Improved code [skip ci]

This reverts commit 32e95a8598.

.github/workflows/build.yml

@@ -8,10 +8,18 @@ jobs:
       fail-fast: false
       matrix:
         include:
+          - postgres: 18
+            os: ubuntu-24.04
+          - postgres: 17
+            os: ubuntu-24.04
           - postgres: 16
             os: ubuntu-24.04-arm
+          - postgres: 15
+            os: ubuntu-22.04
           - postgres: 14
             os: ubuntu-22.04-arm
+          - postgres: 13
+            os: ubuntu-22.04
     steps:
       - uses: actions/checkout@v4
       - uses: ankane/setup-postgres@v1
@@ -20,7 +28,7 @@ jobs:
           dev-files: true
       - run: make
         env:
-          PG_CFLAGS: -DUSE_ASSERT_CHECKING -Wall -Wextra -Werror -Wno-unused-parameter -Wno-sign-compare
+          PG_CFLAGS: ${{ matrix.postgres == 18 && '-Wno-missing-field-initializers' || '' }} -DUSE_ASSERT_CHECKING -Wall -Wextra -Werror -Wno-unused-parameter -Wno-sign-compare
       - run: |
           export PG_CONFIG=`which pg_config`
           sudo --preserve-env=PG_CONFIG make install
@@ -31,3 +39,95 @@ jobs:
           sudo apt-get update
           sudo apt-get install libipc-run-perl
       - run: make prove_installcheck
+  mac:
+    runs-on: ${{ matrix.os }}
+    if: ${{ !startsWith(github.ref_name, 'windows') }}
+    strategy:
+      fail-fast: false
+      matrix:
+        include:
+          - postgres: 17
+            os: macos-15
+          - postgres: 14
+            os: macos-13
+    steps:
+      - uses: actions/checkout@v4
+      - uses: ankane/setup-postgres@v1
+        with:
+          postgres-version: ${{ matrix.postgres }}
+      - run: make
+        env:
+          PG_CFLAGS: -DUSE_ASSERT_CHECKING -Wall -Wextra -Werror -Wno-unused-parameter
+      - run: make install
+      - run: make installcheck
+      - if: ${{ failure() }}
+        run: cat regression.diffs
+      # Homebrew Postgres does not enable TAP tests, so need to download
+      - run: |
+          brew install cpanm
+          cpanm --notest IPC::Run
+          wget -q https://github.com/postgres/postgres/archive/refs/tags/$TAG.tar.gz
+          tar xf $TAG.tar.gz
+          mv postgres-$TAG postgres
+        env:
+          TAG: ${{ matrix.postgres == 17 && 'REL_17_2' || 'REL_14_15' }}
+      - run: make prove_installcheck PROVE_FLAGS="-I ./postgres/src/test/perl -I ./test/perl"
+        env:
+          PERL5LIB: /Users/runner/perl5/lib/perl5
+      - run: make clean && $(brew --prefix llvm@$LLVM_VERSION)/bin/scan-build --status-bugs make
+        env:
+          LLVM_VERSION: ${{ matrix.os == 'macos-15' && 18 || 15 }}
+          PG_CFLAGS: -DUSE_ASSERT_CHECKING
+  windows:
+    runs-on: windows-latest
+    if: ${{ !startsWith(github.ref_name, 'mac') }}
+    steps:
+      - uses: actions/checkout@v4
+      - uses: ankane/setup-postgres@v1
+        with:
+          postgres-version: 14
+      - run: |
+          call "C:\Program Files\Microsoft Visual Studio\2022\Enterprise\VC\Auxiliary\Build\vcvars64.bat" && ^
+          nmake /NOLOGO /F Makefile.win && ^
+          nmake /NOLOGO /F Makefile.win install && ^
+          nmake /NOLOGO /F Makefile.win installcheck && ^
+          nmake /NOLOGO /F Makefile.win clean && ^
+          nmake /NOLOGO /F Makefile.win uninstall
+        shell: cmd
+      - if: ${{ failure() }}
+        run: cat regression.diffs
+  i386:
+    if: ${{ !startsWith(github.ref_name, 'mac') && !startsWith(github.ref_name, 'windows') }}
+    runs-on: ubuntu-latest
+    container:
+      image: debian:12
+      options: --platform linux/386
+    steps:
+      - run: apt-get update && apt-get install -y build-essential git libipc-run-perl postgresql-15 postgresql-server-dev-15 sudo
+      - run: service postgresql start
+      - run: |
+          git clone https://github.com/${{ github.repository }}.git pgvector
+          cd pgvector
+          git fetch origin ${{ github.ref }}
+          git reset --hard FETCH_HEAD
+          make
+          make install
+          chown -R postgres .
+          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
+      - if: ${{ failure() }}
+        run: cat pgvector/regression.diffs
+  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: 17
+          check-ub: yes
+      - run: make OPTFLAGS=""
+      - run: sudo --preserve-env=PG_CONFIG make install
+      - run: make installcheck

