4[], float4[]) RETURNS float8
|
||||
AS 'BEGIN RETURN vector_l2_squared_distance(\$1::vector, \$2::vector); END;'
|
||||
LANGUAGE plpgsql IMMUTABLE STRICT PARALLEL SAFE;
|
||||
|
||||
CREATE OPERATOR <-> (
|
||||
LEFTARG = float4[], RIGHTARG = float4[], PROCEDURE = float4_l2_distance,
|
||||
COMMUTATOR = '<->'
|
||||
);
|
||||
|
||||
CREATE OPERATOR CLASS float4_l2_ops
|
||||
FOR TYPE float4[] USING hnsw AS
|
||||
OPERATOR 1 <-> (float4[], float4[]) FOR ORDER BY float_ops,
|
||||
FUNCTION 1 float4_l2_squared_distance(float4[], float4[]);
|
||||
));
|
||||
|
||||
# Generate queries
|
||||
for (1 .. 20)
|
||||
{
|
||||
my $r1 = rand();
|
||||
my $r2 = rand();
|
||||
my $r3 = rand();
|
||||
push(@queries, "{$r1,$r2,$r3}");
|
||||
}
|
||||
|
||||
# Check each index type
|
||||
my @operators = ("<->");
|
||||
my @opclasses = ("float4_l2_ops");
|
||||
|
||||
for my $i (0 .. $#operators)
|
||||
{
|
||||
my $operator = $operators[$i];
|
||||
my $opclass = $opclasses[$i];
|
||||
|
||||
# Get exact results
|
||||
@expected = ();
|
||||
foreach (@queries)
|
||||
{
|
||||
my $res = $node->safe_psql("postgres", "SELECT i FROM tst ORDER BY v $operator '$_' LIMIT $limit;");
|
||||
push(@expected, $res);
|
||||
}
|
||||
|
||||
# Add index
|
||||
$node->safe_psql("postgres", "CREATE INDEX ON tst USING hnsw (v $opclass);");
|
||||
|
||||
my $min = $operator eq "<#>" ? 0.80 : 0.99;
|
||||
test_recall($min, $operator);
|
||||
}
|
||||
|
||||
done_testing();
|
||||
@@ -1,137 +0,0 @@
|
||||
use strict;
|
||||
use warnings;
|
||||
use PostgresNode;
|
||||
use TestLib;
|
||||
use Test::More;
|
||||
|
||||
my $node;
|
||||
my @queries = ();
|
||||
my @expected;
|
||||
my $limit = 20;
|
||||
my $dim = 52;
|
||||
my $max = 2**$dim;
|
||||
|
||||
sub test_recall
|
||||
{
|
||||
my ($min, $operator) = @_;
|
||||
my $correct = 0;
|
||||
my $total = 0;
|
||||
|
||||
my $explain = $node->safe_psql("postgres", qq(
|
||||
SET enable_seqscan = off;
|
||||
SET hnsw.ef_search = 100;
|
||||
EXPLAIN ANALYZE SELECT i FROM tst ORDER BY v $operator $queries[0] LIMIT $limit;
|
||||
));
|
||||
like($explain, qr/Index Scan/);
|
||||
|
||||
for my $i (0 .. $#queries)
|
||||
{
|
||||
my $actual = $node->safe_psql("postgres", qq(
|
||||
SET enable_seqscan = off;
|
||||
SET hnsw.ef_search = 100;
|
||||
SELECT i FROM tst ORDER BY v $operator $queries[$i] LIMIT $limit;
|
||||
));
|
||||
my @actual_ids = split("\n", $actual);
|
||||
|
||||
my @expected_ids = split("\n", $expected[$i]);
|
||||
my %expected_set = map { $_ => 1 } @expected_ids;
|
||||
|
||||
foreach (@actual_ids)
|
||||
{
|
||||
if (exists($expected_set{$_}))
|
||||
{
|
||||
$correct++;
|
||||
}
|
||||
}
|
||||
|
||||
$total += $limit;
|
||||
}
|
||||
|
||||
cmp_ok($correct / $total, ">=", $min, $operator);
|
||||
}
|
||||
|
||||
# Initialize node
|
||||
$node = get_new_node('node');
|
||||
$node->init;
|
||||
$node->start;
|
||||
|
||||
# Create table
|
||||
$node->safe_psql("postgres", "CREATE EXTENSION vector;");
