Init leader

.github/workflows/build.yml

@@ -8,8 +8,6 @@ jobs:
       fail-fast: false
       matrix:
         include:
-          - postgres: 17
-            os: ubuntu-22.04
           - postgres: 16
             os: ubuntu-22.04
           - postgres: 15
@@ -20,15 +18,17 @@ jobs:
             os: ubuntu-20.04
           - postgres: 12
             os: ubuntu-20.04
+          - 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 }}
           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
@@ -43,13 +43,13 @@ 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
       - 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() }}
@@ -57,23 +57,20 @@ jobs:
       - run: |
           brew install cpanm
           cpanm --notest IPC::Run
-          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"
+          wget -q https://github.com/postgres/postgres/archive/refs/tags/REL_14_5.tar.gz
+          tar xf REL_14_5.tar.gz
+      - run: make prove_installcheck PROVE_FLAGS="-I ./postgres-REL_14_5/src/test/perl" PERL5LIB="/Users/runner/perl5/lib/perl5"
       - run: make clean && /usr/local/opt/llvm@15/bin/scan-build --status-bugs make
-        env:
-          PG_CFLAGS: -DUSE_ASSERT_CHECKING
   windows:
     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
       - 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 && ^
@@ -84,10 +81,10 @@ jobs:
     if: ${{ !startsWith(github.ref_name, 'mac') && !startsWith(github.ref_name, 'windows') }}
     runs-on: ubuntu-latest
     container:
-      image: debian:12
+      image: debian:11
       options: --platform linux/386
     steps:
-      - run: apt-get update && apt-get install -y build-essential git libipc-run-perl postgresql-15 postgresql-server-dev-15 sudo
+      - run: apt-get update && apt-get install -y build-essential git libipc-run-perl postgresql-13 postgresql-server-dev-13 sudo
       - run: service postgresql start
       - run: |
           git clone https://github.com/${{ github.repository }}.git pgvector
@@ -100,15 +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

CHANGELOG.md

@@ -1,38 +1,3 @@
-## 0.7.0 (unreleased)
-
-- Added `sparsevec` type
-
-## 0.6.2 (2024-03-18)
-
-- Reduced lock contention with parallel HNSW index builds
-
-## 0.6.1 (2024-03-04)
-
-- 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)
-
-If upgrading with Postgres 12 or Docker, see [these notes](https://github.com/pgvector/pgvector#060).
-
-- Added support for parallel index builds for HNSW
-- Added validation for GUC parameters
-- Changed storage for vector from `extended` to `external`
-- Improved performance of 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 builds
-- Moved Docker image to `pgvector` org
-- Added Docker tags for each supported version of Postgres
-- Dropped support for Postgres 11
-
-## 0.5.1 (2023-10-10)
-
-- Improved performance of HNSW index builds
-- Added check for MVCC-compliant snapshot for index scans
-
 ## 0.5.0 (2023-08-28)
 
 - Added HNSW index type
@@ -74,7 +39,7 @@ If upgrading with Postgres 12 or Docker, see [these notes](https://github.com/pg
 
 ## 0.4.0 (2023-01-11)
 
-If upgrading with Postgres < 13, see [this note](https://github.com/pgvector/pgvector/blob/v0.4.0/README.md#040).
+If upgrading with Postgres < 13, see [this note](https://github.com/pgvector/pgvector#040).
 
 - Changed text representation for vector elements to match `real`
 - Changed storage for vector from `plain` to `extended`
@@ -91,7 +56,7 @@ If upgrading with Postgres < 13, see [this note](https://github.com/pgvector/pgv
 
 ## 0.3.1 (2022-11-02)
 
-If upgrading from 0.2.7 or 0.3.0, [recreate](https://github.com/pgvector/pgvector/blob/v0.3.1/README.md#031) all `ivfflat` indexes after upgrading to ensure all data is indexed.
+If upgrading from 0.2.7 or 0.3.0, [recreate](https://github.com/pgvector/pgvector#031) all `ivfflat` indexes after upgrading to ensure all data is indexed.
 
 - Fixed issue with inserts silently corrupting `ivfflat` indexes (introduced in 0.2.7)
 - Fixed segmentation fault with index creation when lists > 6500

Dockerfile

@@ -1,4 +1,4 @@
-ARG PG_MAJOR=16
+ARG PG_MAJOR=15
 FROM postgres:$PG_MAJOR
 ARG PG_MAJOR
 

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
 

META.json

@@ -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": "0.5.0",
 	"maintainer": [
 		"Andrew Kane <andrew@ankane.org>"
 	],
@@ -12,7 +12,7 @@
 	"prereqs": {
 		"runtime": {
 			"requires": {
-				"PostgreSQL": "12.0.0"
+				"PostgreSQL": "11.0.0"
 			}
 		}
 	},
@@ -20,7 +20,7 @@
 		"vector": {
 			"file": "sql/vector.sql",
 			"docfile": "README.md",
-			"version": "0.6.2",
+			"version": "0.5.0",
 			"abstract": "Open-source vector similarity search for Postgres"
 		}
 	},

Makefile

@@ -1,14 +1,14 @@
 EXTENSION = vector
-EXTVERSION = 0.6.2
+EXTVERSION = 0.5.0
 
 MODULE_big = vector
 DATA = $(wildcard sql/*--*.sql)
-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/sparsevec.o src/vector.o
-HEADERS = src/sparsevec.h src/vector.h
+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)
 REGRESS = $(patsubst test/sql/%.sql,%,$(TESTS))
-REGRESS_OPTS = --inputdir=test --load-extension=$(EXTENSION)
+REGRESS_OPTS = --inputdir=test --load-extension=vector
 
 OPTFLAGS = -march=native
 
@@ -65,15 +65,13 @@ dist:
 	mkdir -p dist
 	git archive --format zip --prefix=$(EXTENSION)-$(EXTVERSION)/ --output dist/$(EXTENSION)-$(EXTVERSION).zip master
 
-# for Docker
-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 --platform linux/amd64 -t ankane/pgvector:latest .
 
 .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 -t ankane/pgvector:latest .
+	docker buildx build --push --platform linux/amd64,linux/arm64 -t ankane/pgvector:v$(EXTVERSION) .

Makefile.win

@@ -1,11 +1,11 @@
 EXTENSION = vector
-EXTVERSION = 0.6.2
+EXTVERSION = 0.5.0
 
-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\sparsevec.obj src\vector.obj
-HEADERS = src\sparsevec.h src\vector.h
+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 = btree cast copy functions input ivfflat_cosine ivfflat_ip ivfflat_l2 ivfflat_options ivfflat_unlogged
-REGRESS_OPTS = --inputdir=test --load-extension=$(EXTENSION)
+REGRESS_OPTS = --inputdir=test --load-extension=vector
 
 # For /arch flags
 # https://learn.microsoft.com/en-us/cpp/build/reference/arch-minimum-cpu-architecture
@@ -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)

README.md

@@ -10,50 +10,23 @@ 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
 
-[![Build Status](https://github.com/pgvector/pgvector/actions/workflows/build.yml/badge.svg)](https://github.com/pgvector/pgvector/actions)
+[![Build Status](https://github.com/pgvector/pgvector/workflows/build/badge.svg?branch=master)](https://github.com/pgvector/pgvector/actions)
 
 ## Installation
 
-### Linux and Mac
-
-Compile and install the extension (supports Postgres 12+)
+Compile and install the extension (supports Postgres 11+)
 
 ```sh
 cd /tmp
-git clone --branch v0.6.2 https://github.com/pgvector/pgvector.git
+git clone --branch v0.5.0 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).
-
-### Windows
-
-Ensure [C++ support in Visual Studio](https://learn.microsoft.com/en-us/cpp/build/building-on-the-command-line?view=msvc-170#download-and-install-the-tools) is installed, and run:
-
-```cmd
-call "C:\Program Files\Microsoft Visual Studio\2022\Community\VC\Auxiliary\Build\vcvars64.bat"
-```
-
-Note: The exact path will vary depending on your Visual Studio version and edition
-
-Then use `nmake` to build:
-
-```cmd
-set "PGROOT=C:\Program Files\PostgreSQL\16"
-cd %TEMP%
-git clone --branch v0.6.2 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).
+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
 
@@ -105,12 +78,6 @@ 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
@@ -194,107 +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
-
-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
-SET max_parallel_maintenance_workers = 7; -- plus leader
-```
-
-For a large number of workers, you may also need to increase `max_parallel_workers` (8 by default)
-
-### 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
 
@@ -347,32 +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
 
@@ -391,7 +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
@@ -402,62 +317,22 @@ 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
 
-### 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`.
 
@@ -471,7 +346,7 @@ 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).
 
@@ -479,91 +354,23 @@ 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 can take a while for HNSW indexes. Speed it up by reindexing first.
-
-```sql
-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)).
-
-## Sparse Vectors
-
-Create a sparse vector column with 10 dimensions
-
-```sql
-CREATE TABLE items (id bigserial PRIMARY KEY, embedding sparsevec(10));
-```
-
-Insert vectors
-
-```sql
-INSERT INTO items (embedding) VALUES ('{0:1,1:2,2:3}/10'), ('{0:4,1:5,2:6}/10');
-```
-
-Get the nearest neighbors by L2 distance
-
-```sql
-SELECT * FROM items ORDER BY embedding <-> '{0:3,1:1,2:2}/10' LIMIT 5;
-```
-
 ## 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.
 
 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)
@@ -571,7 +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
 
@@ -587,90 +393,11 @@ 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;
-```
-
-#### 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:
-
-```sql
-SELECT pg_size_pretty(pg_relation_size('index_name'));
-```
-
 ## Troubleshooting
 
 #### 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;
@@ -679,8 +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:
@@ -699,12 +424,6 @@ 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.
@@ -713,15 +432,11 @@ 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 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
 
@@ -745,21 +460,21 @@ l1_distance(vector, vector) → double precision | taxicab distance | 0.5.0
 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
 
-## Installation Notes - Linux and Mac
+## Installation Notes
 
 ### Postgres Location
 
 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:
@@ -768,14 +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.
@@ -783,53 +490,49 @@ 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
+Note: Replace `15` with your Postgres server version
 
-### Missing SDK
+### Windows
 
-If compilation fails and the output includes `warning: no such sysroot directory` on Mac, reinstall Xcode Command Line Tools.
+Support for Windows is currently experimental. Ensure [C++ support in Visual Studio](https://learn.microsoft.com/en-us/cpp/build/building-on-the-command-line?view=msvc-170#download-and-install-the-tools) is installed, and run:
 
-### 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=""
+```cmd
+call "C:\Program Files\Microsoft Visual Studio\2022\Community\VC\Auxiliary\Build\vcvars64.bat"
 ```
 
-## Installation Notes - Windows
+Note: The exact path will vary depending on your Visual Studio version and edition
 
-### Missing Header
+Then use `nmake` to build:
 
-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.
+```cmd
+set "PGROOT=C:\Program Files\PostgreSQL\15"
+git clone --branch v0.5.0 https://github.com/pgvector/pgvector.git
+cd pgvector
+nmake /F Makefile.win
+nmake /F Makefile.win install
+```
 
 ## Additional Installation Methods
 
 ### Docker
 
-Get the [Docker image](https://hub.docker.com/r/pgvector/pgvector) with:
+Get the [Docker image](https://hub.docker.com/r/ankane/pgvector) with:
 
 ```sh
-docker pull pgvector/pgvector:pg16
+docker pull ankane/pgvector
 ```
 
-This adds pgvector to the [Postgres image](https://hub.docker.com/_/postgres) (replace `16` with your Postgres server version, and run it the same way).
+This adds pgvector to the [Postgres image](https://hub.docker.com/_/postgres) (run it the same way).
 
 You can also build the image manually:
 
 ```sh
-git clone --branch v0.6.2 https://github.com/pgvector/pgvector.git
+git clone --branch v0.5.0 https://github.com/pgvector/pgvector.git
 cd pgvector
-docker build --build-arg PG_MAJOR=16 -t myuser/pgvector .
+docker build --build-arg PG_MAJOR=15 -t myuser/pgvector .
 ```
 
 ### Homebrew
@@ -855,37 +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
-
-### pkg
-
-Install the FreeBSD package with:
-
-```sh
-pkg install postgresql15-pg_vector
-```
-
-or the port with:
-
-```sh
-cd /usr/ports/databases/pgvector
-make install
-```
+Note: Replace `15` with your Postgres server version
 
 ### conda-forge
 
@@ -907,46 +595,36 @@ 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 and run:
 
 ```sql
 ALTER EXTENSION vector UPDATE;
 ```
 
-You can check the version in the current database with:
-
-```sql
-SELECT extversion FROM pg_extension WHERE extname = 'vector';
-```
-
 ## Upgrade Notes
 
-### 0.6.0
+### 0.4.0
 
-#### Postgres 12
-
-If upgrading with Postgres 12, remove this line from `sql/vector--0.5.1--0.6.0.sql`:
+If upgrading with Postgres < 13, remove this line from `sql/vector--0.3.2--0.4.0.sql`:
 
 ```sql
-ALTER TYPE vector SET (STORAGE = external);
+ALTER TYPE vector SET (STORAGE = extended);
 ```
 
 Then run `make install` and `ALTER EXTENSION vector UPDATE;`.
 
-#### Docker
+### 0.3.1
 
-The Docker image is now published in the `pgvector` org, and there are tags for each supported version of Postgres (rather than a `latest` tag).
+If upgrading from 0.2.7 or 0.3.0, recreate all `ivfflat` indexes after upgrading to ensure all data is indexed.
 
-```sh
-docker pull pgvector/pgvector:pg16
-# or
-docker pull pgvector/pgvector:0.6.0-pg16
-```
+```sql
+-- Postgres 12+
+REINDEX INDEX CONCURRENTLY index_name;
 
-Also, if you’ve increased `maintenance_work_mem`, make sure `--shm-size` is at least that size to avoid an error with parallel HNSW index builds.
-
-```sh
-docker run --shm-size=1g ...
+-- Postgres < 12
+CREATE INDEX CONCURRENTLY temp_name ON table USING ivfflat (column opclass);
+DROP INDEX CONCURRENTLY index_name;
+ALTER INDEX temp_name RENAME TO index_name;
 ```
 
 ## Thanks
@@ -992,32 +670,14 @@ make prove_installcheck  # TAP tests
 To run single tests:
 
 ```sh
-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
+make installcheck REGRESS=functions                    # regression test
+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 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

sql/vector--0.5.0--0.5.1.sql

@@ -1,2 +0,0 @@
--- complain if script is sourced in psql, rather than via CREATE EXTENSION
-\echo Use "ALTER EXTENSION vector UPDATE TO '0.5.1'" to load this file. \quit

sql/vector--0.5.1--0.6.0.sql

@@ -1,5 +0,0 @@
--- complain if script is sourced in psql, rather than via CREATE EXTENSION
-\echo Use "ALTER EXTENSION vector UPDATE TO '0.6.0'" to load this file. \quit
-
--- remove this single line for Postgres < 13
-ALTER TYPE vector SET (STORAGE = external);

sql/vector--0.6.0--0.6.1.sql

@@ -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
-);

sql/vector--0.6.1--0.6.2.sql

@@ -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

sql/vector--0.6.2--0.7.0.sql

@@ -1,95 +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 TYPE sparsevec;
-
-CREATE FUNCTION sparsevec_in(cstring, oid, integer) RETURNS sparsevec
-	AS 'MODULE_PATHNAME' LANGUAGE C IMMUTABLE STRICT PARALLEL SAFE;
-
-CREATE FUNCTION sparsevec_out(sparsevec) RETURNS cstring
-	AS 'MODULE_PATHNAME' LANGUAGE C IMMUTABLE STRICT PARALLEL SAFE;
-
-CREATE FUNCTION sparsevec_typmod_in(cstring[]) RETURNS integer
-	AS 'MODULE_PATHNAME' LANGUAGE C IMMUTABLE STRICT PARALLEL SAFE;
-
-CREATE FUNCTION sparsevec_recv(internal, oid, integer) RETURNS sparsevec
-	AS 'MODULE_PATHNAME' LANGUAGE C IMMUTABLE STRICT PARALLEL SAFE;
-
-CREATE FUNCTION sparsevec_send(sparsevec) RETURNS bytea
-	AS 'MODULE_PATHNAME' LANGUAGE C IMMUTABLE STRICT PARALLEL SAFE;
-
-CREATE TYPE sparsevec (
-	INPUT     = sparsevec_in,
-	OUTPUT    = sparsevec_out,
-	TYPMOD_IN = sparsevec_typmod_in,
-	RECEIVE   = sparsevec_recv,
-	SEND      = sparsevec_send,
-	STORAGE   = external
-);
-
-CREATE FUNCTION l2_distance(sparsevec, sparsevec) RETURNS float8
-	AS 'MODULE_PATHNAME', 'sparsevec_l2_distance' LANGUAGE C IMMUTABLE STRICT PARALLEL SAFE;
-
-CREATE FUNCTION inner_product(sparsevec, sparsevec) RETURNS float8
-	AS 'MODULE_PATHNAME', 'sparsevec_inner_product' LANGUAGE C IMMUTABLE STRICT PARALLEL SAFE;
-
-CREATE FUNCTION cosine_distance(sparsevec, sparsevec) RETURNS float8
-	AS 'MODULE_PATHNAME', 'sparsevec_cosine_distance' LANGUAGE C IMMUTABLE STRICT PARALLEL SAFE;
-
-CREATE FUNCTION sparsevec_norm(sparsevec) RETURNS float8
-	AS 'MODULE_PATHNAME' LANGUAGE C IMMUTABLE STRICT PARALLEL SAFE;
-
-CREATE FUNCTION sparsevec_l2_squared_distance(sparsevec, sparsevec) RETURNS float8
-	AS 'MODULE_PATHNAME' LANGUAGE C IMMUTABLE STRICT PARALLEL SAFE;
-
-CREATE FUNCTION sparsevec_negative_inner_product(sparsevec, sparsevec) RETURNS float8
-	AS 'MODULE_PATHNAME' LANGUAGE C IMMUTABLE STRICT PARALLEL SAFE;
-
-CREATE FUNCTION sparsevec(sparsevec, integer, boolean) RETURNS sparsevec
-	AS 'MODULE_PATHNAME' LANGUAGE C IMMUTABLE STRICT PARALLEL SAFE;
-
-CREATE FUNCTION vector_to_sparsevec(vector, integer, boolean) RETURNS sparsevec
-	AS 'MODULE_PATHNAME' LANGUAGE C IMMUTABLE STRICT PARALLEL SAFE;
-
-CREATE FUNCTION sparsevec_to_vector(sparsevec, integer, boolean) RETURNS vector
-	AS 'MODULE_PATHNAME' LANGUAGE C IMMUTABLE STRICT PARALLEL SAFE;
-
-CREATE CAST (sparsevec AS sparsevec)
-	WITH FUNCTION sparsevec(sparsevec, integer, boolean) AS IMPLICIT;
-
-CREATE CAST (sparsevec AS vector)
-	WITH FUNCTION sparsevec_to_vector(sparsevec, integer, boolean) AS IMPLICIT;
-
-CREATE CAST (vector AS sparsevec)
-	WITH FUNCTION vector_to_sparsevec(vector, integer, boolean) AS IMPLICIT;
-
-CREATE OPERATOR <-> (
-	LEFTARG = sparsevec, RIGHTARG = sparsevec, PROCEDURE = l2_distance,
-	COMMUTATOR = '<->'
-);
-
-CREATE OPERATOR <#> (
-	LEFTARG = sparsevec, RIGHTARG = sparsevec, PROCEDURE = sparsevec_negative_inner_product,
-	COMMUTATOR = '<#>'
-);
-
-CREATE OPERATOR <=> (
-	LEFTARG = sparsevec, RIGHTARG = sparsevec, PROCEDURE = cosine_distance,
-	COMMUTATOR = '<=>'
-);
-
-CREATE OPERATOR CLASS sparsevec_l2_ops
-	FOR TYPE sparsevec USING hnsw AS
-	OPERATOR 1 <-> (sparsevec, sparsevec) FOR ORDER BY float_ops,
-	FUNCTION 1 sparsevec_l2_squared_distance(sparsevec, sparsevec);
-
-CREATE OPERATOR CLASS sparsevec_ip_ops
-	FOR TYPE sparsevec USING hnsw AS
-	OPERATOR 1 <#> (sparsevec, sparsevec) FOR ORDER BY float_ops,
-	FUNCTION 1 sparsevec_negative_inner_product(sparsevec, sparsevec);
-
-CREATE OPERATOR CLASS sparsevec_cosine_ops
-	FOR TYPE sparsevec USING hnsw AS
-	OPERATOR 1 <=> (sparsevec, sparsevec) FOR ORDER BY float_ops,
-	FUNCTION 1 sparsevec_negative_inner_product(sparsevec, sparsevec),
-	FUNCTION 2 sparsevec_norm(sparsevec);

sql/vector.sql

@@ -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;
 
@@ -26,10 +26,10 @@ CREATE TYPE vector (
 	TYPMOD_IN = vector_typmod_in,
 	RECEIVE   = vector_recv,
 	SEND      = vector_send,
-	STORAGE   = external
+	STORAGE   = extended
 );
 
--- vector functions
+-- functions
 
 CREATE FUNCTION l2_distance(vector, vector) RETURNS float8
 	AS 'MODULE_PATHNAME' LANGUAGE C IMMUTABLE STRICT PARALLEL SAFE;
@@ -58,7 +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;
 
--- vector private functions
+-- private functions
 
 CREATE FUNCTION vector_lt(vector, vector) RETURNS bool
 	AS 'MODULE_PATHNAME' LANGUAGE C IMMUTABLE STRICT PARALLEL SAFE;
@@ -99,7 +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;
 
--- vector aggregates
+-- aggregates
 
 CREATE AGGREGATE avg(vector) (
 	SFUNC = vector_accum,
@@ -117,7 +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;
@@ -137,7 +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;
@@ -157,7 +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,
@@ -180,7 +180,8 @@ CREATE OPERATOR + (
 );
 
 CREATE OPERATOR - (
-	LEFTARG = vector, RIGHTARG = vector, PROCEDURE = vector_sub
+	LEFTARG = vector, RIGHTARG = vector, PROCEDURE = vector_sub,
+	COMMUTATOR = -
 );
 
