mirror of
https://github.com/pgvector/pgvector.git
synced 2026-07-22 12:07:34 +08:00
Compare commits
1 Commits
index-limi
...
hnsw-datum
| Author | SHA1 | Date | |
|---|---|---|---|
|
|
f858705293 |
8
.github/workflows/build.yml
vendored
8
.github/workflows/build.yml
vendored
@@ -8,8 +8,6 @@ jobs:
|
||||
fail-fast: false
|
||||
matrix:
|
||||
include:
|
||||
- postgres: 17
|
||||
os: ubuntu-22.04
|
||||
- postgres: 16
|
||||
os: ubuntu-22.04
|
||||
- postgres: 15
|
||||
@@ -23,7 +21,7 @@ jobs:
|
||||
- postgres: 11
|
||||
os: ubuntu-20.04
|
||||
steps:
|
||||
- uses: actions/checkout@v4
|
||||
- uses: actions/checkout@v3
|
||||
- uses: ankane/setup-postgres@v1
|
||||
with:
|
||||
postgres-version: ${{ matrix.postgres }}
|
||||
@@ -45,7 +43,7 @@ jobs:
|
||||
runs-on: macos-latest
|
||||
if: ${{ !startsWith(github.ref_name, 'windows') }}
|
||||
steps:
|
||||
- uses: actions/checkout@v4
|
||||
- uses: actions/checkout@v3
|
||||
- uses: ankane/setup-postgres@v1
|
||||
with:
|
||||
postgres-version: 14
|
||||
@@ -67,7 +65,7 @@ jobs:
|
||||
runs-on: windows-latest
|
||||
if: ${{ !startsWith(github.ref_name, 'mac') }}
|
||||
steps:
|
||||
- uses: actions/checkout@v4
|
||||
- uses: actions/checkout@v3
|
||||
- uses: ankane/setup-postgres@v1
|
||||
with:
|
||||
postgres-version: 14
|
||||
|
||||
@@ -1,12 +1,3 @@
|
||||
## 0.5.2 (unreleased)
|
||||
|
||||
- Improved performance of HNSW
|
||||
- Added support for on-disk parallel index builds for HNSW
|
||||
- Reduced memory usage for HNSW index builds
|
||||
- Reduced WAL generation for HNSW index builds
|
||||
- Fixed error with logical replication
|
||||
- Fixed `invalid memory alloc request size` error with HNSW index build
|
||||
|
||||
## 0.5.1 (2023-10-10)
|
||||
|
||||
- Improved performance of HNSW index builds
|
||||
|
||||
@@ -56,7 +56,7 @@ install:
|
||||
copy $(EXTENSION).control "$(SHAREDIR)\extension"
|
||||
copy sql\$(EXTENSION)--*.sql "$(SHAREDIR)\extension"
|
||||
mkdir "$(INCLUDEDIR_SERVER)\extension\$(EXTENSION)"
|
||||
for %f in ($(HEADERS)) do copy %f "$(INCLUDEDIR_SERVER)\extension\$(EXTENSION)"
|
||||
copy $(HEADERS) "$(INCLUDEDIR_SERVER)\extension\$(EXTENSION)"
|
||||
|
||||
installcheck:
|
||||
"$(BINDIR)\pg_regress" --bindir="$(BINDIR)" $(REGRESS_OPTS) $(REGRESS)
|
||||
|
||||
290
README.md
290
README.md
@@ -10,7 +10,7 @@ Store your vectors with the rest of your data. Supports:
|
||||
|
||||
Plus [ACID](https://en.wikipedia.org/wiki/ACID) compliance, point-in-time recovery, JOINs, and all of the other [great features](https://www.postgresql.org/about/) of Postgres
|
||||
|
||||
[](https://github.com/pgvector/pgvector/actions)
|
||||
[](https://github.com/pgvector/pgvector/actions)
|
||||
|
||||
## Installation
|
||||
|
||||
@@ -26,7 +26,7 @@ make install # may need sudo
|
||||
|
||||
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), 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)
|
||||
|
||||
## Getting Started
|
||||
|
||||
@@ -161,97 +161,8 @@ You can add an index to use approximate nearest neighbor search, which trades so
|
||||
|
||||
Supported index types are:
|
||||
|
||||
- [HNSW](#hnsw) - added in 0.5.0
|
||||
- [IVFFlat](#ivfflat)
|
||||
|
||||
## HNSW
|
||||
|
||||
An HNSW index creates a multilayer graph. It has better query performance than IVFFlat (in terms of speed-recall tradeoff), but has slower build times and uses more memory. Also, an index can be created without any data in the table since there isn’t a training step like IVFFlat.
|
||||
|
||||
Add an index for each distance function you want to use.
|
||||
|
||||
L2 distance
|
||||
|
||||
```sql
|
||||
CREATE INDEX ON items USING hnsw (embedding vector_l2_ops);
|
||||
```
|
||||
|
||||
Inner product
|
||||
|
||||
```sql
|
||||
CREATE INDEX ON items USING hnsw (embedding vector_ip_ops);
|
||||
```
|
||||
|
||||
Cosine distance
|
||||
|
||||
```sql
|
||||
CREATE INDEX ON items USING hnsw (embedding vector_cosine_ops);
|
||||
```
|
||||
|
||||
Vectors with up to 2,000 dimensions can be indexed.
|
||||
|
||||
### Index Options
|
||||
|
||||
Specify HNSW parameters
|
||||
|
||||
- `m` - the max number of connections per layer (16 by default)
|
||||
- `ef_construction` - the size of the dynamic candidate list for constructing the graph (64 by default)
|
||||
|
||||
```sql
|
||||
CREATE INDEX ON items USING hnsw (embedding vector_l2_ops) WITH (m = 16, ef_construction = 64);
|
||||
```
|
||||
|
||||
A higher value of `ef_construction` provides better recall at the cost of index build time / insert speed.
|
||||
|
||||
### Query Options
|
||||
|
||||
Specify the size of the dynamic candidate list for search (40 by default)
|
||||
|
||||
```sql
|
||||
SET hnsw.ef_search = 100;
|
||||
```
|
||||
|
||||
A higher value provides better recall at the cost of speed.
|
||||
|
||||
Use `SET LOCAL` inside a transaction to set it for a single query
|
||||
|
||||
```sql
|
||||
BEGIN;
|
||||
SET LOCAL hnsw.ef_search = 100;
|
||||
SELECT ...
|
||||
COMMIT;
|
||||
```
|
||||
|
||||
### Index Build Time
|
||||
|
||||
Indexes build significantly faster when the graph fits into `maintenance_work_mem`
|
||||
|
||||
```sql
|
||||
SET maintenance_work_mem = '8GB';
|
||||
```
|
||||
|
||||
A notice is shown when the graph no longer fits
|
||||
|
||||
```text
|
||||
NOTICE: hnsw graph no longer fits into maintenance_work_mem after 100000 tuples
|
||||
DETAIL: Building will take significantly more time.
|
||||
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
|
||||
|
||||
### Indexing Progress
|
||||
|
||||
Check [indexing progress](https://www.postgresql.org/docs/current/progress-reporting.html#CREATE-INDEX-PROGRESS-REPORTING) with Postgres 12+
|
||||
|
||||
```sql
|
||||
SELECT phase, round(100.0 * blocks_done / nullif(blocks_total, 0), 1) AS "%" FROM pg_stat_progress_create_index;
|
||||
```
|
||||
|
||||
The phases for HNSW are:
|
||||
|
||||
1. `initializing`
|
||||
2. `loading tuples`
|
||||
- [HNSW](#hnsw) - added in 0.5.0
|
||||
|
||||
## IVFFlat
|
||||
|
||||
@@ -304,32 +215,78 @@ SELECT ...
|
||||
COMMIT;
|
||||
```
|
||||
|
||||
### Index Build Time
|
||||
## HNSW
|
||||
|
||||
Speed up index creation on large tables by increasing the number of parallel workers (2 by default)
|
||||
An HNSW index creates a multilayer graph. It has slower build times and uses more memory than IVFFlat, but has better query performance (in terms of speed-recall tradeoff). There’s no training step like IVFFlat, so the index can be created without any data in the table.
|
||||
|
||||
Add an index for each distance function you want to use.
|
||||
|
||||
L2 distance
|
||||
|
||||
```sql
|
||||
SET max_parallel_maintenance_workers = 7; -- plus leader
|
||||
CREATE INDEX ON items USING hnsw (embedding vector_l2_ops);
|
||||
```
|
||||
|
||||
For a large number of workers, you may also need to increase `max_parallel_workers` (8 by default)
|
||||
Inner product
|
||||
|
||||
### Indexing Progress
|
||||
```sql
|
||||
CREATE INDEX ON items USING hnsw (embedding vector_ip_ops);
|
||||
```
|
||||
|
||||
Cosine distance
|
||||
|
||||
```sql
|
||||
CREATE INDEX ON items USING hnsw (embedding vector_cosine_ops);
|
||||
```
|
||||
|
||||
Vectors with up to 2,000 dimensions can be indexed.
|
||||
|
||||
### Index Options
|
||||
|
||||
Specify HNSW parameters
|
||||
|
||||
- `m` - the max number of connections per layer (16 by default)
|
||||
- `ef_construction` - the size of the dynamic candidate list for constructing the graph (64 by default)
|
||||
|
||||
```sql
|
||||
CREATE INDEX ON items USING hnsw (embedding vector_l2_ops) WITH (m = 16, ef_construction = 64);
|
||||
```
|
||||
|
||||
### Query Options
|
||||
|
||||
Specify the size of the dynamic candidate list for search (40 by default)
|
||||
|
||||
```sql
|
||||
SET hnsw.ef_search = 100;
|
||||
```
|
||||
|
||||
A higher value provides better recall at the cost of speed.
|
||||
|
||||
Use `SET LOCAL` inside a transaction to set it for a single query
|
||||
|
||||
```sql
|
||||
BEGIN;
|
||||
SET LOCAL hnsw.ef_search = 100;
|
||||
SELECT ...
|
||||
COMMIT;
|
||||
```
|
||||
|
||||
## Indexing Progress
|
||||
|
||||
Check [indexing progress](https://www.postgresql.org/docs/current/progress-reporting.html#CREATE-INDEX-PROGRESS-REPORTING) with Postgres 12+
|
||||
|
||||
```sql
|
||||
SELECT phase, round(100.0 * tuples_done / nullif(tuples_total, 0), 1) AS "%" FROM pg_stat_progress_create_index;
|
||||
SELECT phase, tuples_done, tuples_total FROM pg_stat_progress_create_index;
|
||||
```
|
||||
|
||||
The phases for IVFFlat are:
|
||||
The phases are:
|
||||
|
||||
1. `initializing`
|
||||
2. `performing k-means`
|
||||
3. `assigning tuples`
|
||||
2. `performing k-means` - IVFFlat only
|
||||
3. `assigning tuples` - IVFFlat only
|
||||
4. `loading tuples`
|
||||
|
||||
Note: `%` is only populated during the `loading tuples` phase
|
||||
Note: `tuples_done` and `tuples_total` are only populated during the `loading tuples` phase
|
||||
|
||||
## Filtering
|
||||
|
||||
@@ -348,7 +305,8 @@ CREATE INDEX ON items (category_id);
|
||||
Or a [partial index](https://www.postgresql.org/docs/current/indexes-partial.html) on the vector column for approximate search
|
||||
|
||||
```sql
|
||||
CREATE INDEX ON items USING hnsw (embedding vector_l2_ops) WHERE (category_id = 123);
|
||||
CREATE INDEX ON items USING ivfflat (embedding vector_l2_ops) WITH (lists = 100)
|
||||
WHERE (category_id = 123);
|
||||
```
|
||||
|
||||
Use [partitioning](https://www.postgresql.org/docs/current/ddl-partitioning.html) for approximate search on many different values of the `WHERE` columns
|
||||
@@ -359,15 +317,13 @@ CREATE TABLE items (embedding vector(3), category_id int) PARTITION BY LIST(cate
|
||||
|
||||
## Hybrid Search
|
||||
|
||||
Use together with Postgres [full-text search](https://www.postgresql.org/docs/current/textsearch-intro.html) for hybrid search.
|
||||
Use together with Postgres [full-text search](https://www.postgresql.org/docs/current/textsearch-intro.html) for hybrid search ([Python example](https://github.com/pgvector/pgvector-python/blob/master/examples/hybrid_search.py)).
|
||||
|
||||
```sql
|
||||
SELECT id, content FROM items, plainto_tsquery('hello search') query
|
||||
WHERE textsearch @@ query ORDER BY ts_rank_cd(textsearch, query) DESC LIMIT 5;
|
||||
```
|
||||
|
||||
You can use [Reciprocal Rank Fusion](https://github.com/pgvector/pgvector-python/blob/master/examples/hybrid_search_rrf.py) or a [cross-encoder](https://github.com/pgvector/pgvector-python/blob/master/examples/hybrid_search.py) to combine results.
|
||||
|
||||
## Performance
|
||||
|
||||
Use `EXPLAIN ANALYZE` to debug performance.
