Added filtering for text columns

These places called PageIndexTupleOverwrite(), with the new tuple
pointing directly to the original page. The PageIndexTupleOverwrite()
call is unnecessary in these cases, as we have already modified the
tuple on the page directly. Moreover, PageIndexTupleOverWrite() will
call memcpy with same src and dst arguments, which is undefined
behavior.

It's OK to modify the pages on disk directly, and we don't need
critical sections, because we either use the generic xlog functions
which create a temporary copy of the page, or we are building a new
index so if we crash the whole index is invisible and will be dropped
anyway.

CHANGELOG.md

@@ -1,11 +1,19 @@
-## 0.6.0 (unreleased)
+## 0.7.0 (unreleased)
 
-If upgrading with Postgres 12 or Docker, see [these notes](https://github.com/pgvector/pgvector#060-unreleased).
+- Added support for inline filtering with HNSW
+
+## 0.6.1 (unreleased)
+
+- Fixed error with `ANALYZE` and vectors with different dimensions
+
+## 0.6.0 (2024-01-29)
+
+If upgrading with Postgres 12 or Docker, see [these notes](https://github.com/pgvector/pgvector#060).
 
-- Changed storage for vector from `extended` to `external`
-- Improved performance of HNSW
 - 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

Dockerfile

@@ -1,4 +1,4 @@
-ARG PG_MAJOR=15
+ARG PG_MAJOR=16
 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": "0.5.1",
+	"version": "0.6.0",
 	"maintainer": [
 		"Andrew Kane <andrew@ankane.org>"
 	],
@@ -20,7 +20,7 @@
 		"vector": {
 			"file": "sql/vector.sql",
 			"docfile": "README.md",
-			"version": "0.5.1",
+			"version": "0.6.0",
 			"abstract": "Open-source vector similarity search for Postgres"
 		}
 	},

Makefile

@@ -1,5 +1,5 @@
 EXTENSION = vector
-EXTVERSION = 0.5.1
+EXTVERSION = 0.6.0
 
 MODULE_big = vector
 DATA = $(wildcard sql/*--*.sql)

Makefile.win

@@ -1,5 +1,5 @@
 EXTENSION = vector
-EXTVERSION = 0.5.1
+EXTVERSION = 0.6.0
 
 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\vector.obj
 HEADERS = src\vector.h

README.md

@@ -16,11 +16,11 @@ Plus [ACID](https://en.wikipedia.org/wiki/ACID) compliance, point-in-time recove
 
 ### Linux and Mac
 
-Compile and install the extension (supports Postgres 11+)
+Compile and install the extension (supports Postgres 12+)
 
 ```sh
 cd /tmp
-git clone --branch v0.5.1 https://github.com/pgvector/pgvector.git
+git clone --branch v0.6.0 https://github.com/pgvector/pgvector.git
 cd pgvector
 make
 make install # may need sudo
@@ -28,7 +28,7 @@ make install # may need sudo
 
 See the [installation notes](#installation-notes) if you run into issues
 
-You can also install it with [Docker](#docker), [Homebrew](#homebrew), [PGXN](#pgxn), [APT](#apt), [Yum](#yum), or [conda-forge](#conda-forge), and it comes preinstalled with [Postgres.app](#postgresapp) and many [hosted providers](#hosted-postgres). There are also instructions for [GitHub Actions](https://github.com/pgvector/setup-pgvector).
+You can also install it with [Docker](#docker), [Homebrew](#homebrew), [PGXN](#pgxn), [APT](#apt), [Yum](#yum), [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
 
@@ -44,7 +44,7 @@ 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
+git clone --branch v0.6.0 https://github.com/pgvector/pgvector.git
 cd pgvector
 nmake /F Makefile.win
 nmake /F Makefile.win install
@@ -264,7 +264,7 @@ 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 [unreleased], you can also speed up index creation by increasing the number of parallel workers (2 by default)
+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
@@ -377,6 +377,12 @@ Create an index on one [or more](https://www.postgresql.org/docs/current/indexes
 CREATE INDEX ON items (category_id);
 ```
 
+Or a composite HNSW index for approximate search (added in 0.7.0)
+
+```sql
+CREATE INDEX ON items USING hnsw (embedding vector_l2_ops, category_id);
+```
+
 Or a [partial index](https://www.postgresql.org/docs/current/indexes-partial.html) on the vector column for approximate search
 
 ```sql
@@ -430,6 +436,15 @@ 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;
+```
+
 ## 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.
@@ -651,20 +666,20 @@ If compilation fails and the output includes `warning: no such sysroot directory
 
 ### Docker
 
-Get the [Docker image](https://hub.docker.com/r/ankane/pgvector) with:
+Get the [Docker image](https://hub.docker.com/r/pgvector/pgvector) with:
 
 ```sh
-docker pull ankane/pgvector
+docker pull pgvector/pgvector:pg16
 ```
 
-This adds pgvector to the [Postgres image](https://hub.docker.com/_/postgres) (run it the same way).
+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).
 
 You can also build the image manually:
 