 CREATE OPERATOR * (
@@ -194,10 +195,11 @@ CREATE OPERATOR < (
 	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 = (
@@ -212,10 +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 > (
@@ -240,7 +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
@@ -287,110 +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);
-
---- sparsevec type
-
-CREATE TYPE sparsevec;
-
-CREATE FUNCTION sparsevec_in(cstring, oid, integer) RETURNS sparsevec
-	AS 'MODULE_PATHNAME' LANGUAGE C IMMUTABLE STRICT PARALLEL SAFE;
-
-CREATE FUNCTION sparsevec_out(sparsevec) RETURNS cstring
-	AS 'MODULE_PATHNAME' LANGUAGE C IMMUTABLE STRICT PARALLEL SAFE;
-
-CREATE FUNCTION sparsevec_typmod_in(cstring[]) RETURNS integer
-	AS 'MODULE_PATHNAME' LANGUAGE C IMMUTABLE STRICT PARALLEL SAFE;
-
-CREATE FUNCTION sparsevec_recv(internal, oid, integer) RETURNS sparsevec
-	AS 'MODULE_PATHNAME' LANGUAGE C IMMUTABLE STRICT PARALLEL SAFE;
-
-CREATE FUNCTION sparsevec_send(sparsevec) RETURNS bytea
-	AS 'MODULE_PATHNAME' LANGUAGE C IMMUTABLE STRICT PARALLEL SAFE;
-
-CREATE TYPE sparsevec (
-	INPUT     = sparsevec_in,
-	OUTPUT    = sparsevec_out,
-	TYPMOD_IN = sparsevec_typmod_in,
-	RECEIVE   = sparsevec_recv,
-	SEND      = sparsevec_send,
-	STORAGE   = external
-);
-
--- sparsevec functions
-
-CREATE FUNCTION l2_distance(sparsevec, sparsevec) RETURNS float8
-	AS 'MODULE_PATHNAME', 'sparsevec_l2_distance' LANGUAGE C IMMUTABLE STRICT PARALLEL SAFE;
-
-CREATE FUNCTION inner_product(sparsevec, sparsevec) RETURNS float8
-	AS 'MODULE_PATHNAME', 'sparsevec_inner_product' LANGUAGE C IMMUTABLE STRICT PARALLEL SAFE;
-
-CREATE FUNCTION cosine_distance(sparsevec, sparsevec) RETURNS float8
-	AS 'MODULE_PATHNAME', 'sparsevec_cosine_distance' LANGUAGE C IMMUTABLE STRICT PARALLEL SAFE;
-
-CREATE FUNCTION sparsevec_norm(sparsevec) RETURNS float8
-	AS 'MODULE_PATHNAME' LANGUAGE C IMMUTABLE STRICT PARALLEL SAFE;
-
--- sparsevec private functions
-
-CREATE FUNCTION sparsevec_l2_squared_distance(sparsevec, sparsevec) RETURNS float8
-	AS 'MODULE_PATHNAME' LANGUAGE C IMMUTABLE STRICT PARALLEL SAFE;
-
-CREATE FUNCTION sparsevec_negative_inner_product(sparsevec, sparsevec) RETURNS float8
-	AS 'MODULE_PATHNAME' LANGUAGE C IMMUTABLE STRICT PARALLEL SAFE;
-
--- sparsevec cast functions
-
-CREATE FUNCTION sparsevec(sparsevec, integer, boolean) RETURNS sparsevec
-	AS 'MODULE_PATHNAME' LANGUAGE C IMMUTABLE STRICT PARALLEL SAFE;
-
-CREATE FUNCTION vector_to_sparsevec(vector, integer, boolean) RETURNS sparsevec
-	AS 'MODULE_PATHNAME' LANGUAGE C IMMUTABLE STRICT PARALLEL SAFE;
-
-CREATE FUNCTION sparsevec_to_vector(sparsevec, integer, boolean) RETURNS vector
-	AS 'MODULE_PATHNAME' LANGUAGE C IMMUTABLE STRICT PARALLEL SAFE;
-
--- sparsevec casts
-
-CREATE CAST (sparsevec AS sparsevec)
-	WITH FUNCTION sparsevec(sparsevec, integer, boolean) AS IMPLICIT;
-
-CREATE CAST (sparsevec AS vector)
-	WITH FUNCTION sparsevec_to_vector(sparsevec, integer, boolean) AS IMPLICIT;
-
-CREATE CAST (vector AS sparsevec)
-	WITH FUNCTION vector_to_sparsevec(vector, integer, boolean) AS IMPLICIT;
-
--- sparsevec operators
-
-CREATE OPERATOR <-> (
-	LEFTARG = sparsevec, RIGHTARG = sparsevec, PROCEDURE = l2_distance,
-	COMMUTATOR = '<->'
-);
-
-CREATE OPERATOR <#> (
-	LEFTARG = sparsevec, RIGHTARG = sparsevec, PROCEDURE = sparsevec_negative_inner_product,
-	COMMUTATOR = '<#>'
-);
-
-CREATE OPERATOR <=> (
-	LEFTARG = sparsevec, RIGHTARG = sparsevec, PROCEDURE = cosine_distance,
-	COMMUTATOR = '<=>'
-);
-
--- sparsevec opclasses
-
-CREATE OPERATOR CLASS sparsevec_l2_ops
-	FOR TYPE sparsevec USING hnsw AS
-	OPERATOR 1 <-> (sparsevec, sparsevec) FOR ORDER BY float_ops,
-	FUNCTION 1 sparsevec_l2_squared_distance(sparsevec, sparsevec);
-
-CREATE OPERATOR CLASS sparsevec_ip_ops
-	FOR TYPE sparsevec USING hnsw AS
-	OPERATOR 1 <#> (sparsevec, sparsevec) FOR ORDER BY float_ops,
-	FUNCTION 1 sparsevec_negative_inner_product(sparsevec, sparsevec);
-
-CREATE OPERATOR CLASS sparsevec_cosine_ops
-	FOR TYPE sparsevec USING hnsw AS
-	OPERATOR 1 <=> (sparsevec, sparsevec) FOR ORDER BY float_ops,
-	FUNCTION 1 sparsevec_negative_inner_product(sparsevec, sparsevec),
-	FUNCTION 2 sparsevec_norm(sparsevec);

src/hnsw.c

@@ -4,59 +4,24 @@
 #include <math.h>
 
 #include "access/amapi.h"
-#include "access/reloptions.h"
-#include "commands/progress.h"
 #include "commands/vacuum.h"
 #include "hnsw.h"
-#include "miscadmin.h"
 #include "utils/guc.h"
 #include "utils/selfuncs.h"
 
-#if PG_VERSION_NUM < 150000
-#define MarkGUCPrefixReserved(x) EmitWarningsOnPlaceholders(x)
+#if PG_VERSION_NUM >= 120000
+#include "commands/progress.h"
 #endif
 
 int			hnsw_ef_search;
-int			hnsw_lock_tranche_id;
 static relopt_kind hnsw_relopt_kind;
 
-/*
- * Assign a tranche ID for our LWLocks. This only needs to be done by one
- * backend, as the tranche ID is remembered in shared memory.
- *
- * This shared memory area is very small, so we just allocate it from the
- * "slop" that PostgreSQL reserves for small allocations like this. If
- * this grows bigger, we should use a shmem_request_hook and
- * RequestAddinShmemSpace() to pre-reserve space for this.
- */
-void
-HnswInitLockTranche(void)
-{
-	int		   *tranche_ids;
-	bool		found;
-
-	LWLockAcquire(AddinShmemInitLock, LW_EXCLUSIVE);
-	tranche_ids = ShmemInitStruct("hnsw LWLock ids",
-								  sizeof(int) * 1,
-								  &found);
-	if (!found)
-		tranche_ids[0] = LWLockNewTrancheId();
-	hnsw_lock_tranche_id = tranche_ids[0];
-	LWLockRelease(AddinShmemInitLock);
-
-	/* Per-backend registration of the tranche ID */
-	LWLockRegisterTranche(hnsw_lock_tranche_id, "HnswBuild");
-}
-
 /*
  * Initialize index options and variables
  */
 void
 HnswInit(void)
 {
-	if (!process_shared_preload_libraries_in_progress)
-		HnswInitLockTranche();
-
 	hnsw_relopt_kind = add_reloption_kind();
 	add_int_reloption(hnsw_relopt_kind, "m", "Max number of connections",
 					  HNSW_DEFAULT_M, HNSW_MIN_M, HNSW_MAX_M
@@ -74,13 +39,12 @@ 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);
-
-	MarkGUCPrefixReserved("hnsw");
 }
 
 /*
  * Get the name of index build phase
  */
+#if PG_VERSION_NUM >= 120000
 static char *
 hnswbuildphasename(int64 phasenum)
 {
@@ -94,6 +58,7 @@ hnswbuildphasename(int64 phasenum)
 			return NULL;
 	}
 }
+#endif
 
 /*
  * Estimate the cost of an index scan
@@ -108,6 +73,9 @@ hnswcostestimate(PlannerInfo *root, IndexPath *path, double loop_count,
 	int			m;
 	int			entryLevel;
 	Relation	index;
+#if PG_VERSION_NUM < 120000
+	List	   *qinfos;
+#endif
 
 	/* Never use index without order */
 	if (path->indexorderbys == NULL)
@@ -123,7 +91,7 @@ hnswcostestimate(PlannerInfo *root, IndexPath *path, double loop_count,
 	MemSet(&costs, 0, sizeof(costs));
 
 	index = index_open(path->indexinfo->indexoid, NoLock);
-	HnswGetMetaPageInfo(index, &m, NULL);
+	m = HnswGetM(index);
 	index_close(index, NoLock);
 
 	/* Approximate entry level */
@@ -133,7 +101,12 @@ hnswcostestimate(PlannerInfo *root, IndexPath *path, double loop_count,
 	/* Account for number of tuples (or entry level), m, and ef_search */
 	costs.numIndexTuples = (entryLevel + 2) * m;
 
+#if PG_VERSION_NUM >= 120000
 	genericcostestimate(root, path, loop_count, &costs);
+#else
+	qinfos = deconstruct_indexquals(path);
+	genericcostestimate(root, path, loop_count, qinfos, &costs);
+#endif
 
 	/* Use total cost since most work happens before first tuple is returned */
 	*indexStartupCost = costs.indexTotalCost;
@@ -223,11 +196,13 @@ hnswhandler(PG_FUNCTION_ARGS)
 	amroutine->aminsert = hnswinsert;
 	amroutine->ambulkdelete = hnswbulkdelete;
 	amroutine->amvacuumcleanup = hnswvacuumcleanup;
-	amroutine->amcanreturn = NULL;
+	amroutine->amcanreturn = NULL;	/* tuple not included in heapsort */
 	amroutine->amcostestimate = hnswcostestimate;
 	amroutine->amoptions = hnswoptions;
 	amroutine->amproperty = NULL;	/* TODO AMPROP_DISTANCE_ORDERABLE */
+#if PG_VERSION_NUM >= 120000
 	amroutine->ambuildphasename = hnswbuildphasename;
+#endif
 	amroutine->amvalidate = hnswvalidate;
 #if PG_VERSION_NUM >= 140000
 	amroutine->amadjustmembers = NULL;

src/hnsw.h

@@ -3,21 +3,23 @@
 
 #include "postgres.h"
 
-#include "access/genam.h"
+#include "access/generic_xlog.h"
 #include "access/parallel.h"
-#include "lib/pairingheap.h"
+#include "access/reloptions.h"
 #include "nodes/execnodes.h"
 #include "port.h"				/* for random() */
-#include "utils/relptr.h"
 #include "utils/sampling.h"
 #include "vector.h"
 
+#if PG_VERSION_NUM < 110000
+#error "Requires PostgreSQL 11+"
+#endif
+
 #if PG_VERSION_NUM < 120000
-#error "Requires PostgreSQL 12+"
+#include "access/relscan.h"
 #endif
 
 #define HNSW_MAX_DIM 2000
-#define HNSW_MAX_NNZ 1000
 
 /* Support functions */
 #define HNSW_DISTANCE_PROC 1
@@ -56,38 +58,25 @@
 #define HNSW_UPDATE_ENTRY_GREATER 1
 #define HNSW_UPDATE_ENTRY_ALWAYS 2
 
-typedef enum HnswType
-{
-	HNSW_TYPE_VECTOR,
-	HNSW_TYPE_SPARSEVEC
-}			HnswType;
-
 /* Build phases */
 /* PROGRESS_CREATEIDX_SUBPHASE_INITIALIZE is 1 */
 #define PROGRESS_HNSW_PHASE_LOAD		2
 
-#define HNSW_MAX_SIZE (BLCKSZ - MAXALIGN(SizeOfPageHeaderData) - MAXALIGN(sizeof(HnswPageOpaqueData)) - sizeof(ItemIdData))
-#define HNSW_TUPLE_ALLOC_SIZE BLCKSZ
-
-#define HNSW_ELEMENT_TUPLE_SIZE(size)	MAXALIGN(offsetof(HnswElementTupleData, data) + (size))
+#define HNSW_ELEMENT_TUPLE_SIZE(_dim)	MAXALIGN(offsetof(HnswElementTupleData, vec) + VECTOR_SIZE(_dim))
 #define HNSW_NEIGHBOR_TUPLE_SIZE(level, m)	MAXALIGN(offsetof(HnswNeighborTupleData, indextids) + ((level) + 2) * (m) * sizeof(ItemPointerData))
 
-#define HNSW_NEIGHBOR_ARRAY_SIZE(lm)	(offsetof(HnswNeighborArray, items) + sizeof(HnswCandidate) * (lm))
-
 #define HnswPageGetOpaque(page)	((HnswPageOpaque) PageGetSpecialPointer(page))
 #define HnswPageGetMeta(page)	((HnswMetaPageData *) PageGetContents(page))
 
 #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
 #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)
@@ -100,74 +89,39 @@ typedef enum HnswType
 #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 HnswGetValue(base, element) PointerGetDatum(HnswPtrAccess(base, (element)->value))
-
-#if PG_VERSION_NUM < 140005
-#define relptr_offset(rp) ((rp).relptr_off - 1)
-#endif
-
-/* Pointer macros */
-#define HnswPtrAccess(base, hp) ((base) == NULL ? (hp).ptr : relptr_access(base, (hp).relptr))
-#define HnswPtrStore(base, hp, value) ((base) == NULL ? (void) ((hp).ptr = (value)) : (void) relptr_store(base, (hp).relptr, value))
-#define HnswPtrIsNull(base, hp) ((base) == NULL ? (hp).ptr == NULL : relptr_is_null((hp).relptr))
-#define HnswPtrEqual(base, hp1, hp2) ((base) == NULL ? (hp1).ptr == (hp2).ptr : relptr_offset((hp1).relptr) == relptr_offset((hp2).relptr))
-
-/* For code paths dedicated to each type */
-#define HnswPtrPointer(hp) (hp).ptr
-#define HnswPtrOffset(hp) relptr_offset((hp).relptr)
+#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 int	hnsw_lock_tranche_id;
 
-typedef struct HnswElementData HnswElementData;
 typedef struct HnswNeighborArray HnswNeighborArray;
 
-#define HnswPtrDeclare(type, relptrtype, ptrtype) \
-	relptr_declare(type, relptrtype); \
-	typedef union { type *ptr; relptrtype relptr; } ptrtype;
-
-/* Pointers that can be absolute or relative */
-/* Use char for DatumPtr so works with Pointer */
-HnswPtrDeclare(HnswElementData, HnswElementRelptr, HnswElementPtr);
-HnswPtrDeclare(HnswNeighborArray, HnswNeighborArrayRelptr, HnswNeighborArrayPtr);
-HnswPtrDeclare(HnswNeighborArrayPtr, HnswNeighborsRelptr, HnswNeighborsPtr);
-HnswPtrDeclare(char, DatumRelptr, DatumPtr);
-
-struct HnswElementData
+typedef struct HnswElementData
 {
-	HnswElementPtr next;
-	ItemPointerData heaptids[HNSW_HEAPTIDS];
-	uint8		heaptidsLength;
+	List	   *heaptids;
 	uint8		level;
 	uint8		deleted;
-	uint32		hash;
-	HnswNeighborsPtr neighbors;
+	HnswNeighborArray *neighbors;
 	BlockNumber blkno;
 	OffsetNumber offno;
 	OffsetNumber neighborOffno;
 	BlockNumber neighborPage;
-	DatumPtr	value;
-	LWLock		lock;
-};
+	Vector	   *vec;
+}			HnswElementData;
 
 typedef HnswElementData * HnswElement;
 
 typedef struct HnswCandidate
 {
-	HnswElementPtr element;
+	HnswElement element;
 	float		distance;
-	bool		closer;
 }			HnswCandidate;
 
-struct HnswNeighborArray
+typedef struct HnswNeighborArray
 {
 	int			length;
-	bool		closerSet;
-	HnswCandidate items[FLEXIBLE_ARRAY_MEMBER];
-};
+	HnswCandidate *items;
+}			HnswNeighborArray;
 
 typedef struct HnswPairingHeapNode
 {
@@ -183,27 +137,11 @@ typedef struct HnswOptions
 	int			efConstruction; /* size of dynamic candidate list */
 }			HnswOptions;
 
-typedef struct HnswGraph
+typedef struct HnswSpool
 {
-	/* Graph state */
-	slock_t		lock;
-	HnswElementPtr head;
-	double		indtuples;
-
-	/* Entry state */
-	LWLock		entryLock;
-	LWLock		entryWaitLock;
-	HnswElementPtr entryPoint;
-
-	/* Allocations state */
-	LWLock		allocatorLock;
-	long		memoryUsed;
-	long		memoryTotal;
-
-	/* Flushed state */
-	LWLock		flushLock;
-	bool		flushed;
-}			HnswGraph;
+	Relation	heap;
+	Relation	index;
+}			HnswSpool;
 
 typedef struct HnswShared
 {
@@ -211,6 +149,7 @@ typedef struct HnswShared
 	Oid			heaprelid;
 	Oid			indexrelid;
 	bool		isconcurrent;
+	int			scantuplesortstates;
 
 	/* Worker progress */
 	ConditionVariable workersdonecv;
@@ -221,11 +160,17 @@ typedef struct HnswShared
 	/* Mutable state */
 	int			nparticipantsdone;
 	double		reltuples;
-	HnswGraph	graphData;
+	double		indtuples;
+
+#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
 {
@@ -233,15 +178,8 @@ typedef struct HnswLeader
 	int			nparticipanttuplesorts;
 	HnswShared *hnswshared;
 	Snapshot	snapshot;
-	char	   *hnswarea;
 }			HnswLeader;
 
-typedef struct HnswAllocator
-{
-	void	   *(*alloc) (Size size, void *state);
-	void	   *state;
-}			HnswAllocator;
-
 typedef struct HnswBuildState
 {
 	/* Info */
@@ -249,7 +187,6 @@ typedef struct HnswBuildState
 	Relation	index;
 	IndexInfo  *indexInfo;
 	ForkNumber	forkNum;
-	HnswType	type;
 
 	/* Settings */
 	int			dimensions;
@@ -266,20 +203,19 @@ 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;
-	HnswAllocator allocator;
 
 	/* Parallel builds */
 	HnswLeader *hnswleader;
-	HnswShared *hnswshared;
-	char	   *hnswarea;
 }			HnswBuildState;
 
 typedef struct HnswMetaPageData
@@ -315,7 +251,7 @@ typedef struct HnswElementTupleData
 	ItemPointerData heaptids[HNSW_HEAPTIDS];
 	ItemPointerData neighbortid;
 	uint16		unused2;
-	Vector		data;
+	Vector		vec;
 }			HnswElementTupleData;
 
 typedef HnswElementTupleData * HnswElementTuple;
@@ -333,6 +269,7 @@ typedef HnswNeighborTupleData * HnswNeighborTuple;
 typedef struct HnswScanOpaqueData
 {
 	bool		first;
+	Buffer		buf;
 	List	   *w;
 	MemoryContext tmpCtx;
 
@@ -361,7 +298,7 @@ typedef struct HnswVacuumState
 	Oid			collation;
 
 	/* Variables */
-	struct tidhash_hash *deleted;
+	HTAB	   *deleted;
 	BufferAccessStrategy bas;
 	HnswNeighborTuple ntup;
 	HnswElementData highestPoint;
@@ -373,33 +310,32 @@ typedef struct HnswVacuumState
 /* Methods */
 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);
-void		HnswCheckValue(Datum value, HnswType type);
+FmgrInfo   *HnswOptionalProcInfo(Relation rel, 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(char *base, Datum q, List *ep, int ef, int lc, Relation index, FmgrInfo *procinfo, Oid collation, int m, bool inserting, HnswElement skipElement);
+List	   *HnswSearchLayer(Datum q, List *ep, int ef, int lc, Relation index, FmgrInfo *procinfo, Oid collation, bool inserting, HnswElement skipElement);
 HnswElement HnswGetEntryPoint(Relation index);
-void		HnswGetMetaPageInfo(Relation index, int *m, HnswElement * entryPoint);
-void	   *HnswAlloc(HnswAllocator * allocator, Size size);
-HnswElement HnswInitElement(char *base, ItemPointer tid, int m, double ml, int maxLevel, HnswAllocator * alloc);
+HnswElement HnswInitElement(ItemPointer tid, int m, double ml, int maxLevel);
+void		HnswFreeElement(HnswElement element);
 HnswElement HnswInitElementFromBlock(BlockNumber blkno, OffsetNumber offno);
-void		HnswFindElementNeighbors(char *base, HnswElement element, HnswElement entryPoint, Relation index, FmgrInfo *procinfo, Oid collation, int m, int efConstruction, bool existing);
-HnswCandidate *HnswEntryCandidate(char *base, 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		HnswSetNeighborTuple(char *base, HnswNeighborTuple ntup, HnswElement e, int m);
+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);
+void		HnswSetNeighborTuple(HnswNeighborTuple ntup, HnswElement e, int m);
 void		HnswAddHeapTid(HnswElement element, ItemPointer heaptid);
-void		HnswInitNeighbors(char *base, HnswElement element, int m, HnswAllocator * alloc);
-bool		HnswInsertTupleOnDisk(Relation index, Datum value, Datum *values, bool *isnull, ItemPointer heap_tid, bool building);
-void		HnswUpdateNeighborsOnDisk(Relation index, FmgrInfo *procinfo, Oid collation, HnswElement e, int m, bool checkExisting, bool building);
+void		HnswInitNeighbors(HnswElement element, int m);
+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(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);
+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);
 PGDLLEXPORT void HnswParallelBuildMain(dsm_segment *seg, shm_toc *toc);
 
 /* Index access methods */
@@ -418,54 +354,4 @@ void		hnswrescan(IndexScanDesc scan, ScanKey keys, int nkeys, ScanKey orderbys,
 bool		hnswgettuple(IndexScanDesc scan, ScanDirection dir);
 void		hnswendscan(IndexScanDesc scan);
 
-static inline HnswNeighborArray *
-HnswGetNeighbors(char *base, HnswElement element, int lc)
-{
-	HnswNeighborArrayPtr *neighborList = HnswPtrAccess(base, element->neighbors);
-
-	Assert(element->level >= lc);
-
-	return HnswPtrAccess(base, neighborList[lc]);
-}
-
-/* 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"
-
-typedef struct OffsetHashEntry
-{
-	Size		offset;
-	char		status;
-}			OffsetHashEntry;
-
-#define SH_PREFIX offsethash
-#define SH_ELEMENT_TYPE OffsetHashEntry
-#define SH_KEY_TYPE Size
-#define SH_SCOPE extern
-#define SH_DECLARE
-#include "lib/simplehash.h"
-
 #endif

src/hnswinsert.c

@@ -2,11 +2,9 @@
 
 #include <math.h>
 
-#include "access/generic_xlog.h"
 #include "hnsw.h"
 #include "storage/bufmgr.h"
 #include "storage/lmgr.h"
-#include "utils/datum.h"
 #include "utils/memutils.h"
 
 /*
@@ -94,7 +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);
@@ -102,11 +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 */
@@ -117,7 +111,7 @@ HnswInsertAppendPage(Relation index, Buffer *nbuf, Page *npage, GenericXLogState
  * Add to element and neighbor pages
  */
 static void
-AddElementOnDisk(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;
@@ -129,28 +123,28 @@ AddElementOnDisk(Relation index, HnswElement e, int m, BlockNumber insertPage, B
 	Size		minCombinedSize;
 	HnswElementTuple etup;
 	BlockNumber currentPage = insertPage;
+	int			dimensions = e->vec->dim;
 	HnswNeighborTuple ntup;
 	Buffer		nbuf;
 	Page		npage;
 	OffsetNumber freeOffno = InvalidOffsetNumber;
 	OffsetNumber freeNeighborOffno = InvalidOffsetNumber;
 	BlockNumber newInsertPage = InvalidBlockNumber;
-	char	   *base = NULL;
 
 	/* Calculate sizes */
-	etupSize = HNSW_ELEMENT_TUPLE_SIZE(VARSIZE_ANY(HnswPtrAccess(base, e->value)));
+	etupSize = HNSW_ELEMENT_TUPLE_SIZE(dimensions);
 	ntupSize = HNSW_NEIGHBOR_TUPLE_SIZE(e->level, m);
 	combinedSize = etupSize + ntupSize + sizeof(ItemIdData);
-	maxSize = HNSW_MAX_SIZE;
+	maxSize = BLCKSZ - MAXALIGN(SizeOfPageHeaderData) - MAXALIGN(sizeof(HnswPageOpaqueData));
 	minCombinedSize = etupSize + HNSW_NEIGHBOR_TUPLE_SIZE(0, m) + sizeof(ItemIdData);
 
 	/* Prepare element tuple */
 	etup = palloc0(etupSize);
-	HnswSetElementTuple(base, etup, e);
+	HnswSetElementTuple(etup, e);
 
 	/* Prepare neighbor tuple */
 	ntup = palloc0(ntupSize);
-	HnswSetNeighborTuple(base, ntup, e, m);
+	HnswSetNeighborTuple(ntup, e, m);
 