|
||||
@@ -404,21 +360,17 @@ Use pgvector from any language with a Postgres client. You can even generate and
|
||||
|
||||
Language | Libraries / Examples
|
||||
--- | ---
|
||||
C | [pgvector-c](https://github.com/pgvector/pgvector-c)
|
||||
C++ | [pgvector-cpp](https://github.com/pgvector/pgvector-cpp)
|
||||
C#, F#, Visual Basic | [pgvector-dotnet](https://github.com/pgvector/pgvector-dotnet)
|
||||
C# | [pgvector-dotnet](https://github.com/pgvector/pgvector-dotnet)
|
||||
Crystal | [pgvector-crystal](https://github.com/pgvector/pgvector-crystal)
|
||||
Dart | [pgvector-dart](https://github.com/pgvector/pgvector-dart)
|
||||
Elixir | [pgvector-elixir](https://github.com/pgvector/pgvector-elixir)
|
||||
Go | [pgvector-go](https://github.com/pgvector/pgvector-go)
|
||||
Haskell | [pgvector-haskell](https://github.com/pgvector/pgvector-haskell)
|
||||
Java, Kotlin, Groovy, Scala | [pgvector-java](https://github.com/pgvector/pgvector-java)
|
||||
JavaScript, TypeScript | [pgvector-node](https://github.com/pgvector/pgvector-node)
|
||||
Java, Scala | [pgvector-java](https://github.com/pgvector/pgvector-java)
|
||||
Julia | [pgvector-julia](https://github.com/pgvector/pgvector-julia)
|
||||
Lisp | [pgvector-lisp](https://github.com/pgvector/pgvector-lisp)
|
||||
Lua | [pgvector-lua](https://github.com/pgvector/pgvector-lua)
|
||||
Nim | [pgvector-nim](https://github.com/pgvector/pgvector-nim)
|
||||
OCaml | [pgvector-ocaml](https://github.com/pgvector/pgvector-ocaml)
|
||||
Node.js | [pgvector-node](https://github.com/pgvector/pgvector-node)
|
||||
Perl | [pgvector-perl](https://github.com/pgvector/pgvector-perl)
|
||||
PHP | [pgvector-php](https://github.com/pgvector/pgvector-php)
|
||||
Python | [pgvector-python](https://github.com/pgvector/pgvector-python)
|
||||
@@ -426,7 +378,6 @@ R | [pgvector-r](https://github.com/pgvector/pgvector-r)
|
||||
Ruby | [pgvector-ruby](https://github.com/pgvector/pgvector-ruby), [Neighbor](https://github.com/ankane/neighbor)
|
||||
Rust | [pgvector-rust](https://github.com/pgvector/pgvector-rust)
|
||||
Swift | [pgvector-swift](https://github.com/pgvector/pgvector-swift)
|
||||
Zig | [pgvector-zig](https://github.com/pgvector/pgvector-zig)
|
||||
|
||||
## Frequently Asked Questions
|
||||
|
||||
@@ -442,63 +393,6 @@ Yes, pgvector uses the write-ahead log (WAL), which allows for replication and p
|
||||
|
||||
You’ll need to use [dimensionality reduction](https://en.wikipedia.org/wiki/Dimensionality_reduction) at the moment.
|
||||
|
||||
#### Can I store vectors with different dimensions in the same column?
|
||||
|
||||
You can use `vector` as the type (instead of `vector(3)`).
|
||||
|
||||
```sql
|
||||
CREATE TABLE embeddings (model_id bigint, item_id bigint, embedding vector, PRIMARY KEY (model_id, item_id));
|
||||
```
|
||||
|
||||
However, you can only create indexes on rows with the same number of dimensions (using [expression](https://www.postgresql.org/docs/current/indexes-expressional.html) and [partial](https://www.postgresql.org/docs/current/indexes-partial.html) indexing):
|
||||
|
||||
```sql
|
||||
CREATE INDEX ON embeddings USING hnsw ((embedding::vector(3)) vector_l2_ops) WHERE (model_id = 123);
|
||||
```
|
||||
|
||||
and query with:
|
||||
|
||||
```sql
|
||||
SELECT * FROM embeddings WHERE model_id = 123 ORDER BY embedding::vector(3) <-> '[3,1,2]' LIMIT 5;
|
||||
```
|
||||
|
||||
#### Can I store vectors with more precision?
|
||||
|
||||
You can use the `double precision[]` or `numeric[]` type to store vectors with more precision.
|
||||
|
||||
```sql
|
||||
CREATE TABLE items (id bigserial PRIMARY KEY, embedding double precision[]);
|
||||
|
||||
-- use {} instead of [] for Postgres arrays
|
||||
INSERT INTO items (embedding) VALUES ('{1,2,3}'), ('{4,5,6}');
|
||||
```
|
||||
|
||||
Optionally, add a [check constraint](https://www.postgresql.org/docs/current/ddl-constraints.html) to ensure data can be converted to the `vector` type and has the expected dimensions.
|
||||
|
||||
```sql
|
||||
ALTER TABLE items ADD CHECK (vector_dims(embedding::vector) = 3);
|
||||
```
|
||||
|
||||
Use [expression indexing](https://www.postgresql.org/docs/current/indexes-expressional.html) to index (at a lower precision):
|
||||
|
||||
```sql
|
||||
CREATE INDEX ON items USING hnsw ((embedding::vector(3)) vector_l2_ops);
|
||||
```
|
||||
|
||||
and query with:
|
||||
|
||||
```sql
|
||||
SELECT * FROM items ORDER BY embedding::vector(3) <-> '[3,1,2]' LIMIT 5;
|
||||
```
|
||||
|
||||
#### 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:
|
||||
|
||||
```sql
|
||||
SELECT pg_size_pretty(pg_relation_size('index_name'));
|
||||
```
|
||||
|
||||
## Troubleshooting
|
||||
|
||||
#### Why isn’t a query using an index?
|
||||
@@ -512,8 +406,6 @@ SELECT ...
|
||||
COMMIT;
|
||||
```
|
||||
|
||||
Also, if the table is small, a table scan may be faster.
|
||||
|
||||
#### Why isn’t a query using a parallel table scan?
|
||||
|
||||
The planner doesn’t consider [out-of-line storage](https://www.postgresql.org/docs/current/storage-toast.html) in cost estimates, which can make a serial scan look cheaper. You can reduce the cost of a parallel scan for a query with:
|
||||
@@ -582,7 +474,7 @@ sum(vector) → vector | sum | 0.5.0
|
||||
If your machine has multiple Postgres installations, specify the path to [pg_config](https://www.postgresql.org/docs/current/app-pgconfig.html) with:
|
||||
|
||||
```sh
|
||||
export PG_CONFIG=/Library/PostgreSQL/16/bin/pg_config
|
||||
export PG_CONFIG=/Applications/Postgres.app/Contents/Versions/latest/bin/pg_config
|
||||
```
|
||||
|
||||
Then re-run the installation instructions (run `make clean` before `make` if needed). If `sudo` is needed for `make install`, use:
|
||||
@@ -591,14 +483,6 @@ Then re-run the installation instructions (run `make clean` before `make` if nee
|
||||
sudo --preserve-env=PG_CONFIG make install
|
||||
```
|
||||
|
||||
A few common paths on Mac are:
|
||||
|
||||
- EDB installer - `/Library/PostgreSQL/16/bin/pg_config`
|
||||
- Homebrew (arm64) - `/opt/homebrew/opt/postgresql@16/bin/pg_config`
|
||||
- Homebrew (x86-64) - `/usr/local/opt/postgresql@16/bin/pg_config`
|
||||
|
||||
Note: Replace `16` with your Postgres server version
|
||||
|
||||
### Missing Header
|
||||
|
||||
If compilation fails with `fatal error: postgres.h: No such file or directory`, make sure Postgres development files are installed on the server.
|
||||
@@ -606,14 +490,10 @@ If compilation fails with `fatal error: postgres.h: No such file or directory`,
|
||||
For Ubuntu and Debian, use:
|
||||
|
||||
```sh
|
||||
sudo apt install postgresql-server-dev-16
|
||||
sudo apt install postgresql-server-dev-15
|
||||
```
|
||||
|
||||
Note: Replace `16` with your Postgres server version
|
||||
|
||||
### Missing SDK
|
||||
|
||||
If compilation fails and the output includes `warning: no such sysroot directory` on Mac, reinstall Xcode Command Line Tools.
|
||||
Note: Replace `15` with your Postgres server version
|
||||
|
||||
### Windows
|
||||
|
||||
@@ -628,7 +508,7 @@ Note: The exact path will vary depending on your Visual Studio version and editi
|
||||
Then use `nmake` to build:
|
||||
|
||||
```cmd
|
||||
set "PGROOT=C:\Program Files\PostgreSQL\16"
|
||||
set "PGROOT=C:\Program Files\PostgreSQL\15"
|
||||
git clone --branch v0.5.1 https://github.com/pgvector/pgvector.git
|
||||
cd pgvector
|
||||
nmake /F Makefile.win
|
||||
@@ -678,22 +558,22 @@ pgxn install vector
|
||||
Debian and Ubuntu packages are available from the [PostgreSQL APT Repository](https://wiki.postgresql.org/wiki/Apt). Follow the [setup instructions](https://wiki.postgresql.org/wiki/Apt#Quickstart) and run:
|
||||
|
||||
```sh
|
||||
sudo apt install postgresql-16-pgvector
|
||||
sudo apt install postgresql-15-pgvector
|
||||
```
|
||||
|
||||
Note: Replace `16` with your Postgres server version
|
||||
Note: Replace `15` with your Postgres server version
|
||||
|
||||
### Yum
|
||||
|
||||
RPM packages are available from the [PostgreSQL Yum Repository](https://yum.postgresql.org/). Follow the [setup instructions](https://www.postgresql.org/download/linux/redhat/) for your distribution and run:
|
||||
|
||||
```sh
|
||||
sudo yum install pgvector_16
|
||||
sudo yum install pgvector_15
|
||||
# or
|
||||
sudo dnf install pgvector_16
|
||||
sudo dnf install pgvector_15
|
||||
```
|
||||
|
||||
Note: Replace `16` with your Postgres server version
|
||||
Note: Replace `15` with your Postgres server version
|
||||
|
||||
### conda-forge
|
||||
|
||||
@@ -715,7 +595,7 @@ pgvector is available on [these providers](https://github.com/pgvector/pgvector/
|
||||
|
||||
## Upgrading
|
||||
|
||||
[Install](#installation) the latest version (use the same method as the original installation). Then in each database you want to upgrade, run:
|
||||
Install the latest version. Then in each database you want to upgrade, run:
|
||||
|
||||
```sql
|
||||
ALTER EXTENSION vector UPDATE;
|
||||
@@ -803,25 +683,7 @@ make prove_installcheck PROVE_TESTS=test/t/001_wal.pl # TAP test
|
||||
To enable benchmarking:
|
||||
|
||||
```sh
|
||||
make clean && PG_CFLAGS="-DIVFFLAT_BENCH" make && make install
|
||||
```
|
||||
|
||||
To show memory usage:
|
||||
|
||||
```sh
|
||||
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
|
||||
make clean && PG_CFLAGS="-DIVFFLAT_KMEANS_DEBUG" make && make install
|
||||
make clean && PG_CFLAGS=-DIVFFLAT_BENCH make && make install
|
||||
```
|
||||
|
||||
Resources for contributors
|
||||
|
||||
20
src/hnsw.c
20
src/hnsw.c
@@ -14,7 +14,6 @@
|
||||
#endif
|
||||
|
||||
int hnsw_ef_search;
|
||||
bool hnsw_enable_parallel_build;
|
||||
static relopt_kind hnsw_relopt_kind;
|
||||
|
||||
/*
|
||||
@@ -40,11 +39,6 @@ HnswInit(void)
|
||||
DefineCustomIntVariable("hnsw.ef_search", "Sets the size of the dynamic candidate list for search",
|
||||
"Valid range is 1..1000.", &hnsw_ef_search,
|
||||
HNSW_DEFAULT_EF_SEARCH, HNSW_MIN_EF_SEARCH, HNSW_MAX_EF_SEARCH, PGC_USERSET, 0, NULL, NULL, NULL);
|
||||
|
||||
/* Behind a variable for now since can be slower than building in memory */
|
||||
DefineCustomBoolVariable("hnsw.enable_parallel_build", "Enables or disables building indexes in parallel",
|
||||
NULL, &hnsw_enable_parallel_build,
|
||||
false, PGC_USERSET, 0, NULL, NULL, NULL);
|
||||
}
|
||||
|
||||
/*
|
||||
@@ -94,20 +88,6 @@ hnswcostestimate(PlannerInfo *root, IndexPath *path, double loop_count,
|
||||
return;
|
||||
}
|
||||
|
||||
/*
|
||||
* Do not use index if no limit or limit + offset > ef_search unless
|
||||
* enable_seqscan = off
|
||||
*/
|
||||
if (root->limit_tuples < 0 || root->limit_tuples > hnsw_ef_search)
|
||||
{
|
||||
*indexStartupCost = 1.0e10 - 1;
|
||||
*indexTotalCost = 1.0e10 - 1;
|
||||
*indexSelectivity = 0;
|
||||
*indexCorrelation = 0;
|
||||
*indexPages = 0;
|
||||
return;
|
||||
}
|
||||
|
||||
MemSet(&costs, 0, sizeof(costs));
|
||||
|
||||
index = index_open(path->indexinfo->indexoid, NoLock);
|
||||
|
||||
131
src/hnsw.h
131
src/hnsw.h
@@ -4,9 +4,7 @@
|
||||
#include "postgres.h"
|
||||
|
||||
#include "access/generic_xlog.h"
|
||||
#include "access/parallel.h"
|
||||
#include "access/reloptions.h"
|
||||
#include "lib/ilist.h"
|
||||
#include "nodes/execnodes.h"
|
||||
#include "port.h" /* for random() */
|
||||
#include "utils/sampling.h"
|
||||
@@ -16,10 +14,6 @@
|
||||
#error "Requires PostgreSQL 11+"
|
||||
#endif
|
||||
|
||||
#if PG_VERSION_NUM < 120000
|
||||
#include "access/relscan.h"
|
||||
#endif
|
||||
|
||||
#define HNSW_MAX_DIM 2000
|
||||
|
||||
/* Support functions */
|
||||
@@ -65,7 +59,7 @@
|
||||
|
||||
#define HNSW_MAX_SIZE (BLCKSZ - MAXALIGN(SizeOfPageHeaderData) - MAXALIGN(sizeof(HnswPageOpaqueData)) - sizeof(ItemIdData))
|
||||
|
||||
#define HNSW_ELEMENT_TUPLE_SIZE(size) MAXALIGN(offsetof(HnswElementTupleData, data) + (size))
|
||||