 ```sh
-git clone --branch v0.5.1 https://github.com/pgvector/pgvector.git
+git clone --branch v0.6.0 https://github.com/pgvector/pgvector.git
 cd pgvector
-docker build --build-arg PG_MAJOR=15 -t myuser/pgvector .
+docker build --build-arg PG_MAJOR=16 -t myuser/pgvector .
 ```
 
 ### Homebrew
@@ -707,6 +722,21 @@ sudo dnf install pgvector_16
 
 Note: Replace `16` with your Postgres server version
 
+### pkg
+
+Install the FreeBSD package with:
+
+```sh
+pkg install postgresql15-pg_vector
+```
+
+or the port with:
+
+```sh
+cd /usr/ports/databases/pg_vector
+make install
+```
+
 ### conda-forge
 
 With Conda Postgres, install from [conda-forge](https://anaconda.org/conda-forge/pgvector) with:
@@ -741,7 +771,7 @@ SELECT extversion FROM pg_extension WHERE extname = 'vector';
 
 ## Upgrade Notes
 
-### 0.6.0 [unreleased]
+### 0.6.0
 
 #### Postgres 12
 

sql/vector--0.6.1--0.7.0.sql

@@ -0,0 +1,10 @@
+-- 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 OPERATOR CLASS vector_integer_ops
+	DEFAULT FOR TYPE integer USING hnsw AS
+	OPERATOR 2 = (integer, integer);
+
+CREATE OPERATOR CLASS vector_text_ops
+	DEFAULT FOR TYPE text USING hnsw AS
+	OPERATOR 2 = (text, text);

sql/vector.sql

@@ -290,3 +290,13 @@ 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);
+
+-- hnsw attributes
+
+CREATE OPERATOR CLASS vector_integer_ops
+	DEFAULT FOR TYPE integer USING hnsw AS
+	OPERATOR 2 = (integer, integer);
+
+CREATE OPERATOR CLASS vector_text_ops
+	DEFAULT FOR TYPE text USING hnsw AS
+	OPERATOR 2 = (text, text);

src/hnsw.c

@@ -200,7 +200,7 @@ hnswhandler(PG_FUNCTION_ARGS)
 	amroutine->amcanorderbyop = true;
 	amroutine->amcanbackward = false;	/* can change direction mid-scan */
 	amroutine->amcanunique = false;
-	amroutine->amcanmulticol = false;
+	amroutine->amcanmulticol = true;
 	amroutine->amoptionalkey = true;
 	amroutine->amsearcharray = false;
 	amroutine->amsearchnulls = false;

src/hnsw.h

@@ -60,6 +60,7 @@
 #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))
@@ -127,6 +128,7 @@ HnswPtrDeclare(HnswElementData, HnswElementRelptr, HnswElementPtr);
 HnswPtrDeclare(HnswNeighborArray, HnswNeighborArrayRelptr, HnswNeighborArrayPtr);
 HnswPtrDeclare(HnswNeighborArrayPtr, HnswNeighborsRelptr, HnswNeighborsPtr);
 HnswPtrDeclare(char, DatumRelptr, DatumPtr);
+HnswPtrDeclare(IndexTupleData, IndexTupleRelptr, IndexTuplePtr);
 
 typedef struct HnswElementData
 {
@@ -142,6 +144,7 @@ typedef struct HnswElementData
 	OffsetNumber neighborOffno;
 	BlockNumber neighborPage;
 	DatumPtr	value;
+	IndexTuplePtr itup;
 	LWLock		lock;
 }			HnswElementData;
 
@@ -354,6 +357,7 @@ typedef struct HnswVacuumState
 	/* Variables */
 	struct tidhash_hash *deleted;
 	BufferAccessStrategy bas;
+	HnswNeighborTuple ntup;
 	HnswElementData highestPoint;
 
 	/* Memory */
@@ -368,7 +372,7 @@ bool		HnswNormValue(FmgrInfo *procinfo, Oid collation, Datum *value, Vector * re
 Buffer		HnswNewBuffer(Relation index, ForkNumber forkNum);
 void		HnswInitPage(Buffer buf, Page page);
 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(char *base, Datum q, IndexScanDesc scan, List *ep, int ef, int lc, Relation index, FmgrInfo *procinfo, Oid collation, int m, bool inserting, HnswElement skipElement);
 HnswElement HnswGetEntryPoint(Relation index);
 void		HnswGetMetaPageInfo(Relation index, int *m, HnswElement * entryPoint);
 void	   *HnswAlloc(HnswAllocator * allocator, Size size);
@@ -382,11 +386,14 @@ 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		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		HnswLoadElementFromTuple(HnswElement element, HnswElementTuple etup, bool loadHeaptids, bool loadVec, Relation index);
+void		HnswLoadElement(HnswElement element, float *distance, Datum *q, Relation index, FmgrInfo *procinfo, Oid collation, bool loadVec, IndexScanDesc scan, bool *matches);
+void		HnswSetElementTuple(char *base, HnswElementTuple etup, HnswElement element, bool useIndexTuple);
 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);
+TupleDesc	HnswTupleDesc(Relation index);
+IndexTuple	HnswFormIndexTuple(Relation index, TupleDesc tupdesc, Datum value, Datum *values, bool *isnull);
+bool		HnswElementIsDuplicate(char *base, HnswElement a, HnswElement b, Relation index);
 PGDLLEXPORT void HnswParallelBuildMain(dsm_segment *seg, shm_toc *toc);
 
 /* Index access methods */

src/hnswbuild.c

@@ -148,7 +148,6 @@ CreateGraphPages(HnswBuildState * buildstate)
 {
 	Relation	index = buildstate->index;
 	ForkNumber	forkNum = buildstate->forkNum;
-	Size		etupAllocSize;
 	Size		maxSize;
 	HnswElementTuple etup;
 	HnswNeighborTuple ntup;
@@ -158,14 +157,14 @@ CreateGraphPages(HnswBuildState * buildstate)
 	Page		page;
 	HnswElementPtr iter = buildstate->graph->head;
 	char	   *base = buildstate->hnswarea;
+	bool		useIndexTuple = IndexRelationGetNumberOfAttributes(index) > 1;
 
 	/* Calculate sizes */
-	etupAllocSize = BLCKSZ;
 	maxSize = HNSW_MAX_SIZE;
 
 	/* Allocate once */
-	etup = palloc0(etupAllocSize);
-	ntup = palloc0(BLCKSZ);
+	etup = palloc0(HNSW_TUPLE_ALLOC_SIZE);
+	ntup = palloc0(HNSW_TUPLE_ALLOC_SIZE);
 
 	/* Prepare first page */
 	buf = HnswNewBuffer(index, forkNum);
@@ -178,24 +177,25 @@ CreateGraphPages(HnswBuildState * buildstate)
 		Size		etupSize;
 		Size		ntupSize;
 		Size		combinedSize;
