Improved test [skip ci]

.github/workflows/build.yml

@@ -57,9 +57,9 @@ jobs:
       - run: |
           brew install cpanm
           cpanm --notest IPC::Run
-          wget -q https://github.com/postgres/postgres/archive/refs/tags/REL_14_10.tar.gz
+          wget -q https://github.com/postgres/postgres/archive/refs/tags/REL_14_5.tar.gz
-          tar xf REL_14_10.tar.gz
+          tar xf REL_14_5.tar.gz
-      - run: make prove_installcheck PROVE_FLAGS="-I ./postgres-REL_14_10/src/test/perl" PERL5LIB="/Users/runner/perl5/lib/perl5"
+      - 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 PG_CFLAGS="-DUSE_ASSERT_CHECKING"
   windows:
     runs-on: windows-latest

CHANGELOG.md

@@ -1,17 +1,16 @@
-## 0.6.0 (2024-01-29)
+## 0.6.0 (unreleased)
 
-If upgrading with Postgres 12 or Docker, see [these notes](https://github.com/pgvector/pgvector#060).
+If upgrading with Postgres < 13, see [this note](https://github.com/pgvector/pgvector#060).
 
 - Changed storage for vector from `extended` to `external`
-- Added support for parallel index builds for HNSW
-- Added validation for GUC parameters
 - Improved performance of HNSW
+- Added support for parallel index builds for HNSW
+- Added `hamming_distance` function
+- Added validation for GUC parameters
 - 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
+- Fixed `invalid memory alloc request size` error with HNSW index build
-- 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)
@@ -60,7 +59,7 @@ 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`
@@ -77,7 +76,7 @@ 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
 

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": "0.6.0",
+	"version": "0.5.1",
 	"maintainer": [
 		"Andrew Kane <andrew@ankane.org>"
 	],
@@ -20,7 +20,7 @@
 		"vector": {
 			"file": "sql/vector.sql",
 			"docfile": "README.md",
-			"version": "0.6.0",
+			"version": "0.5.1",
 			"abstract": "Open-source vector similarity search for Postgres"
 		}
 	},

Makefile

@@ -1,5 +1,5 @@
 EXTENSION = vector
-EXTVERSION = 0.6.0
+EXTVERSION = 0.5.1
 
 MODULE_big = vector
 DATA = $(wildcard sql/*--*.sql)
@@ -65,17 +65,13 @@ dist:
 	mkdir -p dist
 	git archive --format zip --prefix=$(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) .
+	docker build --pull --no-cache --platform linux/amd64 -t ankane/pgvector:latest .
-	docker build --build-arg PG_MAJOR=$(PG_MAJOR) -t pgvector/pgvector:$(EXTVERSION)-pg$(PG_MAJOR) .
 
 .PHONY: docker-release
 
 docker-release:
-	docker buildx build --push --pull --no-cache --platform linux/amd64,linux/arm64 --build-arg PG_MAJOR=$(PG_MAJOR) -t pgvector/pgvector:pg$(PG_MAJOR) .
+	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 --build-arg PG_MAJOR=$(PG_MAJOR) -t pgvector/pgvector:$(EXTVERSION)-pg$(PG_MAJOR) .
+	docker buildx build --push --platform linux/amd64,linux/arm64 -t ankane/pgvector:v$(EXTVERSION) .

Makefile.win

@@ -1,5 +1,5 @@
 EXTENSION = vector
-EXTVERSION = 0.6.0
+EXTVERSION = 0.5.1
 
 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

README.md

@@ -14,13 +14,11 @@ Plus [ACID](https://en.wikipedia.org/wiki/ACID) compliance, point-in-time recove
 