README.md

@@ -52,8 +52,6 @@ nmake /F Makefile.win
 nmake /F Makefile.win install
 ```
 
-Note: Postgres 17 is not supported with MSVC yet due to an [upstream issue](https://www.postgresql.org/message-id/flat/CAOdR5yF0krWrxycA04rgUKCgKugRvGWzzGLAhDZ9bzNv8g0Lag%40mail.gmail.com)
-
 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).
@@ -84,7 +82,7 @@ Get the nearest neighbors by L2 distance
 SELECT * FROM items ORDER BY embedding <-> '[3,1,2]' LIMIT 5;
 ```
 
-Also supports inner product (`<#>`), cosine distance (`<=>`), and L1 distance (`<+>`, added in 0.7.0)
+Also supports inner product (`<#>`), cosine distance (`<=>`), and L1 distance (`<+>`)
 
 Note: `<#>` returns the negative inner product since Postgres only supports `ASC` order index scans on operators
 
@@ -148,9 +146,9 @@ Supported distance functions are:
 - `<->` - L2 distance
 - `<#>` - (negative) inner product
 - `<=>` - cosine distance
-- `<+>` - L1 distance (added in 0.7.0)
-- `<~>` - Hamming distance (binary vectors, added in 0.7.0)
-- `<%>` - Jaccard distance (binary vectors, added in 0.7.0)
+- `<+>` - L1 distance
+- `<~>` - Hamming distance (binary vectors)
+- `<%>` - Jaccard distance (binary vectors)
 
 Get the nearest neighbors to a row
 
@@ -237,19 +235,19 @@ Cosine distance
 CREATE INDEX ON items USING hnsw (embedding vector_cosine_ops);
 ```
 
-L1 distance - added in 0.7.0
+L1 distance
 
 ```sql
 CREATE INDEX ON items USING hnsw (embedding vector_l1_ops);
 ```
 
-Hamming distance - added in 0.7.0
+Hamming distance
 
 ```sql
 CREATE INDEX ON items USING hnsw (embedding bit_hamming_ops);
 ```
 
-Jaccard distance - added in 0.7.0
+Jaccard distance
 
 ```sql
 CREATE INDEX ON items USING hnsw (embedding bit_jaccard_ops);
@@ -258,9 +256,9 @@ CREATE INDEX ON items USING hnsw (embedding bit_jaccard_ops);
 Supported types are:
 
 - `vector` - up to 2,000 dimensions
-- `halfvec` - up to 4,000 dimensions (added in 0.7.0)
-- `bit` - up to 64,000 dimensions (added in 0.7.0)
-- `sparsevec` - up to 1,000 non-zero elements (added in 0.7.0)
+- `halfvec` - up to 4,000 dimensions
+- `bit` - up to 64,000 dimensions
+- `sparsevec` - up to 1,000 non-zero elements
 
 ### Index Options
 
@@ -314,13 +312,15 @@ Note: Do not set `maintenance_work_mem` so high that it exhausts the memory on t
 
 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)
+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)
+For a large number of workers, you may need to increase `max_parallel_workers` (8 by default)
+
+The [index options](#index-options) also have a significant impact on build time (use the defaults unless seeing low recall)
 
 ### Indexing Progress
 
@@ -367,7 +367,7 @@ Cosine distance
 CREATE INDEX ON items USING ivfflat (embedding vector_cosine_ops) WITH (lists = 100);
 ```
 
-Hamming distance - added in 0.7.0
+Hamming distance
 