-		void	   *valuePtr = HnswPtrAccess(base, element->value);
+		Pointer		valuePtr = HnswPtrAccess(base, element->value);
+		IndexTuple	itup = HnswPtrAccess(base, element->itup);
 
 		/* Update iterator */
 		iter = element->next;
 
 		/* Zero memory for each element */
-		MemSet(etup, 0, etupAllocSize);
+		MemSet(etup, 0, HNSW_TUPLE_ALLOC_SIZE);
 
 		/* Calculate sizes */
-		etupSize = HNSW_ELEMENT_TUPLE_SIZE(VARSIZE_ANY(valuePtr));
+		etupSize = HNSW_ELEMENT_TUPLE_SIZE(useIndexTuple ? IndexTupleSize(itup) : VARSIZE_ANY(valuePtr));
 		ntupSize = HNSW_NEIGHBOR_TUPLE_SIZE(element->level, buildstate->m);
 		combinedSize = etupSize + ntupSize + sizeof(ItemIdData);
 
 		/* Initial size check */
-		if (etupSize > etupAllocSize)
+		if (etupSize > HNSW_TUPLE_ALLOC_SIZE)
 			elog(ERROR, "index tuple too large");
 
-		HnswSetElementTuple(base, etup, element);
+		HnswSetElementTuple(base, etup, element, useIndexTuple);
 
 		/* Keep element and neighbors on the same page if possible */
 		if (PageGetFreeSpace(page) < etupSize || (combinedSize <= maxSize && PageGetFreeSpace(page) < combinedSize))
@@ -254,38 +254,43 @@ WriteNeighborTuples(HnswBuildState * buildstate)
 	int			m = buildstate->m;
 	HnswElementPtr iter = buildstate->graph->head;
 	char	   *base = buildstate->hnswarea;
+	HnswNeighborTuple ntup;
+
+	/* Allocate once */
+	ntup = palloc0(HNSW_TUPLE_ALLOC_SIZE);
 
 	while (!HnswPtrIsNull(base, iter))
 	{
-		HnswElement e = HnswPtrAccess(base, iter);
+		HnswElement element = HnswPtrAccess(base, iter);
 		Buffer		buf;
 		Page		page;
-		ItemId		itemid;
-		HnswNeighborTuple ntup;
+		Size		ntupSize = HNSW_NEIGHBOR_TUPLE_SIZE(element->level, m);
 
 		/* Update iterator */
-		iter = e->next;
+		iter = element->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, e->neighborPage, RBM_NORMAL, NULL);
+		buf = ReadBufferExtended(index, forkNum, element->neighborPage, RBM_NORMAL, NULL);
 		LockBuffer(buf, BUFFER_LOCK_EXCLUSIVE);
 		page = BufferGetPage(buf);
-		itemid = PageGetItemId(page, e->neighborOffno);
-		ntup = (HnswNeighborTuple) PageGetItem(page, itemid);
 
-		/* Check expected size */
-		Assert(ItemIdGetLength(itemid) == HNSW_NEIGHBOR_TUPLE_SIZE(e->level, m));
+		HnswSetNeighborTuple(base, ntup, element, m);
 
-		/* Update page in-place */
-		HnswSetNeighborTuple(base, ntup, e, m);
+		if (!PageIndexTupleOverwrite(page, element->neighborOffno, (Item) ntup, ntupSize))
+			elog(ERROR, "failed to add index item to \"%s\"", RelationGetRelationName(index));
 
 		/* Commit */
 		MarkBufferDirty(buf);
 		UnlockReleaseBuffer(buf);
 	}
+
+	pfree(ntup);
 }
 
 /*
@@ -331,19 +336,17 @@ AddDuplicateInMemory(HnswElement element, HnswElement dup)
  * Find duplicate element
  */
 static bool
-FindDuplicateInMemory(char *base, HnswElement element)
+FindDuplicateInMemory(char *base, HnswElement element, Relation index)
 {
 	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))
+		if (!HnswElementIsDuplicate(base, element, neighborElement, index))
 			return false;
 
 		/* Check for space */
@@ -403,7 +406,7 @@ UpdateGraphInMemory(FmgrInfo *procinfo, Oid collation, HnswElement element, int
 	char	   *base = buildstate->hnswarea;
 
 	/* Look for duplicate */
-	if (FindDuplicateInMemory(base, element))
+	if (FindDuplicateInMemory(base, element, buildstate->index))
 		return;
 
 	/* Add element */
@@ -468,10 +471,14 @@ InsertTuple(Relation index, Datum *values, bool *isnull, ItemPointer heaptid, Hn
 	HnswGraph  *graph = buildstate->graph;
 	HnswElement element;
 	HnswAllocator *allocator = &buildstate->allocator;
-	Size		valueSize;
+	IndexTuple	itup;
+	Size		itupSize;
+	IndexTuple	itupPtr;
 	Pointer		valuePtr;
 	LWLock	   *flushLock = &graph->flushLock;
 	char	   *base = buildstate->hnswarea;
+	TupleDesc	tupdesc = HnswTupleDesc(index);
+	bool		unused;
 
 	/* Detoast once for all calls */
 	Datum		value = PointerGetDatum(PG_DETOAST_DATUM(values[0]));
@@ -484,7 +491,8 @@ InsertTuple(Relation index, Datum *values, bool *isnull, ItemPointer heaptid, Hn
 	}
 
 	/* Get datum size */
-	valueSize = VARSIZE_ANY(DatumGetPointer(value));
+	itup = HnswFormIndexTuple(index, tupdesc, value, values, isnull);
+	itupSize = IndexTupleSize(itup);
 
 	/* Ensure graph not flushed when inserting */
 	LWLockAcquire(flushLock, LW_SHARED);
@@ -531,7 +539,7 @@ InsertTuple(Relation index, Datum *values, bool *isnull, ItemPointer heaptid, Hn
 
 	/* Ok, we can proceed to allocate the element */
 	element = HnswInitElement(base, heaptid, buildstate->m, buildstate->ml, buildstate->maxLevel, allocator);
-	valuePtr = HnswAlloc(allocator, valueSize);
+	itupPtr = HnswAlloc(allocator, itupSize);
 
 	/*
 	 * We have now allocated the space needed for the element, so we don't
@@ -540,8 +548,10 @@ InsertTuple(Relation index, Datum *values, bool *isnull, ItemPointer heaptid, Hn
 	 */
 	LWLockRelease(&graph->allocatorLock);
 
-	/* Copy the datum */
-	memcpy(valuePtr, DatumGetPointer(value), valueSize);
+	/* Copy the index tuple */
+	memcpy(itupPtr, itup, itupSize);
+	HnswPtrStore(base, element->itup, itupPtr);
+	valuePtr = DatumGetPointer(index_getattr(itupPtr, 1, tupdesc, &unused));
 	HnswPtrStore(base, element->value, valuePtr);
 
 	/* Create a lock for the element */
@@ -666,6 +676,19 @@ InitBuildState(HnswBuildState * buildstate, Relation heap, Relation index, Index
 	buildstate->efConstruction = HnswGetEfConstruction(index);
