Version bump to 0.2.6 [skip ci]

.editorconfig

@@ -1,6 +1,6 @@
 root = true
 
-[*.{c,h}]
+[*.{c,h,pl}]
 indent_style = tab
 indent_size = tab
 tab_width = 4

CHANGELOG.md

@@ -1,3 +1,13 @@
+## 0.2.6 (2022-05-22)
+
+- Improved performance of index creation for Postgres < 12
+
+## 0.2.5 (2022-02-11)
+
+- Reduced memory usage during index creation
+- Fixed index creation exceeding `maintenance_work_mem`
+- Fixed error with index creation when lists > 1600
+
 ## 0.2.4 (2022-02-06)
 
 - Added support for parallel vacuum

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": "0.2.4",
+	"version": "0.2.6",
 	"maintainer": [
 		"Andrew Kane <andrew@ankane.org>"
 	],
@@ -20,7 +20,7 @@
 		"vector": {
 			"file": "sql/vector.sql",
 			"docfile": "README.md",
-			"version": "0.2.4",
+			"version": "0.2.6",
 			"abstract": "Open-source vector similarity search for Postgres"
 		}
 	},

Makefile

@@ -1,5 +1,5 @@
 EXTENSION = vector
-EXTVERSION = 0.2.4
+EXTVERSION = 0.2.6
 
 MODULE_big = vector
 DATA = $(wildcard sql/*--*.sql)
@@ -21,7 +21,7 @@ endif
 # For auto-vectorization:
 # - GCC (needs -ftree-vectorize OR -O3) - https://gcc.gnu.org/projects/tree-ssa/vectorization.html
 # - Clang (could use pragma instead) - https://llvm.org/docs/Vectorizers.html
-PG_CFLAGS = $(OPTFLAGS) -ftree-vectorize -fassociative-math -fno-signed-zeros -fno-trapping-math
+PG_CFLAGS += $(OPTFLAGS) -ftree-vectorize -fassociative-math -fno-signed-zeros -fno-trapping-math
 
 # Debug GCC auto-vectorization
 # PG_CFLAGS += -fopt-info-vec

README.md

@@ -17,7 +17,7 @@ Supports L2 distance, inner product, and cosine distance
 Compile and install the extension (supports Postgres 9.6+)
 
 ```sh
-git clone --branch v0.2.4 https://github.com/pgvector/pgvector.git
+git clone --branch v0.2.6 https://github.com/pgvector/pgvector.git
 cd pgvector
 make
 make install # may need sudo
@@ -77,19 +77,7 @@ Cosine distance
 CREATE INDEX ON table USING ivfflat (column vector_cosine_ops);
 ```
 
-Indexes should be created after the table has data for optimal clustering. If the distribution of data changes significantly, you can reindex without downtime:
+Indexes should be created after the table has some data for optimal clustering. Also, unlike typical indexes which only affect performance, you may see different results for queries after adding an approximate index.
-
-```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;
-```
-
-Also, unlike typical indexes which only affect performance, you may see different results for queries after adding an approximate index.
 
 ### Index Options
 
@@ -194,6 +182,7 @@ Libraries that use pgvector:
 