 ## Installation
 
-### Linux and Mac
-
 Compile and install the extension (supports Postgres 11+)
 
 ```sh
 cd /tmp
-git clone --branch v0.6.0 https://github.com/pgvector/pgvector.git
+git clone --branch v0.5.1 https://github.com/pgvector/pgvector.git
 cd pgvector
 make
 make install # may need sudo
@@ -30,28 +28,6 @@ See the [installation notes](#installation-notes) if you run into issues
 
 You can also install it with [Docker](#docker), [Homebrew](#homebrew), [PGXN](#pgxn), [APT](#apt), [Yum](#yum), or [conda-forge](#conda-forge), and it comes preinstalled with [Postgres.app](#postgresapp) and many [hosted providers](#hosted-postgres). There are also instructions for [GitHub Actions](https://github.com/pgvector/setup-pgvector).
 
-### 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"
-git clone --branch v0.6.0 https://github.com/pgvector/pgvector.git
-cd pgvector
-nmake /F Makefile.win
-nmake /F Makefile.win install
-```
-
-You can also install it with [Docker](#docker) or [conda-forge](#conda-forge).
-
 ## Getting Started
 
 Enable the extension (do this once in each database where you want to use it)
@@ -264,14 +240,6 @@ HINT:  Increase maintenance_work_mem to speed up builds.
 
 Note: Do not set `maintenance_work_mem` so high that it exhausts the memory on the server
 
-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+
@@ -647,14 +615,34 @@ Note: Replace `16` with your Postgres server version
 
 If compilation fails and the output includes `warning: no such sysroot directory` on Mac, reinstall Xcode Command Line Tools.
 
+### Windows
+
+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:
+
+```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"
+git clone --branch v0.5.1 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) (run it the same way).
@@ -662,9 +650,9 @@ This adds pgvector to the [Postgres image](https://hub.docker.com/_/postgres) (r
 You can also build the image manually:
 
 ```sh
-git clone --branch v0.6.0 https://github.com/pgvector/pgvector.git
+git clone --branch v0.5.1 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
@@ -741,11 +729,9 @@ SELECT extversion FROM pg_extension WHERE extname = 'vector';
 
 ## Upgrade Notes
 
-### 0.6.0
+### 0.6.0 [unreleased]
 
-#### Postgres 12
+If upgrading with Postgres < 13, remove this line from `sql/vector--0.5.1--0.6.0.sql`:
-
-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);
@@ -753,20 +739,28 @@ ALTER TYPE vector SET (STORAGE = external);
 
 Then run `make install` and `ALTER EXTENSION vector UPDATE;`.
 
-#### Docker
+### 0.4.0
 
-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 with Postgres < 13, remove this line from `sql/vector--0.3.2--0.4.0.sql`:
 
-```sh
+```sql
-docker pull pgvector/pgvector:pg16
+ALTER TYPE vector SET (STORAGE = extended);
-# 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.
+Then run `make install` and `ALTER EXTENSION vector UPDATE;`.
 
-```sh
+### 0.3.1
-docker run --shm-size=1g ...
+
+If upgrading from 0.2.7 or 0.3.0, recreate all `ivfflat` indexes after upgrading to ensure all data is indexed.
+
+```sql
+-- Postgres 12+
+REINDEX INDEX CONCURRENTLY index_name;
+
+-- 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
@@ -812,8 +806,8 @@ make prove_installcheck  # TAP tests
 To run single tests:
 
 ```sh
-make installcheck REGRESS=functions                            # regression test
+make installcheck REGRESS=functions                    # regression test
-make prove_installcheck PROVE_TESTS=test/t/001_ivfflat_wal.pl  # TAP test
+make prove_installcheck PROVE_TESTS=test/t/001_wal.pl  # TAP test
 ```
 
 To enable benchmarking:

sql/vector--0.5.1--0.6.0.sql

@@ -3,3 +3,6 @@
 
 -- remove this single line for Postgres < 13
 ALTER TYPE vector SET (STORAGE = external);
+
+CREATE FUNCTION hamming_distance(bytea, bytea) RETURNS float8
+	AS 'MODULE_PATHNAME' LANGUAGE C IMMUTABLE STRICT PARALLEL SAFE;

sql/vector.sql

@@ -290,3 +290,8 @@ 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);
+
+-- bytea functions
+
+CREATE FUNCTION hamming_distance(bytea, bytea) RETURNS float8
+	AS 'MODULE_PATHNAME' LANGUAGE C IMMUTABLE STRICT PARALLEL SAFE;

src/hnsw.h

@@ -60,13 +60,10 @@
 #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_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))
 
@@ -382,7 +379,7 @@ void		HnswUpdateMetaPage(Relation index, int updateEntry, HnswElement entryPoint
 void		HnswSetNeighborTuple(char *base, 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);
+bool		HnswInsertTupleOnDisk(Relation index, Datum value, Datum *values, bool *isnull, ItemPointer heap_tid, Relation heapRel, bool building);
 void		HnswUpdateNeighborsOnDisk(Relation index, FmgrInfo *procinfo, Oid collation, HnswElement e, int m, bool checkExisting, bool building);
 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);

src/hnswbuild.c

@@ -65,6 +65,8 @@
 #define CALLBACK_ITEM_POINTER HeapTuple hup
 #endif
 
+#define UpdateProgress(index, val) pgstat_progress_update_param(index, val)
+
 #if PG_VERSION_NUM >= 140000
 #include "utils/backend_status.h"
 #include "utils/wait_event.h"
@@ -141,13 +143,14 @@ HnswBuildAppendPage(Relation index, Buffer *buf, Page *page, ForkNumber forkNum)
 }
 