|
||||
$node->safe_psql("postgres", "CREATE TABLE tst (i int4, v bit($dim));");
|
||||
$node->safe_psql("postgres",
|
||||
"INSERT INTO tst SELECT i, (random() * $max)::bigint::bit($dim) FROM generate_series(1, 10000) i;"
|
||||
);
|
||||
|
||||
# Generate queries
|
||||
for (1 .. 20)
|
||||
{
|
||||
my $r = int(rand() * $max);
|
||||
push(@queries, "${r}::bigint::bit($dim)");
|
||||
}
|
||||
|
||||
# Check each index type
|
||||
my @operators = ("<~>", "<\%>");
|
||||
my @opclasses = ("bit_hamming_ops", "bit_jaccard_ops");
|
||||
|
||||
for my $i (0 .. $#operators)
|
||||
{
|
||||
my $operator = $operators[$i];
|
||||
my $opclass = $opclasses[$i];
|
||||
|
||||
# Get exact results
|
||||
@expected = ();
|
||||
foreach (@queries)
|
||||
{
|
||||
# Handle ties
|
||||
my $res = $node->safe_psql("postgres", qq(
|
||||
WITH top AS (
|
||||
SELECT v $operator $_ AS distance FROM tst ORDER BY v $operator $_ LIMIT $limit
|
||||
)
|
||||
SELECT i FROM tst WHERE (v $operator $_) <= (SELECT MAX(distance) FROM top)
|
||||
));
|
||||
push(@expected, $res);
|
||||
}
|
||||
|
||||
# Build index serially
|
||||
$node->safe_psql("postgres", qq(
|
||||
SET max_parallel_maintenance_workers = 0;
|
||||
CREATE INDEX idx ON tst USING hnsw (v $opclass);
|
||||
));
|
||||
|
||||
# Test approximate results
|
||||
my $min = $operator eq "<\%>" ? 0.96 : 0.98;
|
||||
test_recall($min, $operator);
|
||||
|
||||
$node->safe_psql("postgres", "DROP INDEX idx;");
|
||||
|
||||
# Build index in parallel in memory
|
||||
my ($ret, $stdout, $stderr) = $node->psql("postgres", qq(
|
||||
SET client_min_messages = DEBUG;
|
||||
SET min_parallel_table_scan_size = 1;
|
||||
CREATE INDEX idx ON tst USING hnsw (v $opclass);
|
||||
));
|
||||
is($ret, 0, $stderr);
|
||||
like($stderr, qr/using \d+ parallel workers/);
|
||||
|
||||
# Test approximate results
|
||||
test_recall($min, $operator);
|
||||
|
||||
$node->safe_psql("postgres", "DROP INDEX idx;");
|
||||
|
||||
# Build index in parallel on disk
|
||||
# Set parallel_workers on table to use workers with low maintenance_work_mem
|
||||
($ret, $stdout, $stderr) = $node->psql("postgres", qq(
|
||||
ALTER TABLE tst SET (parallel_workers = 2);
|
||||
SET client_min_messages = DEBUG;
|
||||
SET maintenance_work_mem = '4MB';
|
||||
CREATE INDEX idx ON tst USING hnsw (v $opclass);
|
||||
ALTER TABLE tst RESET (parallel_workers);
|
||||
));
|
||||
is($ret, 0, $stderr);
|
||||
like($stderr, qr/using \d+ parallel workers/);
|
||||
like($stderr, qr/hnsw graph no longer fits into maintenance_work_mem/);
|
||||
|
||||
$node->safe_psql("postgres", "DROP INDEX idx;");
|
||||
}
|
||||
|
||||
done_testing();
|
||||
@@ -1,4 +1,4 @@
|
||||
comment = 'vector data type and ivfflat and hnsw access methods'
|
||||
default_version = '0.6.2'
|
||||
default_version = '0.6.0'
|
||||
module_pathname = '$libdir/vector'
|
||||
relocatable = true
|
||||
|
||||
Reference in New Issue
Block a user