#define HNSW_ELEMENT_TUPLE_SIZE(_datum) MAXALIGN(offsetof(HnswElementTupleData, value) + VARSIZE_ANY(_datum))
|
||||
#define HNSW_NEIGHBOR_TUPLE_SIZE(level, m) MAXALIGN(offsetof(HnswNeighborTupleData, indextids) + ((level) + 2) * (m) * sizeof(ItemPointerData))
|
||||
|
||||
#define HnswPageGetOpaque(page) ((HnswPageOpaque) PageGetSpecialPointer(page))
|
||||
@@ -73,10 +67,8 @@
|
||||
|
||||
#if PG_VERSION_NUM >= 150000
|
||||
#define RandomDouble() pg_prng_double(&pg_global_prng_state)
|
||||
#define SeedRandom(seed) pg_prng_seed(&pg_global_prng_state, seed)
|
||||
#else
|
||||
#define RandomDouble() (((double) random()) / MAX_RANDOM_VALUE)
|
||||
#define SeedRandom(seed) srandom(seed)
|
||||
#endif
|
||||
|
||||
#if PG_VERSION_NUM < 130000
|
||||
@@ -94,23 +86,20 @@
|
||||
#define HnswGetMl(m) (1 / log(m))
|
||||
|
||||
/* Ensure fits on page and in uint8 */
|
||||
#define HnswGetMaxLevel(m) Min(((BLCKSZ - MAXALIGN(SizeOfPageHeaderData) - MAXALIGN(sizeof(HnswPageOpaqueData)) - offsetof(HnswNeighborTupleData, indextids) - sizeof(ItemIdData)) / (sizeof(ItemPointerData)) / (m)) - 2, 255)
|
||||
|
||||
#define HnswGetNeighbors(element, lc) (AssertMacro((element)->level >= (lc)), &(element)->neighbors[lc])
|
||||
#define HnswGetMaxLevel(m) Min(((BLCKSZ - MAXALIGN(SizeOfPageHeaderData) - MAXALIGN(sizeof(HnswPageOpaqueData)) - offsetof(HnswNeighborTupleData, indextids) - sizeof(ItemIdData)) / (sizeof(ItemPointerData)) / m) - 2, 255)
|
||||
|
||||
/* Variables */
|
||||
extern int hnsw_ef_search;
|
||||
extern bool hnsw_enable_parallel_build;
|
||||
|
||||
typedef struct HnswNeighborArray HnswNeighborArray;
|
||||
|
||||
typedef struct HnswElementData
|
||||
{
|
||||
slist_node next;
|
||||
ItemPointerData heaptids[HNSW_HEAPTIDS];
|
||||
uint8 heaptidsLength;
|
||||
List *heaptids;
|
||||
uint8 level;
|
||||
uint8 deleted;
|
||||
uint32 hash;
|
||||
struct HnswNeighborArray *neighbors;
|
||||
bool loaded;
|
||||
HnswNeighborArray *neighbors;
|
||||
BlockNumber blkno;
|
||||
OffsetNumber offno;
|
||||
OffsetNumber neighborOffno;
|
||||
@@ -148,59 +137,6 @@ typedef struct HnswOptions
|
||||
int efConstruction; /* size of dynamic candidate list */
|
||||
} HnswOptions;
|
||||
|
||||
typedef struct HnswGraph
|
||||
{
|
||||
slist_head elements;
|
||||
HnswElement entryPoint;
|
||||
long memoryUsed;
|
||||
long memoryTotal;
|
||||
bool flushed;
|
||||
double indtuples;
|
||||
} HnswGraph;
|
||||
|
||||
typedef struct HnswSpool
|
||||
{
|
||||
Relation heap;
|
||||
Relation index;
|
||||
} HnswSpool;
|
||||
|
||||
typedef struct HnswShared
|
||||
{
|
||||
/* Immutable state */
|
||||
Oid heaprelid;
|
||||
Oid indexrelid;
|
||||
bool isconcurrent;
|
||||
int scantuplesortstates;
|
||||
|
||||
/* Worker progress */
|
||||
ConditionVariable workersdonecv;
|
||||
|
||||
/* Mutex for mutable state */
|
||||
slock_t mutex;
|
||||
|
||||
/* Mutable state */
|
||||
int nparticipantsdone;
|
||||
double reltuples;
|
||||
HnswGraph graphData;
|
||||
|
||||
#if PG_VERSION_NUM < 120000
|
||||
ParallelHeapScanDescData heapdesc; /* must come last */
|
||||
#endif
|
||||
} HnswShared;
|
||||
|
||||
#if PG_VERSION_NUM >= 120000
|
||||
#define ParallelTableScanFromHnswShared(shared) \
|
||||
(ParallelTableScanDesc) ((char *) (shared) + BUFFERALIGN(sizeof(HnswShared)))
|
||||
#endif
|
||||
|
||||
typedef struct HnswLeader
|
||||
{
|
||||
ParallelContext *pcxt;
|
||||
int nparticipanttuplesorts;
|
||||
HnswShared *hnswshared;
|
||||
Snapshot snapshot;
|
||||
} HnswLeader;
|
||||
|
||||
typedef struct HnswBuildState
|
||||
{
|
||||
/* Info */
|
||||
@@ -224,19 +160,16 @@ typedef struct HnswBuildState
|
||||
Oid collation;
|
||||
|
||||
/* Variables */
|
||||
HnswGraph graphData;
|
||||
HnswGraph *graph;
|
||||
List *elements;
|
||||
HnswElement entryPoint;
|
||||
double ml;
|
||||
int maxLevel;
|
||||
double maxInMemoryElements;
|
||||
bool flushed;
|
||||
Vector *normvec;
|
||||
|
||||
/* Memory */
|
||||
MemoryContext graphCtx;
|
||||
MemoryContext tmpCtx;
|
||||
|
||||
/* Parallel builds */
|
||||
HnswLeader *hnswleader;
|
||||
HnswShared *hnswshared;
|
||||
} HnswBuildState;
|
||||
|
||||
typedef struct HnswMetaPageData
|
||||
@@ -272,7 +205,7 @@ typedef struct HnswElementTupleData
|
||||
ItemPointerData heaptids[HNSW_HEAPTIDS];
|
||||
ItemPointerData neighbortid;
|
||||
uint16 unused2;
|
||||
Vector data;
|
||||
char value[FLEXIBLE_ARRAY_MEMBER];
|
||||
} HnswElementTupleData;
|
||||
|
||||
typedef HnswElementTupleData * HnswElementTuple;
|
||||
@@ -318,7 +251,7 @@ typedef struct HnswVacuumState
|
||||
Oid collation;
|
||||
|
||||
/* Variables */
|
||||
struct tidhash_hash *deleted;
|
||||
HTAB *deleted;
|
||||
BufferAccessStrategy bas;
|
||||
HnswNeighborTuple ntup;
|
||||
HnswElementData highestPoint;
|
||||
@@ -332,28 +265,31 @@ int HnswGetM(Relation index);
|
||||
int HnswGetEfConstruction(Relation index);
|
||||
FmgrInfo *HnswOptionalProcInfo(Relation index, uint16 procnum);
|
||||
bool HnswNormValue(FmgrInfo *procinfo, Oid collation, Datum *value, Vector * result);
|
||||
void HnswCommitBuffer(Buffer buf, GenericXLogState *state);
|
||||
Buffer HnswNewBuffer(Relation index, ForkNumber forkNum);
|
||||
void HnswInitPage(Buffer buf, Page page);
|
||||
void HnswInitRegisterPage(Relation index, Buffer *buf, Page *page, GenericXLogState **state);
|
||||
void HnswInit(void);
|
||||
List *HnswSearchLayer(Datum q, List *ep, int ef, int lc, Relation index, FmgrInfo *procinfo, Oid collation, int m, bool inserting, HnswElement skipElement);
|
||||
HnswElement HnswGetEntryPoint(Relation index);
|
||||
void HnswGetMetaPageInfo(Relation index, int *m, HnswElement * entryPoint);
|
||||
HnswElement HnswInitElement(ItemPointer tid, int m, double ml, int maxLevel);
|
||||
void HnswFreeElement(HnswElement element);
|
||||
HnswElement HnswInitElementFromBlock(BlockNumber blkno, OffsetNumber offno);
|
||||
void HnswInsertElement(HnswElement element, HnswElement entryPoint, Relation index, FmgrInfo *procinfo, Oid collation, int m, int efConstruction, bool existing);
|
||||
HnswElement HnswFindDuplicate(HnswElement e);
|
||||
HnswCandidate *HnswEntryCandidate(HnswElement em, Datum q, Relation rel, FmgrInfo *procinfo, Oid collation, bool loadVec);
|
||||
void HnswUpdateMetaPage(Relation index, int updateEntry, HnswElement entryPoint, BlockNumber insertPage, ForkNumber forkNum, bool building);
|
||||
void HnswUpdateMetaPage(Relation index, int updateEntry, HnswElement entryPoint, BlockNumber insertPage, ForkNumber forkNum);
|
||||
void HnswSetNeighborTuple(HnswNeighborTuple ntup, HnswElement e, int m);
|
||||
void HnswAddHeapTid(HnswElement element, ItemPointer heaptid);
|
||||
void HnswInitNeighbors(HnswElement element, int m);
|
||||
bool HnswInsertTuple(Relation index, Datum *values, bool *isnull, ItemPointer heap_tid, Relation heapRel, bool building);
|
||||
void HnswUpdateNeighborPages(Relation index, FmgrInfo *procinfo, Oid collation, HnswElement e, int m, bool checkExisting, bool building);
|
||||
bool HnswInsertTuple(Relation index, Datum *values, bool *isnull, ItemPointer heap_tid, Relation heapRel);
|
||||
void HnswUpdateNeighborPages(Relation index, FmgrInfo *procinfo, Oid collation, HnswElement e, int m, bool checkExisting);
|
||||
void HnswLoadElementFromTuple(HnswElement element, HnswElementTuple etup, bool loadHeaptids, bool loadVec);
|
||||
void HnswLoadElement(HnswElement element, float *distance, Datum *q, Relation index, FmgrInfo *procinfo, Oid collation, bool loadVec);
|
||||
void HnswSetElementTuple(HnswElementTuple etup, HnswElement element);
|
||||
void HnswUpdateConnection(HnswElement element, HnswCandidate * hc, int m, int lc, int *updateIdx, Relation index, FmgrInfo *procinfo, Oid collation);
|
||||
void HnswLoadNeighbors(HnswElement element, Relation index, int m);
|
||||
PGDLLEXPORT void HnswParallelBuildMain(dsm_segment *seg, shm_toc *toc);
|
||||
|
||||
/* Index access methods */
|
||||
IndexBuildResult *hnswbuild(Relation heap, Relation index, IndexInfo *indexInfo);
|
||||
@@ -371,31 +307,4 @@ void hnswrescan(IndexScanDesc scan, ScanKey keys, int nkeys, ScanKey orderbys,
|
||||
bool hnswgettuple(IndexScanDesc scan, ScanDirection dir);
|
||||
void hnswendscan(IndexScanDesc scan);
|
||||
|
||||
/* Hash tables */
|
||||
typedef struct TidHashEntry
|
||||
{
|
||||
ItemPointerData tid;
|
||||
char status;
|
||||
} TidHashEntry;
|
||||
|
||||
#define SH_PREFIX tidhash
|
||||
#define SH_ELEMENT_TYPE TidHashEntry
|
||||
#define SH_KEY_TYPE ItemPointerData
|
||||
#define SH_SCOPE extern
|
||||
#define SH_DECLARE
|
||||
#include "lib/simplehash.h"
|
||||
|
||||
typedef struct PointerHashEntry
|
||||
{
|
||||
uintptr_t ptr;
|
||||
char status;
|
||||
} PointerHashEntry;
|
||||
|
||||
#define SH_PREFIX pointerhash
|
||||
#define SH_ELEMENT_TYPE PointerHashEntry
|
||||
#define SH_KEY_TYPE uintptr_t
|
||||
#define SH_SCOPE extern
|
||||
#define SH_DECLARE
|
||||
#include "lib/simplehash.h"
|
||||
|
||||
#endif
|
||||
|
||||
754
src/hnswbuild.c
754
src/hnswbuild.c
File diff suppressed because it is too large
Load Diff
175
src/hnswinsert.c
175
src/hnswinsert.c
@@ -5,7 +5,6 @@
|
||||
#include "hnsw.h"
|
||||
#include "storage/bufmgr.h"
|
||||
#include "storage/lmgr.h"
|
||||
#include "utils/datum.h"
|
||||
#include "utils/memutils.h"
|
||||
|
||||
/*
|
||||
@@ -93,7 +92,7 @@ HnswFreeOffset(Relation index, Buffer buf, Page page, HnswElement element, Size
|
||||
* Add a new page
|
||||
*/
|
||||
static void
|
||||
HnswInsertAppendPage(Relation index, Buffer *nbuf, Page *npage, GenericXLogState *state, Page page, bool building)
|
||||
HnswInsertAppendPage(Relation index, Buffer *nbuf, Page *npage, GenericXLogState *state, Page page)
|
||||
{
|
||||
/* Add a new page */
|
||||
LockRelationForExtension(index, ExclusiveLock);
|
||||
@@ -101,11 +100,7 @@ HnswInsertAppendPage(Relation index, Buffer *nbuf, Page *npage, GenericXLogState
|
||||
UnlockRelationForExtension(index, ExclusiveLock);
|
||||
|
||||
/* Init new page */
|
||||
if (building)
|
||||
*npage = BufferGetPage(*nbuf);
|
||||
else
|
||||
*npage = GenericXLogRegisterBuffer(state, *nbuf, GENERIC_XLOG_FULL_IMAGE);
|
||||
|
||||
*npage = GenericXLogRegisterBuffer(state, *nbuf, GENERIC_XLOG_FULL_IMAGE);
|
||||
HnswInitPage(*nbuf, *npage);
|
||||
|
||||
/* Update previous buffer */
|
||||
@@ -116,7 +111,7 @@ HnswInsertAppendPage(Relation index, Buffer *nbuf, Page *npage, GenericXLogState
|
||||
* Add to element and neighbor pages
|
||||
*/
|
||||
static void
|
||||
WriteNewElementPages(Relation index, HnswElement e, int m, BlockNumber insertPage, BlockNumber *updatedInsertPage, bool building)
|
||||
WriteNewElementPages(Relation index, HnswElement e, int m, BlockNumber insertPage, BlockNumber *updatedInsertPage)
|
||||
{
|
||||
Buffer buf;
|
||||
Page page;
|
||||
@@ -136,7 +131,7 @@ WriteNewElementPages(Relation index, HnswElement e, int m, BlockNumber insertPag
|
||||
BlockNumber newInsertPage = InvalidBlockNumber;
|
||||
|
||||
/* Calculate sizes */
|
||||
etupSize = HNSW_ELEMENT_TUPLE_SIZE(VARSIZE_ANY(DatumGetPointer(e->value)));
|
||||
etupSize = HNSW_ELEMENT_TUPLE_SIZE(e->value);
|
||||
ntupSize = HNSW_NEIGHBOR_TUPLE_SIZE(e->level, m);
|
||||
combinedSize = etupSize + ntupSize + sizeof(ItemIdData);
|
||||
maxSize = HNSW_MAX_SIZE;
|
||||
@@ -156,16 +151,8 @@ WriteNewElementPages(Relation index, HnswElement e, int m, BlockNumber insertPag
|
||||
buf = ReadBuffer(index, currentPage);
|
||||
LockBuffer(buf, BUFFER_LOCK_EXCLUSIVE);
|
||||
|
||||
if (building)
|
||||
{
|
||||
state = NULL;
|
||||
page = BufferGetPage(buf);
|
||||
}
|
||||
else
|
||||
{
|
||||
state = GenericXLogStart(index);
|
||||
page = GenericXLogRegisterBuffer(state, buf, 0);
|
||||
}
|
||||
state = GenericXLogStart(index);
|
||||
page = GenericXLogRegisterBuffer(state, buf, 0);
|
||||
|
||||
/* Keep track of first page where element at level 0 can fit */
|
||||