 	buildstate->dimensions = TupleDescAttr(index->rd_att, 0)->atttypmod;
 
+	/* For now */
+	if (IndexRelationGetNumberOfKeyAttributes(index) > 2)
+		elog(ERROR, "index cannot have more than two columns");
+
+	if (!OidIsValid(index_getprocid(index, 1, HNSW_DISTANCE_PROC)))
+		elog(ERROR, "first column must be a vector");
+
+	for (int i = 1; i < IndexRelationGetNumberOfKeyAttributes(index); i++)
+	{
+		if (OidIsValid(index_getprocid(index, i + 1, HNSW_DISTANCE_PROC)))
+			elog(ERROR, "column %d cannot be a vector", i + 1);
+	}
+
 	/* Require column to have dimensions to be indexed */
 	if (buildstate->dimensions < 0)
 		elog(ERROR, "column does not have dimensions");

src/hnswinsert.c

@@ -136,9 +136,12 @@ AddElementOnDisk(Relation index, HnswElement e, int m, BlockNumber insertPage, B
 	OffsetNumber freeNeighborOffno = InvalidOffsetNumber;
 	BlockNumber newInsertPage = InvalidBlockNumber;
 	char	   *base = NULL;
+	bool		useIndexTuple = IndexRelationGetNumberOfAttributes(index) > 1;
+	Pointer		valuePtr = HnswPtrAccess(base, e->value);
+	IndexTuple	itup = HnswPtrAccess(base, e->itup);
 
 	/* Calculate sizes */
-	etupSize = HNSW_ELEMENT_TUPLE_SIZE(VARSIZE_ANY(HnswPtrAccess(base, e->value)));
+	etupSize = HNSW_ELEMENT_TUPLE_SIZE(useIndexTuple ? IndexTupleSize(itup) : VARSIZE_ANY(valuePtr));
 	ntupSize = HNSW_NEIGHBOR_TUPLE_SIZE(e->level, m);
 	combinedSize = etupSize + ntupSize + sizeof(ItemIdData);
 	maxSize = HNSW_MAX_SIZE;
@@ -146,7 +149,7 @@ AddElementOnDisk(Relation index, HnswElement e, int m, BlockNumber insertPage, B
 
 	/* Prepare element tuple */
 	etup = palloc0(etupSize);
-	HnswSetElementTuple(base, etup, e);
+	HnswSetElementTuple(base, etup, e, useIndexTuple);
 
 	/* Prepare neighbor tuple */
 	ntup = palloc0(ntupSize);
@@ -355,9 +358,7 @@ 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);
@@ -396,9 +397,7 @@ HnswUpdateNeighborsOnDisk(Relation index, FmgrInfo *procinfo, Oid collation, Hns
 			}
 
 			/* Get tuple */
-			itemid = PageGetItemId(page, offno);
-			ntup = (HnswNeighborTuple) PageGetItem(page, itemid);
-			ntupSize = ItemIdGetLength(itemid);
+			ntup = (HnswNeighborTuple) PageGetItem(page, PageGetItemId(page, offno));
 
 			/* Calculate index for update */
 			startIdx = (neighborElement->level - lc) * m;
@@ -427,13 +426,9 @@ HnswUpdateNeighborsOnDisk(Relation index, FmgrInfo *procinfo, Oid collation, Hns
 			{
 				ItemPointer indextid = &ntup->indextids[idx];
 
-				/* Update neighbor */
+				/* Update neighbor on the buffer */
 				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);
@@ -457,9 +452,7 @@ 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 */
@@ -477,9 +470,7 @@ AddDuplicateOnDisk(Relation index, HnswElement element, HnswElement dup, bool bu
 	}
 
 	/* Find space */
-	itemid = PageGetItemId(page, dup->offno);
-	etup = (HnswElementTuple) PageGetItem(page, itemid);
-	etupSize = ItemIdGetLength(itemid);
+	etup = (HnswElementTuple) PageGetItem(page, PageGetItemId(page, dup->offno));
 	for (i = 0; i < HNSW_HEAPTIDS; i++)
 	{
 		if (!ItemPointerIsValid(&etup->heaptids[i]))
@@ -495,13 +486,9 @@ AddDuplicateOnDisk(Relation index, HnswElement element, HnswElement dup, bool bu
 		return false;
 	}
 
-	/* Add heap TID */
+	/* Add heap TID, modifying the tuple on the page directly */
 	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);
@@ -520,16 +507,14 @@ 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))
+		if (!HnswElementIsDuplicate(base, element, neighborElement, index))
 			return false;
 
 		if (AddDuplicateOnDisk(index, element, neighborElement, building))
@@ -580,6 +565,10 @@ HnswInsertTupleOnDisk(Relation index, Datum value, Datum *values, bool *isnull,
 	Oid			collation = index->rd_indcollation[0];
 	LOCKMODE	lockmode = ShareLock;
 	char	   *base = NULL;
+	TupleDesc	tupdesc = HnswTupleDesc(index);
+	bool		unused;
+	IndexTuple	itup;
+	Pointer		valuePtr;
 
 	/*
 	 * Get a shared lock. This allows vacuum to ensure no in-flight inserts
@@ -593,7 +582,10 @@ HnswInsertTupleOnDisk(Relation index, Datum value, Datum *values, bool *isnull,
 
 	/* Create an element */
 	element = HnswInitElement(base, heap_tid, m, HnswGetMl(m), HnswGetMaxLevel(m), NULL);
-	HnswPtrStore(base, element->value, DatumGetPointer(value));
+	itup = HnswFormIndexTuple(index, tupdesc, value, values, isnull);
+	HnswPtrStore(base, element->itup, itup);
+	valuePtr = DatumGetPointer(index_getattr(itup, 1, tupdesc, &unused));
+	HnswPtrStore(base, element->value, valuePtr);
 
 	/* Prevent concurrent inserts when likely updating entry point */
 	if (entryPoint == NULL || element->level > entryPoint->level)

src/hnswscan.c

@@ -33,11 +33,12 @@ GetScanItems(IndexScanDesc scan, Datum q)
 
 	for (int lc = entryPoint->level; lc >= 1; lc--)
 	{
-		w = HnswSearchLayer(base, q, ep, 1, lc, index, procinfo, collation, m, false, NULL);
+		w = HnswSearchLayer(base, q, NULL, ep, 1, lc, index, procinfo, collation, m, false, NULL);
 		ep = w;
 	}
 
-	return HnswSearchLayer(base, q, ep, hnsw_ef_search, 0, index, procinfo, collation, m, false, NULL);
+	/* Only pass scan to check matches for layer 0 */
+	return HnswSearchLayer(base, q, scan, ep, hnsw_ef_search, 0, index, procinfo, collation, m, false, NULL);
 }
 
 /*
@@ -202,7 +203,7 @@ hnswgettuple(IndexScanDesc scan, ScanDirection dir)