 /*
- * Create graph pages
+ * Create element pages
  */
 static void
-CreateGraphPages(HnswBuildState * buildstate)
+CreateElementPages(HnswBuildState * buildstate)
 {
 	Relation	index = buildstate->index;
 	ForkNumber	forkNum = buildstate->forkNum;
+	Size		etupAllocSize;
 	Size		maxSize;
 	HnswElementTuple etup;
 	HnswNeighborTuple ntup;
@@ -159,11 +162,12 @@ CreateGraphPages(HnswBuildState * buildstate)
 	char	   *base = buildstate->hnswarea;
 
 	/* Calculate sizes */
+	etupAllocSize = BLCKSZ;
 	maxSize = HNSW_MAX_SIZE;
 
 	/* Allocate once */
-	etup = palloc0(HNSW_TUPLE_ALLOC_SIZE);
+	etup = palloc0(etupAllocSize);
-	ntup = palloc0(HNSW_TUPLE_ALLOC_SIZE);
+	ntup = palloc0(BLCKSZ);
 
 	/* Prepare first page */
 	buf = HnswNewBuffer(index, forkNum);
@@ -182,7 +186,7 @@ CreateGraphPages(HnswBuildState * buildstate)
 		iter = element->next;
 
 		/* Zero memory for each element */
-		MemSet(etup, 0, HNSW_TUPLE_ALLOC_SIZE);
+		MemSet(etup, 0, etupAllocSize);
 
 		/* Calculate sizes */
 		etupSize = HNSW_ELEMENT_TUPLE_SIZE(VARSIZE_ANY(valuePtr));
@@ -190,7 +194,7 @@ CreateGraphPages(HnswBuildState * buildstate)
 		combinedSize = etupSize + ntupSize + sizeof(ItemIdData);
 
 		/* Initial size check */
-		if (etupSize > HNSW_TUPLE_ALLOC_SIZE)
+		if (etupSize > etupAllocSize)
 			elog(ERROR, "index tuple too large");
 
 		HnswSetElementTuple(base, etup, element);
@@ -242,10 +246,10 @@ CreateGraphPages(HnswBuildState * buildstate)
 }
 
 /*
- * Write neighbor tuples
+ * Create neighbor pages
  */
 static void
-WriteNeighborTuples(HnswBuildState * buildstate)
+CreateNeighborPages(HnswBuildState * buildstate)
 {
 	Relation	index = buildstate->index;
 	ForkNumber	forkNum = buildstate->forkNum;
@@ -255,32 +259,29 @@ WriteNeighborTuples(HnswBuildState * buildstate)
 	HnswNeighborTuple ntup;
 
 	/* Allocate once */
-	ntup = palloc0(HNSW_TUPLE_ALLOC_SIZE);
+	ntup = palloc0(BLCKSZ);
 
 	while (!HnswPtrIsNull(base, iter))
 	{
-		HnswElement element = HnswPtrAccess(base, iter);
+		HnswElement e = HnswPtrAccess(base, iter);
 		Buffer		buf;
 		Page		page;
-		Size		ntupSize = HNSW_NEIGHBOR_TUPLE_SIZE(element->level, m);
+		Size		ntupSize = HNSW_NEIGHBOR_TUPLE_SIZE(e->level, m);
 
 		/* Update iterator */
-		iter = element->next;
+		iter = e->next;
-
-		/* Zero memory for each element */
-		MemSet(ntup, 0, HNSW_TUPLE_ALLOC_SIZE);
 
 		/* Can take a while, so ensure we can interrupt */
 		/* Needs to be called when no buffer locks are held */
 		CHECK_FOR_INTERRUPTS();
 
-		buf = ReadBufferExtended(index, forkNum, element->neighborPage, RBM_NORMAL, NULL);
+		buf = ReadBufferExtended(index, forkNum, e->neighborPage, RBM_NORMAL, NULL);
 		LockBuffer(buf, BUFFER_LOCK_EXCLUSIVE);
 		page = BufferGetPage(buf);
 
-		HnswSetNeighborTuple(base, ntup, element, m);
+		HnswSetNeighborTuple(base, ntup, e, m);
 
-		if (!PageIndexTupleOverwrite(page, element->neighborOffno, (Item) ntup, ntupSize))
+		if (!PageIndexTupleOverwrite(page, e->neighborOffno, (Item) ntup, ntupSize))
 			elog(ERROR, "failed to add index item to \"%s\"", RelationGetRelationName(index));
 