 	/* Find a page (or two if needed) to insert the tuples */
 	for (;;)
@@ -158,16 +152,8 @@ AddElementOnDisk(Relation index, HnswElement e, int m, BlockNumber insertPage, B
 		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)
@@ -187,12 +173,7 @@ AddElementOnDisk(Relation index, HnswElement e, int m, BlockNumber insertPage, B
 		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;
 		}
@@ -201,7 +182,7 @@ AddElementOnDisk(Relation index, HnswElement e, int m, BlockNumber insertPage, B
 		/* 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;
 		}
 
@@ -210,8 +191,7 @@ AddElementOnDisk(Relation index, HnswElement e, int m, BlockNumber insertPage, B
 		if (BlockNumberIsValid(currentPage))
 		{
 			/* Move to next page */
-			if (!building)
-				GenericXLogAbort(state);
+			GenericXLogAbort(state);
 			UnlockReleaseBuffer(buf);
 		}
 		else
@@ -219,33 +199,24 @@ AddElementOnDisk(Relation index, HnswElement e, int m, BlockNumber insertPage, B
 			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);
+			MarkBufferDirty(newbuf);
+			MarkBufferDirty(buf);
+			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;
@@ -299,14 +270,10 @@ AddElementOnDisk(Relation index, HnswElement e, int m, BlockNumber insertPage, B
 	}
 
 	/* Commit */
-	if (building)
-	{
-		MarkBufferDirty(buf);
-		if (nbuf != buf)
-			MarkBufferDirty(nbuf);
-	}
-	else
-		GenericXLogFinish(state);
+	MarkBufferDirty(buf);
+	if (nbuf != buf)
+		MarkBufferDirty(nbuf);
+	GenericXLogFinish(state);
 	UnlockReleaseBuffer(buf);
 	if (nbuf != buf)
 		UnlockReleaseBuffer(nbuf);
@@ -340,14 +307,12 @@ ConnectionExists(HnswElement e, HnswNeighborTuple ntup, int startIdx, int lm)
  * Update neighbors
  */
 void
-HnswUpdateNeighborsOnDisk(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)
 {
-	char	   *base = NULL;
-
 	for (int lc = e->level; lc >= 0; lc--)
 	{
 		int			lm = HnswGetLayerM(m, lc);
-		HnswNeighborArray *neighbors = HnswGetNeighbors(base, e, lc);
+		HnswNeighborArray *neighbors = &e->neighbors[lc];
 
 		for (int i = 0; i < neighbors->length; i++)
 		{
@@ -355,15 +320,16 @@ HnswUpdateNeighborsOnDisk(Relation index, FmgrInfo *procinfo, Oid collation, Hns
 			Buffer		buf;
 			Page		page;
 			GenericXLogState *state;
+			ItemId		itemid;
 			HnswNeighborTuple ntup;
+			Size		ntupSize;
 			int			idx = -1;
 			int			startIdx;
-			HnswElement neighborElement = HnswPtrAccess(base, hc->element);
-			OffsetNumber offno = neighborElement->neighborOffno;
+			OffsetNumber offno = hc->element->neighborOffno;
 
 			/* Get latest neighbors since they may have changed */
 			/* Do not lock yet since selecting neighbors can take time */
-			HnswLoadNeighbors(neighborElement, index, m);
+			HnswLoadNeighbors(hc->element, index);
 
 			/*
 			 * Could improve performance for vacuuming by checking neighbors
@@ -373,31 +339,25 @@ HnswUpdateNeighborsOnDisk(Relation index, FmgrInfo *procinfo, Oid collation, Hns
 			 */
 
 			/* Select neighbors */
-			HnswUpdateConnection(NULL, e, hc, lm, lc, &idx, index, procinfo, collation);
+			HnswUpdateConnection(e, hc, lm, lc, &idx, index, procinfo, collation);
 
 			/* New element was not selected as a neighbor */
 			if (idx == -1)
 				continue;
 
 			/* Register page */
-			buf = ReadBuffer(index, neighborElement->neighborPage);
+			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 */
-			ntup = (HnswNeighborTuple) PageGetItem(page, PageGetItemId(page, offno));
+			itemid = PageGetItemId(page, offno);
+			ntup = (HnswNeighborTuple) PageGetItem(page, itemid);
+			ntupSize = ItemIdGetLength(itemid);
 
 			/* Calculate index for update */
-			startIdx = (neighborElement->level - lc) * m;
+			startIdx = (hc->element->level - lc) * m;
 
 			/* Check for existing connection */
 			if (checkExisting && ConnectionExists(e, ntup, startIdx, lm))
@@ -423,16 +383,18 @@ HnswUpdateNeighborsOnDisk(Relation index, FmgrInfo *procinfo, Oid collation, Hns
 			{
 				ItemPointer indextid = &ntup->indextids[idx];
 
-				/* Update neighbor on the buffer */
+				/* Update neighbor */
 				ItemPointerSet(indextid, e->blkno, e->offno);
 
+				/* Overwrite tuple */
+				if (!PageIndexTupleOverwrite(page, offno, (Item) ntup, ntupSize))
+					elog(ERROR, "failed to add index item to \"%s\"", RelationGetRelationName(index));
+
 				/* Commit */
-				if (building)
-					MarkBufferDirty(buf);
-				else
-					GenericXLogFinish(state);
+				MarkBufferDirty(buf);
+				GenericXLogFinish(state);
 			}
-			else if (!building)
+			else
 				GenericXLogAbort(state);
 
 			UnlockReleaseBuffer(buf);
@@ -444,27 +406,20 @@ HnswUpdateNeighborsOnDisk(Relation index, FmgrInfo *procinfo, Oid collation, Hns
  * Add a heap TID to an existing element
  */
 static bool
-AddDuplicateOnDisk(Relation index, HnswElement element, HnswElement dup, bool building)
+HnswAddDuplicate(Relation index, HnswElement element, HnswElement dup)
 {
 	Buffer		buf;
 	Page		page;
 	GenericXLogState *state;
+	Size		etupSize = HNSW_ELEMENT_TUPLE_SIZE(dup->vec->dim);
 	HnswElementTuple etup;
 	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 */
 	etup = (HnswElementTuple) PageGetItem(page, PageGetItemId(page, dup->offno));
@@ -477,93 +432,88 @@ AddDuplicateOnDisk(Relation index, HnswElement element, HnswElement dup, bool bu
 	/* 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, modifying the tuple on the page directly */
-	etup->heaptids[i] = element->heaptids[0];
+	/* Add heap TID */
+	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);
+	MarkBufferDirty(buf);
+	GenericXLogFinish(state);
 	UnlockReleaseBuffer(buf);
 
 	return true;
 }
 
 /*
- * Find duplicate element
- */
-static bool
-FindDuplicateOnDisk(Relation index, HnswElement element, bool building)
-{
-	char	   *base = NULL;
-	HnswNeighborArray *neighbors = HnswGetNeighbors(base, element, 0);
-	Datum		value = HnswGetValue(base, element);
-
-	for (int i = 0; i < neighbors->length; i++)
-	{
-		HnswCandidate *neighbor = &neighbors->items[i];
-		HnswElement neighborElement = HnswPtrAccess(base, neighbor->element);
-		Datum		neighborValue = HnswGetValue(base, neighborElement);
-
-		/* Exit early since ordered by distance */
-		if (!datumIsEqual(value, neighborValue, false, -1))
-			return false;
-
-		if (AddDuplicateOnDisk(index, element, neighborElement, building))
-			return true;
-	}
-
-	return false;
-}
-
-/*
- * Update graph on disk
+ * Write changes to disk
  */
 static void
-UpdateGraphOnDisk(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 (FindDuplicateOnDisk(index, element, building))
-		return;
+	/* Try to add to existing page */
+	if (dup != NULL)
+	{
+		if (HnswAddDuplicate(index, element, dup))
+			return;
+	}
 
-	/* Add element */
-	AddElementOnDisk(index, element, m, GetInsertPage(index), &newInsertPage, building);
+	/* Write element and neighbor tuples */
+	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 */
-	HnswUpdateNeighborsOnDisk(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
-HnswInsertTupleOnDisk(Relation index, Datum value, Datum *values, bool *isnull, ItemPointer heap_tid, bool building)
+HnswInsertTuple(Relation index, Datum *values, bool *isnull, ItemPointer heap_tid, Relation heapRel)
 {
+	Datum		value;
+	FmgrInfo   *normprocinfo;
 	HnswElement entryPoint;
 	HnswElement element;
-	int			m;
+	int			m = HnswGetM(index);
 	int			efConstruction = HnswGetEfConstruction(index);
+	double		ml = HnswGetMl(m);
 	FmgrInfo   *procinfo = index_getprocinfo(index, 1, HNSW_DISTANCE_PROC);
 	Oid			collation = index->rd_indcollation[0];
+	HnswElement dup;
 	LOCKMODE	lockmode = ShareLock;
-	char	   *base = NULL;
+
+	/* Detoast once for all calls */
+	value = PointerGetDatum(PG_DETOAST_DATUM(values[0]));
+
+	/* Normalize if needed */
+	normprocinfo = HnswOptionalProcInfo(index, HNSW_NORM_PROC);
+	if (normprocinfo != NULL)
+	{
+		if (!HnswNormValue(normprocinfo, collation, &value, NULL))
+			return false;
+	}
+
+	/* Create an element */
+	element = HnswInitElement(heap_tid, m, ml, HnswGetMaxLevel(m));
+	element->vec = DatumGetVector(value);
 
 	/*
 	 * Get a shared lock. This allows vacuum to ensure no in-flight inserts
@@ -572,12 +522,8 @@ HnswInsertTupleOnDisk(Relation index, Datum value, Datum *values, bool *isnull,
 	 */
 	LockPage(index, HNSW_UPDATE_LOCK, lockmode);
 
-	/* Get m and entry point */
-	HnswGetMetaPageInfo(index, &m, &entryPoint);
-
-	/* Create an element */
-	element = HnswInitElement(base, heap_tid, m, HnswGetMl(m), HnswGetMaxLevel(m), NULL);
-	HnswPtrStore(base, element->value, DatumGetPointer(value));
+	/* Get entry point */
+	entryPoint = HnswGetEntryPoint(index);
 
 	/* Prevent concurrent inserts when likely updating entry point */
 	if (entryPoint == NULL || element->level > entryPoint->level)
@@ -593,11 +539,14 @@ HnswInsertTupleOnDisk(Relation index, Datum value, Datum *values, bool *isnull,
 		entryPoint = HnswGetEntryPoint(index);
 	}
 
-	/* Find neighbors for element */
-	HnswFindElementNeighbors(base, element, entryPoint, index, procinfo, collation, m, efConstruction, false);
+	/* Insert element in graph */
+	HnswInsertElement(element, entryPoint, index, procinfo, collation, m, efConstruction, false);
 
-	/* Update graph on disk */
-	UpdateGraphOnDisk(index, procinfo, collation, element, m, efConstruction, entryPoint, building);
+	/* Look for duplicate */
+	dup = HnswFindDuplicate(element);
+
+	/* Write to disk */
+	WriteElement(index, procinfo, collation, element, m, efConstruction, dup, entryPoint);
 
 	/* Release lock */
 	UnlockPage(index, HNSW_UPDATE_LOCK, lockmode);
@@ -605,34 +554,6 @@ HnswInsertTupleOnDisk(Relation index, Datum value, Datum *values, bool *isnull,
 	return true;
 }
 
-/*
- * Insert a tuple into the index
- */
-static void
-HnswInsertTuple(Relation index, Datum *values, bool *isnull, ItemPointer heap_tid)
-{
-	Datum		value;
-	FmgrInfo   *normprocinfo;
-	Oid			collation = index->rd_indcollation[0];
-	HnswType	type = HnswGetType(index);
-
-	/* Detoast once for all calls */
-	value = PointerGetDatum(PG_DETOAST_DATUM(values[0]));
-
-	/* Check value */
-	HnswCheckValue(value, type);
-
-	/* Normalize if needed */
-	normprocinfo = HnswOptionalProcInfo(index, HNSW_NORM_PROC);
-	if (normprocinfo != NULL)
-	{
-		if (!HnswNormValue(normprocinfo, collation, &value, type))
-			return;
-	}
-
-	HnswInsertTupleOnDisk(index, value, values, isnull, heap_tid, false);
-}
-
 /*
  * Insert a tuple into the index
  */
@@ -659,7 +580,7 @@ hnswinsert(Relation index, Datum *values, bool *isnull, ItemPointer heap_tid,
 	oldCtx = MemoryContextSwitchTo(insertCtx);
 
 	/* Insert tuple */
-	HnswInsertTuple(index, values, isnull, heap_tid);
+	HnswInsertTuple(index, values, isnull, heap_tid, heap);
 
 	/* Delete memory context */
 	MemoryContextSwitchTo(oldCtx);

src/hnswscan.c

@@ -19,52 +19,43 @@ GetScanItems(IndexScanDesc scan, Datum q)
 	Oid			collation = so->collation;
 	List	   *ep;
 	List	   *w;
-	int			m;
-	HnswElement entryPoint;
-	char	   *base = NULL;
-
-	/* Get m and entry point */
-	HnswGetMetaPageInfo(index, &m, &entryPoint);
+	HnswElement entryPoint = HnswGetEntryPoint(index);
 
 	if (entryPoint == NULL)
 		return NIL;
 
-	ep = list_make1(HnswEntryCandidate(base, entryPoint, q, index, procinfo, collation, false));
+	ep = list_make1(HnswEntryCandidate(entryPoint, q, index, procinfo, collation, false));
 
 	for (int lc = entryPoint->level; lc >= 1; lc--)
 	{
-		w = HnswSearchLayer(base, q, ep, 1, lc, index, procinfo, collation, m, false, NULL);
+		w = HnswSearchLayer(q, ep, 1, lc, index, procinfo, collation, false, NULL);
 		ep = w;
 	}
 
-	return HnswSearchLayer(base, q, ep, hnsw_ef_search, 0, index, procinfo, collation, m, false, NULL);
+	return HnswSearchLayer(q, ep, hnsw_ef_search, 0, index, procinfo, collation, false, NULL);
 }
 
 /*
- * Get scan value
+ * Get dimensions from metapage
  */
-static Datum
-GetScanValue(IndexScanDesc scan)
+static int
+GetDimensions(Relation index)
 {
-	HnswScanOpaque so = (HnswScanOpaque) scan->opaque;
-	Datum		value;
+	Buffer		buf;
+	Page		page;
+	HnswMetaPage metap;
+	int			dimensions;
 
-	if (scan->orderByData->sk_flags & SK_ISNULL)
-		value = PointerGetDatum(NULL);
-	else
-	{
-		value = scan->orderByData->sk_argument;
+	buf = ReadBuffer(index, HNSW_METAPAGE_BLKNO);
+	LockBuffer(buf, BUFFER_LOCK_SHARE);
+	page = BufferGetPage(buf);
+	metap = HnswPageGetMeta(page);
 
-		/* Value should not be compressed or toasted */
-		Assert(!VARATT_IS_COMPRESSED(DatumGetPointer(value)));
-		Assert(!VARATT_IS_EXTENDED(DatumGetPointer(value)));
+	dimensions = metap->dimensions;
 
-		/* Fine if normalization fails */
-		if (so->normprocinfo != NULL)
-			HnswNormValue(so->normprocinfo, so->collation, &value, HnswGetType(scan->indexRelation));
-	}
+	UnlockReleaseBuffer(buf);
 
-	return value;
+	return dimensions;
 }
 
 /*
@@ -79,6 +70,7 @@ hnswbeginscan(Relation index, int nkeys, int norderbys)
 	scan = RelationGetIndexScan(index, nkeys, norderbys);
 
 	so = (HnswScanOpaque) palloc(sizeof(HnswScanOpaqueData));
+	so->buf = InvalidBuffer;
 	so->first = true;
 	so->tmpCtx = AllocSetContextCreate(CurrentMemoryContext,
 									   "Hnsw scan temporary context",
@@ -138,13 +130,20 @@ hnswgettuple(IndexScanDesc scan, ScanDirection dir)
 		if (scan->orderByData == NULL)
 			elog(ERROR, "cannot scan hnsw index without order");
 
-		/* Requires MVCC-compliant snapshot as not able to maintain a pin */
-		/* https://www.postgresql.org/docs/current/index-locking.html */
-		if (!IsMVCCSnapshot(scan->xs_snapshot))
-			elog(ERROR, "non-MVCC snapshots are not supported with hnsw");
+		if (scan->orderByData->sk_flags & SK_ISNULL)
+			value = PointerGetDatum(InitVector(GetDimensions(scan->indexRelation)));
+		else
+		{
+			value = scan->orderByData->sk_argument;
 
-		/* Get scan value */
-		value = GetScanValue(scan);
+			/* Value should not be compressed or toasted */
+			Assert(!VARATT_IS_COMPRESSED(DatumGetPointer(value)));
+			Assert(!VARATT_IS_EXTENDED(DatumGetPointer(value)));
+
+			/* Fine if normalization fails */
+			if (so->normprocinfo != NULL)
+				HnswNormValue(so->normprocinfo, so->collation, &value, NULL);
+		}
 
 		/*
 		 * Get a shared lock. This allows vacuum to ensure no in-flight scans
@@ -162,24 +161,42 @@ hnswgettuple(IndexScanDesc scan, ScanDirection dir)
 
 	while (list_length(so->w) > 0)
 	{
-		char	   *base = NULL;
 		HnswCandidate *hc = llast(so->w);
-		HnswElement element = HnswPtrAccess(base, hc->element);
-		ItemPointer heaptid;
+		ItemPointer tid;
+		BlockNumber indexblkno;
 
-		/* Move to next element if no valid heap TIDs */
-		if (element->heaptidsLength == 0)
+		/* Move to next element if no valid heap tids */
+		if (list_length(hc->element->heaptids) == 0)
 		{
 			so->w = list_delete_last(so->w);
 			continue;
 		}
 
-		heaptid = &element->heaptids[--element->heaptidsLength];
+		tid = llast(hc->element->heaptids);
+		indexblkno = hc->element->blkno;
+
+		hc->element->heaptids = list_delete_last(hc->element->heaptids);
 
 		MemoryContextSwitchTo(oldCtx);
 
-		scan->xs_heaptid = *heaptid;
-		scan->xs_recheck = false;
+#if PG_VERSION_NUM >= 120000
+		scan->xs_heaptid = *tid;
+#else
+		scan->xs_ctup.t_self = *tid;
+#endif
+
+		/* Unpin buffer */
+		if (BufferIsValid(so->buf))
+			ReleaseBuffer(so->buf);
+
+		/*
+		 * An index scan must maintain a pin on the index page holding the
+		 * item last returned by amgettuple
+		 *
+		 * https://www.postgresql.org/docs/current/index-locking.html
+		 */
+		so->buf = ReadBuffer(scan->indexRelation, indexblkno);
+
 		scan->xs_recheckorderby = false;
 		return true;
 	}
@@ -196,6 +213,10 @@ hnswendscan(IndexScanDesc scan)
 {
 	HnswScanOpaque so = (HnswScanOpaque) scan->opaque;
 
+	/* Release pin */
+	if (BufferIsValid(so->buf))
+		ReleaseBuffer(so->buf);
+
 	MemoryContextDelete(so->tmpCtx);
 
 	pfree(so);

src/hnswvacuum.c

@@ -2,7 +2,6 @@
 
 #include <math.h>
 
-#include "access/generic_xlog.h"
 #include "commands/vacuum.h"
 #include "hnsw.h"
 #include "storage/bufmgr.h"
@@ -13,9 +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;
 }
 
 /*
@@ -91,10 +93,15 @@ RemoveHeapTids(HnswVacuumState * vacuumstate)
 
 				if (itemUpdated)
 				{
+					Size		etupSize = HNSW_ELEMENT_TUPLE_SIZE(etup->vec.dim);
+
 					/* Mark rest as invalid */
 					for (int i = idx; i < HNSW_HEAPTIDS; i++)
 						ItemPointerSetInvalid(&etup->heaptids[i]);
 
+					if (!PageIndexTupleOverwrite(page, offno, (Item) etup, etupSize))
+						elog(ERROR, "failed to add index item to \"%s\"", RelationGetRelationName(index));
+
 					updated = true;
 				}
 			}
@@ -102,13 +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))
 			{
@@ -123,7 +128,10 @@ RemoveHeapTids(HnswVacuumState * vacuumstate)
 		blkno = HnswPageGetOpaque(page)->nextblkno;
 
 		if (updated)
+		{
+			MarkBufferDirty(buf);
 			GenericXLogFinish(state);
+		}
 		else
 			GenericXLogAbort(state);
 
@@ -194,25 +202,21 @@ RepairGraphElement(HnswVacuumState * vacuumstate, HnswElement element, HnswEleme
 	BufferAccessStrategy bas = vacuumstate->bas;
 	HnswNeighborTuple ntup = vacuumstate->ntup;
 	Size		ntupSize = HNSW_NEIGHBOR_TUPLE_SIZE(element->level, m);
-	char	   *base = NULL;
 
 	/* Skip if element is entry point */
 	if (entryPoint != NULL && element->blkno == entryPoint->blkno && element->offno == entryPoint->offno)
 		return;
 
 	/* Init fields */
-	HnswInitNeighbors(base, element, m, NULL);
-	element->heaptidsLength = 0;
+	HnswInitNeighbors(element, m);
+	element->heaptids = NIL;
 
-	/* Find neighbors for element, skipping itself */
-	HnswFindElementNeighbors(base, element, entryPoint, index, procinfo, collation, m, efConstruction, true);
-
-	/* Zero memory for each element */
-	MemSet(ntup, 0, HNSW_TUPLE_ALLOC_SIZE);
+	/* Add element to graph, skipping itself */
+	HnswInsertElement(element, entryPoint, index, procinfo, collation, m, efConstruction, true);
 
 	/* Update neighbor tuple */
 	/* Do this before getting page to minimize locking */
-	HnswSetNeighborTuple(base, ntup, element, m);
+	HnswSetNeighborTuple(ntup, element, m);
 
 	/* Get neighbor page */
 	buf = ReadBufferExtended(index, MAIN_FORKNUM, element->neighborPage, RBM_NORMAL, bas);
@@ -225,11 +229,12 @@ RepairGraphElement(HnswVacuumState * vacuumstate, HnswElement element, HnswEleme
 		elog(ERROR, "failed to add index item to \"%s\"", RelationGetRelationName(index));
 
 	/* Commit */
+	MarkBufferDirty(buf);
 	GenericXLogFinish(state);
 	UnlockReleaseBuffer(buf);
 
 	/* Update neighbors */
-	HnswUpdateNeighborsOnDisk(index, procinfo, collation, element, m, true, false);
+	HnswUpdateNeighborPages(index, procinfo, collation, element, m, true);
 }
 
 /*
@@ -285,7 +290,7 @@ 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
 		{
@@ -300,7 +305,7 @@ RepairGraphEntryPoint(HnswVacuumState * vacuumstate)
 			{
 				/* Reset neighbors from previous update */
 				if (highestPoint != NULL)
-					HnswPtrStore((char *) NULL, highestPoint->neighbors, (HnswNeighborArrayPtr *) NULL);
+					highestPoint->neighbors = NULL;
 
 				RepairGraphElement(vacuumstate, entryPoint, highestPoint);
 			}
@@ -325,10 +330,7 @@ RepairGraph(HnswVacuumState * vacuumstate)
 	BufferAccessStrategy bas = vacuumstate->bas;
 	BlockNumber blkno = HNSW_HEAD_BLKNO;
 
-	/*
-	 * Wait for inserts to complete. Inserts before this point may have
-	 * neighbors about to be deleted. Inserts after this point will not.
-	 */
+	/* Wait for inserts to complete */
 	LockPage(index, HNSW_UPDATE_LOCK, ExclusiveLock);
 	UnlockPage(index, HNSW_UPDATE_LOCK, ExclusiveLock);
 
@@ -418,7 +420,7 @@ 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);
@@ -441,11 +443,7 @@ MarkDeleted(HnswVacuumState * vacuumstate)
 	Relation	index = vacuumstate->index;
 	BufferAccessStrategy bas = vacuumstate->bas;
 