if (!BlockNumberIsValid(newInsertPage) && PageGetFreeSpace(page) >= minCombinedSize)
|
||||
@@ -185,12 +172,7 @@ WriteNewElementPages(Relation index, HnswElement e, int m, BlockNumber insertPag
|
||||
if (HnswFreeOffset(index, buf, page, e, ntupSize, &nbuf, &npage, &freeOffno, &freeNeighborOffno, &newInsertPage))
|
||||
{
|
||||
if (nbuf != buf)
|
||||
{
|
||||
if (building)
|
||||
npage = BufferGetPage(nbuf);
|
||||
else
|
||||
npage = GenericXLogRegisterBuffer(state, nbuf, 0);
|
||||
}
|
||||
npage = GenericXLogRegisterBuffer(state, nbuf, 0);
|
||||
|
||||
break;
|
||||
}
|
||||
@@ -199,7 +181,7 @@ WriteNewElementPages(Relation index, HnswElement e, int m, BlockNumber insertPag
|
||||
/* Skip if both tuples can fit on the same page */
|
||||
if (combinedSize > maxSize && PageGetFreeSpace(page) >= etupSize && !BlockNumberIsValid(HnswPageGetOpaque(page)->nextblkno))
|
||||
{
|
||||
HnswInsertAppendPage(index, &nbuf, &npage, state, page, building);
|
||||
HnswInsertAppendPage(index, &nbuf, &npage, state, page);
|
||||
break;
|
||||
}
|
||||
|
||||
@@ -208,8 +190,7 @@ WriteNewElementPages(Relation index, HnswElement e, int m, BlockNumber insertPag
|
||||
if (BlockNumberIsValid(currentPage))
|
||||
{
|
||||
/* Move to next page */
|
||||
if (!building)
|
||||
GenericXLogAbort(state);
|
||||
GenericXLogAbort(state);
|
||||
UnlockReleaseBuffer(buf);
|
||||
}
|
||||
else
|
||||
@@ -217,33 +198,22 @@ WriteNewElementPages(Relation index, HnswElement e, int m, BlockNumber insertPag
|
||||
Buffer newbuf;
|
||||
Page newpage;
|
||||
|
||||
HnswInsertAppendPage(index, &newbuf, &newpage, state, page, building);
|
||||
HnswInsertAppendPage(index, &newbuf, &newpage, state, page);
|
||||
|
||||
/* Commit */
|
||||
if (building)
|
||||
MarkBufferDirty(buf);
|
||||
else
|
||||
GenericXLogFinish(state);
|
||||
GenericXLogFinish(state);
|
||||
|
||||
/* Unlock previous buffer */
|
||||
UnlockReleaseBuffer(buf);
|
||||
|
||||
/* Prepare new buffer */
|
||||
state = GenericXLogStart(index);
|
||||
buf = newbuf;
|
||||
if (building)
|
||||
{
|
||||
state = NULL;
|
||||
page = BufferGetPage(buf);
|
||||
}
|
||||
else
|
||||
{
|
||||
state = GenericXLogStart(index);
|
||||
page = GenericXLogRegisterBuffer(state, buf, 0);
|
||||
}
|
||||
page = GenericXLogRegisterBuffer(state, buf, 0);
|
||||
|
||||
/* Create new page for neighbors if needed */
|
||||
if (PageGetFreeSpace(page) < combinedSize)
|
||||
HnswInsertAppendPage(index, &nbuf, &npage, state, page, building);
|
||||
HnswInsertAppendPage(index, &nbuf, &npage, state, page);
|
||||
else
|
||||
{
|
||||
nbuf = buf;
|
||||
@@ -297,14 +267,7 @@ WriteNewElementPages(Relation index, HnswElement e, int m, BlockNumber insertPag
|
||||
}
|
||||
|
||||
/* Commit */
|
||||
if (building)
|
||||
{
|
||||
MarkBufferDirty(buf);
|
||||
if (nbuf != buf)
|
||||
MarkBufferDirty(nbuf);
|
||||
}
|
||||
else
|
||||
GenericXLogFinish(state);
|
||||
GenericXLogFinish(state);
|
||||
UnlockReleaseBuffer(buf);
|
||||
if (nbuf != buf)
|
||||
UnlockReleaseBuffer(nbuf);
|
||||
@@ -338,12 +301,12 @@ ConnectionExists(HnswElement e, HnswNeighborTuple ntup, int startIdx, int lm)
|
||||
* Update neighbors
|
||||
*/
|
||||
void
|
||||
HnswUpdateNeighborPages(Relation index, FmgrInfo *procinfo, Oid collation, HnswElement e, int m, bool checkExisting, bool building)
|
||||
HnswUpdateNeighborPages(Relation index, FmgrInfo *procinfo, Oid collation, HnswElement e, int m, bool checkExisting)
|
||||
{
|
||||
for (int lc = e->level; lc >= 0; lc--)
|
||||
{
|
||||
int lm = HnswGetLayerM(m, lc);
|
||||
HnswNeighborArray *neighbors = HnswGetNeighbors(e, lc);
|
||||
HnswNeighborArray *neighbors = &e->neighbors[lc];
|
||||
|
||||
for (int i = 0; i < neighbors->length; i++)
|
||||
{
|
||||
@@ -379,16 +342,8 @@ HnswUpdateNeighborPages(Relation index, FmgrInfo *procinfo, Oid collation, HnswE
|
||||
/* Register page */
|
||||
buf = ReadBuffer(index, hc->element->neighborPage);
|
||||
LockBuffer(buf, BUFFER_LOCK_EXCLUSIVE);
|
||||
if (building)
|
||||
{
|
||||
state = NULL;
|
||||
page = BufferGetPage(buf);
|
||||
}
|
||||
else
|
||||
{
|
||||
state = GenericXLogStart(index);
|
||||
page = GenericXLogRegisterBuffer(state, buf, 0);
|
||||
}
|
||||
state = GenericXLogStart(index);
|
||||
page = GenericXLogRegisterBuffer(state, buf, 0);
|
||||
|
||||
/* Get tuple */
|
||||
itemid = PageGetItemId(page, offno);
|
||||
@@ -430,12 +385,9 @@ HnswUpdateNeighborPages(Relation index, FmgrInfo *procinfo, Oid collation, HnswE
|
||||
elog(ERROR, "failed to add index item to \"%s\"", RelationGetRelationName(index));
|
||||
|
||||
/* Commit */
|
||||
if (building)
|
||||
MarkBufferDirty(buf);
|
||||
else
|
||||
GenericXLogFinish(state);
|
||||
GenericXLogFinish(state);
|
||||
}
|
||||
else if (!building)
|
||||
else
|
||||
GenericXLogAbort(state);
|
||||
|
||||
UnlockReleaseBuffer(buf);
|
||||
@@ -447,34 +399,23 @@ HnswUpdateNeighborPages(Relation index, FmgrInfo *procinfo, Oid collation, HnswE
|
||||
* Add a heap TID to an existing element
|
||||
*/
|
||||
static bool
|
||||
HnswAddDuplicate(Relation index, HnswElement element, HnswElement dup, bool building)
|
||||
HnswAddDuplicate(Relation index, HnswElement element, HnswElement dup)
|
||||
{
|
||||
Buffer buf;
|
||||
Page page;
|
||||
GenericXLogState *state;
|
||||
ItemId itemid;
|
||||
Size etupSize = HNSW_ELEMENT_TUPLE_SIZE(dup->value);
|
||||
HnswElementTuple etup;
|
||||
Size etupSize;
|
||||
int i;
|
||||
|
||||
/* Read page */
|
||||
buf = ReadBuffer(index, dup->blkno);
|
||||
LockBuffer(buf, BUFFER_LOCK_EXCLUSIVE);
|
||||
if (building)
|
||||
{
|
||||
state = NULL;
|
||||
page = BufferGetPage(buf);
|
||||
}
|
||||
else
|
||||
{
|
||||
state = GenericXLogStart(index);
|
||||
page = GenericXLogRegisterBuffer(state, buf, 0);
|
||||
}
|
||||
state = GenericXLogStart(index);
|
||||
page = GenericXLogRegisterBuffer(state, buf, 0);
|
||||
|
||||
/* Find space */
|
||||
itemid = PageGetItemId(page, dup->offno);
|
||||
etup = (HnswElementTuple) PageGetItem(page, itemid);
|
||||
etupSize = ItemIdGetLength(itemid);
|
||||
etup = (HnswElementTuple) PageGetItem(page, PageGetItemId(page, dup->offno));
|
||||
for (i = 0; i < HNSW_HEAPTIDS; i++)
|
||||
{
|
||||
if (!ItemPointerIsValid(&etup->heaptids[i]))
|
||||
@@ -484,84 +425,60 @@ HnswAddDuplicate(Relation index, HnswElement element, HnswElement dup, bool buil
|
||||
/* Either being deleted or we lost our chance to another backend */
|
||||
if (i == 0 || i == HNSW_HEAPTIDS)
|
||||
{
|
||||
if (!building)
|
||||
GenericXLogAbort(state);
|
||||
GenericXLogAbort(state);
|
||||
UnlockReleaseBuffer(buf);
|
||||
return false;
|
||||
}
|
||||
|
||||
/* Add heap TID */
|
||||
etup->heaptids[i] = element->heaptids[0];
|
||||
etup->heaptids[i] = *((ItemPointer) linitial(element->heaptids));
|
||||
|
||||
/* Overwrite tuple */
|
||||
if (!PageIndexTupleOverwrite(page, dup->offno, (Item) etup, etupSize))
|
||||
elog(ERROR, "failed to add index item to \"%s\"", RelationGetRelationName(index));
|
||||
|
||||
/* Commit */
|
||||
if (building)
|
||||
MarkBufferDirty(buf);
|
||||
else
|
||||
GenericXLogFinish(state);
|
||||
GenericXLogFinish(state);
|
||||
UnlockReleaseBuffer(buf);
|
||||
|
||||
return true;
|
||||
}
|
||||
|
||||
/*
|
||||
* Find duplicate element
|
||||
*/
|
||||
static bool
|
||||
HnswFindDuplicate(Relation index, HnswElement element, bool building)
|
||||
{
|
||||
HnswNeighborArray *neighbors = HnswGetNeighbors(element, 0);
|
||||
|
||||
for (int i = 0; i < neighbors->length; i++)
|
||||
{
|
||||
HnswCandidate *neighbor = &neighbors->items[i];
|
||||
|
||||
/* Exit early since ordered by distance */
|
||||
if (!datumIsEqual(element->value, neighbor->element->value, false, -1))
|
||||
return false;
|
||||
|
||||
if (HnswAddDuplicate(index, element, neighbor->element, building))
|
||||
return true;
|
||||
}
|
||||
|
||||
return false;
|
||||
}
|
||||
|
||||
/*
|
||||
* Write changes to disk
|
||||
*/
|
||||
static void
|
||||
WriteElement(Relation index, FmgrInfo *procinfo, Oid collation, HnswElement element, int m, int efConstruction, HnswElement entryPoint, bool building)
|
||||
WriteElement(Relation index, FmgrInfo *procinfo, Oid collation, HnswElement element, int m, int efConstruction, HnswElement dup, HnswElement entryPoint)
|
||||
{
|
||||
BlockNumber newInsertPage = InvalidBlockNumber;
|
||||
|
||||
/* Look for duplicate */
|
||||
if (HnswFindDuplicate(index, element, building))
|
||||
return;
|
||||
/* Try to add to existing page */
|
||||
if (dup != NULL)
|
||||
{
|
||||
if (HnswAddDuplicate(index, element, dup))
|
||||
return;
|
||||
}
|
||||
|
||||
/* Write element and neighbor tuples */
|
||||
WriteNewElementPages(index, element, m, GetInsertPage(index), &newInsertPage, building);
|
||||
WriteNewElementPages(index, element, m, GetInsertPage(index), &newInsertPage);
|
||||
|
||||
/* Update insert page if needed */
|
||||
if (BlockNumberIsValid(newInsertPage))
|
||||
HnswUpdateMetaPage(index, 0, NULL, newInsertPage, MAIN_FORKNUM, building);
|
||||
HnswUpdateMetaPage(index, 0, NULL, newInsertPage, MAIN_FORKNUM);
|
||||
|
||||
/* Update neighbors */
|
||||
HnswUpdateNeighborPages(index, procinfo, collation, element, m, false, building);
|
||||
HnswUpdateNeighborPages(index, procinfo, collation, element, m, false);
|
||||
|
||||
/* Update entry point if needed */
|
||||
/* Update metapage if needed */
|
||||
if (entryPoint == NULL || element->level > entryPoint->level)
|
||||
HnswUpdateMetaPage(index, HNSW_UPDATE_ENTRY_GREATER, element, InvalidBlockNumber, MAIN_FORKNUM, building);
|
||||
HnswUpdateMetaPage(index, HNSW_UPDATE_ENTRY_GREATER, element, InvalidBlockNumber, MAIN_FORKNUM);
|
||||
}
|
||||
|
||||
/*
|
||||
* Insert a tuple into the index
|
||||
*/
|
||||
bool
|
||||
HnswInsertTuple(Relation index, Datum *values, bool *isnull, ItemPointer heap_tid, Relation heapRel, bool building)
|
||||
HnswInsertTuple(Relation index, Datum *values, bool *isnull, ItemPointer heap_tid, Relation heapRel)
|
||||
{
|
||||
Datum value;
|
||||
FmgrInfo *normprocinfo;
|
||||
@@ -571,6 +488,7 @@ HnswInsertTuple(Relation index, Datum *values, bool *isnull, ItemPointer heap_ti
|
||||
int efConstruction = HnswGetEfConstruction(index);
|
||||
FmgrInfo *procinfo = index_getprocinfo(index, 1, HNSW_DISTANCE_PROC);
|
||||
Oid collation = index->rd_indcollation[0];
|
||||
HnswElement dup;
|
||||
LOCKMODE lockmode = ShareLock;
|
||||
|
||||
/* Detoast once for all calls */
|
||||
@@ -615,8 +533,11 @@ HnswInsertTuple(Relation index, Datum *values, bool *isnull, ItemPointer heap_ti
|
||||
/* Insert element in graph */
|
||||
HnswInsertElement(element, entryPoint, index, procinfo, collation, m, efConstruction, false);
|
||||
|
||||
/* Look for duplicate */
|
||||
dup = HnswFindDuplicate(element);
|
||||
|
||||
/* Write to disk */
|
||||
WriteElement(index, procinfo, collation, element, m, efConstruction, entryPoint, building);
|
||||
WriteElement(index, procinfo, collation, element, m, efConstruction, dup, entryPoint);
|
||||
|
||||
/* Release lock */
|
||||
UnlockPage(index, HNSW_UPDATE_LOCK, lockmode);
|
||||
@@ -650,7 +571,7 @@ hnswinsert(Relation index, Datum *values, bool *isnull, ItemPointer heap_tid,
|
||||
oldCtx = MemoryContextSwitchTo(insertCtx);
|
||||
|
||||
/* Insert tuple */
|
||||
HnswInsertTuple(index, values, isnull, heap_tid, heap, false);
|
||||
HnswInsertTuple(index, values, isnull, heap_tid, heap);
|
||||
|
||||
/* Delete memory context */
|
||||
MemoryContextSwitchTo(oldCtx);
|
||||
|
||||
@@ -188,13 +188,15 @@ hnswgettuple(IndexScanDesc scan, ScanDirection dir)
|
||||
ItemPointer heaptid;
|
||||
|
||||
/* Move to next element if no valid heap TIDs */
|
||||
if (hc->element->heaptidsLength == 0)
|
||||
if (list_length(hc->element->heaptids) == 0)
|
||||
{
|
||||
so->w = list_delete_last(so->w);
|
||||
continue;
|
||||
}
|
||||
|
||||
heaptid = &hc->element->heaptids[--hc->element->heaptidsLength];