 		MemoryContextSwitchTo(oldCtx);
 
 		scan->xs_heaptid = *heaptid;
-		scan->xs_recheck = false;
+		scan->xs_recheck = scan->numberOfKeys > 0;
 		scan->xs_recheckorderby = false;
 		return true;
 	}

src/hnswutils.c

@@ -3,6 +3,7 @@
 #include <math.h>
 
 #include "access/generic_xlog.h"
+#include "access/relscan.h"
 #include "hnsw.h"
 #include "lib/pairingheap.h"
 #include "storage/bufmgr.h"
@@ -11,6 +12,10 @@
 #include "utils/rel.h"
 #include "vector.h"
 
+#if PG_VERSION_NUM < 130000
+#define TYPSTORAGE_PLAIN 'p'
+#endif
+
 #if PG_VERSION_NUM >= 130000
 #include "common/hashfn.h"
 #else
@@ -295,6 +300,82 @@ HnswInitElementFromBlock(BlockNumber blkno, OffsetNumber offno)
 	return element;
 }
 
+/*
+ * Get the tuple descriptor
+ */
+TupleDesc
+HnswTupleDesc(Relation index)
+{
+	TupleDesc	tupdesc = CreateTupleDescCopyConstr(RelationGetDescr(index));
+
+	/* Prevent compression */
+	TupleDescAttr(tupdesc, 0)->attstorage = TYPSTORAGE_PLAIN;
+
+	return tupdesc;
+}
+
+/*
+ * Form an index tuple
+ */
+IndexTuple
+HnswFormIndexTuple(Relation index, TupleDesc tupdesc, Datum value, Datum *values, bool *isnull)
+{
+	Size		size = sizeof(Datum) * IndexRelationGetNumberOfAttributes(index);
+	Datum	   *newValues = palloc(size);
+
+	memcpy(newValues, values, size);
+	newValues[0] = value;
+
+	return index_form_tuple(tupdesc, newValues, isnull);
+}
+
+/*
+ * Check if elements are duplicates
+ */
+bool
+HnswElementIsDuplicate(char *base, HnswElement a, HnswElement b, Relation index)
+{
+	if (IndexRelationGetNumberOfAttributes(index) == 1)
+	{
+		Datum		value = HnswGetValue(base, a);
+		Datum		value2 = HnswGetValue(base, b);
+
+		return datumIsEqual(value, value2, false, -1);
+	}
+	else
+	{
+		TupleDesc	tupdesc = RelationGetDescr(index);
+		IndexTuple	itup = HnswPtrAccess(base, a->itup);
+		IndexTuple	itup2 = HnswPtrAccess(base, b->itup);
+
+		for (int i = 0; i < tupdesc->natts; i++)
+		{
+			Datum		value;
+			Datum		value2;
+			bool		isnull;
+			bool		isnull2;
+
+			value = index_getattr(itup, i + 1, tupdesc, &isnull);
+			value2 = index_getattr(itup2, i + 1, tupdesc, &isnull2);
+
+			if (isnull || isnull2)
+			{
+				if (isnull != isnull2)
+					return false;
+			}
+			else
+			{
+				Form_pg_attribute att = TupleDescAttr(tupdesc, i);
+
+				if (!datumIsEqual(value, value2, att->attbyval, att->attlen))
+					return false;
+			}
+		}
+
+		return true;
+	}
+}
+
 /*
  * Get the metapage info
  */
@@ -404,10 +485,8 @@ HnswUpdateMetaPage(Relation index, int updateEntry, HnswElement entryPoint, Bloc
  * Set element tuple, except for neighbor info
  */
 void
-HnswSetElementTuple(char *base, HnswElementTuple etup, HnswElement element)
+HnswSetElementTuple(char *base, HnswElementTuple etup, HnswElement element, bool useIndexTuple)
 {
-	Pointer		valuePtr = HnswPtrAccess(base, element->value);
-
 	etup->type = HNSW_ELEMENT_TUPLE_TYPE;
 	etup->level = element->level;
 	etup->deleted = 0;
@@ -418,7 +497,19 @@ HnswSetElementTuple(char *base, HnswElementTuple etup, HnswElement element)
 		else
 			ItemPointerSetInvalid(&etup->heaptids[i]);
 	}
-	memcpy(&etup->data, valuePtr, VARSIZE_ANY(valuePtr));
+
+	if (useIndexTuple)
+	{
+		IndexTuple	itup = HnswPtrAccess(base, element->itup);
+
+		memcpy(&etup->data, itup, IndexTupleSize(itup));
+	}
+	else
+	{
+		Pointer		valuePtr = HnswPtrAccess(base, element->value);
+
+		memcpy(&etup->data, valuePtr, VARSIZE_ANY(valuePtr));
+	}
 }
 
 /*
@@ -522,7 +613,7 @@ HnswLoadNeighbors(HnswElement element, Relation index, int m)
  * Load an element from a tuple
  */
 void
-HnswLoadElementFromTuple(HnswElement element, HnswElementTuple etup, bool loadHeaptids, bool loadVec)
+HnswLoadElementFromTuple(HnswElement element, HnswElementTuple etup, bool loadHeaptids, bool loadVec, Relation index)
 {
 	element->level = etup->level;
 	element->deleted = etup->deleted;
@@ -545,17 +636,64 @@ HnswLoadElementFromTuple(HnswElement element, HnswElementTuple etup, bool loadHe
 	if (loadVec)
 	{
 		char	   *base = NULL;
-		Datum		value = datumCopy(PointerGetDatum(&etup->data), false, -1);
 
-		HnswPtrStore(base, element->value, DatumGetPointer(value));
+		if (IndexRelationGetNumberOfAttributes(index) > 1)
+		{
+			TupleDesc	tupdesc = RelationGetDescr(index);
+			bool		unused;
+			IndexTuple	itup = CopyIndexTuple((IndexTuple) &etup->data);
+			Datum		value = index_getattr(itup, 1, tupdesc, &unused);
+
+			HnswPtrStore(base, element->itup, itup);
+			HnswPtrStore(base, element->value, DatumGetPointer(value));
+		}
+		else
+		{
+			Datum		value = datumCopy(PointerGetDatum(&etup->data), false, -1);
+
+			HnswPtrStore(base, element->value, DatumGetPointer(value));
+		}
 	}
 }
 
+/*
+ * Check if an element matches
+ */
+static bool
+HnswCheckMatches(Relation index, HnswElementTuple etup, IndexScanDesc scan)
+{
+	if (scan == NULL)
+		return true;
+
+	for (int i = 0; i < scan->numberOfKeys; i++)
+	{
+		IndexTuple	itup = (IndexTuple) &etup->data;
+		TupleDesc	tupdesc = RelationGetDescr(index);
+		ScanKey		key = &scan->keyData[i];
+		bool		isnull;
+		Datum		value = index_getattr(itup, key->sk_attno, tupdesc, &isnull);
+		bool		attnull = key->sk_flags & SK_ISNULL;
+
+		if (isnull || attnull)
+		{
+			if (isnull != attnull)
+				return false;
+		}
+		else
+		{
+			if (!DatumGetBool(FunctionCall2Coll(&key->sk_func, key->sk_collation, value, key->sk_argument)))