-	/*
-	 * Wait for index scans to complete. Scans before this point may contain
-	 * tuples about to be deleted. Scans after this point will not, since the
-	 * graph has been repaired.
-	 */
+	/* Wait for selects to complete */
 	LockPage(index, HNSW_SCAN_LOCK, ExclusiveLock);
 	UnlockPage(index, HNSW_SCAN_LOCK, ExclusiveLock);
 
@@ -478,6 +476,8 @@ MarkDeleted(HnswVacuumState * vacuumstate)
 		{
 			HnswElementTuple etup = (HnswElementTuple) PageGetItem(page, PageGetItemId(page, offno));
 			HnswNeighborTuple ntup;
+			Size		etupSize;
+			Size		ntupSize;
 			Buffer		nbuf;
 			Page		npage;
 			BlockNumber neighborPage;
@@ -501,6 +501,10 @@ MarkDeleted(HnswVacuumState * vacuumstate)
 			if (ItemPointerIsValid(&etup->heaptids[0]))
 				continue;
 
+			/* Calculate sizes */
+			etupSize = HNSW_ELEMENT_TUPLE_SIZE(etup->vec.dim);
+			ntupSize = HNSW_NEIGHBOR_TUPLE_SIZE(etup->level, vacuumstate->m);
+
 			/* Get neighbor page */
 			neighborPage = ItemPointerGetBlockNumber(&etup->neighbortid);
 			neighborOffno = ItemPointerGetOffsetNumber(&etup->neighbortid);
@@ -521,18 +525,24 @@ MarkDeleted(HnswVacuumState * vacuumstate)
 
 			/* Overwrite element */
 			etup->deleted = 1;
-			MemSet(&etup->data, 0, VARSIZE_ANY(&etup->data));
+			MemSet(&etup->vec.x, 0, etup->vec.dim * sizeof(float));
 
 			/* Overwrite neighbors */
 			for (int i = 0; i < ntup->count; i++)
 				ItemPointerSetInvalid(&ntup->indextids[i]);
 
-			/*
-			 * We modified the tuples in place, no need to call
-			 * PageIndexTupleOverwrite
-			 */
+			/* Overwrite element tuple */
+			if (!PageIndexTupleOverwrite(page, offno, (Item) etup, etupSize))
+				elog(ERROR, "failed to add index item to \"%s\"", RelationGetRelationName(index));
+
+			/* Overwrite neighbor tuple */
+			if (!PageIndexTupleOverwrite(npage, neighborOffno, (Item) ntup, ntupSize))
+				elog(ERROR, "failed to add index item to \"%s\"", RelationGetRelationName(index));
 
 			/* Commit */
+			MarkBufferDirty(buf);
+			if (nbuf != buf)
+				MarkBufferDirty(nbuf);
 			GenericXLogFinish(state);
 			if (nbuf != buf)
 				UnlockReleaseBuffer(nbuf);
@@ -553,7 +563,7 @@ 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);
 }
 
 /*
@@ -563,6 +573,7 @@ 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));
@@ -571,20 +582,21 @@ InitVacuumState(HnswVacuumState * vacuumstate, IndexVacuumInfo *info, IndexBulkD
 	vacuumstate->stats = stats;
 	vacuumstate->callback = callback;
 	vacuumstate->callback_state = callback_state;
+	vacuumstate->m = HnswGetM(index);
 	vacuumstate->efConstruction = HnswGetEfConstruction(index);
 	vacuumstate->bas = GetAccessStrategy(BAS_BULKREAD);
 	vacuumstate->procinfo = index_getprocinfo(index, 1, HNSW_DISTANCE_PROC);
 	vacuumstate->collation = index->rd_indcollation[0];
-	vacuumstate->ntup = palloc0(HNSW_TUPLE_ALLOC_SIZE);
+	vacuumstate->ntup = palloc0(BLCKSZ);
 	vacuumstate->tmpCtx = AllocSetContextCreate(CurrentMemoryContext,
 												"Hnsw vacuum temporary context",
 												ALLOCSET_DEFAULT_SIZES);
 
-	/* Get m from metapage */
-	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);
 }
 
 /*
@@ -593,7 +605,7 @@ 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);

src/ivfbuild.c

@@ -2,38 +2,58 @@
 
 #include <float.h>
 
-#include "access/table.h"
-#include "access/tableam.h"
 #include "access/parallel.h"
 #include "access/xact.h"
 #include "catalog/index.h"
 #include "catalog/pg_operator_d.h"
 #include "catalog/pg_type_d.h"
-#include "commands/progress.h"
 #include "ivfflat.h"
 #include "miscadmin.h"
-#include "optimizer/optimizer.h"
 #include "storage/bufmgr.h"
-#include "tcop/tcopprot.h"
 #include "utils/memutils.h"
+#include "tcop/tcopprot.h"
 
 #if PG_VERSION_NUM >= 140000
 #include "utils/backend_progress.h"
-#else
+#elif PG_VERSION_NUM >= 120000
 #include "pgstat.h"
 #endif
 
+#if PG_VERSION_NUM >= 120000
+#include "access/tableam.h"
+#include "commands/progress.h"
+#else
+#define PROGRESS_CREATEIDX_SUBPHASE 0
+#define PROGRESS_CREATEIDX_TUPLES_TOTAL 0
+#define PROGRESS_CREATEIDX_TUPLES_DONE 0
+#endif
+
 #if PG_VERSION_NUM >= 130000
 #define CALLBACK_ITEM_POINTER ItemPointer tid
 #else
 #define CALLBACK_ITEM_POINTER HeapTuple hup
 #endif
 
+#if PG_VERSION_NUM >= 120000
+#define UpdateProgress(index, val) pgstat_progress_update_param(index, val)
+#else
+#define UpdateProgress(index, val) ((void)val)
+#endif
+
 #if PG_VERSION_NUM >= 140000
 #include "utils/backend_status.h"
 #include "utils/wait_event.h"
 #endif
 
+#if PG_VERSION_NUM >= 120000
+#include "access/table.h"
+#include "optimizer/optimizer.h"
+#else
+#include "access/heapam.h"
+#include "optimizer/planner.h"
+#include "pgstat.h"
+#endif
+
 #define PARALLEL_KEY_IVFFLAT_SHARED		UINT64CONST(0xA000000000000001)
 #define PARALLEL_KEY_TUPLESORT			UINT64CONST(0xA000000000000002)
 #define PARALLEL_KEY_IVFFLAT_CENTERS	UINT64CONST(0xA000000000000003)
@@ -57,7 +77,7 @@ 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 +125,7 @@ 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);
@@ -130,8 +150,13 @@ SampleRows(IvfflatBuildState * buildstate)
 	{
 		BlockNumber targblock = BlockSampler_Next(&buildstate->bs);
 
+#if PG_VERSION_NUM >= 120000
 		table_index_build_range_scan(buildstate->heap, buildstate->index, buildstate->indexInfo,
 									 false, true, false, targblock, 1, SampleCallback, (void *) buildstate, NULL);
+#else
+		IndexBuildHeapRangeScan(buildstate->heap, buildstate->index, buildstate->indexInfo,
+								false, true, targblock, 1, SampleCallback, (void *) buildstate, NULL);
+#endif
 	}
 }
 
@@ -153,7 +178,7 @@ 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;
 	}
 
@@ -257,12 +282,16 @@ InsertTuples(Relation index, IvfflatBuildState * buildstate, ForkNumber forkNum)
 	IndexTuple	itup = NULL;	/* silence compiler warning */
 	int64		inserted = 0;
 
+#if PG_VERSION_NUM >= 120000
 	TupleTableSlot *slot = MakeSingleTupleTableSlot(buildstate->tupdesc, &TTSOpsMinimalTuple);
+#else
+	TupleTableSlot *slot = MakeSingleTupleTableSlot(buildstate->tupdesc);
+#endif
 	TupleDesc	tupdesc = RelationGetDescr(index);
 
-	pgstat_progress_update_param(PROGRESS_CREATEIDX_SUBPHASE, PROGRESS_IVFFLAT_PHASE_LOAD);
+	UpdateProgress(PROGRESS_CREATEIDX_SUBPHASE, PROGRESS_IVFFLAT_PHASE_LOAD);
 
-	pgstat_progress_update_param(PROGRESS_CREATEIDX_TUPLES_TOTAL, buildstate->indtuples);
+	UpdateProgress(PROGRESS_CREATEIDX_TUPLES_TOTAL, buildstate->indtuples);
 
 	GetNextTuple(buildstate->sortstate, tupdesc, slot, &itup, &list);
 
@@ -298,7 +327,7 @@ InsertTuples(Relation index, IvfflatBuildState * buildstate, ForkNumber forkNum)
 
 			pfree(itup);
 
-			pgstat_progress_update_param(PROGRESS_CREATEIDX_TUPLES_DONE, ++inserted);
+			UpdateProgress(PROGRESS_CREATEIDX_TUPLES_DONE, ++inserted);
 
 			GetNextTuple(buildstate->sortstate, tupdesc, slot, &itup, &list);
 		}
@@ -346,16 +375,27 @@ InitBuildState(IvfflatBuildState * buildstate, Relation heap, Relation index, In
 		elog(ERROR, "dimensions must be greater than one for this opclass");
 
 	/* Create tuple description for sorting */
+#if PG_VERSION_NUM >= 120000
 	buildstate->tupdesc = CreateTemplateTupleDesc(3);
+#else
+	buildstate->tupdesc = CreateTemplateTupleDesc(3, false);
+#endif
 	TupleDescInitEntry(buildstate->tupdesc, (AttrNumber) 1, "list", INT4OID, -1, 0);
 	TupleDescInitEntry(buildstate->tupdesc, (AttrNumber) 2, "tid", TIDOID, -1, 0);
 	TupleDescInitEntry(buildstate->tupdesc, (AttrNumber) 3, "vector", RelationGetDescr(index)->attrs[0].atttypid, -1, 0);
 
+#if PG_VERSION_NUM >= 120000
 	buildstate->slot = MakeSingleTupleTableSlot(buildstate->tupdesc, &TTSOpsVirtual);
+#else
+	buildstate->slot = MakeSingleTupleTableSlot(buildstate->tupdesc);
+#endif
 
 	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 +417,7 @@ FreeBuildState(IvfflatBuildState * buildstate)
 {
 	VectorArrayFree(buildstate->centers);
 	pfree(buildstate->listInfo);
+	pfree(buildstate->normvec);
 
 #ifdef IVFFLAT_KMEANS_DEBUG
 	pfree(buildstate->listSums);
@@ -394,7 +435,7 @@ ComputeCenters(IvfflatBuildState * buildstate)
 {
 	int			numSamples;
 
-	pgstat_progress_update_param(PROGRESS_CREATEIDX_SUBPHASE, PROGRESS_IVFFLAT_PHASE_KMEANS);
+	UpdateProgress(PROGRESS_CREATEIDX_SUBPHASE, PROGRESS_IVFFLAT_PHASE_KMEANS);
 
 	/* Target 50 samples per list, with at least 10000 samples */
 	/* The number of samples has a large effect on index build time */
@@ -465,30 +506,29 @@ CreateListPages(Relation index, VectorArray centers, int dimensions,
 	Buffer		buf;
 	Page		page;
 	GenericXLogState *state;
-	Size		listSize;
+	OffsetNumber offno;
+	Size		itemsz;
 	IvfflatList list;
 
-	listSize = MAXALIGN(IVFFLAT_LIST_SIZE(dimensions));
-	list = palloc0(listSize);
+	itemsz = MAXALIGN(IVFFLAT_LIST_SIZE(dimensions));
+	list = palloc(itemsz);
 
 	buf = IvfflatNewBuffer(index, forkNum);
 	IvfflatInitRegisterPage(index, &buf, &page, &state);
 
 	for (int i = 0; i < lists; i++)
 	{
-		OffsetNumber offno;
-
 		/* Load list */
 		list->startPage = InvalidBlockNumber;
 		list->insertPage = InvalidBlockNumber;
 		memcpy(&list->center, VectorArrayGet(centers, i), VECTOR_SIZE(dimensions));
 
 		/* Ensure free space */
-		if (PageGetFreeSpace(page) < listSize)
+		if (PageGetFreeSpace(page) < itemsz)
 			IvfflatAppendPage(index, &buf, &page, &state, forkNum);
 
 		/* Add the item */
-		offno = PageAddItem(page, (Item) list, listSize, InvalidOffsetNumber, false, false);
+		offno = PageAddItem(page, (Item) list, itemsz, InvalidOffsetNumber, false, false);
 		if (offno == InvalidOffsetNumber)
 			elog(ERROR, "failed to add index item to \"%s\"", RelationGetRelationName(index));
 
@@ -502,10 +542,10 @@ 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)
 {
@@ -590,7 +630,11 @@ IvfflatParallelScanAndSort(IvfflatSpool * ivfspool, IvfflatShared * ivfshared, S
 {
 	SortCoordinate coordinate;
 	IvfflatBuildState buildstate;
+#if PG_VERSION_NUM >= 120000
 	TableScanDesc scan;
+#else
+	HeapScanDesc scan;
+#endif
 	double		reltuples;
 	IndexInfo  *indexInfo;
 
@@ -614,11 +658,18 @@ IvfflatParallelScanAndSort(IvfflatSpool * ivfspool, IvfflatShared * ivfshared, S
 	buildstate.centers->length = buildstate.centers->maxlen;
 	ivfspool->sortstate = tuplesort_begin_heap(buildstate.tupdesc, 1, attNums, sortOperators, sortCollations, nullsFirstFlags, sortmem, coordinate, false);
 	buildstate.sortstate = ivfspool->sortstate;
+#if PG_VERSION_NUM >= 120000
 	scan = table_beginscan_parallel(ivfspool->heap,
 									ParallelTableScanFromIvfflatShared(ivfshared));
 	reltuples = table_index_build_scan(ivfspool->heap, ivfspool->index, indexInfo,
 									   true, progress, BuildCallback,
 									   (void *) &buildstate, scan);
+#else
+	scan = heap_beginscan_parallel(ivfspool->heap, &ivfshared->heapdesc);
+	reltuples = IndexBuildHeapScan(ivfspool->heap, ivfspool->index, indexInfo,
+								   true, BuildCallback,
+								   (void *) &buildstate, scan);
+#endif
 
 	/* Execute this worker's part of the sort */
 	tuplesort_performsort(ivfspool->sortstate);
@@ -688,7 +739,11 @@ IvfflatParallelBuildMain(dsm_segment *seg, shm_toc *toc)
 	}
 
 	/* Open relations within worker */
+#if PG_VERSION_NUM >= 120000
 	heapRel = table_open(ivfshared->heaprelid, heapLockmode);
+#else
+	heapRel = heap_open(ivfshared->heaprelid, heapLockmode);
+#endif
 	indexRel = index_open(ivfshared->indexrelid, indexLockmode);
 
 	/* Initialize worker's own spool */
@@ -708,7 +763,11 @@ IvfflatParallelBuildMain(dsm_segment *seg, shm_toc *toc)
 
 	/* Close relations within worker */
 	index_close(indexRel, indexLockmode);
+#if PG_VERSION_NUM >= 120000
 	table_close(heapRel, heapLockmode);
+#else
+	heap_close(heapRel, heapLockmode);
+#endif
 }
 
 /*
@@ -733,7 +792,19 @@ IvfflatEndParallel(IvfflatLeader * ivfleader)
 static Size
 ParallelEstimateShared(Relation heap, Snapshot snapshot)
 {
+#if PG_VERSION_NUM >= 120000
 	return add_size(BUFFERALIGN(sizeof(IvfflatShared)), table_parallelscan_estimate(heap, snapshot));
+#else
+	if (!IsMVCCSnapshot(snapshot))
+	{
+		Assert(snapshot == SnapshotAny);
+		return sizeof(IvfflatShared);
+	}
+
+	return add_size(offsetof(IvfflatShared, heapdesc) +
+					offsetof(ParallelHeapScanDescData, phs_snapshot_data),
+					EstimateSnapshotSpace(snapshot));
+#endif
 }
 
 /*
@@ -784,7 +855,11 @@ IvfflatBeginParallel(IvfflatBuildState * buildstate, bool isconcurrent, int requ
 	/* Enter parallel mode and create context */
 	EnterParallelMode();
 	Assert(request > 0);
+#if PG_VERSION_NUM >= 120000
 	pcxt = CreateParallelContext("vector", "IvfflatParallelBuildMain", request);
+#else
+	pcxt = CreateParallelContext("vector", "IvfflatParallelBuildMain", request, true);
+#endif
 
 	scantuplesortstates = leaderparticipates ? request + 1 : request;
 
@@ -842,9 +917,13 @@ IvfflatBeginParallel(IvfflatBuildState * buildstate, bool isconcurrent, int requ
 #ifdef IVFFLAT_KMEANS_DEBUG
 	ivfshared->inertia = 0;
 #endif
+#if PG_VERSION_NUM >= 120000
 	table_parallelscan_initialize(buildstate->heap,
 								  ParallelTableScanFromIvfflatShared(ivfshared),
 								  snapshot);
+#else
+	heap_parallelscan_initialize(&ivfshared->heapdesc, buildstate->heap, snapshot);
+#endif
 
 	/* Store shared tuplesort-private state, for which we reserved space */
 	sharedsort = (Sharedsort *) shm_toc_allocate(pcxt->toc, estsort);
@@ -915,7 +994,7 @@ AssignTuples(IvfflatBuildState * buildstate)
 	Oid			sortCollations[] = {InvalidOid};
 	bool		nullsFirstFlags[] = {false};
 
-	pgstat_progress_update_param(PROGRESS_CREATEIDX_SUBPHASE, PROGRESS_IVFFLAT_PHASE_ASSIGN);
+	UpdateProgress(PROGRESS_CREATEIDX_SUBPHASE, PROGRESS_IVFFLAT_PHASE_ASSIGN);
 
 	/* Calculate parallel workers */
 	if (buildstate->heap != NULL)
@@ -943,8 +1022,15 @@ AssignTuples(IvfflatBuildState * buildstate)
 		if (buildstate->ivfleader)
 			buildstate->reltuples = ParallelHeapScan(buildstate);
 		else
+		{
+#if PG_VERSION_NUM >= 120000
 			buildstate->reltuples = table_index_build_scan(buildstate->heap, buildstate->index, buildstate->indexInfo,
 														   true, true, BuildCallback, (void *) buildstate, NULL);
+#else
+			buildstate->reltuples = IndexBuildHeapScan(buildstate->heap, buildstate->index, buildstate->indexInfo,
+													   true, BuildCallback, (void *) buildstate, NULL);
+#endif
+		}
 
 #ifdef IVFFLAT_KMEANS_DEBUG
 		PrintKmeansMetrics(buildstate);

src/ivfflat.c

@@ -3,16 +3,14 @@
 #include <float.h>
 
 #include "access/amapi.h"
-#include "access/reloptions.h"
-#include "commands/progress.h"
 #include "commands/vacuum.h"
 #include "ivfflat.h"
 #include "utils/guc.h"
 #include "utils/selfuncs.h"
 #include "utils/spccache.h"
 
-#if PG_VERSION_NUM < 150000
-#define MarkGUCPrefixReserved(x) EmitWarningsOnPlaceholders(x)
+#if PG_VERSION_NUM >= 120000
+#include "commands/progress.h"
 #endif
 
 int			ivfflat_probes;
@@ -35,13 +33,12 @@ IvfflatInit(void)
 	DefineCustomIntVariable("ivfflat.probes", "Sets the number of probes",
 							"Valid range is 1..lists.", &ivfflat_probes,
 							IVFFLAT_DEFAULT_PROBES, IVFFLAT_MIN_LISTS, IVFFLAT_MAX_LISTS, PGC_USERSET, 0, NULL, NULL, NULL);
-
-	MarkGUCPrefixReserved("ivfflat");
 }
 
 /*
  * Get the name of index build phase
  */
+#if PG_VERSION_NUM >= 120000
 static char *
 ivfflatbuildphasename(int64 phasenum)
 {
@@ -59,6 +56,7 @@ ivfflatbuildphasename(int64 phasenum)
 			return NULL;
 	}
 }
+#endif
 
 /*
  * Estimate the cost of an index scan
@@ -73,7 +71,10 @@ ivfflatcostestimate(PlannerInfo *root, IndexPath *path, double loop_count,
 	int			lists;
 	double		ratio;
 	double		spc_seq_page_cost;
-	Relation	index;
+	Relation	indexRel;
+#if PG_VERSION_NUM < 120000
+	List	   *qinfos;
+#endif
 
 	/* Never use index without order */
 	if (path->indexorderbys == NULL)
@@ -88,9 +89,9 @@ ivfflatcostestimate(PlannerInfo *root, IndexPath *path, double loop_count,
 
 	MemSet(&costs, 0, sizeof(costs));
 
-	index = index_open(path->indexinfo->indexoid, NoLock);
-	IvfflatGetMetaPageInfo(index, &lists, NULL);
-	index_close(index, NoLock);
+	indexRel = index_open(path->indexinfo->indexoid, NoLock);
+	lists = IvfflatGetLists(indexRel);
+	index_close(indexRel, NoLock);
 
 	/* Get the ratio of lists that we need to visit */
 	ratio = ((double) ivfflat_probes) / lists;
@@ -104,7 +105,12 @@ ivfflatcostestimate(PlannerInfo *root, IndexPath *path, double loop_count,
 	 */
 	costs.numIndexTuples = path->indexinfo->tuples * ratio;
 
+#if PG_VERSION_NUM >= 120000
 	genericcostestimate(root, path, loop_count, &costs);
+#else
+	qinfos = deconstruct_indexquals(path);
+	genericcostestimate(root, path, loop_count, qinfos, &costs);
+#endif
 
 	get_tablespace_page_costs(path->indexinfo->reltablespace, NULL, &spc_seq_page_cost);
 
@@ -221,7 +227,9 @@ ivfflathandler(PG_FUNCTION_ARGS)
 	amroutine->amcostestimate = ivfflatcostestimate;
 	amroutine->amoptions = ivfflatoptions;
 	amroutine->amproperty = NULL;	/* TODO AMPROP_DISTANCE_ORDERABLE */
+#if PG_VERSION_NUM >= 120000
 	amroutine->ambuildphasename = ivfflatbuildphasename;
+#endif
 	amroutine->amvalidate = ivfflatvalidate;
 #if PG_VERSION_NUM >= 140000
 	amroutine->amadjustmembers = NULL;

src/ivfflat.h

@@ -3,10 +3,9 @@
 
 #include "postgres.h"
 
-#include "access/genam.h"
 #include "access/generic_xlog.h"
 #include "access/parallel.h"
-#include "lib/pairingheap.h"
+#include "access/reloptions.h"
 #include "nodes/execnodes.h"
 #include "port.h"				/* for random() */
 #include "utils/sampling.h"
@@ -17,6 +16,10 @@
 #include "common/pg_prng.h"
 #endif
 
+#if PG_VERSION_NUM < 120000
+#include "access/relscan.h"
+#endif
+
 #ifdef IVFFLAT_BENCH
 #include "portability/instr_time.h"
 #endif
@@ -132,10 +135,16 @@ typedef struct IvfflatShared
 #ifdef IVFFLAT_KMEANS_DEBUG
 	double		inertia;
 #endif
+
+#if PG_VERSION_NUM < 120000
+	ParallelHeapScanDescData heapdesc;	/* must come last */
+#endif
 }			IvfflatShared;
 
+#if PG_VERSION_NUM >= 120000
 #define ParallelTableScanFromIvfflatShared(shared) \
 	(ParallelTableScanDesc) ((char *) (shared) + BUFFERALIGN(sizeof(IvfflatShared)))
+#endif
 
 typedef struct IvfflatLeader
 {
@@ -172,6 +181,7 @@ typedef struct IvfflatBuildState
 	VectorArray samples;
 	VectorArray centers;
 	ListInfo   *listInfo;
+	Vector	   *normvec;
 
 #ifdef IVFFLAT_KMEANS_DEBUG
 	double		inertia;
@@ -234,8 +244,8 @@ typedef struct IvfflatScanList
 typedef struct IvfflatScanOpaqueData
 {
 	int			probes;
-	int			dimensions;
 	bool		first;
+	Buffer		buf;
 