 		/* Commit */
@@ -302,8 +303,8 @@ FlushPages(HnswBuildState * buildstate)
 #endif
 
 	CreateMetaPage(buildstate);
-	CreateGraphPages(buildstate);
+	CreateElementPages(buildstate);
-	WriteNeighborTuples(buildstate);
+	CreateNeighborPages(buildstate);
 
 	buildstate->graph->flushed = true;
 	MemoryContextReset(buildstate->graphCtx);
@@ -497,7 +498,7 @@ InsertTuple(Relation index, Datum *values, bool *isnull, ItemPointer heaptid, Hn
 	{
 		LWLockRelease(flushLock);
 
-		return HnswInsertTupleOnDisk(index, value, values, isnull, heaptid, true);
+		return HnswInsertTupleOnDisk(index, value, values, isnull, heaptid, buildstate->heap, true);
 	}
 
 	/*
@@ -529,7 +530,7 @@ InsertTuple(Relation index, Datum *values, bool *isnull, ItemPointer heaptid, Hn
 
 		LWLockRelease(flushLock);
 
-		return HnswInsertTupleOnDisk(index, value, values, isnull, heaptid, true);
+		return HnswInsertTupleOnDisk(index, value, values, isnull, heaptid, buildstate->heap, true);
 	}
 
 	/* Ok, we can proceed to allocate the element */
@@ -586,7 +587,7 @@ BuildCallback(Relation index, CALLBACK_ITEM_POINTER, Datum *values,
 	{
 		/* Update progress */
 		SpinLockAcquire(&graph->lock);
-		pgstat_progress_update_param(PROGRESS_CREATEIDX_TUPLES_DONE, ++graph->indtuples);
+		UpdateProgress(PROGRESS_CREATEIDX_TUPLES_DONE, ++graph->indtuples);
 		SpinLockRelease(&graph->lock);
 	}
 
@@ -1047,7 +1048,7 @@ BuildGraph(HnswBuildState * buildstate, ForkNumber forkNum)
 {
 	int			parallel_workers = 0;
 
-	pgstat_progress_update_param(PROGRESS_CREATEIDX_SUBPHASE, PROGRESS_HNSW_PHASE_LOAD);
+	UpdateProgress(PROGRESS_CREATEIDX_SUBPHASE, PROGRESS_HNSW_PHASE_LOAD);
 
 	/* Calculate parallel workers */
 	if (buildstate->heap != NULL)

src/hnswinsert.c

@@ -355,7 +355,9 @@ 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);
@@ -394,7 +396,9 @@ HnswUpdateNeighborsOnDisk(Relation index, FmgrInfo *procinfo, Oid collation, Hns
 			}
 
 			/* 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;
@@ -423,9 +427,13 @@ 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);
@@ -449,7 +457,9 @@ AddDuplicateOnDisk(Relation index, HnswElement element, HnswElement dup, bool bu
 	Buffer		buf;
 	Page		page;
 	GenericXLogState *state;
+	ItemId		itemid;
 	HnswElementTuple etup;
+	Size		etupSize;
 	int			i;
 
 	/* Read page */
@@ -467,7 +477,9 @@ AddDuplicateOnDisk(Relation index, HnswElement element, HnswElement dup, bool bu
 	}
 
 	/* Find space */
-	etup = (HnswElementTuple) PageGetItem(page, PageGetItemId(page, dup->offno));
+	itemid = PageGetItemId(page, dup->offno);
+	etup = (HnswElementTuple) PageGetItem(page, itemid);
+	etupSize = ItemIdGetLength(itemid);
 	for (i = 0; i < HNSW_HEAPTIDS; i++)
 	{
 		if (!ItemPointerIsValid(&etup->heaptids[i]))
@@ -483,9 +495,13 @@ AddDuplicateOnDisk(Relation index, HnswElement element, HnswElement dup, bool bu
 		return false;
 	}
 
-	/* Add heap TID, modifying the tuple on the page directly */
+	/* Add heap TID */
 	etup->heaptids[i] = element->heaptids[0];
 