+				return false;
+		}
+	}
+
+	return true;
+}
+
 /*
  * Load an element and optionally get its distance from q
  */
 void
-HnswLoadElement(HnswElement element, float *distance, Datum *q, Relation index, FmgrInfo *procinfo, Oid collation, bool loadVec)
+HnswLoadElement(HnswElement element, float *distance, Datum *q, Relation index, FmgrInfo *procinfo, Oid collation, bool loadVec, IndexScanDesc scan, bool *matches)
 {
 	Buffer		buf;
 	Page		page;
@@ -571,11 +709,29 @@ HnswLoadElement(HnswElement element, float *distance, Datum *q, Relation index,
 	Assert(HnswIsElementTuple(etup));
 
 	/* Load element */
-	HnswLoadElementFromTuple(element, etup, true, loadVec);
+	HnswLoadElementFromTuple(element, etup, true, loadVec, index);
 
 	/* Calculate distance */
 	if (distance != NULL)
-		*distance = (float) DatumGetFloat8(FunctionCall2Coll(procinfo, collation, *q, PointerGetDatum(&etup->data)));
+	{
+		Datum		value;
+
+		if (IndexRelationGetNumberOfAttributes(index) > 1)
+		{
+			IndexTuple	itup = (IndexTuple) &etup->data;
+			TupleDesc	tupdesc = RelationGetDescr(index);
+			bool		unused;
+
+			value = index_getattr(itup, 1, tupdesc, &unused);
+		}
+		else
+			value = PointerGetDatum(&etup->data);
+
+		*distance = (float) DatumGetFloat8(FunctionCall2Coll(procinfo, collation, *q, value));
+	}
+
+	if (matches != NULL)
+		*matches = HnswCheckMatches(index, etup, scan);
 
 	UnlockReleaseBuffer(buf);
 }
@@ -604,7 +760,7 @@ HnswEntryCandidate(char *base, HnswElement entryPoint, Datum q, Relation index,
 	if (index == NULL)
 		hc->distance = GetCandidateDistance(base, hc, q, procinfo, collation);
 	else
-		HnswLoadElement(entryPoint, &hc->distance, &q, index, procinfo, collation, loadVec);
+		HnswLoadElement(entryPoint, &hc->distance, &q, index, procinfo, collation, loadVec, NULL, NULL);
 	return hc;
 }
 
@@ -722,12 +878,14 @@ CountElement(char *base, HnswElement skipElement, HnswCandidate * hc)
  * Algorithm 2 from paper
  */
 List *
-HnswSearchLayer(char *base, Datum q, List *ep, int ef, int lc, Relation index, FmgrInfo *procinfo, Oid collation, int m, bool inserting, HnswElement skipElement)
+HnswSearchLayer(char *base, Datum q, IndexScanDesc scan, List *ep, int ef, int lc, Relation index, FmgrInfo *procinfo, Oid collation, int m, bool inserting, HnswElement skipElement)
 {
 	List	   *w = NIL;
 	pairingheap *C = pairingheap_allocate(CompareNearestCandidates, NULL);
 	pairingheap *W = pairingheap_allocate(CompareFurthestCandidates, NULL);
 	int			wlen = 0;
+	uint64		additional = 0;
+	uint64		maxAdditional = (scan != NULL && scan->numberOfKeys > 0) ? 100 * ef : 0;
 	visited_hash v;
 	ListCell   *lc2;
 	HnswNeighborArray *neighborhoodData = NULL;
@@ -799,14 +957,18 @@ HnswSearchLayer(char *base, Datum q, List *ep, int ef, int lc, Relation index, F
 			if (!visited)
 			{
 				float		eDistance;
+				bool		eMatches;
 				HnswElement eElement = HnswPtrAccess(base, e->element);
 
 				f = ((HnswPairingHeapNode *) pairingheap_first(W))->inner;
 
 				if (index == NULL)
+				{
 					eDistance = GetCandidateDistance(base, e, q, procinfo, collation);
+					eMatches = true;
+				}
 				else
-					HnswLoadElement(eElement, &eDistance, &q, index, procinfo, collation, inserting);
+					HnswLoadElement(eElement, &eDistance, &q, index, procinfo, collation, inserting, scan, &eMatches);
 
 				Assert(!eElement->deleted);
 
@@ -825,6 +987,16 @@ HnswSearchLayer(char *base, Datum q, List *ep, int ef, int lc, Relation index, F
 					pairingheap_add(C, &(CreatePairingHeapNode(ec)->ph_node));
 					pairingheap_add(W, &(CreatePairingHeapNode(ec)->ph_node));
 
+					/*
+					 * Do not count elements that do not match filter towards
+					 * ef
+					 */
+					if (!eMatches)
+					{
+						if ((++additional) <= maxAdditional)
+							continue;
+					}
+
 					/*
 					 * Do not count elements being deleted towards ef when
 					 * vacuuming. It would be ideal to do this for inserts as
@@ -1099,7 +1271,7 @@ HnswUpdateConnection(char *base, HnswElement element, HnswCandidate * hc, int lm
 				HnswElement hc3Element = HnswPtrAccess(base, hc3->element);
 
 				if (HnswPtrIsNull(base, hc3Element->value))
-					HnswLoadElement(hc3Element, &hc3->distance, &q, index, procinfo, collation, true);
+					HnswLoadElement(hc3Element, &hc3->distance, &q, index, procinfo, collation, true, NULL, NULL);
 				else
 					hc3->distance = GetCandidateDistance(base, hc3, q, procinfo, collation);
 
@@ -1221,7 +1393,7 @@ HnswFindElementNeighbors(char *base, HnswElement element, HnswElement entryPoint
 	/* 1st phase: greedy search to insert level */
 	for (int lc = entryLevel; lc >= level + 1; lc--)
 	{
-		w = HnswSearchLayer(base, q, ep, 1, lc, index, procinfo, collation, m, true, skipElement);
+		w = HnswSearchLayer(base, q, NULL, ep, 1, lc, index, procinfo, collation, m, true, skipElement);
 		ep = w;
 	}
 
@@ -1239,7 +1411,7 @@ HnswFindElementNeighbors(char *base, HnswElement element, HnswElement entryPoint
 		List	   *neighbors;
 		List	   *lw;
 
-		w = HnswSearchLayer(base, q, ep, efConstruction, lc, index, procinfo, collation, m, true, skipElement);
+		w = HnswSearchLayer(base, q, NULL, ep, efConstruction, lc, index, procinfo, collation, m, true, skipElement);
 
 		/* Elements being deleted or skipped can help with search */