 	/* Sorting */
 	Tuplesortstate *sortstate;
@@ -265,10 +275,9 @@ VectorArray VectorArrayInit(int maxlen, int dimensions);
 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);
+FmgrInfo   *IvfflatOptionalProcInfo(Relation rel, uint16 procnum);
+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);
 void		IvfflatCommitBuffer(Buffer buf, GenericXLogState *state);
 void		IvfflatAppendPage(Relation index, Buffer *buf, Page *page, GenericXLogState **state, ForkNumber forkNum);

src/ivfinsert.c

@@ -2,7 +2,6 @@
 
 #include <float.h>
 
-#include "access/generic_xlog.h"
 #include "ivfflat.h"
 #include "storage/bufmgr.h"
 #include "storage/lmgr.h"
@@ -12,37 +11,36 @@
  * Find the list that minimizes the distance function
  */
 static void
-FindInsertPage(Relation index, Datum *values, BlockNumber *insertPage, ListInfo * listInfo)
+FindInsertPage(Relation rel, Datum *values, BlockNumber *insertPage, ListInfo * listInfo)
 {
+	Buffer		cbuf;
+	Page		cpage;
+	IvfflatList list;
+	double		distance;
 	double		minDistance = DBL_MAX;
 	BlockNumber nextblkno = IVFFLAT_HEAD_BLKNO;
 	FmgrInfo   *procinfo;
 	Oid			collation;
+	OffsetNumber offno;
+	OffsetNumber maxoffno;
 
 	/* Avoid compiler warning */
 	listInfo->blkno = nextblkno;
 	listInfo->offno = FirstOffsetNumber;
 
-	procinfo = index_getprocinfo(index, 1, IVFFLAT_DISTANCE_PROC);
-	collation = index->rd_indcollation[0];
+	procinfo = index_getprocinfo(rel, 1, IVFFLAT_DISTANCE_PROC);
+	collation = rel->rd_indcollation[0];
 
 	/* Search all list pages */
 	while (BlockNumberIsValid(nextblkno))
 	{
-		Buffer		cbuf;
-		Page		cpage;
-		OffsetNumber maxoffno;
-
-		cbuf = ReadBuffer(index, nextblkno);
+		cbuf = ReadBuffer(rel, nextblkno);
 		LockBuffer(cbuf, BUFFER_LOCK_SHARE);
 		cpage = BufferGetPage(cbuf);
 		maxoffno = PageGetMaxOffsetNumber(cpage);
 
-		for (OffsetNumber offno = FirstOffsetNumber; offno <= maxoffno; offno = OffsetNumberNext(offno))
+		for (offno = FirstOffsetNumber; offno <= maxoffno; offno = OffsetNumberNext(offno))
 		{
-			IvfflatList list;
-			double		distance;
-
 			list = (IvfflatList) PageGetItem(cpage, PageGetItemId(cpage, offno));
 			distance = DatumGetFloat8(FunctionCall2Coll(procinfo, collation, values[0], PointerGetDatum(&list->center)));
 
@@ -65,7 +63,7 @@ FindInsertPage(Relation index, Datum *values, BlockNumber *insertPage, ListInfo
  * Insert a tuple into the index
  */
 static void
-InsertTuple(Relation index, Datum *values, bool *isnull, ItemPointer heap_tid, Relation heapRel)
+InsertTuple(Relation rel, Datum *values, bool *isnull, ItemPointer heap_tid, Relation heapRel)
 {
 	IndexTuple	itup;
 	Datum		value;
@@ -82,33 +80,33 @@ InsertTuple(Relation index, Datum *values, bool *isnull, ItemPointer heap_tid, R
 	value = PointerGetDatum(PG_DETOAST_DATUM(values[0]));
 
 	/* Normalize if needed */
-	normprocinfo = IvfflatOptionalProcInfo(index, IVFFLAT_NORM_PROC);
+	normprocinfo = IvfflatOptionalProcInfo(rel, IVFFLAT_NORM_PROC);
 	if (normprocinfo != NULL)
 	{
-		if (!IvfflatNormValue(normprocinfo, index->rd_indcollation[0], &value))
+		if (!IvfflatNormValue(normprocinfo, rel->rd_indcollation[0], &value, NULL))
 			return;
 	}
 
 	/* Find the insert page - sets the page and list info */
-	FindInsertPage(index, values, &insertPage, &listInfo);
+	FindInsertPage(rel, values, &insertPage, &listInfo);
 	Assert(BlockNumberIsValid(insertPage));
 	originalInsertPage = insertPage;
 
 	/* Form tuple */
-	itup = index_form_tuple(RelationGetDescr(index), &value, isnull);
+	itup = index_form_tuple(RelationGetDescr(rel), &value, isnull);
 	itup->t_tid = *heap_tid;
 
 	/* Get tuple size */
 	itemsz = MAXALIGN(IndexTupleSize(itup));
-	Assert(itemsz <= BLCKSZ - MAXALIGN(SizeOfPageHeaderData) - MAXALIGN(sizeof(IvfflatPageOpaqueData)) - sizeof(ItemIdData));
+	Assert(itemsz <= BLCKSZ - MAXALIGN(SizeOfPageHeaderData) - MAXALIGN(sizeof(IvfflatPageOpaqueData)));
 
 	/* Find a page to insert the item */
 	for (;;)
 	{
-		buf = ReadBuffer(index, insertPage);
+		buf = ReadBuffer(rel, insertPage);
 		LockBuffer(buf, BUFFER_LOCK_EXCLUSIVE);
 
-		state = GenericXLogStart(index);
+		state = GenericXLogStart(rel);
 		page = GenericXLogRegisterBuffer(state, buf, 0);
 
 		if (PageGetFreeSpace(page) >= itemsz)
@@ -128,9 +126,9 @@ InsertTuple(Relation index, Datum *values, bool *isnull, ItemPointer heap_tid, R
 			Page		newpage;
 
 			/* Add a new page */
-			LockRelationForExtension(index, ExclusiveLock);
-			newbuf = IvfflatNewBuffer(index, MAIN_FORKNUM);
-			UnlockRelationForExtension(index, ExclusiveLock);
+			LockRelationForExtension(rel, ExclusiveLock);
+			newbuf = IvfflatNewBuffer(rel, MAIN_FORKNUM);
+			UnlockRelationForExtension(rel, ExclusiveLock);
 
 			/* Init new page */
 			newpage = GenericXLogRegisterBuffer(state, newbuf, GENERIC_XLOG_FULL_IMAGE);
@@ -143,13 +141,15 @@ InsertTuple(Relation index, Datum *values, bool *isnull, ItemPointer heap_tid, R
 			IvfflatPageGetOpaque(page)->nextblkno = insertPage;
 
 			/* Commit */
+			MarkBufferDirty(newbuf);
+			MarkBufferDirty(buf);
 			GenericXLogFinish(state);
 
 			/* Unlock previous buffer */
 			UnlockReleaseBuffer(buf);
 
 			/* Prepare new buffer */
-			state = GenericXLogStart(index);
+			state = GenericXLogStart(rel);
 			buf = newbuf;
 			page = GenericXLogRegisterBuffer(state, buf, 0);
 			break;
@@ -158,13 +158,13 @@ InsertTuple(Relation index, Datum *values, bool *isnull, ItemPointer heap_tid, R
 
 	/* Add to next offset */
 	if (PageAddItem(page, (Item) itup, itemsz, InvalidOffsetNumber, false, false) == InvalidOffsetNumber)
-		elog(ERROR, "failed to add index item to \"%s\"", RelationGetRelationName(index));
+		elog(ERROR, "failed to add index item to \"%s\"", RelationGetRelationName(rel));
 
 	IvfflatCommitBuffer(buf, state);
 
 	/* Update the insert page */
 	if (insertPage != originalInsertPage)
-		IvfflatUpdateList(index, listInfo, insertPage, originalInsertPage, InvalidBlockNumber, MAIN_FORKNUM);
+		IvfflatUpdateList(rel, listInfo, insertPage, originalInsertPage, InvalidBlockNumber, MAIN_FORKNUM);
 }
 
 /*

src/ivfkmeans.c

@@ -6,10 +6,6 @@
 #include "ivfflat.h"
 #include "miscadmin.h"
 
-#ifdef IVFFLAT_MEMORY
-#include "utils/memutils.h"
-#endif
-
 /*
  * Initialize with kmeans++
  *
@@ -21,6 +17,10 @@ InitCenters(Relation index, VectorArray samples, VectorArray centers, float *low
 	FmgrInfo   *procinfo;
 	Oid			collation;
 	int64		j;
+	double		distance;
+	double		sum;
+	double		choice;
+	Vector	   *vec;
 	float	   *weight = palloc(samples->length * sizeof(float));
 	int			numCenters = centers->maxlen;
 	int			numSamples = samples->length;
@@ -33,21 +33,17 @@ InitCenters(Relation index, VectorArray samples, VectorArray centers, float *low
 	centers->length++;
 
 	for (j = 0; j < numSamples; j++)
-		weight[j] = FLT_MAX;
+		weight[j] = DBL_MAX;
 
 	for (int i = 0; i < numCenters; i++)
 	{
-		double		sum;
-		double		choice;
-
 		CHECK_FOR_INTERRUPTS();
 
 		sum = 0.0;
 
 		for (j = 0; j < numSamples; j++)
 		{
-			Vector	   *vec = VectorArrayGet(samples, j);
-			double		distance;
+			vec = VectorArrayGet(samples, j);
 
 			/* Only need to compute distance for new center */
 			/* TODO Use triangle inequality to reduce distance calculations */
@@ -116,6 +112,7 @@ CompareVectors(const void *a, const void *b)
 static void
 QuickCenters(Relation index, VectorArray samples, VectorArray centers)
 {
+	Vector	   *vec;
 	int			dimensions = centers->dim;
 	Oid			collation = index->rd_indcollation[0];
 	FmgrInfo   *normprocinfo = IvfflatOptionalProcInfo(index, IVFFLAT_KMEANS_NORM_PROC);
@@ -126,7 +123,7 @@ QuickCenters(Relation index, VectorArray samples, VectorArray centers)
 		qsort(samples->items, samples->length, VECTOR_SIZE(samples->dim), CompareVectors);
 		for (int i = 0; i < samples->length; i++)
 		{
-			Vector	   *vec = VectorArrayGet(samples, i);
+			vec = VectorArrayGet(samples, i);
 
 			if (i == 0 || CompareVectors(vec, VectorArrayGet(samples, i - 1)) != 0)
 			{
@@ -139,7 +136,7 @@ QuickCenters(Relation index, VectorArray samples, VectorArray centers)
 	/* Fill remaining with random data */
 	while (centers->length < centers->maxlen)
 	{
-		Vector	   *vec = VectorArrayGet(centers, centers->length);
+		vec = VectorArrayGet(centers, centers->length);
 
 		SET_VARSIZE(vec, VECTOR_SIZE(dimensions));
 		vec->dim = dimensions;
@@ -155,23 +152,6 @@ 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.
  *
@@ -188,6 +168,7 @@ ElkanKmeans(Relation index, VectorArray samples, VectorArray centers)
 	Oid			collation;
 	Vector	   *vec;
 	Vector	   *newCenter;
+	int			iteration;
 	int64		j;
 	int64		k;
 	int			dimensions = centers->dim;
@@ -201,6 +182,14 @@ ElkanKmeans(Relation index, VectorArray samples, VectorArray centers)
 	float	   *s;
 	float	   *halfcdist;
 	float	   *newcdist;
+	int			changes;
+	double		minDistance;
+	int			closestCenter;
+	double		distance;
+	bool		rj;
+	bool		rjreset;
+	double		dxcx;
+	double		dxc;
 
 	/* Calculate allocation sizes */
 	Size		samplesSize = VECTOR_ARRAY_SIZE(samples->maxlen, samples->dim);
@@ -252,24 +241,20 @@ ElkanKmeans(Relation index, VectorArray samples, VectorArray centers)
 		vec->dim = dimensions;
 	}
 
-#ifdef IVFFLAT_MEMORY
-	ShowMemoryUsage(totalSize);
-#endif
-
 	/* Pick initial centers */
 	InitCenters(index, samples, centers, lowerBound);
 
 	/* Assign each x to its closest initial center c(x) = argmin d(x,c) */
 	for (j = 0; j < numSamples; j++)
 	{
-		float		minDistance = FLT_MAX;
-		int			closestCenter = 0;
+		minDistance = DBL_MAX;
+		closestCenter = 0;
 
 		/* Find closest center */
 		for (k = 0; k < numCenters; k++)
 		{
 			/* TODO Use Lemma 1 in k-means++ initialization */
-			float		distance = lowerBound[j * numCenters + k];
+			distance = lowerBound[j * numCenters + k];
 
 			if (distance < minDistance)
 			{
@@ -283,14 +268,13 @@ ElkanKmeans(Relation index, VectorArray samples, VectorArray centers)
 	}
 
 	/* Give 500 iterations to converge */
-	for (int iteration = 0; iteration < 500; iteration++)
+	for (iteration = 0; iteration < 500; iteration++)
 	{
-		int			changes = 0;
-		bool		rjreset;
-
 		/* Can take a while, so ensure we can interrupt */
 		CHECK_FOR_INTERRUPTS();
 
+		changes = 0;
+
 		/* Step 1: For all centers, compute distance */
 		for (j = 0; j < numCenters; j++)
 		{
@@ -298,8 +282,7 @@ ElkanKmeans(Relation index, VectorArray samples, VectorArray centers)
 
 			for (k = j + 1; k < numCenters; k++)
 			{
-				float		distance = 0.5 * DatumGetFloat8(FunctionCall2Coll(procinfo, collation, PointerGetDatum(vec), PointerGetDatum(VectorArrayGet(centers, k))));
-
+				distance = 0.5 * DatumGetFloat8(FunctionCall2Coll(procinfo, collation, PointerGetDatum(vec), PointerGetDatum(VectorArrayGet(centers, k))));
 				halfcdist[j * numCenters + k] = distance;
 				halfcdist[k * numCenters + j] = distance;
 			}
@@ -308,12 +291,10 @@ ElkanKmeans(Relation index, VectorArray samples, VectorArray centers)
 		/* For all centers c, compute s(c) */
 		for (j = 0; j < numCenters; j++)
 		{
-			float		minDistance = FLT_MAX;
+			minDistance = DBL_MAX;
 
 			for (k = 0; k < numCenters; k++)
 			{
-				float		distance;
-
 				if (j == k)
 					continue;
 
@@ -329,8 +310,6 @@ ElkanKmeans(Relation index, VectorArray samples, VectorArray centers)
 
 		for (j = 0; j < numSamples; j++)
 		{
-			bool		rj;
-
 			/* Step 2: Identify all points x such that u(x) <= s(c(x)) */
 			if (upperBound[j] <= s[closestCenters[j]])
 				continue;
@@ -339,8 +318,6 @@ ElkanKmeans(Relation index, VectorArray samples, VectorArray centers)
 
 			for (k = 0; k < numCenters; k++)
 			{
-				float		dxcx;
-
 				/* Step 3: For all remaining points x and centers c */
 				if (k == closestCenters[j])
 					continue;
@@ -370,7 +347,7 @@ ElkanKmeans(Relation index, VectorArray samples, VectorArray centers)
 				/* Step 3b */
 				if (dxcx > lowerBound[j * numCenters + k] || dxcx > halfcdist[closestCenters[j] * numCenters + k])
 				{
-					float		dxc = DatumGetFloat8(FunctionCall2Coll(procinfo, collation, PointerGetDatum(vec), PointerGetDatum(VectorArrayGet(centers, k))));
+					dxc = DatumGetFloat8(FunctionCall2Coll(procinfo, collation, PointerGetDatum(vec), PointerGetDatum(VectorArrayGet(centers, k))));
 
 					/* d(x,c) calculated */
 					lowerBound[j * numCenters + k] = dxc;
@@ -384,6 +361,7 @@ ElkanKmeans(Relation index, VectorArray samples, VectorArray centers)
 
 						changes++;
 					}
+
 				}
 			}
 		}
@@ -400,8 +378,6 @@ ElkanKmeans(Relation index, VectorArray samples, VectorArray centers)
 
 		for (j = 0; j < numSamples; j++)
 		{
-			int			closestCenter;
-
 			vec = VectorArrayGet(samples, j);
 			closestCenter = closestCenters[j];
 
@@ -450,7 +426,7 @@ ElkanKmeans(Relation index, VectorArray samples, VectorArray centers)
 		{
 			for (k = 0; k < numCenters; k++)
 			{
-				float		distance = lowerBound[j * numCenters + k] - newcdist[k];
+				distance = lowerBound[j * numCenters + k] - newcdist[k];
 
 				if (distance < 0)
 					distance = 0;
@@ -466,7 +442,7 @@ ElkanKmeans(Relation index, VectorArray samples, VectorArray centers)
 
 		/* Step 7 */
 		for (j = 0; j < numCenters; j++)
-			VectorArraySet(centers, j, VectorArrayGet(newCenters, j));
+			memcpy(VectorArrayGet(centers, j), VectorArrayGet(newCenters, j), VECTOR_SIZE(dimensions));
 
 		if (changes == 0 && iteration != 0)
 			break;
@@ -489,6 +465,9 @@ static void
 CheckCenters(Relation index, VectorArray centers)
 {
 	FmgrInfo   *normprocinfo;
+	Oid			collation;
+	Vector	   *vec;
+	double		norm;
 
 	if (centers->length != centers->maxlen)
 		elog(ERROR, "Not enough centers. Please report a bug.");
@@ -496,7 +475,7 @@ CheckCenters(Relation index, VectorArray centers)
 	/* Ensure no NaN or infinite values */
 	for (int i = 0; i < centers->length; i++)
 	{
-		Vector	   *vec = VectorArrayGet(centers, i);
+		vec = VectorArrayGet(centers, i);
 
 		for (int j = 0; j < vec->dim; j++)
 		{
@@ -522,12 +501,11 @@ CheckCenters(Relation index, VectorArray centers)
 	normprocinfo = IvfflatOptionalProcInfo(index, IVFFLAT_NORM_PROC);
 	if (normprocinfo != NULL)
 	{
-		Oid			collation = index->rd_indcollation[0];
+		collation = index->rd_indcollation[0];
 
 		for (int i = 0; i < centers->length; i++)
 		{
-			double		norm = DatumGetFloat8(FunctionCall1Coll(normprocinfo, collation, PointerGetDatum(VectorArrayGet(centers, i))));
-
+			norm = DatumGetFloat8(FunctionCall1Coll(normprocinfo, collation, PointerGetDatum(VectorArrayGet(centers, i))));
 			if (norm == 0)
 				elog(ERROR, "Zero norm detected. Please report a bug.");
 		}

src/ivfscan.c

@@ -5,7 +5,6 @@
 #include "access/relscan.h"
 #include "catalog/pg_operator_d.h"
 #include "catalog/pg_type_d.h"
-#include "lib/pairingheap.h"
 #include "ivfflat.h"
 #include "miscadmin.h"
 #include "pgstat.h"
@@ -32,36 +31,36 @@ CompareLists(const pairingheap_node *a, const pairingheap_node *b, void *arg)
 static void
 GetScanLists(IndexScanDesc scan, Datum value)
 {
-	IvfflatScanOpaque so = (IvfflatScanOpaque) scan->opaque;
+	Buffer		cbuf;
+	Page		cpage;
+	IvfflatList list;
+	OffsetNumber offno;
+	OffsetNumber maxoffno;
 	BlockNumber nextblkno = IVFFLAT_HEAD_BLKNO;
 	int			listCount = 0;
+	IvfflatScanOpaque so = (IvfflatScanOpaque) scan->opaque;
+	double		distance;
+	IvfflatScanList *scanlist;
 	double		maxDistance = DBL_MAX;
 
 	/* Search all list pages */
 	while (BlockNumberIsValid(nextblkno))
 	{
-		Buffer		cbuf;
-		Page		cpage;
-		OffsetNumber maxoffno;
-
 		cbuf = ReadBuffer(scan->indexRelation, nextblkno);
 		LockBuffer(cbuf, BUFFER_LOCK_SHARE);
 		cpage = BufferGetPage(cbuf);
 
 		maxoffno = PageGetMaxOffsetNumber(cpage);
 
-		for (OffsetNumber offno = FirstOffsetNumber; offno <= maxoffno; offno = OffsetNumberNext(offno))
+		for (offno = FirstOffsetNumber; offno <= maxoffno; offno = OffsetNumberNext(offno))
 		{
-			IvfflatList list = (IvfflatList) PageGetItem(cpage, PageGetItemId(cpage, offno));
-			double		distance;
+			list = (IvfflatList) PageGetItem(cpage, PageGetItemId(cpage, offno));
 
 			/* Use procinfo from the index instead of scan key for performance */
 			distance = DatumGetFloat8(FunctionCall2Coll(so->procinfo, so->collation, PointerGetDatum(&list->center), value));
 
 			if (listCount < so->probes)
 			{
-				IvfflatScanList *scanlist;
-
 				scanlist = &so->lists[listCount];
 				scanlist->startPage = list->startPage;
 				scanlist->distance = distance;
@@ -76,8 +75,6 @@ GetScanLists(IndexScanDesc scan, Datum value)
 			}
 			else if (distance < maxDistance)
 			{
-				IvfflatScanList *scanlist;
-
 				/* Remove */
 				scanlist = (IvfflatScanList *) pairingheap_remove_first(so->listQueue);
 
@@ -104,9 +101,22 @@ static void
 GetScanItems(IndexScanDesc scan, Datum value)
 {
 	IvfflatScanOpaque so = (IvfflatScanOpaque) scan->opaque;
+	Buffer		buf;
+	Page		page;
+	IndexTuple	itup;
+	BlockNumber searchPage;
+	OffsetNumber offno;
+	OffsetNumber maxoffno;
+	Datum		datum;
+	bool		isnull;
 	TupleDesc	tupdesc = RelationGetDescr(scan->indexRelation);
 	double		tuples = 0;
+
+#if PG_VERSION_NUM >= 120000
 	TupleTableSlot *slot = MakeSingleTupleTableSlot(so->tupdesc, &TTSOpsVirtual);
+#else
+	TupleTableSlot *slot = MakeSingleTupleTableSlot(so->tupdesc);
+#endif
 
 	/*
 	 * Reuse same set of shared buffers for scan
@@ -118,28 +128,19 @@ GetScanItems(IndexScanDesc scan, Datum value)
 	/* Search closest probes lists */
 	while (!pairingheap_is_empty(so->listQueue))
 	{
-		BlockNumber searchPage = ((IvfflatScanList *) pairingheap_remove_first(so->listQueue))->startPage;
+		searchPage = ((IvfflatScanList *) pairingheap_remove_first(so->listQueue))->startPage;
 
 		/* Search all entry pages for list */
 		while (BlockNumberIsValid(searchPage))
 		{
-			Buffer		buf;
-			Page		page;
-			OffsetNumber maxoffno;
-
 			buf = ReadBufferExtended(scan->indexRelation, MAIN_FORKNUM, searchPage, RBM_NORMAL, bas);
 			LockBuffer(buf, BUFFER_LOCK_SHARE);
 			page = BufferGetPage(buf);
 			maxoffno = PageGetMaxOffsetNumber(page);
 
-			for (OffsetNumber offno = FirstOffsetNumber; offno <= maxoffno; offno = OffsetNumberNext(offno))
+			for (offno = FirstOffsetNumber; offno <= maxoffno; offno = OffsetNumberNext(offno))
 			{
-				IndexTuple	itup;
-				Datum		datum;
-				bool		isnull;
-				ItemId		itemid = PageGetItemId(page, offno);
-
-				itup = (IndexTuple) PageGetItem(page, itemid);
+				itup = (IndexTuple) PageGetItem(page, PageGetItemId(page, offno));
 				datum = index_getattr(itup, 1, tupdesc, &isnull);
 
 				/*
@@ -153,6 +154,8 @@ GetScanItems(IndexScanDesc scan, Datum value)
 				slot->tts_isnull[0] = false;
 				slot->tts_values[1] = PointerGetDatum(&itup->t_tid);
 				slot->tts_isnull[1] = false;
+				slot->tts_values[2] = Int32GetDatum((int) searchPage);
+				slot->tts_isnull[2] = false;
 				ExecStoreVirtualTuple(slot);
 
 				tuplesort_puttupleslot(so->sortstate, slot);
@@ -177,6 +180,29 @@ GetScanItems(IndexScanDesc scan, Datum value)
 	tuplesort_performsort(so->sortstate);
 }
 