 - [pgvector-python](https://github.com/pgvector/pgvector-python) (Python)
 - [Neighbor](https://github.com/ankane/neighbor) (Ruby)
+- [pgvector-ruby](https://github.com/pgvector/pgvector-ruby) (Ruby)
 - [pgvector-node](https://github.com/pgvector/pgvector-node) (Node.js)
 - [pgvector-go](https://github.com/pgvector/pgvector-go) (Go)
 - [pgvector-rust](https://github.com/pgvector/pgvector-rust) (Rust)
@@ -231,7 +220,7 @@ This adds pgvector to the [Postgres image](https://hub.docker.com/_/postgres).
 You can also build the image manually
 
 ```sh
-git clone --branch v0.2.4 https://github.com/pgvector/pgvector.git
+git clone --branch v0.2.6 https://github.com/pgvector/pgvector.git
 cd pgvector
 docker build -t pgvector .
 ```
@@ -315,6 +304,12 @@ make installcheck REGRESS=functions                    # regression test
 make prove_installcheck PROVE_TESTS=test/t/001_wal.pl  # TAP test
 ```
 
+To enable benchmarking:
+
+```sh
+make clean && PG_CFLAGS=-DIVFFLAT_BENCH make && make install
+```
+
 Resources for contributors
 
 - [Extension Building Infrastructure](https://www.postgresql.org/docs/current/extend-pgxs.html)

sql/vector--0.2.4--0.2.5.sql

@@ -0,0 +1,2 @@
+-- complain if script is sourced in psql, rather than via CREATE EXTENSION
+\echo Use "ALTER EXTENSION vector UPDATE TO '0.2.5'" to load this file. \quit

sql/vector--0.2.5--0.2.6.sql

@@ -0,0 +1,2 @@
+-- complain if script is sourced in psql, rather than via CREATE EXTENSION
+\echo Use "ALTER EXTENSION vector UPDATE TO '0.2.6'" to load this file. \quit

src/ivfbuild.c

@@ -36,16 +36,11 @@
 #define CALLBACK_ITEM_POINTER HeapTuple hup
 #endif
 
-/*
- * Update build phase progress
- */
-static inline void
-UpdateProgress(int index, int64 val)
-{
 #if PG_VERSION_NUM >= 120000
-	pgstat_progress_update_param(index, val);
+#define UpdateProgress(index, val) pgstat_progress_update_param(index, val)
+#else
+#define UpdateProgress(index, val) ((void)val)
 #endif
-}
 
 /*
  * Callback for sampling
@@ -117,13 +112,13 @@ SampleRows(IvfflatBuildState * buildstate)
 
 #if PG_VERSION_NUM >= 120000
 		table_index_build_range_scan(buildstate->heap, buildstate->index, buildstate->indexInfo,
-									 false, true, true, targblock, 1, SampleCallback, (void *) buildstate, NULL);
+									 false, true, false, targblock, 1, SampleCallback, (void *) buildstate, NULL);
 #elif PG_VERSION_NUM >= 110000
 		IndexBuildHeapRangeScan(buildstate->heap, buildstate->index, buildstate->indexInfo,
-								true, true, targblock, 1, SampleCallback, (void *) buildstate, NULL);
+								false, true, targblock, 1, SampleCallback, (void *) buildstate, NULL);
 #else
 		IndexBuildHeapRangeScan(buildstate->heap, buildstate->index, buildstate->indexInfo,
-								true, true, targblock, 1, SampleCallback, (void *) buildstate);
+								false, true, targblock, 1, SampleCallback, (void *) buildstate);
 #endif
 	}
 }
@@ -170,16 +165,20 @@ BuildCallback(Relation index, CALLBACK_ITEM_POINTER, Datum *values,
 		}
 	}
 
+#ifdef IVFFLAT_KMEANS_DEBUG
+	buildstate->inertia += minDistance;
+	buildstate->listSums[closestCenter] += minDistance;
+	buildstate->listCounts[closestCenter]++;
+#endif
+
 	/* Create a virtual tuple */
 	ExecClearTuple(slot);