 		/* but should be removed before selecting neighbors */

src/hnswvacuum.c

@@ -60,8 +60,7 @@ RemoveHeapTids(HnswVacuumState * vacuumstate)
 		/* Iterate over nodes */
 		for (offno = FirstOffsetNumber; offno <= maxoffno; offno = OffsetNumberNext(offno))
 		{
-			ItemId		itemid = PageGetItemId(page, offno);
-			HnswElementTuple etup = (HnswElementTuple) PageGetItem(page, itemid);
+			HnswElementTuple etup = (HnswElementTuple) PageGetItem(page, PageGetItemId(page, offno));
 			int			idx = 0;
 			bool		itemUpdated = false;
 
@@ -92,15 +91,10 @@ 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;
 				}
 			}
@@ -198,8 +192,8 @@ RepairGraphElement(HnswVacuumState * vacuumstate, HnswElement element, HnswEleme
 	FmgrInfo   *procinfo = vacuumstate->procinfo;
 	Oid			collation = vacuumstate->collation;
 	BufferAccessStrategy bas = vacuumstate->bas;
-	ItemId		itemid;
-	HnswNeighborTuple ntup;
+	HnswNeighborTuple ntup = vacuumstate->ntup;
+	Size		ntupSize = HNSW_NEIGHBOR_TUPLE_SIZE(element->level, m);
 	char	   *base = NULL;
 
 	/* Skip if element is entry point */
@@ -213,23 +207,23 @@ 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);
+
 	/* Get neighbor page */
 	buf = ReadBufferExtended(index, MAIN_FORKNUM, element->neighborPage, RBM_NORMAL, bas);
 	LockBuffer(buf, BUFFER_LOCK_EXCLUSIVE);
 	state = GenericXLogStart(index);
 	page = GenericXLogRegisterBuffer(state, buf, 0);
 
-	/* Get neighbor tuple */
-	itemid = PageGetItemId(page, element->neighborOffno);
-	ntup = (HnswNeighborTuple) PageGetItem(page, itemid);
-
-	/* Check expected size */
-	if (ItemIdGetLength(itemid) != HNSW_NEIGHBOR_TUPLE_SIZE(element->level, m))
+	/* Overwrite tuple */
+	if (!PageIndexTupleOverwrite(page, element->neighborOffno, (Item) ntup, ntupSize))
 		elog(ERROR, "failed to add index item to \"%s\"", RelationGetRelationName(index));
 
-	/* Update page in-place */
-	HnswSetNeighborTuple(base, ntup, element, m);
-
 	/* Commit */
 	GenericXLogFinish(state);
 	UnlockReleaseBuffer(buf);
@@ -262,7 +256,7 @@ RepairGraphEntryPoint(HnswVacuumState * vacuumstate)
 		LockPage(index, HNSW_UPDATE_LOCK, ShareLock);
 
 		/* Load element */
-		HnswLoadElement(highestPoint, NULL, NULL, index, vacuumstate->procinfo, vacuumstate->collation, true);
+		HnswLoadElement(highestPoint, NULL, NULL, index, vacuumstate->procinfo, vacuumstate->collation, true, NULL, NULL);
 
 		/* Repair if needed */
 		if (NeedsUpdated(vacuumstate, highestPoint))
@@ -300,7 +294,7 @@ RepairGraphEntryPoint(HnswVacuumState * vacuumstate)
 			 * is outdated, this can remove connections at higher levels in
 			 * the graph until they are repaired, but this should be fine.
 			 */
-			HnswLoadElement(entryPoint, NULL, NULL, index, vacuumstate->procinfo, vacuumstate->collation, true);
+			HnswLoadElement(entryPoint, NULL, NULL, index, vacuumstate->procinfo, vacuumstate->collation, true, NULL, NULL);
 
 			if (NeedsUpdated(vacuumstate, entryPoint))
 			{
@@ -376,7 +370,7 @@ RepairGraph(HnswVacuumState * vacuumstate)
 
 			/* Create an element */
 			element = HnswInitElementFromBlock(blkno, offno);
-			HnswLoadElementFromTuple(element, etup, false, true);
+			HnswLoadElementFromTuple(element, etup, false, true, index);
 
 			elements = lappend(elements, element);
 		}
@@ -446,6 +440,7 @@ MarkDeleted(HnswVacuumState * vacuumstate)
 	BlockNumber insertPage = InvalidBlockNumber;
 	Relation	index = vacuumstate->index;
 	BufferAccessStrategy bas = vacuumstate->bas;
+	bool		useIndexTuple = IndexRelationGetNumberOfAttributes(index);
 
 	/*
 	 * Wait for index scans to complete. Scans before this point may contain
@@ -482,11 +477,8 @@ MarkDeleted(HnswVacuumState * vacuumstate)
 		/* Update element and neighbors together */
 		for (offno = FirstOffsetNumber; offno <= maxoffno; offno = OffsetNumberNext(offno))
 		{
-			ItemId		itemid = PageGetItemId(page, offno);
-			HnswElementTuple etup = (HnswElementTuple) PageGetItem(page, itemid);
+			HnswElementTuple etup = (HnswElementTuple) PageGetItem(page, PageGetItemId(page, offno));
 			HnswNeighborTuple ntup;
-			Size		etupSize;
-			Size		ntupSize;
 			Buffer		nbuf;
 			Page		npage;
 			BlockNumber neighborPage;
@@ -510,10 +502,6 @@ 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);
@@ -534,19 +522,27 @@ MarkDeleted(HnswVacuumState * vacuumstate)
 
 			/* Overwrite element */
 			etup->deleted = 1;
-			MemSet(&etup->data, 0, VARSIZE_ANY(&etup->data));
+			if (useIndexTuple)
+			{
+				IndexTuple	itup = (IndexTuple) &etup->data;
+
+				MemSet(itup, 0, IndexTupleSize(itup));
+			}
+			else
+			{
+				Vector	   *vec = (Vector *) (&etup->data);
+
+				MemSet(vec, 0, VARSIZE_ANY(vec));
+			}
 
 			/* Overwrite neighbors */
 			for (int i = 0; i < ntup->count; i++)
 				ItemPointerSetInvalid(&ntup->indextids[i]);
 
-			/* 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));
+			/*
+			 * We modified the tuples in place, no need to call
+			 * PageIndexTupleOverwrite
+			 */
 
 			/* Commit */
 			GenericXLogFinish(state);
@@ -591,6 +587,7 @@ 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->tmpCtx = AllocSetContextCreate(CurrentMemoryContext,
 												"Hnsw vacuum temporary context",
 												ALLOCSET_DEFAULT_SIZES);
@@ -610,6 +607,7 @@ FreeVacuumState(HnswVacuumState * vacuumstate)
 {
 	tidhash_destroy(vacuumstate->deleted);