+/*
+ * Get dimensions from metapage
+ */
+static int
+GetDimensions(Relation index)
+{
+	Buffer		buf;
+	Page		page;
+	IvfflatMetaPage metap;
+	int			dimensions;
+
+	buf = ReadBuffer(index, IVFFLAT_METAPAGE_BLKNO);
+	LockBuffer(buf, BUFFER_LOCK_SHARE);
+	page = BufferGetPage(buf);
+	metap = IvfflatPageGetMeta(page);
+
+	dimensions = metap->dimensions;
+
+	UnlockReleaseBuffer(buf);
+
+	return dimensions;
+}
+
 /*
  * Prepare for an index scan
  */
@@ -186,7 +212,6 @@ ivfflatbeginscan(Relation index, int nkeys, int norderbys)
 	IndexScanDesc scan;
 	IvfflatScanOpaque so;
 	int			lists;
-	int			dimensions;
 	AttrNumber	attNums[] = {1};
 	Oid			sortOperators[] = {Float8LessOperator};
 	Oid			sortCollations[] = {InvalidOid};
@@ -194,17 +219,15 @@ ivfflatbeginscan(Relation index, int nkeys, int norderbys)
 	int			probes = ivfflat_probes;
 
 	scan = RelationGetIndexScan(index, nkeys, norderbys);
-
-	/* Get lists and dimensions from metapage */
-	IvfflatGetMetaPageInfo(index, &lists, &dimensions);
+	lists = IvfflatGetLists(scan->indexRelation);
 
 	if (probes > lists)
 		probes = lists;
 
 	so = (IvfflatScanOpaque) palloc(offsetof(IvfflatScanOpaqueData, lists) + probes * sizeof(IvfflatScanList));
+	so->buf = InvalidBuffer;
 	so->first = true;
 	so->probes = probes;
-	so->dimensions = dimensions;
 
 	/* Set support functions */
 	so->procinfo = index_getprocinfo(index, 1, IVFFLAT_DISTANCE_PROC);
@@ -212,14 +235,23 @@ ivfflatbeginscan(Relation index, int nkeys, int norderbys)
 	so->collation = index->rd_indcollation[0];
 
 	/* Create tuple description for sorting */
-	so->tupdesc = CreateTemplateTupleDesc(2);
+#if PG_VERSION_NUM >= 120000
+	so->tupdesc = CreateTemplateTupleDesc(3);
+#else
+	so->tupdesc = CreateTemplateTupleDesc(3, false);
+#endif
 	TupleDescInitEntry(so->tupdesc, (AttrNumber) 1, "distance", FLOAT8OID, -1, 0);
-	TupleDescInitEntry(so->tupdesc, (AttrNumber) 2, "heaptid", TIDOID, -1, 0);
+	TupleDescInitEntry(so->tupdesc, (AttrNumber) 2, "tid", TIDOID, -1, 0);
+	TupleDescInitEntry(so->tupdesc, (AttrNumber) 3, "indexblkno", INT4OID, -1, 0);
 
 	/* Prep sort */
 	so->sortstate = tuplesort_begin_heap(so->tupdesc, 1, attNums, sortOperators, sortCollations, nullsFirstFlags, work_mem, NULL, false);
 
+#if PG_VERSION_NUM >= 120000
 	so->slot = MakeSingleTupleTableSlot(so->tupdesc, &TTSOpsMinimalTuple);
+#else
+	so->slot = MakeSingleTupleTableSlot(so->tupdesc);
+#endif
 
 	so->listQueue = pairingheap_allocate(CompareLists, scan);
 
@@ -276,13 +308,8 @@ ivfflatgettuple(IndexScanDesc scan, ScanDirection dir)
 		if (scan->orderByData == NULL)
 			elog(ERROR, "cannot scan ivfflat index without order");
 
-		/* Requires MVCC-compliant snapshot as not able to pin during sorting */
-		/* https://www.postgresql.org/docs/current/index-locking.html */
-		if (!IsMVCCSnapshot(scan->xs_snapshot))
-			elog(ERROR, "non-MVCC snapshots are not supported with ivfflat");
-
 		if (scan->orderByData->sk_flags & SK_ISNULL)
-			value = PointerGetDatum(InitVector(so->dimensions));
+			value = PointerGetDatum(InitVector(GetDimensions(scan->indexRelation)));
 		else
 		{
 			value = scan->orderByData->sk_argument;
@@ -293,7 +320,7 @@ 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));
@@ -307,10 +334,27 @@ ivfflatgettuple(IndexScanDesc scan, ScanDirection dir)
 
 	if (tuplesort_gettupleslot(so->sortstate, true, false, so->slot, NULL))
 	{
-		ItemPointer heaptid = (ItemPointer) DatumGetPointer(slot_getattr(so->slot, 2, &so->isnull));
+		ItemPointer tid = (ItemPointer) DatumGetPointer(slot_getattr(so->slot, 2, &so->isnull));
+		BlockNumber indexblkno = DatumGetInt32(slot_getattr(so->slot, 3, &so->isnull));
+
+#if PG_VERSION_NUM >= 120000
+		scan->xs_heaptid = *tid;
+#else
+		scan->xs_ctup.t_self = *tid;
+#endif
+
+		/* Unpin buffer */
+		if (BufferIsValid(so->buf))
+			ReleaseBuffer(so->buf);
+
+		/*
+		 * An index scan must maintain a pin on the index page holding the
+		 * item last returned by amgettuple
+		 *
+		 * https://www.postgresql.org/docs/current/index-locking.html
+		 */
+		so->buf = ReadBuffer(scan->indexRelation, indexblkno);
 
-		scan->xs_heaptid = *heaptid;
-		scan->xs_recheck = false;
 		scan->xs_recheckorderby = false;
 		return true;
 	}
@@ -326,6 +370,10 @@ ivfflatendscan(IndexScanDesc scan)
 {
 	IvfflatScanOpaque so = (IvfflatScanOpaque) scan->opaque;
 
+	/* Release pin */
+	if (BufferIsValid(so->buf))
+		ReleaseBuffer(so->buf);
+
 	pairingheap_free(so->listQueue);
 	tuplesort_end(so->sortstate);
 

src/ivfutils.c

@@ -1,6 +1,5 @@
 #include "postgres.h"
 
-#include "access/generic_xlog.h"
 #include "ivfflat.h"
 #include "storage/bufmgr.h"
 #include "vector.h"
@@ -58,12 +57,12 @@ IvfflatGetLists(Relation index)
  * Get proc
  */
 FmgrInfo *
-IvfflatOptionalProcInfo(Relation index, uint16 procnum)
+IvfflatOptionalProcInfo(Relation rel, uint16 procnum)
 {
-	if (!OidIsValid(index_getprocid(index, 1, procnum)))
+	if (!OidIsValid(index_getprocid(rel, 1, procnum)))
 		return NULL;
 
-	return index_getprocinfo(index, 1, procnum);
+	return index_getprocinfo(rel, 1, procnum);
 }
 
 /*
@@ -75,14 +74,16 @@ 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;
@@ -135,6 +136,7 @@ IvfflatInitRegisterPage(Relation index, Buffer *buf, Page *page, GenericXLogStat
 void
 IvfflatCommitBuffer(Buffer buf, GenericXLogState *state)
 {
+	MarkBufferDirty(buf);
 	GenericXLogFinish(state);
 	UnlockReleaseBuffer(buf);
 }
@@ -158,6 +160,8 @@ IvfflatAppendPage(Relation index, Buffer *buf, Page *page, GenericXLogState **st
 	IvfflatInitPage(newbuf, newpage);
 
 	/* Commit */
+	MarkBufferDirty(*buf);
+	MarkBufferDirty(newbuf);
 	GenericXLogFinish(*state);
 
 	/* Unlock */
@@ -168,29 +172,6 @@ IvfflatAppendPage(Relation index, Buffer *buf, Page *page, GenericXLogState **st
 	*buf = newbuf;
 }
 
-/*
- * Get the metapage info
- */
-void
-IvfflatGetMetaPageInfo(Relation index, int *lists, int *dimensions)
-{
-	Buffer		buf;
-	Page		page;
-	IvfflatMetaPage metap;
-
-	buf = ReadBuffer(index, IVFFLAT_METAPAGE_BLKNO);
-	LockBuffer(buf, BUFFER_LOCK_SHARE);
-	page = BufferGetPage(buf);
-	metap = IvfflatPageGetMeta(page);
-
-	*lists = metap->lists;
-
-	if (dimensions != NULL)
-		*dimensions = metap->dimensions;
-
-	UnlockReleaseBuffer(buf);
-}
-
 /*
  * Update the start or insert page of a list
  */

src/ivfvacuum.c

@@ -1,6 +1,5 @@
 #include "postgres.h"
 
-#include "access/generic_xlog.h"
 #include "commands/vacuum.h"
 #include "ivfflat.h"
 #include "storage/bufmgr.h"
@@ -108,6 +107,7 @@ ivfflatbulkdelete(IndexVacuumInfo *info, IndexBulkDeleteResult *stats,
 				{
 					/* Delete tuples */
 					PageIndexMultiDelete(page, deletable, ndeletable);
+					MarkBufferDirty(buf);
 					GenericXLogFinish(state);
 				}
 				else

src/sparsevec.c

@@ -1,779 +0,0 @@
-#include "postgres.h"
-
-#include <limits.h>
-#include <math.h>
-
-#include "fmgr.h"
-#include "libpq/pqformat.h"
-#include "sparsevec.h"
-#include "utils/array.h"
-#include "utils/builtins.h"
-#include "vector.h"
-
-#if PG_VERSION_NUM >= 120000
-#include "common/shortest_dec.h"
-#include "utils/float.h"
-#else
-#include <float.h>
-#include "utils/builtins.h"
-#endif
-
-/*
- * Ensure same dimensions
- */
-static inline void
-CheckDims(SparseVector * a, SparseVector * b)
-{
-	if (a->dim != b->dim)
-		ereport(ERROR,
-				(errcode(ERRCODE_DATA_EXCEPTION),
-				 errmsg("different sparsevec dimensions %d and %d", a->dim, b->dim)));
-}
-
-/*
- * Ensure expected dimensions
- */
-static inline void
-CheckExpectedDim(int32 typmod, int dim)
-{
-	if (typmod != -1 && typmod != dim)
-		ereport(ERROR,
-				(errcode(ERRCODE_DATA_EXCEPTION),
-				 errmsg("expected %d dimensions, not %d", typmod, dim)));
-}
-
-/*
- * Ensure valid dimensions
- */
-static inline void
-CheckDim(int dim)
-{
-	if (dim < 1)
-		ereport(ERROR,
-				(errcode(ERRCODE_DATA_EXCEPTION),
-				 errmsg("sparsevec must have at least 1 dimension")));
-
-	if (dim > SPARSEVEC_MAX_DIM)
-		ereport(ERROR,
-				(errcode(ERRCODE_PROGRAM_LIMIT_EXCEEDED),
-				 errmsg("sparsevec cannot have more than %d dimensions", SPARSEVEC_MAX_DIM)));
-}
-
-/*
- * Ensure valid nnz
- */
-static inline void
-CheckNnz(int nnz, int dim)
-{
-	if (nnz < 0)
-		ereport(ERROR,
-				(errcode(ERRCODE_DATA_EXCEPTION),
-				 errmsg("sparsevec must have at least one element")));
-
-	if (nnz > dim)
-		ereport(ERROR,
-				(errcode(ERRCODE_PROGRAM_LIMIT_EXCEEDED),
-				 errmsg("sparsevec cannot have more elements than dimensions")));
-}
-
-/*
- * Ensure valid index
- */
-static inline void
-CheckIndex(int32 *indices, int i, int dim)
-{
-	int32		index = indices[i];
-
-	if (index < 0)
-		ereport(ERROR,
-				(errcode(ERRCODE_DATA_EXCEPTION),
-				 errmsg("index must not be negative")));
-
-	if (index >= dim)
-		ereport(ERROR,
-				(errcode(ERRCODE_DATA_EXCEPTION),
-				 errmsg("index must be less than dimensions")));
-
-	if (i > 0)
-	{
-		if (index < indices[i - 1])
-			ereport(ERROR,
-					(errcode(ERRCODE_DATA_EXCEPTION),
-					 errmsg("indexes must be in ascending order")));
-
-		if (index == indices[i - 1])
-			ereport(ERROR,
-					(errcode(ERRCODE_DATA_EXCEPTION),
-					 errmsg("indexes must not contain duplicates")));
-	}
-}
-
-/*
- * Ensure finite element
- */
-static inline void
-CheckElement(float value)
-{
-	if (isnan(value))
-		ereport(ERROR,
-				(errcode(ERRCODE_DATA_EXCEPTION),
-				 errmsg("NaN not allowed in sparsevec")));
-
-	if (isinf(value))
-		ereport(ERROR,
-				(errcode(ERRCODE_DATA_EXCEPTION),
-				 errmsg("infinite value not allowed in sparsevec")));
-}
-
-/*
- * Allocate and initialize a new sparse vector
- */
-SparseVector *
-InitSparseVector(int dim, int nnz)
-{
-	SparseVector *result;
-	int			size;
-
-	size = SPARSEVEC_SIZE(nnz);
-	result = (SparseVector *) palloc0(size);
-	SET_VARSIZE(result, size);
-	result->dim = dim;
-	result->nnz = nnz;
-
-	return result;
-}
-
-/*
- * Check for whitespace, since array_isspace() is static
- */
-static inline bool
-sparsevec_isspace(char ch)
-{
-	if (ch == ' ' ||
-		ch == '\t' ||
-		ch == '\n' ||
-		ch == '\r' ||
-		ch == '\v' ||
-		ch == '\f')
-		return true;
-	return false;
-}
-
-/*
- * Convert textual representation to internal representation
- */
-PGDLLEXPORT PG_FUNCTION_INFO_V1(sparsevec_in);
-Datum
-sparsevec_in(PG_FUNCTION_ARGS)
-{
-	char	   *lit = PG_GETARG_CSTRING(0);
-	int32		typmod = PG_GETARG_INT32(2);
-	int			dim;
-	char	   *pt;
-	char	   *stringEnd;
-	SparseVector *result;
-	float	   *rvalues;
-	char	   *litcopy = pstrdup(lit);
-	char	   *str = litcopy;
-	int32	   *indices;
-	float	   *values;
-	int			maxNnz;
-	int			nnz = 0;
-
-	maxNnz = 1;
-	pt = str;
-	while (*pt != '\0')
-	{
-		if (*pt == ',')
-			maxNnz++;
-
-		pt++;
-	}
-
-	indices = palloc(maxNnz * sizeof(int32));
-	values = palloc(maxNnz * sizeof(float));
-
-	while (sparsevec_isspace(*str))
-		str++;
-
-	if (*str != '{')
-		ereport(ERROR,
-				(errcode(ERRCODE_INVALID_TEXT_REPRESENTATION),
-				 errmsg("malformed sparsevec literal: \"%s\"", lit),
-				 errdetail("Vector contents must start with \"{\".")));
-
-	str++;
-	pt = strtok(str, ",");
-	stringEnd = pt;
-
-	while (pt != NULL && *stringEnd != '}')
-	{
-		long		index;
-		float		value;
-
-		/* TODO Better error */
-		if (nnz == maxNnz)
-			ereport(ERROR,
-					(errcode(ERRCODE_INVALID_TEXT_REPRESENTATION),
-					 errmsg("ran out of buffer: \"%s\"", lit)));
-
-		while (sparsevec_isspace(*pt))
-			pt++;
-
-		/* Check for empty string like float4in */
-		if (*pt == '\0')
-			ereport(ERROR,
-					(errcode(ERRCODE_INVALID_TEXT_REPRESENTATION),
-					 errmsg("invalid input syntax for type sparsevec: \"%s\"", lit)));
-
-		/* Use similar logic as int2vectorin */
-		errno = 0;
-		index = strtol(pt, &stringEnd, 10);
-
-		if (stringEnd == pt)
-			ereport(ERROR,
-					(errcode(ERRCODE_INVALID_TEXT_REPRESENTATION),
-					 errmsg("invalid input syntax for type sparsevec: \"%s\"", lit)));
-
-		if (errno == ERANGE || index < 0 || index > INT_MAX)
-			ereport(ERROR,
-					(errcode(ERRCODE_NUMERIC_VALUE_OUT_OF_RANGE),
-					 errmsg("index \"%ld\" is out of range for type sparsevec", index)));
-
-		if (stringEnd == pt)
-			ereport(ERROR,
-					(errcode(ERRCODE_INVALID_TEXT_REPRESENTATION),
-					 errmsg("invalid input syntax for type sparsevec: \"%s\"", lit)));
-
-		while (sparsevec_isspace(*stringEnd))
-			stringEnd++;
-
-		if (*stringEnd != ':')
-			ereport(ERROR,
-					(errcode(ERRCODE_INVALID_TEXT_REPRESENTATION),
-					 errmsg("invalid input syntax for type sparsevec: \"%s\"", lit)));
-
-		stringEnd++;
-
-		while (sparsevec_isspace(*stringEnd))
-			stringEnd++;
-
-		errno = 0;
-		pt = stringEnd;
-		value = strtof(pt, &stringEnd);
-
-		if (stringEnd == pt)
-			ereport(ERROR,
-					(errcode(ERRCODE_INVALID_TEXT_REPRESENTATION),
-					 errmsg("invalid input syntax for type sparsevec: \"%s\"", lit)));
-
-		/* Check for range error like float4in */
-		if (errno == ERANGE && (value == 0 || isinf(value)))
-			ereport(ERROR,
-					(errcode(ERRCODE_NUMERIC_VALUE_OUT_OF_RANGE),
-					 errmsg("\"%s\" is out of range for type sparsevec", pt)));
-
-		/* TODO Decide whether to store zero values */
-		if (value != 0)
-		{
-			indices[nnz] = index;
-			values[nnz] = value;
-			nnz++;
-		}
-
-		if (*stringEnd != '\0' && *stringEnd != '}')
-			ereport(ERROR,
-					(errcode(ERRCODE_INVALID_TEXT_REPRESENTATION),
-					 errmsg("invalid input syntax for type sparsevec: \"%s\"", lit)));
-
-		pt = strtok(NULL, ",");
-	}
-
-	if (stringEnd == NULL || *stringEnd != '}')
-		ereport(ERROR,
-				(errcode(ERRCODE_INVALID_TEXT_REPRESENTATION),
-				 errmsg("malformed sparsevec literal: \"%s\"", lit),
-				 errdetail("Unexpected end of input.")));
-
-	stringEnd++;
-
-	if (*stringEnd != '/')
-		ereport(ERROR,
-				(errcode(ERRCODE_INVALID_TEXT_REPRESENTATION),
-				 errmsg("malformed sparsevec literal: \"%s\"", lit),
-				 errdetail("Unexpected end of input.")));
-
-	stringEnd++;
-
-	/* Use similar logic as int2vectorin */
-	errno = 0;
-	pt = stringEnd;
-	dim = strtol(pt, &stringEnd, 10);
-
-	if (stringEnd == pt)
-		ereport(ERROR,
-				(errcode(ERRCODE_INVALID_TEXT_REPRESENTATION),
-				 errmsg("invalid input syntax for type sparsevec: \"%s\"", lit)));
-
-	/* Only whitespace is allowed after the closing brace */
-	while (sparsevec_isspace(*stringEnd))
-		stringEnd++;
-
-	if (*stringEnd != '\0')
-		ereport(ERROR,
-				(errcode(ERRCODE_INVALID_TEXT_REPRESENTATION),
-				 errmsg("malformed sparsevec literal: \"%s\"", lit),
-				 errdetail("Junk after closing.")));
-
-	pfree(litcopy);
-
-	CheckDim(dim);
-	CheckExpectedDim(typmod, dim);
-
-	result = InitSparseVector(dim, nnz);
-	rvalues = SPARSEVEC_VALUES(result);
-	for (int i = 0; i < nnz; i++)
-	{
-		result->indices[i] = indices[i];
-		rvalues[i] = values[i];
-
-		CheckIndex(result->indices, i, dim);
-		CheckElement(rvalues[i]);
-	}
-
-	PG_RETURN_POINTER(result);
-}
-
-#define AppendChar(ptr, c) (*(ptr)++ = (c))
-#define AppendFloat(ptr, f) ((ptr) += float_to_shortest_decimal_bufn((f), (ptr)))
-
-#if PG_VERSION_NUM >= 140000
-#define AppendInt(ptr, i) ((ptr) += pg_ltoa((i), (ptr)))
-#else
-#define AppendInt(ptr, i) \
-	do { \
-		pg_ltoa(i, ptr); \
-		while (*ptr != '\0') \
-			ptr++; \
-	} while (0)
-#endif
-
-/*
- * Convert internal representation to textual representation
- */
-PGDLLEXPORT PG_FUNCTION_INFO_V1(sparsevec_out);
-Datum
-sparsevec_out(PG_FUNCTION_ARGS)
-{
-	SparseVector *sparsevec = PG_GETARG_SPARSEVEC_P(0);
-	float	   *values = SPARSEVEC_VALUES(sparsevec);
-	char	   *buf;
-	char	   *ptr;
-
-	/*
-	 * Need:
-	 *
-	 * nnz * 10 bytes for index (positive integer)
-	 *
-	 * nnz bytes for :
-	 *
-	 * nnz * (FLOAT_SHORTEST_DECIMAL_LEN - 1) bytes for
-	 * float_to_shortest_decimal_bufn
-	 *
-	 * nnz - 1 bytes for ,
-	 *
-	 * 10 bytes for dimensions
-	 *
-	 * 4 bytes for {, }, /, and \0
-	 */
-	buf = (char *) palloc((11 + FLOAT_SHORTEST_DECIMAL_LEN) * sparsevec->nnz + 13);
-	ptr = buf;
-
-	AppendChar(ptr, '{');
-
-	for (int i = 0; i < sparsevec->nnz; i++)
-	{
-		if (i > 0)
-			AppendChar(ptr, ',');
-
-		AppendInt(ptr, sparsevec->indices[i]);
-		AppendChar(ptr, ':');
-		AppendFloat(ptr, values[i]);
-	}
-
-	AppendChar(ptr, '}');
-	AppendChar(ptr, '/');
-	AppendInt(ptr, sparsevec->dim);
-
-	*ptr = '\0';
-
-	PG_FREE_IF_COPY(sparsevec, 0);
-	PG_RETURN_CSTRING(buf);
-}
-
-/*
- * Convert type modifier
- */
-PGDLLEXPORT PG_FUNCTION_INFO_V1(sparsevec_typmod_in);
-Datum
-sparsevec_typmod_in(PG_FUNCTION_ARGS)
-{
-	ArrayType  *ta = PG_GETARG_ARRAYTYPE_P(0);
-	int32	   *tl;
-	int			n;
-
-	tl = ArrayGetIntegerTypmods(ta, &n);
-
-	if (n != 1)
-		ereport(ERROR,
-				(errcode(ERRCODE_INVALID_PARAMETER_VALUE),
-				 errmsg("invalid type modifier")));
-
-	if (*tl < 1)
-		ereport(ERROR,
-				(errcode(ERRCODE_INVALID_PARAMETER_VALUE),
-				 errmsg("dimensions for type sparsevec must be at least 1")));
-
-	if (*tl > SPARSEVEC_MAX_DIM)
-		ereport(ERROR,
-				(errcode(ERRCODE_INVALID_PARAMETER_VALUE),
-				 errmsg("dimensions for type sparsevec cannot exceed %d", SPARSEVEC_MAX_DIM)));
-
-	PG_RETURN_INT32(*tl);
-}
-
-/*
- * Convert external binary representation to internal representation
- */
-PGDLLEXPORT PG_FUNCTION_INFO_V1(sparsevec_recv);
-Datum
-sparsevec_recv(PG_FUNCTION_ARGS)
-{
-	StringInfo	buf = (StringInfo) PG_GETARG_POINTER(0);
-	int32		typmod = PG_GETARG_INT32(2);
-	SparseVector *result;
-	int32		dim;
-	int32		nnz;
-	int32		unused;
-	float	   *values;
-
-	dim = pq_getmsgint(buf, sizeof(int32));
-	nnz = pq_getmsgint(buf, sizeof(int32));
-	unused = pq_getmsgint(buf, sizeof(int32));
-
-	CheckDim(dim);
-	CheckNnz(nnz, dim);
-	CheckExpectedDim(typmod, dim);
-
-	if (unused != 0)
-		ereport(ERROR,
-				(errcode(ERRCODE_DATA_EXCEPTION),
-				 errmsg("expected unused to be 0, not %d", unused)));
-
-	result = InitSparseVector(dim, nnz);
-	values = SPARSEVEC_VALUES(result);
-
-	for (int i = 0; i < nnz; i++)
-	{
-		result->indices[i] = pq_getmsgint(buf, sizeof(int32));
-		CheckIndex(result->indices, i, dim);
-	}
-
-	for (int i = 0; i < nnz; i++)
-	{
-		values[i] = pq_getmsgfloat4(buf);
-		CheckElement(values[i]);
-	}
-
-	PG_RETURN_POINTER(result);
-}
-
-/*
- * Convert internal representation to the external binary representation
- */
-PGDLLEXPORT PG_FUNCTION_INFO_V1(sparsevec_send);
-Datum
-sparsevec_send(PG_FUNCTION_ARGS)
-{
-	SparseVector *svec = PG_GETARG_SPARSEVEC_P(0);
-	float	   *values = SPARSEVEC_VALUES(svec);
-	StringInfoData buf;
-
-	pq_begintypsend(&buf);
-	pq_sendint(&buf, svec->dim, sizeof(int32));
-	pq_sendint(&buf, svec->nnz, sizeof(int32));
-	pq_sendint(&buf, svec->unused, sizeof(int32));
-	for (int i = 0; i < svec->nnz; i++)
-		pq_sendint(&buf, svec->indices[i], sizeof(int32));
-	for (int i = 0; i < svec->nnz; i++)
-		pq_sendfloat4(&buf, values[i]);
-
-	PG_RETURN_BYTEA_P(pq_endtypsend(&buf));
-}
-
-/*
- * Convert sparse vector to sparse vector
- * This is needed to check the type modifier
- */
-PGDLLEXPORT PG_FUNCTION_INFO_V1(sparsevec);
-Datum
-sparsevec(PG_FUNCTION_ARGS)
-{
-	SparseVector *svec = PG_GETARG_SPARSEVEC_P(0);
-	int32		typmod = PG_GETARG_INT32(1);
-
-	CheckExpectedDim(typmod, svec->dim);
-
-	PG_RETURN_POINTER(svec);
-}
-
-/*
- * Convert dense vector to sparse vector
- */
-PGDLLEXPORT PG_FUNCTION_INFO_V1(vector_to_sparsevec);
-Datum
-vector_to_sparsevec(PG_FUNCTION_ARGS)
-{
-	Vector	   *vec = PG_GETARG_VECTOR_P(0);
-	int32		typmod = PG_GETARG_INT32(1);
-	SparseVector *result;
-	int			dim = vec->dim;
-	int			nnz = 0;
-	float	   *values;
-	int			j = 0;
-
-	CheckDim(dim);
-	CheckExpectedDim(typmod, dim);
-
-	for (int i = 0; i < dim; i++)
-	{
-		if (vec->x[i] != 0)
-			nnz++;
-	}
-
-	result = InitSparseVector(dim, nnz);
-	values = SPARSEVEC_VALUES(result);
-	for (int i = 0; i < dim; i++)
-	{
-		if (vec->x[i] != 0)
-		{
-			/* Safety check */
-			if (j == nnz)
-				elog(ERROR, "safety check failed");
-
-			result->indices[j] = i;
-			values[j] = vec->x[i];
-			j++;
-		}
-	}
-
-	PG_RETURN_POINTER(result);
-}
-
-/*
- *  Get the L2 squared distance between sparse vectors
- */
-static double
-l2_distance_squared_internal(SparseVector * a, SparseVector * b)
-{
-	float	   *ax = SPARSEVEC_VALUES(a);
-	float	   *bx = SPARSEVEC_VALUES(b);
-	double		distance = 0.0;
-	int			bpos = 0;
-
-	for (int i = 0; i < a->nnz; i++)
-	{
-		int			ai = a->indices[i];
-		int			bi = -1;
-
-		for (int j = bpos; j < b->nnz; j++)
-		{
-			bi = b->indices[j];
-
-			if (ai == bi)
-			{
-				double		diff = ax[i] - bx[j];
-
-				distance += diff * diff;
-			}
-			else if (ai > bi)
-				distance += bx[j] * bx[j];
-
-			/* Update start for next iteration */
-			if (ai >= bi)
-				bpos = j + 1;
-
-			/* Found or passed it */
-			if (bi >= ai)
-				break;
-		}
-
-		if (ai != bi)
-			distance += ax[i] * ax[i];
-	}
-
-	for (int j = bpos; j < b->nnz; j++)
-		distance += bx[j] * bx[j];
-
-	return distance;
-}
-
-/*
- * Get the L2 distance between sparse vectors
- */
-PGDLLEXPORT PG_FUNCTION_INFO_V1(sparsevec_l2_distance);
-Datum
-sparsevec_l2_distance(PG_FUNCTION_ARGS)
-{
-	SparseVector *a = PG_GETARG_SPARSEVEC_P(0);
-	SparseVector *b = PG_GETARG_SPARSEVEC_P(1);
-
-	CheckDims(a, b);
-
-	PG_RETURN_FLOAT8(sqrt(l2_distance_squared_internal(a, b)));
-}
-
-/*
- * Get the L2 squared distance between sparse vectors
- * This saves a sqrt calculation
- */
-PGDLLEXPORT PG_FUNCTION_INFO_V1(sparsevec_l2_squared_distance);
-Datum
-sparsevec_l2_squared_distance(PG_FUNCTION_ARGS)
-{
-	SparseVector *a = PG_GETARG_SPARSEVEC_P(0);
-	SparseVector *b = PG_GETARG_SPARSEVEC_P(1);
-
-	CheckDims(a, b);
-
-	PG_RETURN_FLOAT8(l2_distance_squared_internal(a, b));
-}
-
-/*
- * Get the inner product of two sparse vectors
- */
-static double
-inner_product_internal(SparseVector * a, SparseVector * b)
-{
-	float	   *ax = SPARSEVEC_VALUES(a);
-	float	   *bx = SPARSEVEC_VALUES(b);
-	double		distance = 0.0;
-	int			bpos = 0;
-
-	for (int i = 0; i < a->nnz; i++)
-	{
-		int			ai = a->indices[i];
-
-		for (int j = bpos; j < b->nnz; j++)
-		{
-			int			bi = b->indices[j];
-
-			/* Only update when the same index */
-			if (ai == bi)
-				distance += ax[i] * bx[j];
-
-			/* Update start for next iteration */
-			if (ai >= bi)
-				bpos = j + 1;
-
-			/* Found or passed it */
-			if (bi >= ai)
-				break;
-		}
-	}
-
-	return distance;
-}
-
-/*
- * Get the inner product of two sparse vectors
- */
-PGDLLEXPORT PG_FUNCTION_INFO_V1(sparsevec_inner_product);
-Datum
-sparsevec_inner_product(PG_FUNCTION_ARGS)
-{
-	SparseVector *a = PG_GETARG_SPARSEVEC_P(0);
-	SparseVector *b = PG_GETARG_SPARSEVEC_P(1);
-
-	CheckDims(a, b);
-
-	PG_RETURN_FLOAT8(inner_product_internal(a, b));
-}
-
-/*
- * Get the negative inner product of two sparse vectors
- */
-PGDLLEXPORT PG_FUNCTION_INFO_V1(sparsevec_negative_inner_product);
-Datum
-sparsevec_negative_inner_product(PG_FUNCTION_ARGS)
-{
-	SparseVector *a = PG_GETARG_SPARSEVEC_P(0);
-	SparseVector *b = PG_GETARG_SPARSEVEC_P(1);
-
-	CheckDims(a, b);
-
-	PG_RETURN_FLOAT8(-inner_product_internal(a, b));
-}
-
-/*
- * Get the cosine distance between two sparse vectors
- */
-PGDLLEXPORT PG_FUNCTION_INFO_V1(sparsevec_cosine_distance);
-Datum
-sparsevec_cosine_distance(PG_FUNCTION_ARGS)
-{
-	SparseVector *a = PG_GETARG_SPARSEVEC_P(0);
-	SparseVector *b = PG_GETARG_SPARSEVEC_P(1);
-	float	   *ax = SPARSEVEC_VALUES(a);
-	float	   *bx = SPARSEVEC_VALUES(b);
-	float		norma = 0.0;
-	float		normb = 0.0;
-	double		similarity;
-
-	CheckDims(a, b);
-
-	similarity = inner_product_internal(a, b);
-
-	/* Auto-vectorized */
-	for (int i = 0; i < a->nnz; i++)
-		norma += ax[i] * ax[i];
-
-	/* Auto-vectorized */
-	for (int i = 0; i < b->nnz; i++)
-		normb += bx[i] * bx[i];
-
-	/* Use sqrt(a * b) over sqrt(a) * sqrt(b) */
-	similarity /= sqrt((double) norma * (double) normb);
-
-#ifdef _MSC_VER
-	/* /fp:fast may not propagate NaN */
-	if (isnan(similarity))
-		PG_RETURN_FLOAT8(NAN);
-#endif
-
-	/* Keep in range */
-	if (similarity > 1)
-		similarity = 1.0;
-	else if (similarity < -1)
-		similarity = -1.0;
-
-	PG_RETURN_FLOAT8(1.0 - similarity);
-}
-
-/*
- * Get the L2 norm of a sparse vector
- */
-PGDLLEXPORT PG_FUNCTION_INFO_V1(sparsevec_norm);
-Datum
-sparsevec_norm(PG_FUNCTION_ARGS)
-{
-	SparseVector *a = PG_GETARG_SPARSEVEC_P(0);
-	float	   *ax = SPARSEVEC_VALUES(a);
-	double		norm = 0.0;
-
-	/* Auto-vectorized */
-	for (int i = 0; i < a->nnz; i++)
-		norm += (double) ax[i] * (double) ax[i];
-
-	PG_RETURN_FLOAT8(sqrt(norm));
-}