|
||||
heaptid = llast(hc->element->heaptids);
|
||||
|
||||
hc->element->heaptids = list_delete_last(hc->element->heaptids);
|
||||
|
||||
MemoryContextSwitchTo(oldCtx);
|
||||
|
||||
|
||||
281
src/hnswutils.c
281
src/hnswutils.c
@@ -7,89 +7,6 @@
|
||||
#include "utils/datum.h"
|
||||
#include "vector.h"
|
||||
|
||||
#if PG_VERSION_NUM >= 130000
|
||||
#include "common/hashfn.h"
|
||||
#else
|
||||
#include "utils/hashutils.h"
|
||||
#endif
|
||||
|
||||
#if PG_VERSION_NUM < 170000
|
||||
static inline uint64
|
||||
murmurhash64(uint64 data)
|
||||
{
|
||||
uint64 h = data;
|
||||
|
||||
h ^= h >> 33;
|
||||
h *= 0xff51afd7ed558ccd;
|
||||
h ^= h >> 33;
|
||||
h *= 0xc4ceb9fe1a85ec53;
|
||||
h ^= h >> 33;
|
||||
|
||||
return h;
|
||||
}
|
||||
#endif
|
||||
|
||||
/* TID hash table */
|
||||
static uint32
|
||||
hash_tid(ItemPointerData tid)
|
||||
{
|
||||
union
|
||||
{
|
||||
uint64 i;
|
||||
ItemPointerData tid;
|
||||
} x;
|
||||
|
||||
/* Initialize unused bytes */
|
||||
x.i = 0;
|
||||
x.tid = tid;
|
||||
|
||||
return murmurhash64(x.i);
|
||||
}
|
||||
|
||||
#define SH_PREFIX tidhash
|
||||
#define SH_ELEMENT_TYPE TidHashEntry
|
||||
#define SH_KEY_TYPE ItemPointerData
|
||||
#define SH_KEY tid
|
||||
#define SH_HASH_KEY(tb, key) hash_tid(key)
|
||||
#define SH_EQUAL(tb, a, b) ItemPointerEquals(&a, &b)
|
||||
#define SH_SCOPE extern
|
||||
#define SH_DEFINE
|
||||
#include "lib/simplehash.h"
|
||||
|
||||
/* Needed to include simplehash.h twice */
|
||||
#if PG_VERSION_NUM < 120000
|
||||
#undef SH_EQUAL
|
||||
#define sh_log2 pointerhash_sh_log2
|
||||
#define sh_pow2 pointerhash_sh_pow2
|
||||
#endif
|
||||
|
||||
/* Pointer hash table */
|
||||
static uint32
|
||||
hash_pointer(uintptr_t ptr)
|
||||
{
|
||||
#if SIZEOF_VOID_P == 8
|
||||
return murmurhash64((uint64) ptr);
|
||||
#else
|
||||
return murmurhash32((uint32) ptr);
|
||||
#endif
|
||||
}
|
||||
|
||||
#define SH_PREFIX pointerhash
|
||||
#define SH_ELEMENT_TYPE PointerHashEntry
|
||||
#define SH_KEY_TYPE uintptr_t
|
||||
#define SH_KEY ptr
|
||||
#define SH_HASH_KEY(tb, key) hash_pointer(key)
|
||||
#define SH_EQUAL(tb, a, b) (a == b)
|
||||
#define SH_SCOPE extern
|
||||
#define SH_DEFINE
|
||||
#include "lib/simplehash.h"
|
||||
|
||||
typedef union
|
||||
{
|
||||
pointerhash_hash *pointers;
|
||||
tidhash_hash *tids;
|
||||
} visited_hash;
|
||||
|
||||
/*
|
||||
* Get the max number of connections in an upper layer for each element in the index
|
||||
*/
|
||||
@@ -184,6 +101,27 @@ HnswInitPage(Buffer buf, Page page)
|
||||
HnswPageGetOpaque(page)->page_id = HNSW_PAGE_ID;
|
||||
}
|
||||
|
||||
/*
|
||||
* Init and register page
|
||||
*/
|
||||
void
|
||||
HnswInitRegisterPage(Relation index, Buffer *buf, Page *page, GenericXLogState **state)
|
||||
{
|
||||
*state = GenericXLogStart(index);
|
||||
*page = GenericXLogRegisterBuffer(*state, *buf, GENERIC_XLOG_FULL_IMAGE);
|
||||
HnswInitPage(*buf, *page);
|
||||
}
|
||||
|
||||
/*
|
||||
* Commit buffer
|
||||
*/
|
||||
void
|
||||
HnswCommitBuffer(Buffer buf, GenericXLogState *state)
|
||||
{
|
||||
GenericXLogFinish(state);
|
||||
UnlockReleaseBuffer(buf);
|
||||
}
|
||||
|
||||
/*
|
||||
* Allocate neighbors
|
||||
*/
|
||||
@@ -206,6 +144,17 @@ HnswInitNeighbors(HnswElement element, int m)
|
||||
}
|
||||
}
|
||||
|
||||
/*
|
||||
* Free neighbors
|
||||
*/
|
||||
static void
|
||||
HnswFreeNeighbors(HnswElement element)
|
||||
{
|
||||
for (int lc = 0; lc <= element->level; lc++)
|
||||
pfree(element->neighbors[lc].items);
|
||||
pfree(element->neighbors);
|
||||
}
|
||||
|
||||
/*
|
||||
* Allocate an element
|
||||
*/
|
||||
@@ -220,7 +169,7 @@ HnswInitElement(ItemPointer heaptid, int m, double ml, int maxLevel)
|
||||
if (level > maxLevel)
|
||||
level = maxLevel;
|
||||
|
||||
element->heaptidsLength = 0;
|
||||
element->heaptids = NIL;
|
||||
HnswAddHeapTid(element, heaptid);
|
||||
|
||||
element->level = level;
|
||||
@@ -228,18 +177,32 @@ HnswInitElement(ItemPointer heaptid, int m, double ml, int maxLevel)
|
||||
|
||||
HnswInitNeighbors(element, m);
|
||||
|
||||
element->value = PointerGetDatum(NULL);
|
||||
|
||||
return element;
|
||||
}
|
||||
|
||||
/*
|
||||
* Free an element
|
||||
*/
|
||||
void
|
||||
HnswFreeElement(HnswElement element)
|
||||
{
|
||||
HnswFreeNeighbors(element);
|
||||
list_free_deep(element->heaptids);
|
||||
if (element->loaded)
|
||||
pfree(DatumGetPointer(element->value));
|
||||
pfree(element);
|
||||
}
|
||||
|
||||
/*
|
||||
* Add a heap TID to an element
|
||||
*/
|
||||
void
|
||||
HnswAddHeapTid(HnswElement element, ItemPointer heaptid)
|
||||
{
|
||||
element->heaptids[element->heaptidsLength++] = *heaptid;
|
||||
ItemPointer copy = palloc(sizeof(ItemPointerData));
|
||||
|
||||
ItemPointerCopy(heaptid, copy);
|
||||
element->heaptids = lappend(element->heaptids, copy);
|
||||
}
|
||||
|
||||
/*
|
||||
@@ -253,7 +216,7 @@ HnswInitElementFromBlock(BlockNumber blkno, OffsetNumber offno)
|
||||
element->blkno = blkno;
|
||||
element->offno = offno;
|
||||
element->neighbors = NULL;
|
||||
element->value = PointerGetDatum(NULL);
|
||||
element->loaded = false;
|
||||
return element;
|
||||
}
|
||||
|
||||
@@ -331,7 +294,7 @@ HnswUpdateMetaPageInfo(Page page, int updateEntry, HnswElement entryPoint, Block
|
||||
* Update the metapage
|
||||
*/
|
||||
void
|
||||
HnswUpdateMetaPage(Relation index, int updateEntry, HnswElement entryPoint, BlockNumber insertPage, ForkNumber forkNum, bool building)
|
||||
HnswUpdateMetaPage(Relation index, int updateEntry, HnswElement entryPoint, BlockNumber insertPage, ForkNumber forkNum)
|
||||
{
|
||||
Buffer buf;
|
||||
Page page;
|
||||
@@ -339,24 +302,12 @@ HnswUpdateMetaPage(Relation index, int updateEntry, HnswElement entryPoint, Bloc
|
||||
|
||||
buf = ReadBufferExtended(index, forkNum, HNSW_METAPAGE_BLKNO, RBM_NORMAL, NULL);
|
||||
LockBuffer(buf, BUFFER_LOCK_EXCLUSIVE);
|
||||
if (building)
|
||||
{
|
||||
state = NULL;
|
||||
page = BufferGetPage(buf);
|
||||
}
|
||||
else
|
||||
{
|
||||
state = GenericXLogStart(index);
|
||||
page = GenericXLogRegisterBuffer(state, buf, 0);
|
||||
}
|
||||
state = GenericXLogStart(index);
|
||||
page = GenericXLogRegisterBuffer(state, buf, 0);
|
||||
|
||||
HnswUpdateMetaPageInfo(page, updateEntry, entryPoint, insertPage);
|
||||
|
||||
if (building)
|
||||
MarkBufferDirty(buf);
|
||||
else
|
||||
GenericXLogFinish(state);
|
||||
UnlockReleaseBuffer(buf);
|
||||
HnswCommitBuffer(buf, state);
|
||||
}
|
||||
|
||||
/*
|
||||
@@ -370,12 +321,12 @@ HnswSetElementTuple(HnswElementTuple etup, HnswElement element)
|
||||
etup->deleted = 0;
|
||||
for (int i = 0; i < HNSW_HEAPTIDS; i++)
|
||||
{
|
||||
if (i < element->heaptidsLength)
|
||||
etup->heaptids[i] = element->heaptids[i];
|
||||
if (i < list_length(element->heaptids))
|
||||
etup->heaptids[i] = *((ItemPointer) list_nth(element->heaptids, i));
|
||||
else
|
||||
ItemPointerSetInvalid(&etup->heaptids[i]);
|
||||
}
|
||||
memcpy(&etup->data, DatumGetPointer(element->value), VARSIZE_ANY(DatumGetPointer(element->value)));
|
||||
memcpy(&etup->value, DatumGetPointer(element->value), VARSIZE_ANY(element->value));
|
||||
}
|
||||
|
||||
/*
|
||||
@@ -390,7 +341,7 @@ HnswSetNeighborTuple(HnswNeighborTuple ntup, HnswElement e, int m)
|
||||
|
||||
for (int lc = e->level; lc >= 0; lc--)
|
||||
{
|
||||
HnswNeighborArray *neighbors = HnswGetNeighbors(e, lc);
|
||||
HnswNeighborArray *neighbors = &e->neighbors[lc];
|
||||
int lm = HnswGetLayerM(m, lc);
|
||||
|
||||
for (int i = 0; i < lm; i++)
|
||||
@@ -448,7 +399,7 @@ LoadNeighborsFromPage(HnswElement element, Relation index, Page page, int m)
|
||||
if (level < 0)
|
||||
level = 0;
|
||||
|
||||
neighbors = HnswGetNeighbors(element, level);
|
||||
neighbors = &element->neighbors[level];
|
||||
hc = &neighbors->items[neighbors->length++];
|
||||
hc->element = e;
|
||||
}
|
||||
@@ -482,7 +433,7 @@ HnswLoadElementFromTuple(HnswElement element, HnswElementTuple etup, bool loadHe
|
||||
element->deleted = etup->deleted;
|
||||
element->neighborPage = ItemPointerGetBlockNumber(&etup->neighbortid);
|
||||
element->neighborOffno = ItemPointerGetOffsetNumber(&etup->neighbortid);
|
||||
element->heaptidsLength = 0;
|
||||
element->heaptids = NIL;
|
||||
|
||||
if (loadHeaptids)
|
||||
{
|
||||
@@ -497,7 +448,12 @@ HnswLoadElementFromTuple(HnswElement element, HnswElementTuple etup, bool loadHe
|
||||
}
|
||||
|
||||
if (loadVec)
|
||||
element->value = datumCopy(PointerGetDatum(&etup->data), false, -1);
|
||||
{
|
||||
Datum value = PointerGetDatum(&etup->value);
|
||||
|
||||
element->value = datumCopy(value, false, -1);
|
||||
element->loaded = true;
|
||||
}
|
||||
}
|
||||
|
||||
/*
|
||||
@@ -524,7 +480,7 @@ HnswLoadElement(HnswElement element, float *distance, Datum *q, Relation index,
|
||||
|
||||
/* Calculate distance */
|
||||
if (distance != NULL)
|
||||
*distance = (float) DatumGetFloat8(FunctionCall2Coll(procinfo, collation, *q, PointerGetDatum(&etup->data)));
|
||||
*distance = (float) DatumGetFloat8(FunctionCall2Coll(procinfo, collation, *q, PointerGetDatum(&etup->value)));
|
||||
|
||||
UnlockReleaseBuffer(buf);
|
||||
}
|
||||
@@ -600,22 +556,16 @@ CreatePairingHeapNode(HnswCandidate * c)
|
||||
* Add to visited
|
||||
*/
|
||||
static inline void
|
||||
AddToVisited(visited_hash v, HnswCandidate * hc, Relation index, bool *found)
|
||||
AddToVisited(HTAB *v, HnswCandidate * hc, Relation index, bool *found)
|
||||
{
|
||||
if (index == NULL)
|
||||
{
|
||||
#if PG_VERSION_NUM >= 130000
|
||||
pointerhash_insert_hash(v.pointers, (uintptr_t) hc->element, hc->element->hash, found);
|
||||
#else
|
||||
pointerhash_insert(v.pointers, (uintptr_t) hc->element, found);
|
||||
#endif
|
||||
}
|
||||
hash_search(v, &hc->element, HASH_ENTER, found);
|
||||
else
|
||||
{
|
||||
ItemPointerData indextid;
|
||||
|
||||
ItemPointerSet(&indextid, hc->element->blkno, hc->element->offno);
|
||||
tidhash_insert(v.tids, indextid, found);
|
||||
hash_search(v, &indextid, HASH_ENTER, found);
|
||||
}
|
||||
}
|
||||
|
||||
@@ -625,26 +575,36 @@ AddToVisited(visited_hash v, HnswCandidate * hc, Relation index, bool *found)
|
||||
List *
|
||||
HnswSearchLayer(Datum q, List *ep, int ef, int lc, Relation index, FmgrInfo *procinfo, Oid collation, int m, bool inserting, HnswElement skipElement)
|
||||
{
|
||||
ListCell *lc2;
|
||||
|
||||
List *w = NIL;
|
||||
pairingheap *C = pairingheap_allocate(CompareNearestCandidates, NULL);
|
||||
pairingheap *W = pairingheap_allocate(CompareFurthestCandidates, NULL);
|
||||
int wlen = 0;
|
||||
visited_hash v;
|
||||
ListCell *lc2;
|
||||
HASHCTL hash_ctl;
|
||||
HTAB *v;
|
||||
|
||||
/* Create hash table */
|
||||
if (index == NULL)
|
||||
v.pointers = pointerhash_create(CurrentMemoryContext, ef * m * 2, NULL);
|
||||
{
|
||||
hash_ctl.keysize = sizeof(HnswElement *);
|
||||
hash_ctl.entrysize = sizeof(HnswElement *);
|
||||
}
|
||||
else
|
||||
v.tids = tidhash_create(CurrentMemoryContext, ef * m * 2, NULL);
|
||||
{
|
||||
hash_ctl.keysize = sizeof(ItemPointerData);
|
||||
hash_ctl.entrysize = sizeof(ItemPointerData);
|
||||
}
|
||||
|
||||
hash_ctl.hcxt = CurrentMemoryContext;
|
||||
v = hash_create("hnsw visited", 256, &hash_ctl, HASH_ELEM | HASH_BLOBS | HASH_CONTEXT);
|
||||
|
||||
/* Add entry points to v, C, and W */
|
||||
foreach(lc2, ep)
|
||||
{
|
||||
HnswCandidate *hc = (HnswCandidate *) lfirst(lc2);
|
||||
bool found;
|
||||
|
||||
AddToVisited(v, hc, index, &found);
|
||||
AddToVisited(v, hc, index, NULL);
|
||||
|
||||
pairingheap_add(C, &(CreatePairingHeapNode(hc)->ph_node));
|
||||
pairingheap_add(W, &(CreatePairingHeapNode(hc)->ph_node));
|
||||
@@ -654,7 +614,7 @@ HnswSearchLayer(Datum q, List *ep, int ef, int lc, Relation index, FmgrInfo *pro
|
||||
* would be ideal to do this for inserts as well, but this could
|
||||
* affect insert performance.