+	/* 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);
@@ -554,7 +570,7 @@ UpdateGraphOnDisk(Relation index, FmgrInfo *procinfo, Oid collation, HnswElement
  * Insert a tuple into the index
  */
 bool
-HnswInsertTupleOnDisk(Relation index, Datum value, Datum *values, bool *isnull, ItemPointer heap_tid, bool building)
+HnswInsertTupleOnDisk(Relation index, Datum value, Datum *values, bool *isnull, ItemPointer heap_tid, Relation heapRel, bool building)
 {
 	HnswElement entryPoint;
 	HnswElement element;
@@ -609,7 +625,7 @@ HnswInsertTupleOnDisk(Relation index, Datum value, Datum *values, bool *isnull,
  * Insert a tuple into the index
  */
 static void
-HnswInsertTuple(Relation index, Datum *values, bool *isnull, ItemPointer heap_tid)
+HnswInsertTuple(Relation index, Datum *values, bool *isnull, ItemPointer heap_tid, Relation heapRel)
 {
 	Datum		value;
 	FmgrInfo   *normprocinfo;
@@ -626,7 +642,7 @@ HnswInsertTuple(Relation index, Datum *values, bool *isnull, ItemPointer heap_ti
 			return;
 	}
 
-	HnswInsertTupleOnDisk(index, value, values, isnull, heap_tid, false);
+	HnswInsertTupleOnDisk(index, value, values, isnull, heap_tid, heapRel, false);
 }
 
 /*
@@ -655,7 +671,7 @@ 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

@@ -202,7 +202,6 @@ hnswgettuple(IndexScanDesc scan, ScanDirection dir)
 		MemoryContextSwitchTo(oldCtx);
 
 		scan->xs_heaptid = *heaptid;
-		scan->xs_recheck = false;
 		scan->xs_recheckorderby = false;
 		return true;
 	}

src/hnswutils.c

@@ -7,7 +7,6 @@
 #include "lib/pairingheap.h"
 #include "storage/bufmgr.h"
 #include "utils/datum.h"
-#include "utils/memdebug.h"
 #include "utils/rel.h"
 #include "vector.h"
 
@@ -203,19 +202,6 @@ HnswInitPage(Buffer buf, Page page)
 	HnswPageGetOpaque(page)->page_id = HNSW_PAGE_ID;
 }
 
-/*
- * Allocate a neighbor array
- */
-static HnswNeighborArray *
-HnswInitNeighborArray(int lm, HnswAllocator * allocator)
-{
-	HnswNeighborArray *a = HnswAlloc(allocator, HNSW_NEIGHBOR_ARRAY_SIZE(lm));
-
-	a->length = 0;
-	a->closerSet = false;
-	return a;
-}
-
 /*
  * Allocate neighbors
  */
@@ -223,12 +209,22 @@ void
 HnswInitNeighbors(char *base, HnswElement element, int m, HnswAllocator * allocator)
 {
 	int			level = element->level;
+
 	HnswNeighborArrayPtr *neighborList = (HnswNeighborArrayPtr *) HnswAlloc(allocator, sizeof(HnswNeighborArrayPtr) * (level + 1));
 
 	HnswPtrStore(base, element->neighbors, neighborList);
 
 	for (int lc = 0; lc <= level; lc++)
-		HnswPtrStore(base, neighborList[lc], HnswInitNeighborArray(HnswGetLayerM(m, lc), allocator));
+	{
+		HnswNeighborArray *a;
+		int			lm = HnswGetLayerM(m, lc);
+
+		HnswPtrStore(base, neighborList[lc], (HnswNeighborArray *) HnswAlloc(allocator, offsetof(HnswNeighborArray, items) + sizeof(HnswCandidate) * lm));
+
+		a = HnswGetNeighbors(base, element, lc);
+		a->length = 0;
+		a->closerSet = false;
+	}
 }
 
 /*
@@ -316,10 +312,7 @@ HnswGetMetaPageInfo(Relation index, int *m, HnswElement * entryPoint)
 	if (entryPoint != NULL)
 	{
 		if (BlockNumberIsValid(metap->entryBlkno))
-		{
 			*entryPoint = HnswInitElementFromBlock(metap->entryBlkno, metap->entryOffno);
-			(*entryPoint)->level = metap->entryLevel;
-		}
 		else
 			*entryPoint = NULL;
 	}
@@ -738,7 +731,7 @@ HnswSearchLayer(char *base, Datum q, List *ep, int ef, int lc, Relation index, F
 	/* Create local memory for neighborhood if needed */
 	if (index == NULL)
 	{
-		neighborhoodSize = HNSW_NEIGHBOR_ARRAY_SIZE(HnswGetLayerM(m, lc));
+		neighborhoodSize = offsetof(HnswNeighborArray, items) + sizeof(HnswCandidate) * HnswGetLayerM(m, lc);
 		neighborhoodData = palloc(neighborhoodSize);
 	}
 