src/sparsevec.h

@@ -1,23 +0,0 @@
-#ifndef SPARSEVEC_H
-#define SPARSEVEC_H
-
-#define SPARSEVEC_MAX_DIM 100000
-
-#define SPARSEVEC_SIZE(_nnz)		(offsetof(SparseVector, indices) + MAXALIGN((_nnz) * sizeof(int32)) + (_nnz * sizeof(float)))
-#define SPARSEVEC_VALUES(x)		((float *) (((char *) (x)) + offsetof(SparseVector, indices) + MAXALIGN((x)->nnz * sizeof(int32))))
-#define DatumGetSparseVector(x)		((SparseVector *) PG_DETOAST_DATUM(x))
-#define PG_GETARG_SPARSEVEC_P(x)	DatumGetSparseVector(PG_GETARG_DATUM(x))
-#define PG_RETURN_SPARSEVEC_P(x)	PG_RETURN_POINTER(x)
-
-typedef struct SparseVector
-{
-	int32		vl_len_;		/* varlena header (do not touch directly!) */
-	int32		dim;			/* number of dimensions */
-	int32		nnz;
-	int32		unused;
-	int32		indices[FLEXIBLE_ARRAY_MEMBER];
-}			SparseVector;
-
-SparseVector *InitSparseVector(int dim, int nnz);
-
-#endif

src/vector.c

@@ -3,17 +3,14 @@
 #include <math.h>
 
 #include "catalog/pg_type.h"
-#include "common/shortest_dec.h"
 #include "fmgr.h"
 #include "hnsw.h"
 #include "ivfflat.h"
 #include "lib/stringinfo.h"
 #include "libpq/pqformat.h"
 #include "port.h"				/* for strtof() */
-#include "sparsevec.h"
 #include "utils/array.h"
 #include "utils/builtins.h"
-#include "utils/float.h"
 #include "utils/lsyscache.h"
 #include "utils/numeric.h"
 #include "vector.h"
@@ -22,6 +19,13 @@
 #include "varatt.h"
 #endif
 
+#if PG_VERSION_NUM >= 120000
+#include "common/shortest_dec.h"
+#include "utils/float.h"
+#else
+#include <float.h>
+#endif
+
 #if PG_VERSION_NUM < 130000
 #define TYPALIGN_DOUBLE 'd'
 #define TYPALIGN_INT 'i'
@@ -85,7 +89,7 @@ CheckDim(int dim)
 }
 
 /*
- * Ensure finite element
+ * Ensure finite elements
  */
 static inline void
 CheckElement(float value)
@@ -173,15 +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++;
@@ -265,7 +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);
 
@@ -289,6 +292,15 @@ vector_out(PG_FUNCTION_ARGS)
 	char	   *ptr;
 	int			n;
 
+#if PG_VERSION_NUM < 120000
+	int			ndig = FLT_DIG + extra_float_digits;
+
+	if (ndig < 1)
+		ndig = 1;
+
+#define FLOAT_SHORTEST_DECIMAL_LEN (ndig + 10)
+#endif
+
 	/*
 	 * Need:
 	 *
@@ -312,7 +324,11 @@ vector_out(PG_FUNCTION_ARGS)
 			ptr++;
 		}
 
+#if PG_VERSION_NUM >= 120000
 		n = float_to_shortest_decimal_bufn(vector->x[i], ptr);
+#else
+		n = sprintf(ptr, "%.*g", ndig, vector->x[i]);
+#endif
 		ptr += n;
 	}
 	*ptr = ']';
@@ -421,18 +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);
 }
 
 /*
@@ -449,6 +464,7 @@ array_to_vector(PG_FUNCTION_ARGS)
 	bool		typbyval;
 	char		typalign;
 	Datum	   *elemsp;
+	bool	   *nullsp;
 	int			nelemsp;
 
 	if (ARR_NDIM(array) > 1)
@@ -462,7 +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);
@@ -496,12 +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]);
@@ -685,8 +695,6 @@ vector_spherical_distance(PG_FUNCTION_ARGS)
 {
 	Vector	   *a = PG_GETARG_VECTOR_P(0);
 	Vector	   *b = PG_GETARG_VECTOR_P(1);
-	float	   *ax = a->x;
-	float	   *bx = b->x;
 	float		dp = 0.0;
 	double		distance;
 
@@ -694,7 +702,7 @@ vector_spherical_distance(PG_FUNCTION_ARGS)
 
 	/* Auto-vectorized */
 	for (int i = 0; i < a->dim; i++)
-		dp += ax[i] * bx[i];
+		dp += a->x[i] * b->x[i];
 
 	distance = (double) dp;
 
@@ -708,7 +716,7 @@ 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
@@ -867,10 +875,9 @@ vector_mul(PG_FUNCTION_ARGS)
 int
 vector_cmp_internal(Vector * a, Vector * b)
 {
-	int			dim = Min(a->dim, b->dim);
+	CheckDims(a, b);
 
-	/* Check values before dimensions to be consistent with Postgres arrays */
-	for (int i = 0; i < dim; i++)
+	for (int i = 0; i < a->dim; i++)
 	{
 		if (a->x[i] < b->x[i])
 			return -1;
@@ -878,13 +885,6 @@ vector_cmp_internal(Vector * a, Vector * b)
 		if (a->x[i] > b->x[i])
 			return 1;
 	}
-
-	if (a->dim < b->dim)
-		return -1;
-
-	if (a->dim > b->dim)
-		return 1;
-
 	return 0;
 }
 
@@ -895,11 +895,8 @@ 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);
-
-	/* TODO Remove in 0.7.0 */
-	CheckDims(a, b);
+	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);
 }
@@ -911,11 +908,8 @@ 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);
-
-	/* TODO Remove in 0.7.0 */
-	CheckDims(a, b);
+	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);
 }
@@ -927,11 +921,8 @@ 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);
-
-	/* TODO Remove in 0.7.0 */
-	CheckDims(a, b);
+	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);
 }
@@ -943,11 +934,8 @@ 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);
-
-	/* TODO Remove in 0.7.0 */
-	CheckDims(a, b);
+	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);
 }
@@ -959,11 +947,8 @@ 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);
-
-	/* TODO Remove in 0.7.0 */
-	CheckDims(a, b);
+	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);
 }
@@ -975,11 +960,8 @@ 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);
-
-	/* TODO Remove in 0.7.0 */
-	CheckDims(a, b);
+	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);
 }
@@ -991,8 +973,8 @@ 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));
 }
@@ -1161,26 +1143,3 @@ vector_avg(PG_FUNCTION_ARGS)
 
 	PG_RETURN_POINTER(result);
 }
-
-/*
- * Convert sparse vector to dense vector
- */
-PGDLLEXPORT PG_FUNCTION_INFO_V1(sparsevec_to_vector);
-Datum
-sparsevec_to_vector(PG_FUNCTION_ARGS)
-{
-	SparseVector *svec = PG_GETARG_SPARSEVEC_P(0);
-	int32		typmod = PG_GETARG_INT32(1);
-	Vector	   *result;
-	int			dim = svec->dim;
-	float	   *values = SPARSEVEC_VALUES(svec);
-
-	CheckDim(dim);
-	CheckExpectedDim(typmod, dim);
-
-	result = InitVector(dim);
-	for (int i = 0; i < svec->nnz; i++)
-		result->x[svec->indices[i]] = values[i];
-
-	PG_RETURN_POINTER(result);
-}

test/expected/functions.out

@@ -24,56 +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 = '[1,2,3]';
- ?column? 
-----------
- t
-(1 row)
-
-SELECT '[1,2,3]'::vector = '[1,2]';
-ERROR:  different vector dimensions 3 and 2
-SELECT vector_cmp('[1,2,3]', '[1,2,3]');
- vector_cmp 
-------------
-          0
-(1 row)
-
-SELECT vector_cmp('[1,2,3]', '[0,0,0]');
- vector_cmp 
-------------
-          1
-(1 row)
-
-SELECT vector_cmp('[0,0,0]', '[1,2,3]');
- vector_cmp 
-------------
-         -1
-(1 row)
-
-SELECT vector_cmp('[1,2]', '[1,2,3]');
- vector_cmp 
-------------
-         -1
-(1 row)
-
-SELECT vector_cmp('[1,2,3]', '[1,2]');
- vector_cmp 
-------------
-          1
-(1 row)
-
-SELECT vector_cmp('[1,2]', '[2,3,4]');
- vector_cmp 
-------------
-         -1
-(1 row)
-
-SELECT vector_cmp('[2,3]', '[1,2,3]');
- vector_cmp 
-------------
-          1
-(1 row)
-
 SELECT vector_dims('[1,2,3]');
  vector_dims 
 -------------
@@ -104,105 +54,99 @@ 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]');
- 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

test/expected/hnsw_l2.out

@@ -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 

test/expected/hnsw_sparsevec_cosine.out

@@ -1,26 +0,0 @@
-SET enable_seqscan = off;
-CREATE TABLE t (val sparsevec(3));
-INSERT INTO t (val) VALUES ('{}/3'), ('{0:1,1:2,2:3}/3'), ('{0:1,1:1,2:1}/3'), (NULL);
-CREATE INDEX ON t USING hnsw (val sparsevec_cosine_ops);
-INSERT INTO t (val) VALUES ('{0:1,1:2,2:4}/3');
-SELECT * FROM t ORDER BY val <=> '{0:3,1:3,2:3}/3';
-       val       
------------------
- {0:1,1:1,2:1}/3
- {0:1,1:2,2:3}/3
- {0:1,1:2,2:4}/3
-(3 rows)
-
-SELECT COUNT(*) FROM (SELECT * FROM t ORDER BY val <=> '{}/3') t2;
- count 
--------
-     3
-(1 row)
-
-SELECT COUNT(*) FROM (SELECT * FROM t ORDER BY val <=> (SELECT NULL::sparsevec)) t2;
- count 
--------
-     3
-(1 row)
-
-DROP TABLE t;

test/expected/hnsw_sparsevec_ip.out

@@ -1,21 +0,0 @@
-SET enable_seqscan = off;
-CREATE TABLE t (val sparsevec(3));
-INSERT INTO t (val) VALUES ('{}/3'), ('{0:1,1:2,2:3}/3'), ('{0:1,1:1,2:1}/3'), (NULL);
-CREATE INDEX ON t USING hnsw (val sparsevec_ip_ops);
-INSERT INTO t (val) VALUES ('{0:1,1:2,2:4}/3');
-SELECT * FROM t ORDER BY val <#> '{0:3,1:3,2:3}/3';
-       val       
------------------
- {0:1,1:2,2:4}/3
- {0:1,1:2,2:3}/3
- {0:1,1:1,2:1}/3
- {}/3
-(4 rows)
-
-SELECT COUNT(*) FROM (SELECT * FROM t ORDER BY val <#> (SELECT NULL::sparsevec)) t2;
- count 
--------
-     4
-(1 row)
-
-DROP TABLE t;

test/expected/hnsw_sparsevec_l2.out

@@ -1,43 +0,0 @@
-SET enable_seqscan = off;
-CREATE TABLE t (val sparsevec(3));
-INSERT INTO t (val) VALUES ('{}/3'), ('{0:1,1:2,2:3}/3'), ('{0:1,1:1,2:1}/3'), (NULL);
-CREATE INDEX ON t USING hnsw (val sparsevec_l2_ops);
-INSERT INTO t (val) VALUES ('{0:1,1:2,2:4}/3');
-SELECT * FROM t ORDER BY val <-> '{0:3,1:3,2:3}/3';
-       val       
------------------
- {0:1,1:2,2:3}/3
- {0:1,1:2,2:4}/3
- {0:1,1:1,2:1}/3
- {}/3
-(4 rows)
-
-SELECT COUNT(*) FROM (SELECT * FROM t ORDER BY val <-> (SELECT NULL::sparsevec)) t2;
- count 
--------
-     4
-(1 row)
-
-SELECT COUNT(*) FROM t;
- count 
--------
-     5
-(1 row)
-
-TRUNCATE t;
-SELECT * FROM t ORDER BY val <-> '{0:3,1:3,2:3}/3';
- val 
------
-(0 rows)
-
-DROP TABLE t;
--- TODO move
-CREATE TABLE t (val sparsevec(1001));
-INSERT INTO t (val) VALUES (array_fill(1, ARRAY[1001])::vector::sparsevec);
-CREATE INDEX ON t USING hnsw (val sparsevec_l2_ops);
-ERROR:  sparsevec cannot have more than 1000 non-zero elements for hnsw index
-TRUNCATE t;
-CREATE INDEX ON t USING hnsw (val sparsevec_l2_ops);
-INSERT INTO t (val) VALUES (array_fill(1, ARRAY[1001])::vector::sparsevec);
-ERROR:  sparsevec cannot have more than 1000 non-zero elements for hnsw index
-DROP TABLE t;

test/expected/input.out

@@ -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  
 ---------

test/expected/sparsevec_functions.out

@@ -1,62 +0,0 @@
-SELECT l2_distance('{}/2'::sparsevec, '{0:3,1:4}/2');
- l2_distance 
--------------
-           5
-(1 row)
-
-SELECT l2_distance('{}/2'::sparsevec, '{1:1}/2');
- l2_distance 
--------------
-           1
-(1 row)
-
-SELECT '{}/2'::sparsevec <-> '{0:3,1:4}/2';
- ?column? 
-----------
-        5
-(1 row)
-
-SELECT inner_product('{0:1,1:2}/2'::sparsevec, '{0:2,1:4}/2');
- inner_product 
----------------
-            10
-(1 row)
-
-SELECT sparsevec_negative_inner_product('{0:1,1:2}/2', '{0:2,1:4}/2');
- sparsevec_negative_inner_product 
-----------------------------------
-                              -10
-(1 row)
-
-SELECT cosine_distance('{0:1,1:2}/2'::sparsevec, '{0:2,1:4}/2');
- cosine_distance 
------------------
-               0
-(1 row)
-
-SELECT cosine_distance('{0:1,1:2}/2'::sparsevec, '{}/2');
- cosine_distance 
------------------
-             NaN
-(1 row)
-
-SELECT cosine_distance('{0:1,1:1}/2'::sparsevec, '{0:-1,1:-1}/2');
- cosine_distance 
------------------
-               2
-(1 row)
-
-SELECT cosine_distance('{0:1}/2'::sparsevec, '{1:2}/2');
- cosine_distance 
------------------
-               1
-(1 row)
-
-SELECT cosine_distance('{}/1'::sparsevec, '{}/1');
- cosine_distance 
------------------
-             NaN
-(1 row)
-
-SELECT cosine_distance('{0:1}/2'::sparsevec, '{0:1}/3');
-ERROR:  different sparsevec dimensions 2 and 3