 	slot->tts_values[0] = Int32GetDatum(closestCenter);
 	slot->tts_isnull[0] = false;
-	slot->tts_values[1] = Int32GetDatum(ItemPointerGetBlockNumberNoCheck(tid));
+	slot->tts_values[1] = PointerGetDatum(tid);
 	slot->tts_isnull[1] = false;
-	slot->tts_values[2] = Int32GetDatum(ItemPointerGetOffsetNumberNoCheck(tid));
+	slot->tts_values[2] = value;
 	slot->tts_isnull[2] = false;
-	slot->tts_values[3] = value;
-	slot->tts_isnull[3] = false;
 	ExecStoreVirtualTuple(slot);
 
 	/*
@@ -201,8 +200,6 @@ GetNextTuple(Tuplesortstate *sortstate, TupleDesc tupdesc, TupleTableSlot *slot,
 {
 	Datum		value;
 	bool		isnull;
-	int			tupblk;
-	int			tupoff;
 
 #if PG_VERSION_NUM >= 100000
 	if (tuplesort_gettupleslot(sortstate, true, false, slot, NULL))
@@ -211,13 +208,11 @@ GetNextTuple(Tuplesortstate *sortstate, TupleDesc tupdesc, TupleTableSlot *slot,
 #endif
 	{
 		*list = DatumGetInt32(slot_getattr(slot, 1, &isnull));
-		tupblk = DatumGetInt32(slot_getattr(slot, 2, &isnull));
+		value = slot_getattr(slot, 3, &isnull);
-		tupoff = DatumGetInt32(slot_getattr(slot, 3, &isnull));
-		value = slot_getattr(slot, 4, &isnull);
 
 		/* Form the index tuple */
 		*itup = index_form_tuple(tupdesc, &value, &isnull);
-		ItemPointerSet(&(*itup)->t_tid, tupblk, tupoff);
+		(*itup)->t_tid = *((ItemPointer) DatumGetPointer(slot_getattr(slot, 2, &isnull)));
 	}
 	else
 		*list = -1;
@@ -326,17 +321,16 @@ InitBuildState(IvfflatBuildState * buildstate, Relation heap, Relation index, In
 
 	/* Create tuple description for sorting */
 #if PG_VERSION_NUM >= 120000
-	buildstate->tupdesc = CreateTemplateTupleDesc(4);
+	buildstate->tupdesc = CreateTemplateTupleDesc(3);
 #else
-	buildstate->tupdesc = CreateTemplateTupleDesc(4, false);
+	buildstate->tupdesc = CreateTemplateTupleDesc(3, false);
 #endif
 	TupleDescInitEntry(buildstate->tupdesc, (AttrNumber) 1, "list", INT4OID, -1, 0);
-	TupleDescInitEntry(buildstate->tupdesc, (AttrNumber) 2, "blkno", INT4OID, -1, 0);
+	TupleDescInitEntry(buildstate->tupdesc, (AttrNumber) 2, "tid", TIDOID, -1, 0);
-	TupleDescInitEntry(buildstate->tupdesc, (AttrNumber) 3, "offset", INT4OID, -1, 0);
 #if PG_VERSION_NUM >= 110000
-	TupleDescInitEntry(buildstate->tupdesc, (AttrNumber) 4, "vector", RelationGetDescr(index)->attrs[0].atttypid, -1, 0);
+	TupleDescInitEntry(buildstate->tupdesc, (AttrNumber) 3, "vector", RelationGetDescr(index)->attrs[0].atttypid, -1, 0);
 #else
-	TupleDescInitEntry(buildstate->tupdesc, (AttrNumber) 4, "vector", RelationGetDescr(index)->attrs[0]->atttypid, -1, 0);
+	TupleDescInitEntry(buildstate->tupdesc, (AttrNumber) 3, "vector", RelationGetDescr(index)->attrs[0]->atttypid, -1, 0);
 #endif
 
 #if PG_VERSION_NUM >= 120000
@@ -350,6 +344,12 @@ InitBuildState(IvfflatBuildState * buildstate, Relation heap, Relation index, In
 
 	/* Reuse for each tuple */
 	buildstate->normvec = InitVector(buildstate->dimensions);
+
+#ifdef IVFFLAT_KMEANS_DEBUG
+	buildstate->inertia = 0;
+	buildstate->listSums = palloc0(sizeof(double) * buildstate->lists);
+	buildstate->listCounts = palloc0(sizeof(int) * buildstate->lists);
+#endif
 }
 
 /*
@@ -361,6 +361,11 @@ FreeBuildState(IvfflatBuildState * buildstate)
 	pfree(buildstate->centers);