 ```sql
 CREATE INDEX ON items USING ivfflat (embedding bit_hamming_ops) WITH (lists = 100);
@@ -376,8 +376,8 @@ CREATE INDEX ON items USING ivfflat (embedding bit_hamming_ops) WITH (lists = 10
 Supported types are:
 
 - `vector` - up to 2,000 dimensions
-- `halfvec` - up to 4,000 dimensions (added in 0.7.0)
-- `bit` - up to 64,000 dimensions (added in 0.7.0)
+- `halfvec` - up to 4,000 dimensions
+- `bit` - up to 64,000 dimensions
 
 ### Query Options
 
@@ -477,8 +477,6 @@ CREATE TABLE items (embedding vector(3), category_id int) PARTITION BY LIST(cate
 
 ## Iterative Index Scans
 
-*Added in 0.8.0*
-
 With approximate indexes, queries with filtering can return less results since filtering is applied *after* the index is scanned. Starting with 0.8.0, you can enable iterative index scans, which will automatically scan more of the index until enough results are found (or it reaches `hnsw.max_scan_tuples` or `ivfflat.max_probes`).
 
 Iterative scans can use strict or relaxed ordering.
@@ -549,8 +547,6 @@ Note: If this is lower than `ivfflat.probes`, `ivfflat.probes` will be used
 
 ## Half-Precision Vectors
 
-*Added in 0.7.0*
-
 Use the `halfvec` type to store half-precision vectors
 
 ```sql
@@ -559,8 +555,6 @@ CREATE TABLE items (id bigserial PRIMARY KEY, embedding halfvec(3));
 
 ## Half-Precision Indexing
 
-*Added in 0.7.0*
-
 Index vectors at half precision for smaller indexes
 
 ```sql
@@ -582,24 +576,16 @@ CREATE TABLE items (id bigserial PRIMARY KEY, embedding bit(3));
 INSERT INTO items (embedding) VALUES ('000'), ('111');
 ```
 
-Get the nearest neighbors by Hamming distance (added in 0.7.0)
+Get the nearest neighbors by Hamming distance
 
 ```sql
 SELECT * FROM items ORDER BY embedding <~> '101' LIMIT 5;
 ```
 
-Or (before 0.7.0)
-
-```sql
-SELECT * FROM items ORDER BY bit_count(embedding # '101') LIMIT 5;
-```
-
 Also supports Jaccard distance (`<%>`)
 
 ## Binary Quantization
 
-*Added in 0.7.0*
-
 Use expression indexing for binary quantization
 
 ```sql
@@ -622,8 +608,6 @@ SELECT * FROM (
 
 ## Sparse Vectors
 
-*Added in 0.7.0*
-
 Use the `sparsevec` type to store sparse vectors
 
 ```sql
@@ -657,8 +641,6 @@ You can use [Reciprocal Rank Fusion](https://github.com/pgvector/pgvector-python
 
 ## Indexing Subvectors
 
-*Added in 0.7.0*
-
 Use expression indexing to index subvectors
 
 ```sql
@@ -772,8 +754,6 @@ SELECT query, calls, ROUND((total_plan_time + total_exec_time) / calls) AS avg_t
     FROM pg_stat_statements ORDER BY total_plan_time + total_exec_time DESC LIMIT 20;
 ```
 
-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
@@ -842,7 +822,7 @@ You can use [half-precision indexing](#half-precision-indexing) to index up to 4
 
 #### Can I store vectors with different dimensions in the same column?
 
-You can use `vector` as the type (instead of `vector(3)`).
+You can use `vector` as the type (instead of `vector(n)`).
 
 ```sql
 CREATE TABLE embeddings (model_id bigint, item_id bigint, embedding vector, PRIMARY KEY (model_id, item_id));
@@ -1121,7 +1101,13 @@ Note: Replace `17` with your Postgres server version
 
 ### Missing SDK
 
-If compilation fails and the output includes `warning: no such sysroot directory` on Mac, reinstall Xcode Command Line Tools.
+If compilation fails and the output includes `warning: no such sysroot directory` on Mac, your Postgres installation points to a path that no longer exists.
+
+```sh
+pg_config --cppflags
+```
+
+Reinstall Postgres to fix this.
 
 ### Portability
 
@@ -1139,6 +1125,14 @@ make OPTFLAGS=""
 
 If compilation fails with `Cannot open include file: 'postgres.h': No such file or directory`, make sure `PGROOT` is correct.
 