test/expected/sparsevec_input.out

@@ -1,62 +0,0 @@
-SELECT '{0:1.5,2:3.5}/5'::sparsevec;
-    sparsevec    
------------------
- {0:1.5,2:3.5}/5
-(1 row)
-
-SELECT '{0:1.5,2:3.5}/5'::sparsevec::vector;
-     vector      
------------------
- [1.5,0,3.5,0,0]
-(1 row)
-
-SELECT '{0:1.5,2:3.5}/5'::sparsevec::vector(5);
-     vector      
------------------
- [1.5,0,3.5,0,0]
-(1 row)
-
-SELECT '{0:1.5,2:3.5}/5'::sparsevec::vector(4);
-ERROR:  expected 4 dimensions, not 5
-SELECT '[0,1.5,0,3.5,0]'::vector::sparsevec;
-    sparsevec    
------------------
- {1:1.5,3:3.5}/5
-(1 row)
-
-SELECT '{0:0,1:1,2:0}/3'::sparsevec;
- sparsevec 
------------
- {1:1}/3
-(1 row)
-
-SELECT '{1:1,0:1}/2'::sparsevec;
-ERROR:  indexes must be in ascending order
-LINE 1: SELECT '{1:1,0:1}/2'::sparsevec;
-               ^
-SELECT '{}/5'::sparsevec;
- sparsevec 
------------
- {}/5
-(1 row)
-
-SELECT '{}/-1'::sparsevec;
-ERROR:  sparsevec must have at least 1 dimension
-LINE 1: SELECT '{}/-1'::sparsevec;
-               ^
-SELECT '{}/100001'::sparsevec;
-ERROR:  sparsevec cannot have more than 100000 dimensions
-LINE 1: SELECT '{}/100001'::sparsevec;
-               ^
-SELECT '{}/16001'::sparsevec::vector;
-ERROR:  vector cannot have more than 16000 dimensions
-SELECT '{-1:1}/1'::sparsevec;
-ERROR:  index "-1" is out of range for type sparsevec
-LINE 1: SELECT '{-1:1}/1'::sparsevec;
-               ^
-SELECT '{1:1}/1'::sparsevec;
-ERROR:  index must be less than dimensions
-LINE 1: SELECT '{1:1}/1'::sparsevec;
-               ^
-SELECT '{}/1'::sparsevec(2);
-ERROR:  expected 2 dimensions, not 1

test/sql/functions.sql

@@ -6,17 +6,6 @@ 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);
@@ -24,29 +13,28 @@ 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 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]'::vector, '[3,4]');
-SELECT inner_product('[1,2]'::vector, '[3]');
-SELECT inner_product('[3e38]'::vector, '[3e38]');
+SELECT inner_product('[1,2]', '[3,4]');
+SELECT inner_product('[1,2]', '[3]');
+SELECT inner_product('[3e38]', '[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 cosine_distance('[1,2]', '[2,4]');
+SELECT cosine_distance('[1,2]', '[0,0]');
+SELECT cosine_distance('[1,1]', '[1,1]');
+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]'::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 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;

test/sql/hnsw_l2.sql

@@ -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;

test/sql/hnsw_sparsevec_cosine.sql

@@ -1,13 +0,0 @@
-SET enable_seqscan = off;
-
-CREATE TABLE t (val sparsevec(3));
-INSERT INTO t (val) VALUES ('{}/3'), ('{0:1,1:2,2:3}/3'), ('{0:1,1:1,2:1}/3'), (NULL);
-CREATE INDEX ON t USING hnsw (val sparsevec_cosine_ops);
-
-INSERT INTO t (val) VALUES ('{0:1,1:2,2:4}/3');
-
-SELECT * FROM t ORDER BY val <=> '{0:3,1:3,2:3}/3';
-SELECT COUNT(*) FROM (SELECT * FROM t ORDER BY val <=> '{}/3') t2;
-SELECT COUNT(*) FROM (SELECT * FROM t ORDER BY val <=> (SELECT NULL::sparsevec)) t2;
-
-DROP TABLE t;

test/sql/hnsw_sparsevec_ip.sql

@@ -1,12 +0,0 @@
-SET enable_seqscan = off;
-
-CREATE TABLE t (val sparsevec(3));
-INSERT INTO t (val) VALUES ('{}/3'), ('{0:1,1:2,2:3}/3'), ('{0:1,1:1,2:1}/3'), (NULL);
-CREATE INDEX ON t USING hnsw (val sparsevec_ip_ops);
-
-INSERT INTO t (val) VALUES ('{0:1,1:2,2:4}/3');
-
-SELECT * FROM t ORDER BY val <#> '{0:3,1:3,2:3}/3';
-SELECT COUNT(*) FROM (SELECT * FROM t ORDER BY val <#> (SELECT NULL::sparsevec)) t2;
-
-DROP TABLE t;

test/sql/hnsw_sparsevec_l2.sql

@@ -1,25 +0,0 @@
-SET enable_seqscan = off;
-
-CREATE TABLE t (val sparsevec(3));
-INSERT INTO t (val) VALUES ('{}/3'), ('{0:1,1:2,2:3}/3'), ('{0:1,1:1,2:1}/3'), (NULL);
-CREATE INDEX ON t USING hnsw (val sparsevec_l2_ops);
-
-INSERT INTO t (val) VALUES ('{0:1,1:2,2:4}/3');
-
-SELECT * FROM t ORDER BY val <-> '{0:3,1:3,2:3}/3';
-SELECT COUNT(*) FROM (SELECT * FROM t ORDER BY val <-> (SELECT NULL::sparsevec)) t2;
-SELECT COUNT(*) FROM t;
-
-TRUNCATE t;
-SELECT * FROM t ORDER BY val <-> '{0:3,1:3,2:3}/3';
-
-DROP TABLE t;
-
--- TODO move
-CREATE TABLE t (val sparsevec(1001));
-INSERT INTO t (val) VALUES (array_fill(1, ARRAY[1001])::vector::sparsevec);
-CREATE INDEX ON t USING hnsw (val sparsevec_l2_ops);
-TRUNCATE t;
-CREATE INDEX ON t USING hnsw (val sparsevec_l2_ops);
-INSERT INTO t (val) VALUES (array_fill(1, ARRAY[1001])::vector::sparsevec);
-DROP TABLE t;

test/sql/input.sql

@@ -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)[];

test/sql/sparsevec_functions.sql

@@ -1,13 +0,0 @@
-SELECT l2_distance('{}/2'::sparsevec, '{0:3,1:4}/2');
-SELECT l2_distance('{}/2'::sparsevec, '{1:1}/2');
-SELECT '{}/2'::sparsevec <-> '{0:3,1:4}/2';
-
-SELECT inner_product('{0:1,1:2}/2'::sparsevec, '{0:2,1:4}/2');
-SELECT sparsevec_negative_inner_product('{0:1,1:2}/2', '{0:2,1:4}/2');
-
-SELECT cosine_distance('{0:1,1:2}/2'::sparsevec, '{0:2,1:4}/2');
-SELECT cosine_distance('{0:1,1:2}/2'::sparsevec, '{}/2');
-SELECT cosine_distance('{0:1,1:1}/2'::sparsevec, '{0:-1,1:-1}/2');
-SELECT cosine_distance('{0:1}/2'::sparsevec, '{1:2}/2');
-SELECT cosine_distance('{}/1'::sparsevec, '{}/1');
-SELECT cosine_distance('{0:1}/2'::sparsevec, '{0:1}/3');

test/sql/sparsevec_input.sql

@@ -1,19 +0,0 @@
-SELECT '{0:1.5,2:3.5}/5'::sparsevec;
-SELECT '{0:1.5,2:3.5}/5'::sparsevec::vector;
-SELECT '{0:1.5,2:3.5}/5'::sparsevec::vector(5);
-SELECT '{0:1.5,2:3.5}/5'::sparsevec::vector(4);
-SELECT '[0,1.5,0,3.5,0]'::vector::sparsevec;
-
-SELECT '{0:0,1:1,2:0}/3'::sparsevec;
-
-SELECT '{1:1,0:1}/2'::sparsevec;
-
-SELECT '{}/5'::sparsevec;
-SELECT '{}/-1'::sparsevec;
-SELECT '{}/100001'::sparsevec;
-SELECT '{}/16001'::sparsevec::vector;
-
-SELECT '{-1:1}/1'::sparsevec;
-SELECT '{1:1}/1'::sparsevec;
-
-SELECT '{}/1'::sparsevec(2);

test/t/001_ivfflat_wal.pl

@@ -19,6 +19,8 @@ sub test_index_replay
 
 	# Wait for replica to catch up
 	my $applname = $node_replica->name;
+
+	my $server_version_num = $node_primary->safe_psql("postgres", "SHOW server_version_num");
 	my $caughtup_query = "SELECT pg_current_wal_lsn() <= replay_lsn FROM pg_stat_replication WHERE application_name = '$applname';";
 	$node_primary->poll_query_until('postgres', $caughtup_query)
 	  or die "Timed out while waiting for replica 1 to catch up";

test/t/003_ivfflat_recall.pl

@@ -94,7 +94,7 @@ for my $i (0 .. $#operators)
 	# Test approximate results
 	if ($operator ne "<#>")
 	{
-		# TODO Fix test (uniform random vectors all have similar inner product)
+		# TODO fix test
 		test_recall(1, 0.71, $operator);
 		test_recall(10, 0.95, $operator);
 	}
@@ -115,7 +115,7 @@ for my $i (0 .. $#operators)
 	# Test approximate results
 	if ($operator ne "<#>")
 	{
-		# TODO Fix test (uniform random vectors all have similar inner product)
+		# TODO fix test
 		test_recall(1, 0.71, $operator);
 		test_recall(10, 0.95, $operator);
 	}

test/t/010_hnsw_wal.pl

@@ -19,6 +19,8 @@ sub test_index_replay
 
 	# Wait for replica to catch up
 	my $applname = $node_replica->name;
+
+	my $server_version_num = $node_primary->safe_psql("postgres", "SHOW server_version_num");
 	my $caughtup_query = "SELECT pg_current_wal_lsn() <= replay_lsn FROM pg_stat_replication WHERE application_name = '$applname';";
 	$node_primary->poll_query_until('postgres', $caughtup_query)
 	  or die "Timed out while waiting for replica 1 to catch up";

test/t/012_hnsw_build_recall.pl

@@ -95,7 +95,7 @@ for my $i (0 .. $#operators)
 
 	$node->safe_psql("postgres", "DROP INDEX idx;");
 
-	# Build index in parallel in memory
+	# Build index in parallel
 	my ($ret, $stdout, $stderr) = $node->psql("postgres", qq(
 		SET client_min_messages = DEBUG;
 		SET min_parallel_table_scan_size = 1;
@@ -108,21 +108,6 @@ for my $i (0 .. $#operators)
 	test_recall($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();

test/t/016_hnsw_vacuum_recall.pl

@@ -86,10 +86,10 @@ 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
+# TODO test concurrent inserts with vacuum
 $node->safe_psql("postgres", "VACUUM tst;");
 
 test_recall(0.95, $limit, "after vacuum");

test/t/017_ivfflat_insert_recall.pl

@@ -1,117 +0,0 @@
-use strict;
-use warnings;
-use PostgresNode;
-use TestLib;
-use Test::More;
-
-my $node;
-my @queries = ();
-my @expected;
-my $limit = 20;
-
-sub test_recall
-{
-	my ($probes, $min, $operator) = @_;
-	my $correct = 0;
-	my $total = 0;
-
-	my $explain = $node->safe_psql("postgres", qq(
-		SET enable_seqscan = off;
-		SET ivfflat.probes = $probes;
-		EXPLAIN ANALYZE SELECT i FROM tst ORDER BY v $operator '$queries[0]' LIMIT $limit;
-	));
-	like($explain, qr/Index Scan using idx on tst/);
-
-	for my $i (0 .. $#queries)
-	{
-		my $actual = $node->safe_psql("postgres", qq(
-			SET enable_seqscan = off;
-			SET ivfflat.probes = $probes;
-			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 serial, v vector(3));");
-
-# 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 = ("vector_l2_ops", "vector_ip_ops", "vector_cosine_ops");
-
-for my $i (0 .. $#operators)
-{
-	my $operator = $operators[$i];
-	my $opclass = $opclasses[$i];
-
-	# Add index
-	$node->safe_psql("postgres", "CREATE INDEX idx ON tst USING ivfflat (v $opclass);");
-
-	# Use concurrent inserts
-	$node->pgbench(
-		"--no-vacuum --client=10 --transactions=1000",
-		0,
-		[qr{actually processed}],
-		[qr{^$}],
-		"concurrent INSERTs",
-		{
-			"017_ivfflat_insert_recall_$opclass" => "INSERT INTO tst (v) SELECT ARRAY[random(), random(), random()] FROM generate_series(1, 10) i;"
-		}
-	);
-
-	# Get exact results
-	@expected = ();
-	foreach (@queries)
-	{
-		my $res = $node->safe_psql("postgres", qq(
-			SET enable_indexscan = off;
-			SELECT i FROM tst ORDER BY v $operator '$_' LIMIT $limit;
-		));
-		push(@expected, $res);
-	}
-
-	# Test approximate results
-	if ($operator ne "<#>")
-	{
-		# TODO Fix test (uniform random vectors all have similar inner product)
-		test_recall(1, 0.71, $operator);
-		test_recall(10, 0.95, $operator);
-	}
-	# Account for equal distances
-	test_recall(100, 0.9925, $operator);
-
-	$node->safe_psql("postgres", "DROP INDEX idx;");
-	$node->safe_psql("postgres", "TRUNCATE tst;");
-}
-
-done_testing();

test/t/018_hnsw_filtering.pl

@@ -1,114 +0,0 @@
-use strict;
-use warnings;
-use PostgresNode;
-use TestLib;
-use Test::More;
-
-my $dim = 3;
-my $nc = 50;
-my $limit = 20;
-
-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;");
-$node->safe_psql("postgres", "CREATE TABLE tst (i int4, v vector($dim), c int4, t text);");
-$node->safe_psql("postgres",
-	"INSERT INTO tst SELECT i, ARRAY[$array_sql], i % $nc, 'test ' || i FROM generate_series(1, 10000) i;"
-);
-$node->safe_psql("postgres", "CREATE INDEX idx ON tst USING hnsw (v vector_l2_ops);");
-$node->safe_psql("postgres", "ANALYZE tst;");
-
-# Generate query
-my @r = ();
-for (1 .. $dim)
-{
-	push(@r, rand());
-}
-my $query = "[" . join(",", @r) . "]";
-my $c = int(rand() * $nc);
-
-# Test attribute filtering
-my $explain = $node->safe_psql("postgres", qq(
-	EXPLAIN ANALYZE SELECT i FROM tst WHERE c = $c ORDER BY v <-> '$query' LIMIT $limit;
-));
-# TODO Do not use index
-like($explain, qr/Index Scan using idx/);
-
-# Test attribute filtering with few rows removed
-$explain = $node->safe_psql("postgres", qq(
-	EXPLAIN ANALYZE SELECT i FROM tst WHERE c != $c ORDER BY v <-> '$query' LIMIT $limit;
-));
-like($explain, qr/Index Scan using idx/);
-
-# Test attribute filtering with few rows removed comparison
-$explain = $node->safe_psql("postgres", qq(
-	EXPLAIN ANALYZE SELECT i FROM tst WHERE c >= 1 ORDER BY v <-> '$query' LIMIT $limit;
-));
-like($explain, qr/Index Scan using idx/);
-
-# Test attribute filtering with many rows removed comparison
-$explain = $node->safe_psql("postgres", qq(
-	EXPLAIN ANALYZE SELECT i FROM tst WHERE c < 1 ORDER BY v <-> '$query' LIMIT $limit;
-));
-# TODO Do not use index
-like($explain, qr/Index Scan using idx/);
-
-# Test attribute filtering with few rows removed like
-$explain = $node->safe_psql("postgres", qq(
-	EXPLAIN ANALYZE SELECT i FROM tst WHERE t LIKE '%%test%%' ORDER BY v <-> '$query' LIMIT $limit;
-));
-like($explain, qr/Index Scan using idx/);
-
-# Test attribute filtering with many rows removed like
-$explain = $node->safe_psql("postgres", qq(
-	EXPLAIN ANALYZE SELECT i FROM tst WHERE t LIKE '%%other%%' ORDER BY v <-> '$query' LIMIT $limit;
-));
-like($explain, qr/Seq Scan/);
-
-# Test distance filtering
-$explain = $node->safe_psql("postgres", qq(
-	EXPLAIN ANALYZE SELECT i FROM tst WHERE v <-> '$query' < 1 ORDER BY v <-> '$query' LIMIT $limit;
-));
-like($explain, qr/Index Scan using idx/);
-
-# Test distance filtering greater than distance
-$explain = $node->safe_psql("postgres", qq(
-	EXPLAIN ANALYZE SELECT i FROM tst WHERE v <-> '$query' > 1 ORDER BY v <-> '$query' LIMIT $limit;
-));
-# TODO Do not use index
-like($explain, qr/Index Scan using idx/);
-
-# Test distance filtering without order
-$explain = $node->safe_psql("postgres", qq(
-	EXPLAIN ANALYZE SELECT i FROM tst WHERE v <-> '$query' < 1;
-));
-like($explain, qr/Seq Scan/);
-
-# Test distance filtering without limit
-$explain = $node->safe_psql("postgres", qq(
-	EXPLAIN ANALYZE SELECT i FROM tst WHERE v <-> '$query' < 1 ORDER BY v <-> '$query';
-));
-like($explain, qr/Seq Scan/);
-
-# Test attribute index
-$node->safe_psql("postgres", "CREATE INDEX attribute_idx ON tst (c);");
-$explain = $node->safe_psql("postgres", qq(
-	EXPLAIN ANALYZE SELECT i FROM tst WHERE c = $c ORDER BY v <-> '$query' LIMIT $limit;
-));
-# TODO Use attribute index
-like($explain, qr/Index Scan using idx/);
-
-# Test partial index
-$node->safe_psql("postgres", "CREATE INDEX partial_idx ON tst USING hnsw (v vector_l2_ops) WHERE (c = $c);");
-$explain = $node->safe_psql("postgres", qq(
-	EXPLAIN ANALYZE SELECT i FROM tst WHERE c = $c ORDER BY v <-> '$query' LIMIT $limit;
-));
-like($explain, qr/Index Scan using partial_idx/);
-
-done_testing();

test/t/019_ivfflat_filtering.pl

@@ -1,116 +0,0 @@
-use strict;
-use warnings;
-use PostgresNode;
-use TestLib;
-use Test::More;
-
-my $dim = 3;
-my $nc = 50;
-my $limit = 20;
-
-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;");
-$node->safe_psql("postgres", "CREATE TABLE tst (i int4, v vector($dim), c int4, t text);");
-$node->safe_psql("postgres",
-	"INSERT INTO tst SELECT i, ARRAY[$array_sql], i % $nc, 'test ' || i FROM generate_series(1, 10000) i;"
-);
-$node->safe_psql("postgres", "CREATE INDEX idx ON tst USING ivfflat (v vector_l2_ops) WITH (lists = 100);");
-$node->safe_psql("postgres", "ANALYZE tst;");
-
-# Generate query
-my @r = ();
-for (1 .. $dim)
-{
-	push(@r, rand());
-}
-my $query = "[" . join(",", @r) . "]";
-my $c = int(rand() * $nc);
-
-# Test attribute filtering
-my $explain = $node->safe_psql("postgres", qq(
-	EXPLAIN ANALYZE SELECT i FROM tst WHERE c = $c ORDER BY v <-> '$query' LIMIT $limit;
-));
-# TODO Do not use index
-like($explain, qr/Index Scan using idx/);
-
-# Test attribute filtering with few rows removed
-$explain = $node->safe_psql("postgres", qq(
-	EXPLAIN ANALYZE SELECT i FROM tst WHERE c != $c ORDER BY v <-> '$query' LIMIT $limit;
-));
-like($explain, qr/Index Scan using idx/);
-
-# Test attribute filtering with few rows removed comparison
-$explain = $node->safe_psql("postgres", qq(
-	EXPLAIN ANALYZE SELECT i FROM tst WHERE c >= 1 ORDER BY v <-> '$query' LIMIT $limit;
-));
-like($explain, qr/Index Scan using idx/);
-
-# Test attribute filtering with many rows removed comparison
-$explain = $node->safe_psql("postgres", qq(
-	EXPLAIN ANALYZE SELECT i FROM tst WHERE c < 1 ORDER BY v <-> '$query' LIMIT $limit;
-));
-# TODO Do not use index
-like($explain, qr/Index Scan using idx/);
-
-# Test attribute filtering with few rows removed like
-$explain = $node->safe_psql("postgres", qq(
-	EXPLAIN ANALYZE SELECT i FROM tst WHERE t LIKE '%%test%%' ORDER BY v <-> '$query' LIMIT $limit;
-));
-like($explain, qr/Index Scan using idx/);
-
-# Test attribute filtering with many rows removed like
-$explain = $node->safe_psql("postgres", qq(
-	EXPLAIN ANALYZE SELECT i FROM tst WHERE t LIKE '%%other%%' ORDER BY v <-> '$query' LIMIT $limit;
-));
-like($explain, qr/Seq Scan/);
-
-# Test distance filtering
-$explain = $node->safe_psql("postgres", qq(
-	EXPLAIN ANALYZE SELECT i FROM tst WHERE v <-> '$query' < 1 ORDER BY v <-> '$query' LIMIT $limit;
-));
-like($explain, qr/Index Scan using idx/);
-
-# Test distance filtering greater than distance
-$explain = $node->safe_psql("postgres", qq(
-	EXPLAIN ANALYZE SELECT i FROM tst WHERE v <-> '$query' > 1 ORDER BY v <-> '$query' LIMIT $limit;
-));
-# TODO Do not use index
-like($explain, qr/Index Scan using idx/);
-
-# Test distance filtering without order
-$explain = $node->safe_psql("postgres", qq(
-	EXPLAIN ANALYZE SELECT i FROM tst WHERE v <-> '$query' < 1;
-));
-like($explain, qr/Seq Scan/);
-
-# Test distance filtering without limit
-$explain = $node->safe_psql("postgres", qq(
-	EXPLAIN ANALYZE SELECT i FROM tst WHERE v <-> '$query' < 1 ORDER BY v <-> '$query';
-));
-# TODO Do not use index
-like($explain, qr/Index Scan using idx/);
-
-# Test attribute index
-$node->safe_psql("postgres", "CREATE INDEX attribute_idx ON tst (c);");
-$explain = $node->safe_psql("postgres", qq(
-	EXPLAIN ANALYZE SELECT i FROM tst WHERE c = $c ORDER BY v <-> '$query' LIMIT $limit;
-));
-# TODO Use attribute index
-like($explain, qr/Index Scan using idx/);
-
-# Test partial index
-$node->safe_psql("postgres", "CREATE INDEX partial_idx ON tst USING ivfflat (v vector_l2_ops) WITH (lists = 5) WHERE (c = $c);");
-$explain = $node->safe_psql("postgres", qq(
-	EXPLAIN ANALYZE SELECT i FROM tst WHERE c = $c ORDER BY v <-> '$query' LIMIT $limit;
-));
-# TODO Use partial index
-like($explain, qr/Index Scan using idx/);
-
-done_testing();

vector.control

@@ -1,4 +1,4 @@
-comment = 'vector data type and ivfflat and hnsw access methods'
-default_version = '0.6.2'
+comment = 'vector data type and ivfflat access method'
+default_version = '0.5.0'
 module_pathname = '$libdir/vector'
 relocatable = true