@@ -914,10 +907,40 @@ CompareCandidateDistancesOffset(const void *a, const void *b)
  * Calculate the distance between elements
  */
 static float
-HnswGetDistance(char *base, HnswElement a, HnswElement b, FmgrInfo *procinfo, Oid collation)
+HnswGetDistance(char *base, HnswElement a, HnswElement b, int lc, FmgrInfo *procinfo, Oid collation)
 {
-	Datum		aValue = HnswGetValue(base, a);
+	Datum		aValue;
-	Datum		bValue = HnswGetValue(base, b);
+	Datum		bValue;
+
+	/* Look for cached distance */
+	if (!HnswPtrIsNull(base, a->neighbors))
+	{
+		HnswNeighborArray *neighbors = HnswGetNeighbors(base, a, lc);
+
+		for (int i = 0; i < neighbors->length; i++)
+		{
+			HnswElement element = HnswPtrAccess(base, neighbors->items[i].element);
+
+			if (element == b)
+				return neighbors->items[i].distance;
+		}
+	}
+
+	if (!HnswPtrIsNull(base, b->neighbors))
+	{
+		HnswNeighborArray *neighbors = HnswGetNeighbors(base, b, lc);
+
+		for (int i = 0; i < neighbors->length; i++)
+		{
+			HnswElement element = HnswPtrAccess(base, neighbors->items[i].element);
+
+			if (element == a)
+				return neighbors->items[i].distance;
+		}
+	}
+
+	aValue = HnswGetValue(base, a);
+	bValue = HnswGetValue(base, b);
 
 	return DatumGetFloat8(FunctionCall2Coll(procinfo, collation, aValue, bValue));
 }
@@ -926,7 +949,7 @@ HnswGetDistance(char *base, HnswElement a, HnswElement b, FmgrInfo *procinfo, Oi
  * Check if an element is closer to q than any element from R
  */
 static bool
-CheckElementCloser(char *base, HnswCandidate * e, List *r, FmgrInfo *procinfo, Oid collation)
+CheckElementCloser(char *base, HnswCandidate * e, List *r, int lc, FmgrInfo *procinfo, Oid collation)
 {
 	HnswElement eElement = HnswPtrAccess(base, e->element);
 	ListCell   *lc2;
@@ -935,7 +958,7 @@ CheckElementCloser(char *base, HnswCandidate * e, List *r, FmgrInfo *procinfo, O
 	{
 		HnswCandidate *ri = lfirst(lc2);
 		HnswElement riElement = HnswPtrAccess(base, ri->element);
-		float		distance = HnswGetDistance(base, eElement, riElement, procinfo, collation);
+		float		distance = HnswGetDistance(base, eElement, riElement, lc, procinfo, collation);
 
 		if (distance <= e->distance)
 			return false;
@@ -981,20 +1004,16 @@ SelectNeighbors(char *base, List *c, int lm, int lc, FmgrInfo *procinfo, Oid col
 
 		/* Use previous state of r and wd to skip work when possible */
 		if (mustCalculate)
-			e->closer = CheckElementCloser(base, e, r, procinfo, collation);
+			e->closer = CheckElementCloser(base, e, r, lc, procinfo, collation);
 		else if (list_length(added) > 0)
 		{
-			/* Keep Valgrind happy for in-memory, parallel builds */
-			if (base != NULL)
-				VALGRIND_MAKE_MEM_DEFINED(&e->closer, 1);
-
 			/*
 			 * If the current candidate was closer, we only need to compare it
 			 * with the other candidates that we have added.
 			 */
 			if (e->closer)
 			{
-				e->closer = CheckElementCloser(base, e, added, procinfo, collation);
+				e->closer = CheckElementCloser(base, e, added, lc, procinfo, collation);
 
 				if (!e->closer)
 					removedAny = true;
@@ -1007,7 +1026,7 @@ SelectNeighbors(char *base, List *c, int lm, int lc, FmgrInfo *procinfo, Oid col
 				 */
 				if (removedAny)
 				{
-					e->closer = CheckElementCloser(base, e, r, procinfo, collation);
+					e->closer = CheckElementCloser(base, e, r, lc, procinfo, collation);
 					if (e->closer)
 						added = lappend(added, e);
 				}
@@ -1015,15 +1034,11 @@ SelectNeighbors(char *base, List *c, int lm, int lc, FmgrInfo *procinfo, Oid col
 		}
 		else if (e == newCandidate)
 		{
-			e->closer = CheckElementCloser(base, e, r, procinfo, collation);
+			e->closer = CheckElementCloser(base, e, r, lc, procinfo, collation);
 			if (e->closer)
 				added = lappend(added, e);
 		}
 