|
||||
*/
|
||||
if (skipElement == NULL || hc->element->heaptidsLength != 0)
|
||||
if (skipElement == NULL || list_length(hc->element->heaptids) != 0)
|
||||
wlen++;
|
||||
}
|
||||
|
||||
@@ -671,7 +631,7 @@ HnswSearchLayer(Datum q, List *ep, int ef, int lc, Relation index, FmgrInfo *pro
|
||||
HnswLoadNeighbors(c->element, index, m);
|
||||
|
||||
/* Get the neighborhood at layer lc */
|
||||
neighborhood = HnswGetNeighbors(c->element, lc);
|
||||
neighborhood = &c->element->neighbors[lc];
|
||||
|
||||
for (int i = 0; i < neighborhood->length; i++)
|
||||
{
|
||||
@@ -713,7 +673,7 @@ HnswSearchLayer(Datum q, List *ep, int ef, int lc, Relation index, FmgrInfo *pro
|
||||
* vacuuming. It would be ideal to do this for inserts as
|
||||
* well, but this could affect insert performance.
|
||||
*/
|
||||
if (skipElement == NULL || e->element->heaptidsLength != 0)
|
||||
if (skipElement == NULL || list_length(e->element->heaptids) != 0)
|
||||
{
|
||||
wlen++;
|
||||
|
||||
@@ -774,23 +734,23 @@ HnswGetDistance(HnswElement a, HnswElement b, int lc, FmgrInfo *procinfo, Oid co
|
||||
/* Look for cached distance */
|
||||
if (a->neighbors != NULL)
|
||||
{
|
||||
HnswNeighborArray *neighbors = HnswGetNeighbors(a, lc);
|
||||
Assert(a->level >= lc);
|
||||
|
||||
for (int i = 0; i < neighbors->length; i++)
|
||||
for (int i = 0; i < a->neighbors[lc].length; i++)
|
||||
{
|
||||
if (neighbors->items[i].element == b)
|
||||
return neighbors->items[i].distance;
|
||||
if (a->neighbors[lc].items[i].element == b)
|
||||
return a->neighbors[lc].items[i].distance;
|
||||
}
|
||||
}
|
||||
|
||||
if (b->neighbors != NULL)
|
||||
{
|
||||
HnswNeighborArray *neighbors = HnswGetNeighbors(b, lc);
|
||||
Assert(b->level >= lc);
|
||||
|
||||
for (int i = 0; i < neighbors->length; i++)
|
||||
for (int i = 0; i < b->neighbors[lc].length; i++)
|
||||
{
|
||||
if (neighbors->items[i].element == a)
|
||||
return neighbors->items[i].distance;
|
||||
if (b->neighbors[lc].items[i].element == a)
|
||||
return b->neighbors[lc].items[i].distance;
|
||||
}
|
||||
}
|
||||
|
||||
@@ -826,8 +786,7 @@ SelectNeighbors(List *c, int m, int lc, FmgrInfo *procinfo, Oid collation, HnswE
|
||||
List *r = NIL;
|
||||
List *w = list_copy(c);
|
||||
pairingheap *wd;
|
||||
HnswNeighborArray *neighbors = HnswGetNeighbors(e2, lc);
|
||||
bool mustCalculate = !neighbors->closerSet;
|
||||
bool mustCalculate = !e2->neighbors[lc].closerSet;
|
||||
List *added = NIL;
|
||||
bool removedAny = false;
|
||||
|
||||
@@ -891,7 +850,7 @@ SelectNeighbors(List *c, int m, int lc, FmgrInfo *procinfo, Oid collation, HnswE
|
||||
}
|
||||
|
||||
/* Cached value can only be used in future if sorted deterministically */
|
||||
neighbors->closerSet = sortCandidates;
|
||||
e2->neighbors[lc].closerSet = sortCandidates;
|
||||
|
||||
/* Keep pruned connections */
|
||||
while (!pairingheap_is_empty(wd) && list_length(r) < m)
|
||||
@@ -909,6 +868,30 @@ SelectNeighbors(List *c, int m, int lc, FmgrInfo *procinfo, Oid collation, HnswE
|
||||
return r;
|
||||
}
|
||||
|
||||
/*
|
||||
* Find duplicate element
|
||||
*/
|
||||
HnswElement
|
||||
HnswFindDuplicate(HnswElement e)
|
||||
{
|
||||
HnswNeighborArray *neighbors = &e->neighbors[0];
|
||||
|
||||
for (int i = 0; i < neighbors->length; i++)
|
||||
{
|
||||
HnswCandidate *neighbor = &neighbors->items[i];
|
||||
|
||||
/* Exit early since ordered by distance */
|
||||
if (!datumIsEqual(e->value, neighbor->element->value, false, -1))
|
||||
break;
|
||||
|
||||
/* Check for space */
|
||||
if (list_length(neighbor->element->heaptids) < HNSW_HEAPTIDS)
|
||||
return neighbor->element;
|
||||
}
|
||||
|
||||
return NULL;
|
||||
}
|
||||
|
||||
/*
|
||||
* Add connections
|
||||
*/
|
||||
@@ -916,7 +899,7 @@ static void
|
||||
AddConnections(HnswElement element, List *neighbors, int m, int lc)
|
||||
{
|
||||
ListCell *lc2;
|
||||
HnswNeighborArray *a = HnswGetNeighbors(element, lc);
|
||||
HnswNeighborArray *a = &element->neighbors[lc];
|
||||
|
||||
foreach(lc2, neighbors)
|
||||
a->items[a->length++] = *((HnswCandidate *) lfirst(lc2));
|
||||
@@ -928,7 +911,7 @@ AddConnections(HnswElement element, List *neighbors, int m, int lc)
|
||||
void
|
||||
HnswUpdateConnection(HnswElement element, HnswCandidate * hc, int m, int lc, int *updateIdx, Relation index, FmgrInfo *procinfo, Oid collation)
|
||||
{
|
||||
HnswNeighborArray *currentNeighbors = HnswGetNeighbors(hc->element, lc);
|
||||
HnswNeighborArray *currentNeighbors = &hc->element->neighbors[lc];
|
||||
|
||||
HnswCandidate hc2;
|
||||
|
||||
@@ -957,13 +940,13 @@ HnswUpdateConnection(HnswElement element, HnswCandidate * hc, int m, int lc, int
|
||||
{
|
||||
HnswCandidate *hc3 = ¤tNeighbors->items[i];
|
||||
|
||||
if (DatumGetPointer(hc3->element->value) == NULL)
|
||||
if (!hc3->element->loaded)
|
||||
HnswLoadElement(hc3->element, &hc3->distance, &q, index, procinfo, collation, true);
|
||||
else
|
||||
hc3->distance = GetCandidateDistance(hc3, q, procinfo, collation);
|
||||
|
||||
/* Prune element if being deleted */
|
||||
if (hc3->element->heaptidsLength == 0)
|
||||
if (list_length(hc3->element->heaptids) == 0)
|
||||
{
|
||||
pruned = ¤tNeighbors->items[i];
|
||||
break;
|
||||
@@ -1021,7 +1004,7 @@ RemoveElements(List *w, HnswElement skipElement)
|
||||
if (skipElement != NULL && hc->element->blkno == skipElement->blkno && hc->element->offno == skipElement->offno)
|
||||
continue;
|
||||
|
||||
if (hc->element->heaptidsLength != 0)
|
||||
if (list_length(hc->element->heaptids) != 0)
|
||||
w2 = lappend(w2, hc);
|
||||
}
|
||||
|
||||
@@ -1041,12 +1024,6 @@ HnswInsertElement(HnswElement element, HnswElement entryPoint, Relation index, F
|
||||
Datum q = element->value;
|
||||
HnswElement skipElement = existing ? element : NULL;
|
||||
|
||||
#if PG_VERSION_NUM >= 130000
|
||||
/* Precompute hash */
|
||||
if (index == NULL)
|
||||
element->hash = hash_pointer((uintptr_t) element);
|
||||
#endif
|
||||
|
||||
/* No neighbors if no entry point */
|
||||
if (entryPoint == NULL)
|
||||
return;
|
||||
|
||||
@@ -12,9 +12,12 @@
|
||||
* Check if deleted list contains an index TID
|
||||
*/
|
||||
static bool
|
||||
DeletedContains(tidhash_hash * deleted, ItemPointer indextid)
|
||||
DeletedContains(HTAB *deleted, ItemPointer indextid)
|
||||
{
|
||||
return tidhash_lookup(deleted, *indextid) != NULL;
|
||||
bool found;
|
||||
|
||||
hash_search(deleted, indextid, HASH_FIND, &found);
|
||||
return found;
|
||||
}
|
||||
|
||||
/*
|
||||
@@ -59,8 +62,7 @@ RemoveHeapTids(HnswVacuumState * vacuumstate)
|
||||
/* Iterate over nodes */
|
||||
for (offno = FirstOffsetNumber; offno <= maxoffno; offno = OffsetNumberNext(offno))
|
||||
{
|
||||
ItemId itemid = PageGetItemId(page, offno);
|
||||
HnswElementTuple etup = (HnswElementTuple) PageGetItem(page, itemid);
|
||||
HnswElementTuple etup = (HnswElementTuple) PageGetItem(page, PageGetItemId(page, offno));
|
||||
int idx = 0;
|
||||
bool itemUpdated = false;
|
||||
|
||||
@@ -91,7 +93,7 @@ RemoveHeapTids(HnswVacuumState * vacuumstate)
|
||||
|
||||
if (itemUpdated)
|
||||
{
|
||||
Size etupSize = ItemIdGetLength(itemid);
|
||||
Size etupSize = HNSW_ELEMENT_TUPLE_SIZE(PointerGetDatum(&etup->value));
|
||||
|
||||
/* Mark rest as invalid */
|
||||
for (int i = idx; i < HNSW_HEAPTIDS; i++)
|
||||
@@ -107,13 +109,11 @@ RemoveHeapTids(HnswVacuumState * vacuumstate)
|
||||
if (!ItemPointerIsValid(&etup->heaptids[0]))
|
||||
{
|
||||
ItemPointerData ip;
|
||||
bool found;
|
||||
|
||||
/* Add to deleted list */
|
||||
ItemPointerSet(&ip, blkno, offno);
|
||||
|
||||
tidhash_insert(vacuumstate->deleted, ip, &found);
|
||||
Assert(!found);
|
||||
(void) hash_search(vacuumstate->deleted, &ip, HASH_ENTER, NULL);
|
||||
}
|
||||
else if (etup->level > highestLevel && !(entryPoint != NULL && blkno == entryPoint->blkno && offno == entryPoint->offno))
|
||||
{
|
||||
@@ -206,7 +206,7 @@ RepairGraphElement(HnswVacuumState * vacuumstate, HnswElement element, HnswEleme
|
||||
|
||||
/* Init fields */
|
||||
HnswInitNeighbors(element, m);
|
||||
element->heaptidsLength = 0;
|
||||
element->heaptids = NIL;
|
||||
|
||||
/* Add element to graph, skipping itself */
|
||||
HnswInsertElement(element, entryPoint, index, procinfo, collation, m, efConstruction, true);
|
||||
@@ -230,7 +230,7 @@ RepairGraphElement(HnswVacuumState * vacuumstate, HnswElement element, HnswEleme
|
||||
UnlockReleaseBuffer(buf);
|
||||
|
||||
/* Update neighbors */
|
||||
HnswUpdateNeighborPages(index, procinfo, collation, element, m, true, false);
|
||||
HnswUpdateNeighborPages(index, procinfo, collation, element, m, true);
|
||||
}
|
||||
|
||||
/*
|
||||
@@ -286,7 +286,7 @@ RepairGraphEntryPoint(HnswVacuumState * vacuumstate)
|
||||
* point is outdated and empty, the entry point will be empty
|
||||
* until an element is repaired.