+### Mismatched Architecture
+
+If compilation fails with `error C2196: case value '4' already used`, make sure `vcvars64.bat` was called. Then run `nmake /F Makefile.win clean` and re-run the installation instructions.
+
+### Missing Symbol
+
+If linking fails with `unresolved external symbol float_to_shortest_decimal_bufn` with Postgres 17.0-17.2, upgrade to Postgres 17.3+.
+
 ### Permissions
 
 If installation fails with `Access is denied`, re-run the installation instructions as an administrator.
@@ -1163,6 +1157,12 @@ cd pgvector
 docker build --pull --build-arg PG_MAJOR=17 -t myuser/pgvector .
 ```
 
+If you increase `maintenance_work_mem`, make sure `--shm-size` is at least that size to avoid an error with parallel HNSW index builds.
+
+```sh
+docker run --shm-size=1g ...
+```
+
 ### Homebrew
 
 With Homebrew Postgres, you can use:
@@ -1250,36 +1250,6 @@ You can check the version in the current database with:
 SELECT extversion FROM pg_extension WHERE extname = 'vector';
 ```
 
-## Upgrade Notes
-
-### 0.6.0
-
-#### Postgres 12
-
-If upgrading with Postgres 12, remove this line from `sql/vector--0.5.1--0.6.0.sql`:
-
-```sql
-ALTER TYPE vector SET (STORAGE = external);
-```
-
-Then run `make install` and `ALTER EXTENSION vector UPDATE;`.
-
-#### Docker
-
-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).
-
-```sh
-docker pull pgvector/pgvector:pg16
-# or
-docker pull pgvector/pgvector:0.6.0-pg16
-```
-
-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 ...
-```
-
 ## Thanks
 
 Thanks to:

src/hnsw.c

@@ -259,6 +259,11 @@ hnswhandler(PG_FUNCTION_ARGS)
 	amroutine->amoptsprocnum = 0;
 	amroutine->amcanorder = false;
 	amroutine->amcanorderbyop = true;
+#if PG_VERSION_NUM >= 180000
+	amroutine->amcanhash = false;
+	amroutine->amconsistentequality = false;
+	amroutine->amconsistentordering = false;
+#endif
 	amroutine->amcanbackward = false;	/* can change direction mid-scan */
 	amroutine->amcanunique = false;
 	amroutine->amcanmulticol = false;
@@ -291,6 +296,9 @@ hnswhandler(PG_FUNCTION_ARGS)
 	amroutine->amvacuumcleanup = hnswvacuumcleanup;
 	amroutine->amcanreturn = NULL;
 	amroutine->amcostestimate = hnswcostestimate;
+#if PG_VERSION_NUM >= 180000
+	amroutine->amgettreeheight = NULL;
+#endif
 	amroutine->amoptions = hnswoptions;
 	amroutine->amproperty = NULL;	/* TODO AMPROP_DISTANCE_ORDERABLE */
 	amroutine->ambuildphasename = hnswbuildphasename;
@@ -311,5 +319,10 @@ hnswhandler(PG_FUNCTION_ARGS)
 	amroutine->aminitparallelscan = NULL;
 	amroutine->amparallelrescan = NULL;
 
+#if PG_VERSION_NUM >= 180000
+	amroutine->amtranslatestrategy = NULL;
+	amroutine->amtranslatecmptype = NULL;
+#endif
+
 	PG_RETURN_POINTER(amroutine);
 }

src/hnswutils.c

@@ -805,6 +805,12 @@ HnswLoadUnvisitedFromDisk(HnswElement element, HnswUnvisited * unvisited, int *u
 		if (!found)
 			unvisited[(*unvisitedLength)++].indextid = *indextid;
 	}
+
+#ifdef USE_PREFETCH
+	/* TODO limit by get_tablespace_io_concurrency */
+	for (int i = 0; i < *unvisitedLength; i++)
+		PrefetchBuffer(index, MAIN_FORKNUM, ItemPointerGetBlockNumber(&unvisited[i].indextid));
+#endif
 }
 
 /*

src/hnswvacuum.c

@@ -9,6 +9,10 @@
 #include "storage/lmgr.h"
 #include "utils/memutils.h"
 