-		/* Keep Valgrind happy for in-memory, parallel builds */
-		if (base != NULL)
-			VALGRIND_MAKE_MEM_DEFINED(&e->closer, 1);
-
 		if (e->closer)
 			r = lappend(r, e);
 		else

src/hnswvacuum.c

@@ -60,7 +60,8 @@ RemoveHeapTids(HnswVacuumState * vacuumstate)
 		/* Iterate over nodes */
 		for (offno = FirstOffsetNumber; offno <= maxoffno; offno = OffsetNumberNext(offno))
 		{
-			HnswElementTuple etup = (HnswElementTuple) PageGetItem(page, PageGetItemId(page, offno));
+			ItemId		itemid = PageGetItemId(page, offno);
+			HnswElementTuple etup = (HnswElementTuple) PageGetItem(page, itemid);
 			int			idx = 0;
 			bool		itemUpdated = false;
 
@@ -91,10 +92,15 @@ RemoveHeapTids(HnswVacuumState * vacuumstate)
 
 				if (itemUpdated)
 				{
+					Size		etupSize = ItemIdGetLength(itemid);
+
 					/* 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;
 				}
 			}
@@ -207,9 +213,6 @@ RepairGraphElement(HnswVacuumState * vacuumstate, HnswElement element, HnswEleme
 	/* 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);
-
 	/* Update neighbor tuple */
 	/* Do this before getting page to minimize locking */
 	HnswSetNeighborTuple(base, ntup, element, m);
@@ -476,8 +479,11 @@ MarkDeleted(HnswVacuumState * vacuumstate)
 		/* Update element and neighbors together */
 		for (offno = FirstOffsetNumber; offno <= maxoffno; offno = OffsetNumberNext(offno))
 		{
-			HnswElementTuple etup = (HnswElementTuple) PageGetItem(page, PageGetItemId(page, offno));
+			ItemId		itemid = PageGetItemId(page, offno);
+			HnswElementTuple etup = (HnswElementTuple) PageGetItem(page, itemid);
 			HnswNeighborTuple ntup;
+			Size		etupSize;
+			Size		ntupSize;
 			Buffer		nbuf;
 			Page		npage;
 			BlockNumber neighborPage;
@@ -501,6 +507,10 @@ MarkDeleted(HnswVacuumState * vacuumstate)
 			if (ItemPointerIsValid(&etup->heaptids[0]))
 				continue;
 
+			/* Calculate sizes */
+			etupSize = ItemIdGetLength(itemid);
+			ntupSize = HNSW_NEIGHBOR_TUPLE_SIZE(etup->level, vacuumstate->m);
+
 			/* Get neighbor page */
 			neighborPage = ItemPointerGetBlockNumber(&etup->neighbortid);
 			neighborOffno = ItemPointerGetOffsetNumber(&etup->neighbortid);
@@ -527,10 +537,13 @@ MarkDeleted(HnswVacuumState * vacuumstate)
 			for (int i = 0; i < ntup->count; i++)
 				ItemPointerSetInvalid(&ntup->indextids[i]);
 
-			/*
+			/* Overwrite element tuple */
-			 * We modified the tuples in place, no need to call
+			if (!PageIndexTupleOverwrite(page, offno, (Item) etup, etupSize))
-			 * PageIndexTupleOverwrite
+				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 */
 			GenericXLogFinish(state);
@@ -575,7 +588,7 @@ InitVacuumState(HnswVacuumState * vacuumstate, IndexVacuumInfo *info, IndexBulkD
 	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);

src/ivfbuild.c

@@ -29,6 +29,8 @@
 #define CALLBACK_ITEM_POINTER HeapTuple hup
 #endif
 
+#define UpdateProgress(index, val) pgstat_progress_update_param(index, val)
+
 #if PG_VERSION_NUM >= 140000
 #include "utils/backend_status.h"
 #include "utils/wait_event.h"
@@ -260,9 +262,9 @@ InsertTuples(Relation index, IvfflatBuildState * buildstate, ForkNumber forkNum)
 	TupleTableSlot *slot = MakeSingleTupleTableSlot(buildstate->tupdesc, &TTSOpsMinimalTuple);
 	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 +300,7 @@ 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);
 		}
@@ -398,7 +400,7 @@ 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 */
@@ -473,7 +475,7 @@ CreateListPages(Relation index, VectorArray centers, int dimensions,
 	IvfflatList list;
 
 	listSize = MAXALIGN(IVFFLAT_LIST_SIZE(dimensions));
-	list = palloc0(listSize);
+	list = palloc(listSize);
 
 	buf = IvfflatNewBuffer(index, forkNum);
 	IvfflatInitRegisterPage(index, &buf, &page, &state);
@@ -919,7 +921,7 @@ 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)