|
||||
*/
|
||||
HnswUpdateMetaPage(index, HNSW_UPDATE_ENTRY_ALWAYS, highestPoint, InvalidBlockNumber, MAIN_FORKNUM, false);
|
||||
HnswUpdateMetaPage(index, HNSW_UPDATE_ENTRY_ALWAYS, highestPoint, InvalidBlockNumber, MAIN_FORKNUM);
|
||||
}
|
||||
else
|
||||
{
|
||||
@@ -419,7 +419,7 @@ RepairGraph(HnswVacuumState * vacuumstate)
|
||||
* was replaced and highest point was outdated.
|
||||
*/
|
||||
if (entryPoint == NULL || element->level > entryPoint->level)
|
||||
HnswUpdateMetaPage(index, HNSW_UPDATE_ENTRY_GREATER, element, InvalidBlockNumber, MAIN_FORKNUM, false);
|
||||
HnswUpdateMetaPage(index, HNSW_UPDATE_ENTRY_GREATER, element, InvalidBlockNumber, MAIN_FORKNUM);
|
||||
|
||||
/* Release lock */
|
||||
UnlockPage(index, HNSW_UPDATE_LOCK, lockmode);
|
||||
@@ -477,11 +477,11 @@ MarkDeleted(HnswVacuumState * vacuumstate)
|
||||
/* Update element and neighbors together */
|
||||
for (offno = FirstOffsetNumber; offno <= maxoffno; offno = OffsetNumberNext(offno))
|
||||
{
|
||||
ItemId itemid = PageGetItemId(page, offno);
|
||||
HnswElementTuple etup = (HnswElementTuple) PageGetItem(page, itemid);
|
||||
HnswElementTuple etup = (HnswElementTuple) PageGetItem(page, PageGetItemId(page, offno));
|
||||
HnswNeighborTuple ntup;
|
||||
Size etupSize;
|
||||
Size ntupSize;
|
||||
Datum value;
|
||||
Buffer nbuf;
|
||||
Page npage;
|
||||
BlockNumber neighborPage;
|
||||
@@ -505,8 +505,11 @@ MarkDeleted(HnswVacuumState * vacuumstate)
|
||||
if (ItemPointerIsValid(&etup->heaptids[0]))
|
||||
continue;
|
||||
|
||||
/* Get datum */
|
||||
value = PointerGetDatum(&etup->value);
|
||||
|
||||
/* Calculate sizes */
|
||||
etupSize = ItemIdGetLength(itemid);
|
||||
etupSize = HNSW_ELEMENT_TUPLE_SIZE(value);
|
||||
ntupSize = HNSW_NEIGHBOR_TUPLE_SIZE(etup->level, vacuumstate->m);
|
||||
|
||||
/* Get neighbor page */
|
||||
@@ -529,7 +532,7 @@ MarkDeleted(HnswVacuumState * vacuumstate)
|
||||
|
||||
/* Overwrite element */
|
||||
etup->deleted = 1;
|
||||
MemSet(&etup->data, 0, VARSIZE_ANY(&etup->data));
|
||||
MemSet(&etup->value, 0, VARSIZE_ANY(value));
|
||||
|
||||
/* Overwrite neighbors */
|
||||
for (int i = 0; i < ntup->count; i++)
|
||||
@@ -564,7 +567,7 @@ MarkDeleted(HnswVacuumState * vacuumstate)
|
||||
}
|
||||
|
||||
/* Update insert page last, after everything has been marked as deleted */
|
||||
HnswUpdateMetaPage(index, 0, NULL, insertPage, MAIN_FORKNUM, false);
|
||||
HnswUpdateMetaPage(index, 0, NULL, insertPage, MAIN_FORKNUM);
|
||||
}
|
||||
|
||||
/*
|
||||
@@ -574,6 +577,7 @@ static void
|
||||
InitVacuumState(HnswVacuumState * vacuumstate, IndexVacuumInfo *info, IndexBulkDeleteResult *stats, IndexBulkDeleteCallback callback, void *callback_state)
|
||||
{
|
||||
Relation index = info->index;
|
||||
HASHCTL hash_ctl;
|
||||
|
||||
if (stats == NULL)
|
||||
stats = (IndexBulkDeleteResult *) palloc0(sizeof(IndexBulkDeleteResult));
|
||||
@@ -595,7 +599,10 @@ InitVacuumState(HnswVacuumState * vacuumstate, IndexVacuumInfo *info, IndexBulkD
|
||||
HnswGetMetaPageInfo(index, &vacuumstate->m, NULL);
|
||||
|
||||
/* Create hash table */
|
||||
vacuumstate->deleted = tidhash_create(CurrentMemoryContext, 256, NULL);
|
||||
hash_ctl.keysize = sizeof(ItemPointerData);
|
||||
hash_ctl.entrysize = sizeof(ItemPointerData);
|
||||
hash_ctl.hcxt = CurrentMemoryContext;
|
||||
vacuumstate->deleted = hash_create("hnswbulkdelete indextids", 256, &hash_ctl, HASH_ELEM | HASH_BLOBS | HASH_CONTEXT);
|
||||
}
|
||||
|
||||
/*
|
||||
@@ -604,7 +611,7 @@ InitVacuumState(HnswVacuumState * vacuumstate, IndexVacuumInfo *info, IndexBulkD
|
||||
static void
|
||||
FreeVacuumState(HnswVacuumState * vacuumstate)
|
||||
{
|
||||
tidhash_destroy(vacuumstate->deleted);
|
||||
hash_destroy(vacuumstate->deleted);
|
||||
FreeAccessStrategy(vacuumstate->bas);
|
||||
pfree(vacuumstate->ntup);
|
||||
MemoryContextDelete(vacuumstate->tmpCtx);
|
||||
|
||||
@@ -543,10 +543,10 @@ CreateListPages(Relation index, VectorArray centers, int dimensions,
|
||||
pfree(list);
|
||||
}
|
||||
|
||||
#ifdef IVFFLAT_KMEANS_DEBUG
|
||||
/*
|
||||
* Print k-means metrics
|
||||
*/
|
||||
#ifdef IVFFLAT_KMEANS_DEBUG
|
||||
static void
|
||||
PrintKmeansMetrics(IvfflatBuildState * buildstate)
|
||||
{
|
||||
|
||||
@@ -105,20 +105,6 @@ ivfflatcostestimate(PlannerInfo *root, IndexPath *path, double loop_count,
|
||||
*/
|
||||
costs.numIndexTuples = path->indexinfo->tuples * ratio;
|
||||
|
||||
/*
|
||||
* Do not use index if no limit or limit + offset > expected tuples unless
|
||||
* enable_seqscan = off
|
||||
*/
|
||||
if (root->limit_tuples < 0 || root->limit_tuples > costs.numIndexTuples)
|
||||
{
|
||||
*indexStartupCost = 1.0e10 - 1;
|
||||
*indexTotalCost = 1.0e10 - 1;
|
||||
*indexSelectivity = 0;
|
||||
*indexCorrelation = 0;
|
||||
*indexPages = 0;
|
||||
return;
|
||||
}
|
||||
|
||||
#if PG_VERSION_NUM >= 120000
|
||||
genericcostestimate(root, path, loop_count, &costs);
|
||||
#else
|
||||
|
||||
@@ -6,10 +6,6 @@
|
||||
#include "ivfflat.h"
|
||||
#include "miscadmin.h"
|
||||
|
||||
#ifdef IVFFLAT_MEMORY
|
||||
#include "utils/memutils.h"
|
||||
#endif
|
||||
|
||||
/*
|
||||
* Initialize with kmeans++
|
||||
*
|
||||
@@ -155,23 +151,6 @@ QuickCenters(Relation index, VectorArray samples, VectorArray centers)
|
||||
}
|
||||
}
|
||||
|
||||
#ifdef IVFFLAT_MEMORY
|
||||
/*
|
||||
* Show memory usage
|
||||
*/
|
||||
static void
|
||||
ShowMemoryUsage(Size estimatedSize)
|
||||
{
|
||||
#if PG_VERSION_NUM >= 130000
|
||||
elog(INFO, "total memory: %zu MB",
|
||||
MemoryContextMemAllocated(CurrentMemoryContext, true) / (1024 * 1024));
|
||||
#else
|
||||
MemoryContextStats(CurrentMemoryContext);
|
||||
#endif
|
||||
elog(INFO, "estimated memory: %zu MB", estimatedSize / (1024 * 1024));
|
||||
}
|
||||
#endif
|
||||
|
||||
/*
|
||||
* Use Elkan for performance. This requires distance function to satisfy triangle inequality.
|
||||
*
|
||||
@@ -252,10 +231,6 @@ ElkanKmeans(Relation index, VectorArray samples, VectorArray centers)
|
||||
vec->dim = dimensions;
|
||||
}
|
||||
|
||||
#ifdef IVFFLAT_MEMORY
|
||||
ShowMemoryUsage(totalSize);
|
||||
#endif
|
||||
|
||||
/* Pick initial centers */
|
||||
InitCenters(index, samples, centers, lowerBound);
|
||||
|
||||
|
||||
55
src/vector.c
55
src/vector.c
@@ -89,7 +89,7 @@ CheckDim(int dim)
|
||||
}
|
||||
|
||||
/*
|
||||
* Ensure finite element
|
||||
* Ensure finite elements
|
||||
*/
|
||||
static inline void
|
||||
CheckElement(float value)
|
||||
@@ -177,15 +177,14 @@ PGDLLEXPORT PG_FUNCTION_INFO_V1(vector_in);
|
||||
Datum
|
||||
vector_in(PG_FUNCTION_ARGS)
|
||||
{
|
||||
char *lit = PG_GETARG_CSTRING(0);
|
||||
char *str = PG_GETARG_CSTRING(0);
|
||||
int32 typmod = PG_GETARG_INT32(2);
|
||||
float x[VECTOR_MAX_DIM];
|
||||
int dim = 0;
|
||||
char *pt;
|
||||
char *stringEnd;
|
||||
Vector *result;
|
||||
char *litcopy = pstrdup(lit);
|
||||
char *str = litcopy;
|
||||
char *lit = pstrdup(str);
|
||||
|
||||
while (vector_isspace(*str))
|
||||
str++;
|
||||
@@ -269,7 +268,7 @@ vector_in(PG_FUNCTION_ARGS)
|
||||
(errcode(ERRCODE_DATA_EXCEPTION),
|
||||
errmsg("vector must have at least 1 dimension")));
|
||||
|
||||
pfree(litcopy);
|
||||
pfree(lit);
|
||||
|
||||
CheckExpectedDim(typmod, dim);
|
||||
|
||||
@@ -438,18 +437,17 @@ vector_send(PG_FUNCTION_ARGS)
|
||||
|
||||
/*
|
||||
* Convert vector to vector
|
||||
* This is needed to check the type modifier
|
||||
*/
|
||||
PGDLLEXPORT PG_FUNCTION_INFO_V1(vector);
|
||||
Datum
|
||||
vector(PG_FUNCTION_ARGS)
|
||||
{
|
||||
Vector *vec = PG_GETARG_VECTOR_P(0);
|
||||
Vector *arg = PG_GETARG_VECTOR_P(0);
|
||||
int32 typmod = PG_GETARG_INT32(1);
|
||||
|
||||
CheckExpectedDim(typmod, vec->dim);
|
||||
CheckExpectedDim(typmod, arg->dim);
|
||||
|
||||
PG_RETURN_POINTER(vec);
|
||||
PG_RETURN_POINTER(arg);
|
||||
}
|
||||
|
||||
/*
|
||||
@@ -466,6 +464,7 @@ array_to_vector(PG_FUNCTION_ARGS)
|
||||
bool typbyval;
|
||||
char typalign;
|
||||
Datum *elemsp;
|
||||
bool *nullsp;
|
||||
int nelemsp;
|
||||
|
||||
if (ARR_NDIM(array) > 1)
|
||||
@@ -479,7 +478,7 @@ array_to_vector(PG_FUNCTION_ARGS)
|
||||
errmsg("array must not contain nulls")));
|
||||
|
||||
get_typlenbyvalalign(ARR_ELEMTYPE(array), &typlen, &typbyval, &typalign);
|
||||
deconstruct_array(array, ARR_ELEMTYPE(array), typlen, typbyval, typalign, &elemsp, NULL, &nelemsp);
|
||||
deconstruct_array(array, ARR_ELEMTYPE(array), typlen, typbyval, typalign, &elemsp, &nullsp, &nelemsp);
|
||||
|
||||
CheckDim(nelemsp);
|
||||
CheckExpectedDim(typmod, nelemsp);
|
||||
@@ -513,12 +512,6 @@ array_to_vector(PG_FUNCTION_ARGS)
|
||||
errmsg("unsupported array type")));
|
||||
}
|
||||
|
||||
/*
|
||||
* Free allocation from deconstruct_array. Do not free individual elements
|
||||
* when pass-by-reference since they point to original array.