+#if PG_VERSION_NUM >= 180000
+#define vacuum_delay_point() vacuum_delay_point(false)
+#endif
+
 /*
  * Check if deleted list contains an index TID
  */

src/ivfflat.c

@@ -186,6 +186,11 @@ ivfflathandler(PG_FUNCTION_ARGS)
 	amroutine->amoptsprocnum = 0;
 	amroutine->amcanorder = false;
 	amroutine->amcanorderbyop = true;
+#if PG_VERSION_NUM >= 180000
+	amroutine->amcanhash = false;
+	amroutine->amconsistentequality = false;
+	amroutine->amconsistentordering = false;
+#endif
 	amroutine->amcanbackward = false;	/* can change direction mid-scan */
 	amroutine->amcanunique = false;
 	amroutine->amcanmulticol = false;
@@ -218,6 +223,9 @@ ivfflathandler(PG_FUNCTION_ARGS)
 	amroutine->amvacuumcleanup = ivfflatvacuumcleanup;
 	amroutine->amcanreturn = NULL;	/* tuple not included in heapsort */
 	amroutine->amcostestimate = ivfflatcostestimate;
+#if PG_VERSION_NUM >= 180000
+	amroutine->amgettreeheight = NULL;
+#endif
 	amroutine->amoptions = ivfflatoptions;
 	amroutine->amproperty = NULL;	/* TODO AMPROP_DISTANCE_ORDERABLE */
 	amroutine->ambuildphasename = ivfflatbuildphasename;
@@ -238,5 +246,10 @@ ivfflathandler(PG_FUNCTION_ARGS)
 	amroutine->aminitparallelscan = NULL;
 	amroutine->amparallelrescan = NULL;
 
+#if PG_VERSION_NUM >= 180000
+	amroutine->amtranslatestrategy = NULL;
+	amroutine->amtranslatecmptype = NULL;
+#endif
+
 	PG_RETURN_POINTER(amroutine);
 }

src/ivfvacuum.c

@@ -5,6 +5,10 @@
 #include "ivfflat.h"
 #include "storage/bufmgr.h"
 
+#if PG_VERSION_NUM >= 180000
+#define vacuum_delay_point() vacuum_delay_point(false)
+#endif
+
 /*
  * Bulk delete tuples from the index
  */

test/expected/hnsw_vector.out

@@ -123,6 +123,12 @@ SELECT * FROM t ORDER BY val <-> '[3,3,3]';
  [0,0,0]
 (3 rows)
 
+TRUNCATE t;
+SELECT * FROM t ORDER BY val <-> '[3,3,3]';
+ val 
+-----
+(0 rows)
+
 RESET hnsw.iterative_scan;
 RESET hnsw.ef_search;
 DROP TABLE t;

test/expected/ivfflat_vector.out

@@ -110,6 +110,15 @@ SELECT * FROM t ORDER BY val <-> '[3,3,3]';
  [1,1,1]
 (2 rows)
 
+TRUNCATE t;
+NOTICE:  ivfflat index created with little data
+DETAIL:  This will cause low recall.
+HINT:  Drop the index until the table has more data.
+SELECT * FROM t ORDER BY val <-> '[3,3,3]';
+ val 
+-----
+(0 rows)
+
 RESET ivfflat.iterative_scan;
 RESET ivfflat.max_probes;
 DROP TABLE t;

test/sql/hnsw_vector.sql

@@ -70,6 +70,9 @@ SELECT * FROM t ORDER BY val <-> '[3,3,3]';
 SET hnsw.iterative_scan = relaxed_order;
 SELECT * FROM t ORDER BY val <-> '[3,3,3]';
 
+TRUNCATE t;
+SELECT * FROM t ORDER BY val <-> '[3,3,3]';
+
 RESET hnsw.iterative_scan;
 RESET hnsw.ef_search;
 DROP TABLE t;

test/sql/ivfflat_vector.sql

@@ -59,6 +59,9 @@ SELECT * FROM t ORDER BY val <-> '[3,3,3]';
 SET ivfflat.max_probes = 2;
 SELECT * FROM t ORDER BY val <-> '[3,3,3]';
 
+TRUNCATE t;
+SELECT * FROM t ORDER BY val <-> '[3,3,3]';
+
 RESET ivfflat.iterative_scan;
 RESET ivfflat.max_probes;
 DROP TABLE t;