 	FreeAccessStrategy(vacuumstate->bas);
+	pfree(vacuumstate->ntup);
 	MemoryContextDelete(vacuumstate->tmpCtx);
 }
 

src/vector.c

@@ -866,9 +866,10 @@ vector_mul(PG_FUNCTION_ARGS)
 int
 vector_cmp_internal(Vector * a, Vector * b)
 {
-	CheckDims(a, b);
+	int			dim = Min(a->dim, b->dim);
 
-	for (int i = 0; i < a->dim; i++)
+	/* Check values before dimensions to be consistent with Postgres arrays */
+	for (int i = 0; i < dim; i++)
 	{
 		if (a->x[i] < b->x[i])
 			return -1;
@@ -876,6 +877,13 @@ 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;
 }
 
@@ -889,6 +897,9 @@ vector_lt(PG_FUNCTION_ARGS)
 	Vector	   *a = PG_GETARG_VECTOR_P(0);
 	Vector	   *b = PG_GETARG_VECTOR_P(1);
 
+	/* TODO Remove in 0.7.0 */
+	CheckDims(a, b);
+
 	PG_RETURN_BOOL(vector_cmp_internal(a, b) < 0);
 }
 
@@ -902,6 +913,9 @@ vector_le(PG_FUNCTION_ARGS)
 	Vector	   *a = PG_GETARG_VECTOR_P(0);
 	Vector	   *b = PG_GETARG_VECTOR_P(1);
 
+	/* TODO Remove in 0.7.0 */
+	CheckDims(a, b);
+
 	PG_RETURN_BOOL(vector_cmp_internal(a, b) <= 0);
 }
 
@@ -915,6 +929,9 @@ vector_eq(PG_FUNCTION_ARGS)
 	Vector	   *a = PG_GETARG_VECTOR_P(0);
 	Vector	   *b = PG_GETARG_VECTOR_P(1);
 
+	/* TODO Remove in 0.7.0 */
+	CheckDims(a, b);
+
 	PG_RETURN_BOOL(vector_cmp_internal(a, b) == 0);
 }
 
@@ -928,6 +945,9 @@ vector_ne(PG_FUNCTION_ARGS)
 	Vector	   *a = PG_GETARG_VECTOR_P(0);
 	Vector	   *b = PG_GETARG_VECTOR_P(1);
 
+	/* TODO Remove in 0.7.0 */
+	CheckDims(a, b);
+
 	PG_RETURN_BOOL(vector_cmp_internal(a, b) != 0);
 }
 
@@ -941,6 +961,9 @@ vector_ge(PG_FUNCTION_ARGS)
 	Vector	   *a = PG_GETARG_VECTOR_P(0);
 	Vector	   *b = PG_GETARG_VECTOR_P(1);
 
+	/* TODO Remove in 0.7.0 */
+	CheckDims(a, b);
+
 	PG_RETURN_BOOL(vector_cmp_internal(a, b) >= 0);
 }
 
@@ -954,6 +977,9 @@ vector_gt(PG_FUNCTION_ARGS)
 	Vector	   *a = PG_GETARG_VECTOR_P(0);
 	Vector	   *b = PG_GETARG_VECTOR_P(1);
 
+	/* TODO Remove in 0.7.0 */
+	CheckDims(a, b);
+
 	PG_RETURN_BOOL(vector_cmp_internal(a, b) > 0);
 }
 

test/expected/functions.out

@@ -24,6 +24,56 @@ 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 
 -------------

test/sql/functions.sql

@@ -6,6 +6,17 @@ 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);

test/t/020_hnsw_filtering.pl

@@ -0,0 +1,113 @@
+use strict;
+use warnings;
+use PostgresNode;
+use TestLib;
+use Test::More;
+
+my $node;
+my @queries = ();
+my @cs = ();
+my @expected;
+my $limit = 20;
+my $dim = 3;
+my $array_sql = join(",", ('random()') x $dim);
+my $nc = 50;
+
+sub test_recall
+{
+	my ($min, $operator) = @_;
+	my $correct = 0;
+	my $total = 0;
+
+	my $explain = $node->safe_psql("postgres", qq(
+		SET enable_seqscan = off;
+		EXPLAIN ANALYZE SELECT i FROM tst WHERE c = '$cs[0]' ORDER BY v $operator '$queries[0]' LIMIT $limit;
+	));
+	like($explain, qr/Index Cond/);
+
+	for my $i (0 .. $#queries)
+	{
+		my $actual = $node->safe_psql("postgres", qq(
+			SET enable_seqscan = off;
+			SELECT i FROM tst WHERE c = '$cs[$i]' ORDER BY v $operator '$queries[$i]' LIMIT $limit;
+		));
+		my @actual_ids = split("\n", $actual);
+		my %actual_set = map { $_ => 1 } @actual_ids;
+
+		is(scalar(@actual_ids), $limit);
+
+		my @expected_ids = split("\n", $expected[$i]);
+
+		foreach (@expected_ids)
+		{
+			if (exists($actual_set{$_}))
+			{
+				$correct++;
+			}
+			$total++;
+		}
+	}
+
+	cmp_ok($correct / $total, ">=", $min, $operator);
+}
+
+# Initialize node
+$node = get_new_node('node');
+$node->init;
+$node->start;
+
+# Create table
+$node->safe_psql("postgres", "CREATE EXTENSION vector;");
+$node->safe_psql("postgres", "CREATE TABLE tst (i int4, v vector($dim), c text);");
+$node->safe_psql("postgres",
+	"INSERT INTO tst SELECT i, ARRAY[$array_sql], i % $nc FROM generate_series(1, 10000) i;"
+);
+$node->safe_psql("postgres", "CREATE INDEX ON tst USING hnsw (v vector_l2_ops, c);");
+$node->safe_psql("postgres",
+	"INSERT INTO tst SELECT i, ARRAY[$array_sql], (i % $nc)::text FROM generate_series(1, 10000) i;"
+);
+
+# Generate queries
+for (1 .. 20)
+{
+	my @r = ();
+	for (1 .. $dim)
+	{
+		push(@r, rand());
+	}
+	push(@queries, "[" . join(",", @r) . "]");
+	push(@cs, int(rand() * $nc));
+}
+
+# Get exact results
+@expected = ();
+for my $i (0 .. $#queries)
+{
+	my $res = $node->safe_psql("postgres", qq(
+		SET enable_indexscan = off;
+		SELECT i FROM tst WHERE c = '$cs[$i]' ORDER BY v <-> '$queries[$i]' LIMIT $limit;
+	));
+	push(@expected, $res);
+}
+
+# Test recall
+test_recall(0.99, '<->');
+
+# Test vacuum
+$node->safe_psql("postgres", "DELETE FROM tst WHERE c > '5';");
+$node->safe_psql("postgres", "VACUUM tst;");
+
+# Test columns
+my ($ret, $stdout, $stderr) = $node->psql("postgres", "CREATE INDEX ON tst USING hnsw (c);");
+like($stderr, qr/first column must be a vector/);
+
+($ret, $stdout, $stderr) = $node->psql("postgres", "CREATE INDEX ON tst USING hnsw (c, v vector_l2_ops);");
+like($stderr, qr/first column must be a vector/);
+
+($ret, $stdout, $stderr) = $node->psql("postgres", "CREATE INDEX ON tst USING hnsw (v vector_l2_ops, c, c);");
+like($stderr, qr/index cannot have more than two columns/);
+
+($ret, $stdout, $stderr) = $node->psql("postgres", "CREATE INDEX ON tst USING hnsw (v vector_l2_ops, v vector_l2_ops);");
+like($stderr, qr/column 2 cannot be a vector/);
+
+done_testing();

vector.control

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