|
||||
*/
|
||||
pfree(elemsp);
|
||||
|
||||
/* Check elements */
|
||||
for (int i = 0; i < result->dim; i++)
|
||||
CheckElement(result->x[i]);
|
||||
@@ -725,7 +718,7 @@ vector_spherical_distance(PG_FUNCTION_ARGS)
|
||||
}
|
||||
|
||||
/*
|
||||
* Get the L1 distance between two vectors
|
||||
* Get the L1 distance between vectors
|
||||
*/
|
||||
PGDLLEXPORT PG_FUNCTION_INFO_V1(l1_distance);
|
||||
Datum
|
||||
@@ -904,8 +897,8 @@ PGDLLEXPORT PG_FUNCTION_INFO_V1(vector_lt);
|
||||
Datum
|
||||
vector_lt(PG_FUNCTION_ARGS)
|
||||
{
|
||||
Vector *a = PG_GETARG_VECTOR_P(0);
|
||||
Vector *b = PG_GETARG_VECTOR_P(1);
|
||||
Vector *a = (Vector *) PG_GETARG_VECTOR_P(0);
|
||||
Vector *b = (Vector *) PG_GETARG_VECTOR_P(1);
|
||||
|
||||
PG_RETURN_BOOL(vector_cmp_internal(a, b) < 0);
|
||||
}
|
||||
@@ -917,8 +910,8 @@ PGDLLEXPORT PG_FUNCTION_INFO_V1(vector_le);
|
||||
Datum
|
||||
vector_le(PG_FUNCTION_ARGS)
|
||||
{
|
||||
Vector *a = PG_GETARG_VECTOR_P(0);
|
||||
Vector *b = PG_GETARG_VECTOR_P(1);
|
||||
Vector *a = (Vector *) PG_GETARG_VECTOR_P(0);
|
||||
Vector *b = (Vector *) PG_GETARG_VECTOR_P(1);
|
||||
|
||||
PG_RETURN_BOOL(vector_cmp_internal(a, b) <= 0);
|
||||
}
|
||||
@@ -930,8 +923,8 @@ PGDLLEXPORT PG_FUNCTION_INFO_V1(vector_eq);
|
||||
Datum
|
||||
vector_eq(PG_FUNCTION_ARGS)
|
||||
{
|
||||
Vector *a = PG_GETARG_VECTOR_P(0);
|
||||
Vector *b = PG_GETARG_VECTOR_P(1);
|
||||
Vector *a = (Vector *) PG_GETARG_VECTOR_P(0);
|
||||
Vector *b = (Vector *) PG_GETARG_VECTOR_P(1);
|
||||
|
||||
PG_RETURN_BOOL(vector_cmp_internal(a, b) == 0);
|
||||
}
|
||||
@@ -943,8 +936,8 @@ PGDLLEXPORT PG_FUNCTION_INFO_V1(vector_ne);
|
||||
Datum
|
||||
vector_ne(PG_FUNCTION_ARGS)
|
||||
{
|
||||
Vector *a = PG_GETARG_VECTOR_P(0);
|
||||
Vector *b = PG_GETARG_VECTOR_P(1);
|
||||
Vector *a = (Vector *) PG_GETARG_VECTOR_P(0);
|
||||
Vector *b = (Vector *) PG_GETARG_VECTOR_P(1);
|
||||
|
||||
PG_RETURN_BOOL(vector_cmp_internal(a, b) != 0);
|
||||
}
|
||||
@@ -956,8 +949,8 @@ PGDLLEXPORT PG_FUNCTION_INFO_V1(vector_ge);
|
||||
Datum
|
||||
vector_ge(PG_FUNCTION_ARGS)
|
||||
{
|
||||
Vector *a = PG_GETARG_VECTOR_P(0);
|
||||
Vector *b = PG_GETARG_VECTOR_P(1);
|
||||
Vector *a = (Vector *) PG_GETARG_VECTOR_P(0);
|
||||
Vector *b = (Vector *) PG_GETARG_VECTOR_P(1);
|
||||
|
||||
PG_RETURN_BOOL(vector_cmp_internal(a, b) >= 0);
|
||||
}
|
||||
@@ -969,8 +962,8 @@ PGDLLEXPORT PG_FUNCTION_INFO_V1(vector_gt);
|
||||
Datum
|
||||
vector_gt(PG_FUNCTION_ARGS)
|
||||
{
|
||||
Vector *a = PG_GETARG_VECTOR_P(0);
|
||||
Vector *b = PG_GETARG_VECTOR_P(1);
|
||||
Vector *a = (Vector *) PG_GETARG_VECTOR_P(0);
|
||||
Vector *b = (Vector *) PG_GETARG_VECTOR_P(1);
|
||||
|
||||
PG_RETURN_BOOL(vector_cmp_internal(a, b) > 0);
|
||||
}
|
||||
@@ -982,8 +975,8 @@ PGDLLEXPORT PG_FUNCTION_INFO_V1(vector_cmp);
|
||||
Datum
|
||||
vector_cmp(PG_FUNCTION_ARGS)
|
||||
{
|
||||
Vector *a = PG_GETARG_VECTOR_P(0);
|
||||
Vector *b = PG_GETARG_VECTOR_P(1);
|
||||
Vector *a = (Vector *) PG_GETARG_VECTOR_P(0);
|
||||
Vector *b = (Vector *) PG_GETARG_VECTOR_P(1);
|
||||
|
||||
PG_RETURN_INT32(vector_cmp_internal(a, b));
|
||||
}
|
||||
|
||||
@@ -83,32 +83,11 @@ for my $i (0 .. $#operators)
|
||||
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);
|
||||
));
|
||||
# Add index
|
||||
$node->safe_psql("postgres", "CREATE INDEX ON tst USING hnsw (v $opclass);");
|
||||
|
||||
# Test approximate results
|
||||
my $min = $operator eq "<#>" ? 0.80 : 0.99;
|
||||
test_recall($min, $operator);
|
||||
|
||||
$node->safe_psql("postgres", "DROP INDEX idx;");
|
||||
|
||||
# Build index in parallel
|
||||
my ($ret, $stdout, $stderr) = $node->psql("postgres", qq(
|
||||
SET client_min_messages = DEBUG;
|
||||
SET min_parallel_table_scan_size = 1;
|
||||
SET hnsw.enable_parallel_build = on;
|
||||
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;");
|
||||
}
|
||||
|
||||
done_testing();
|
||||
|
||||
43
test/t/018_ivfflat_deletes.pl
Normal file
43
test/t/018_ivfflat_deletes.pl
Normal file
@@ -0,0 +1,43 @@
|
||||
use strict;
|
||||
use warnings;
|
||||
use PostgresNode;
|
||||
use TestLib;
|
||||
use Test::More;
|
||||
|
||||
my $dim = 3;
|
||||
|
||||
my $array_sql = join(",", ('random()') x $dim);
|
||||
|
||||
# Initialize node
|
||||
my $node = get_new_node('node');
|
||||
$node->init;
|
||||
$node->start;
|
||||
|
||||
# Create table and index
|
||||
$node->safe_psql("postgres", "CREATE EXTENSION vector;");
|
||||
$node->safe_psql("postgres", "CREATE TABLE tst (i serial, v vector($dim));");
|
||||
$node->safe_psql("postgres",
|
||||
"INSERT INTO tst (v) SELECT ARRAY[$array_sql] FROM generate_series(1, 10000) i;"
|
||||
);
|
||||
$node->safe_psql("postgres", "CREATE INDEX ON tst USING ivfflat (v vector_l2_ops);");
|
||||
|
||||
# Delete data
|
||||
$node->safe_psql("postgres", "DELETE FROM tst WHERE i % 100 != 0;");
|
||||
|
||||
my $exp = $node->safe_psql("postgres", qq(
|
||||
SET enable_indexscan = off;
|
||||
SELECT i FROM tst ORDER BY v <-> '[0,0,0]';
|
||||
));
|
||||
|
||||
# Run twice to make sure correct tuples marked as dead
|
||||
for (1 .. 2)
|
||||
{
|
||||
my $res = $node->safe_psql("postgres", qq(
|
||||
SET enable_seqscan = off;
|
||||
SET ivfflat.probes = 100;
|
||||
SELECT i FROM tst ORDER BY v <-> '[0,0,0]';
|
||||
));
|
||||
is($res, $exp);
|
||||
}
|
||||
|
||||
done_testing();
|
||||
113
test/t/019_hnsw_array.pl
Normal file
113
test/t/019_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();
|
||||
@@ -1,64 +0,0 @@
|
||||
use strict;
|
||||
use warnings;
|
||||
use PostgresNode;
|
||||
use TestLib;
|
||||
use Test::More;
|
||||
|
||||
# Initialize node
|
||||
my $node = get_new_node('node');
|
||||
$node->init;
|
||||
$node->start;
|
||||
|
||||
# Create table and index
|
||||
$node->safe_psql("postgres", "CREATE EXTENSION vector;");
|
||||
$node->safe_psql("postgres", "CREATE TABLE tst (v vector(3));");
|
||||
$node->safe_psql("postgres",
|
||||
"INSERT INTO tst SELECT ARRAY[random(), random(), random()] FROM generate_series(1, 1000) i;"
|
||||
);
|
||||
$node->safe_psql("postgres", "CREATE INDEX ON tst USING ivfflat (v vector_l2_ops) WITH (lists = 10);");
|
||||
|
||||
# Test limit
|
||||
my $explain = $node->safe_psql("postgres", qq(
|
||||
EXPLAIN ANALYZE SELECT * FROM tst ORDER BY v <-> '[1,2,3]' LIMIT 100;
|
||||
));
|
||||
like($explain, qr/Index Scan/);
|
||||
|
||||
# Test limit with probes
|
||||
$explain = $node->safe_psql("postgres", qq(
|
||||
SET ivfflat.probes = 2;
|
||||
EXPLAIN ANALYZE SELECT * FROM tst ORDER BY v <-> '[1,2,3]' LIMIT 200;
|
||||
));
|
||||
like($explain, qr/Index Scan/);
|
||||
|
||||
# Test limit + offset
|
||||
$explain = $node->safe_psql("postgres", qq(
|
||||
EXPLAIN ANALYZE SELECT * FROM tst ORDER BY v <-> '[1,2,3]' LIMIT 90 OFFSET 10;
|
||||
));
|
||||
like($explain, qr/Index Scan/);
|
||||
|
||||
# Test limit > expected tuples
|
||||
$explain = $node->safe_psql("postgres", qq(
|
||||
EXPLAIN ANALYZE SELECT * FROM tst ORDER BY v <-> '[1,2,3]' LIMIT 101;
|
||||
));
|
||||
like($explain, qr/Seq Scan/);
|
||||
|
||||
# Test limit > expected tuples with probes
|
||||
$explain = $node->safe_psql("postgres", qq(
|
||||
SET ivfflat.probes = 2;
|
||||
EXPLAIN ANALYZE SELECT * FROM tst ORDER BY v <-> '[1,2,3]' LIMIT 201;
|
||||
));
|
||||
like($explain, qr/Seq Scan/);
|
||||
|
||||
# Test limit + offset > expected tuples
|
||||
$explain = $node->safe_psql("postgres", qq(
|
||||
EXPLAIN ANALYZE SELECT * FROM tst ORDER BY v <-> '[1,2,3]' LIMIT 91 OFFSET 10;
|
||||
));
|
||||
like($explain, qr/Seq Scan/);
|
||||
|
||||
# Test no limit
|
||||
$explain = $node->safe_psql("postgres", qq(
|
||||
EXPLAIN ANALYZE SELECT * FROM tst ORDER BY v <-> '[1,2,3]';
|
||||
));
|
||||
like($explain, qr/Seq Scan/);
|
||||
|
||||
done_testing();
|
||||
@@ -1,62 +0,0 @@
|
||||
use strict;
|
||||
use warnings;
|
||||
use PostgresNode;
|
||||
use TestLib;
|
||||
use Test::More;
|
||||
|
||||
# Initialize node
|
||||
my $node = get_new_node('node');
|
||||
$node->init;
|
||||
$node->start;
|
||||
|
||||
# Create table and index
|
||||
$node->safe_psql("postgres", "CREATE EXTENSION vector;");
|
||||
$node->safe_psql("postgres", "CREATE TABLE tst (v vector(3));");
|
||||
$node->safe_psql("postgres",
|
||||
"INSERT INTO tst SELECT ARRAY[random(), random(), random()] FROM generate_series(1, 1000) i;"
|
||||
);
|
||||
$node->safe_psql("postgres", "CREATE INDEX ON tst USING hnsw (v vector_l2_ops);");
|
||||
|
||||
# Test limit
|
||||
my $explain = $node->safe_psql("postgres", qq(
|
||||
EXPLAIN ANALYZE SELECT * FROM tst ORDER BY v <-> '[1,2,3]' LIMIT 40;
|
||||
));
|
||||
like($explain, qr/Index Scan/);
|
||||
|
||||
# Test limit with CTE
|
||||
$explain = $node->safe_psql("postgres", qq(
|
||||
EXPLAIN ANALYZE WITH cte AS (SELECT * FROM tst ORDER BY v <-> '[1,2,3]' LIMIT 40) SELECT * FROM cte;
|
||||
));
|
||||
like($explain, qr/Index Scan/);
|
||||
|
||||
# Test limit + offset
|
||||
$explain = $node->safe_psql("postgres", qq(
|
||||
EXPLAIN ANALYZE SELECT * FROM tst ORDER BY v <-> '[1,2,3]' LIMIT 30 OFFSET 10;
|
||||
));
|
||||
like($explain, qr/Index Scan/);
|
||||
|
||||
# Test limit > ef_search
|
||||
$explain = $node->safe_psql("postgres", qq(
|
||||
EXPLAIN ANALYZE SELECT * FROM tst ORDER BY v <-> '[1,2,3]' LIMIT 41;
|
||||
));
|
||||
like($explain, qr/Seq Scan/);
|
||||
|
||||
# Test limit > ef_search with CTE
|
||||
$explain = $node->safe_psql("postgres", qq(
|
||||
EXPLAIN ANALYZE WITH cte AS (SELECT * FROM tst ORDER BY v <-> '[1,2,3]' LIMIT 41) SELECT * FROM cte;
|
||||
));
|
||||
like($explain, qr/Seq Scan/);
|
||||
|
||||
# Test limit + offset > ef_search
|
||||
$explain = $node->safe_psql("postgres", qq(
|
||||
EXPLAIN ANALYZE SELECT * FROM tst ORDER BY v <-> '[1,2,3]' LIMIT 31 OFFSET 10;
|
||||
));
|
||||
like($explain, qr/Seq Scan/);
|
||||
|
||||
# Test no limit
|
||||
$explain = $node->safe_psql("postgres", qq(
|
||||
EXPLAIN ANALYZE SELECT * FROM tst ORDER BY v <-> '[1,2,3]';
|
||||
));
|
||||
like($explain, qr/Seq Scan/);
|
||||
|
||||
done_testing();
|
||||
Reference in New Issue
Block a user