 	pfree(buildstate->listInfo);
 	pfree(buildstate->normvec);
+
+#ifdef IVFFLAT_KMEANS_DEBUG
+	pfree(buildstate->listSums);
+	pfree(buildstate->listCounts);
+#endif
 }
 
 /*
@@ -377,14 +382,19 @@ ComputeCenters(IvfflatBuildState * buildstate)
 	if (numSamples < 10000)
 		numSamples = 10000;
 
-	/* Sample samples */
+	/* Skip samples for unlogged table */
+	if (buildstate->heap == NULL)
+		numSamples = 1;
+
+	/* Sample rows */
+	/* TODO Ensure within maintenance_work_mem */
 	buildstate->samples = VectorArrayInit(numSamples, buildstate->dimensions);
 	if (buildstate->heap != NULL)
 		SampleRows(buildstate);
 
 	/* Calculate centers */
 	UpdateProgress(PROGRESS_CREATEIDX_SUBPHASE, PROGRESS_IVFFLAT_PHASE_KMEANS);
-	IvfflatKmeans(buildstate->index, buildstate->samples, buildstate->centers);
+	IvfflatBench("k-means", IvfflatKmeans(buildstate->index, buildstate->samples, buildstate->centers));
 
 	/* Free samples before we allocate more memory */
 	pfree(buildstate->samples);
@@ -463,6 +473,51 @@ CreateListPages(Relation index, VectorArray centers, int dimensions,
 	pfree(list);
 }
 
+/*
+ * Print k-means metrics
+ */
+#ifdef IVFFLAT_KMEANS_DEBUG
+static void
+PrintKmeansMetrics(IvfflatBuildState * buildstate)
+{
+	elog(INFO, "inertia: %.3e", buildstate->inertia);
+
+	/* Calculate Davies-Bouldin index */
+	if (buildstate->lists > 1)
+	{
+		double		db = 0.0;
+
+		/* Calculate average distance */
+		for (int i = 0; i < buildstate->lists; i++)
+		{
+			if (buildstate->listCounts[i] > 0)
+				buildstate->listSums[i] /= buildstate->listCounts[i];
+		}
+
+		for (int i = 0; i < buildstate->lists; i++)
+		{
+			double		max = 0.0;
+			double		distance;
+
+			for (int j = 0; j < buildstate->lists; j++)
+			{
+				if (j == i)
+					continue;
+
+				distance = DatumGetFloat8(FunctionCall2Coll(buildstate->procinfo, buildstate->collation, PointerGetDatum(VectorArrayGet(buildstate->centers, i)), PointerGetDatum(VectorArrayGet(buildstate->centers, j))));
+				distance = (buildstate->listSums[i] + buildstate->listSums[j]) / distance;
+
+				if (distance > max)
+					max = distance;
+			}
+			db += max;
+		}
+		db /= buildstate->lists;
+		elog(INFO, "davies-bouldin: %.3f", db);
+	}
+}
+#endif
+
 /*
  * Create entry pages
  */
@@ -497,8 +552,14 @@ CreateEntryPages(IvfflatBuildState * buildstate, ForkNumber forkNum)
 #endif
 	}
 
-	/* Sort and insert */
+	/* Sort */
 	tuplesort_performsort(buildstate->sortstate);
+
+#ifdef IVFFLAT_KMEANS_DEBUG
+	PrintKmeansMetrics(buildstate);
+#endif
+
+	/* Insert */
 	InsertTuples(buildstate->index, buildstate, forkNum);
 	tuplesort_end(buildstate->sortstate);
 }
@@ -517,7 +578,7 @@ BuildIndex(Relation heap, Relation index, IndexInfo *indexInfo,
 	/* Create pages */
 	CreateMetaPage(index, buildstate->dimensions, buildstate->lists, forkNum);
 	CreateListPages(index, buildstate->centers, buildstate->dimensions, buildstate->lists, forkNum, &buildstate->listInfo);
-	CreateEntryPages(buildstate, forkNum);
+	IvfflatBench("CreateEntryPages", CreateEntryPages(buildstate, forkNum));
 
 	FreeBuildState(buildstate);
 }

src/ivfflat.c

@@ -159,6 +159,11 @@ ivfflatvalidate(Oid opclassoid)
 	return true;
 }
 
+/*
+ * Define index handler
+ *