src/ivfscan.c

@@ -310,7 +310,6 @@ ivfflatgettuple(IndexScanDesc scan, ScanDirection dir)
 		ItemPointer heaptid = (ItemPointer) DatumGetPointer(slot_getattr(so->slot, 2, &so->isnull));
 
 		scan->xs_heaptid = *heaptid;
-		scan->xs_recheck = false;
 		scan->xs_recheckorderby = false;
 		return true;
 	}

src/vector.c

@@ -10,6 +10,7 @@
 #include "lib/stringinfo.h"
 #include "libpq/pqformat.h"
 #include "port.h"				/* for strtof() */
+#include "port/pg_bitutils.h"
 #include "utils/array.h"
 #include "utils/builtins.h"
 #include "utils/float.h"
@@ -1134,3 +1135,43 @@ vector_avg(PG_FUNCTION_ARGS)
 
 	PG_RETURN_POINTER(result);
 }
+
+/*
+ * Ensure same number of bytes
+ */
+static inline void
+CheckByteLengths(uint32 aLen, uint32 bLen)
+{
+	if (aLen != bLen)
+		ereport(ERROR,
+				(errcode(ERRCODE_DATA_EXCEPTION),
+				 errmsg("different byte lengths %u and %u", aLen, bLen)));
+}
+
+/*
+ * Get the Hamming distance between two binary strings
+ */
+PGDLLEXPORT PG_FUNCTION_INFO_V1(hamming_distance);
+Datum
+hamming_distance(PG_FUNCTION_ARGS)
+{
+	bytea	   *a = PG_GETARG_BYTEA_PP(0);
+	bytea	   *b = PG_GETARG_BYTEA_PP(1);
+	char	   *ax = VARDATA_ANY(a);
+	char	   *bx = VARDATA_ANY(b);
+	uint32		aLen = VARSIZE_ANY_EXHDR(a);
+	uint32		bLen = VARSIZE_ANY_EXHDR(b);
+	uint64		distance = 0;
+
+	CheckByteLengths(aLen, bLen);
+
+	for (uint32 i = 0; i < aLen; i++)
+	{
+		unsigned char diff = (unsigned char) (ax[i] ^ bx[i]);
+
+		distance += pg_number_of_ones[diff];
+	}
+
+	/* TODO Decide on return type */
+	PG_RETURN_FLOAT8((double) distance);
+}

test/expected/functions.out

@@ -158,6 +158,32 @@ SELECT l1_distance('[3e38]', '[-3e38]');
     Infinity
 (1 row)
 
+SELECT hamming_distance('\xFFFF', '\xFFFF');
+ hamming_distance 
+------------------
+                0
+(1 row)
+
+SELECT hamming_distance('\xFFFF', '\xFFFE');
+ hamming_distance 
+------------------
+                1
+(1 row)
+
+SELECT hamming_distance('\xFFFF', '\xFFFC');
+ hamming_distance 
+------------------
+                2
+(1 row)
+
+SELECT hamming_distance('\xFFFF', '\x0000');
+ hamming_distance 
+------------------
+               16
+(1 row)
+
+SELECT hamming_distance('\xFFFF', '\x00');
+ERROR:  different byte lengths 2 and 1
 SELECT avg(v) FROM unnest(ARRAY['[1,2,3]'::vector, '[3,5,7]']) v;
     avg    
 -----------

test/sql/functions.sql

@@ -37,6 +37,12 @@ SELECT l1_distance('[0,0]', '[0,1]');
 SELECT l1_distance('[1,2]', '[3]');
 SELECT l1_distance('[3e38]', '[-3e38]');
 
+SELECT hamming_distance('\xFFFF', '\xFFFF');
+SELECT hamming_distance('\xFFFF', '\xFFFE');
+SELECT hamming_distance('\xFFFF', '\xFFFC');
+SELECT hamming_distance('\xFFFF', '\x0000');
+SELECT hamming_distance('\xFFFF', '\x00');
+
 SELECT avg(v) FROM unnest(ARRAY['[1,2,3]'::vector, '[3,5,7]']) v;
 SELECT avg(v) FROM unnest(ARRAY['[1,2,3]'::vector, '[3,5,7]', NULL]) v;
 SELECT avg(v) FROM unnest(ARRAY[]::vector[]) v;

vector.control

@@ -1,4 +1,4 @@
 comment = 'vector data type and ivfflat and hnsw access methods'
-default_version = '0.6.0'
+default_version = '0.5.1'
 module_pathname = '$libdir/vector'
 relocatable = true