+ * See https://www.postgresql.org/docs/current/index-api.html
+ */
 PG_FUNCTION_INFO_V1(ivfflathandler);
 Datum
 ivfflathandler(PG_FUNCTION_ARGS)
@@ -193,6 +198,7 @@ ivfflathandler(PG_FUNCTION_ARGS)
 #endif
 	amroutine->amkeytype = InvalidOid;
 
+	/* Interface functions */
 	amroutine->ambuild = ivfflatbuild;
 	amroutine->ambuildempty = ivfflatbuildempty;
 	amroutine->aminsert = ivfflatinsert;
@@ -206,6 +212,9 @@ ivfflathandler(PG_FUNCTION_ARGS)
 	amroutine->ambuildphasename = ivfflatbuildphasename;
 #endif
 	amroutine->amvalidate = ivfflatvalidate;
+#if PG_VERSION_NUM >= 140000
+	amroutine->amadjustmembers = NULL;
+#endif
 	amroutine->ambeginscan = ivfflatbeginscan;
 	amroutine->amrescan = ivfflatrescan;
 	amroutine->amgettuple = ivfflatgettuple;
@@ -213,6 +222,8 @@ ivfflathandler(PG_FUNCTION_ARGS)
 	amroutine->amendscan = ivfflatendscan;
 	amroutine->ammarkpos = NULL;
 	amroutine->amrestrpos = NULL;
+
+	/* Interface functions to support parallel index scans */
 #if PG_VERSION_NUM >= 100000
 	amroutine->amestimateparallelscan = NULL;
 	amroutine->aminitparallelscan = NULL;

src/ivfflat.h

@@ -10,6 +10,14 @@
 #include "utils/tuplesort.h"
 #include "vector.h"
 
+#ifdef IVFFLAT_BENCH
+#include "portability/instr_time.h"
+#endif
+
+#if PG_VERSION_NUM < 90600
+#error "Requires PostgreSQL 9.6+"
+#endif
+
 /* Support functions */
 #define IVFFLAT_DISTANCE_PROC 1
 #define IVFFLAT_NORM_PROC 2
@@ -39,9 +47,19 @@
 #define IvfflatPageGetOpaque(page)	((IvfflatPageOpaque) PageGetSpecialPointer(page))
 #define IvfflatPageGetMeta(page)	((IvfflatMetaPageData *) PageGetContents(page))
 
-#if PG_VERSION_NUM < 100000
+#ifdef IVFFLAT_BENCH
-#define ItemPointerGetBlockNumberNoCheck ItemPointerGetBlockNumber
+#define IvfflatBench(name, code) \
-#define ItemPointerGetOffsetNumberNoCheck ItemPointerGetOffsetNumber
+	do { \
+		instr_time	start; \
+		instr_time	duration; \
+		INSTR_TIME_SET_CURRENT(start); \
+		(code); \
+		INSTR_TIME_SET_CURRENT(duration); \
+		INSTR_TIME_SUBTRACT(duration, start); \
+		elog(INFO, "%s: %.3f ms", name, INSTR_TIME_GET_MILLISEC(duration)); \
+	} while (0)
+#else
+#define IvfflatBench(name, code) (code)
 #endif
 
 /* Variables */
@@ -97,6 +115,12 @@ typedef struct IvfflatBuildState
 	ListInfo   *listInfo;
 	Vector	   *normvec;
 
+#ifdef IVFFLAT_KMEANS_DEBUG
+	double		inertia;
+	double	   *listSums;
+	int		   *listCounts;
+#endif
+
 	/* Sampling */
 	BlockSamplerData bs;
 	ReservoirStateData rstate;
@@ -138,6 +162,7 @@ typedef IvfflatListData * IvfflatList;
 
 typedef struct IvfflatScanList
 {
+	pairingheap_node ph_node;
 	BlockNumber startPage;
 	double		distance;
 }			IvfflatScanList;
@@ -159,6 +184,8 @@ typedef struct IvfflatScanOpaqueData
 	FmgrInfo   *normprocinfo;
 	Oid			collation;
 
+	/* Lists */
+	pairingheap *listQueue;
 	IvfflatScanList lists[FLEXIBLE_ARRAY_MEMBER];	/* must come last */
 }			IvfflatScanOpaqueData;
 

src/ivfkmeans.c

@@ -11,7 +11,7 @@
  * https://theory.stanford.edu/~sergei/papers/kMeansPP-soda.pdf
  */
 static void
-InitCenters(Relation index, VectorArray samples, VectorArray centers, double *lowerBound)
+InitCenters(Relation index, VectorArray samples, VectorArray centers, float *lowerBound)
 {
 	FmgrInfo   *procinfo;
 	Oid			collation;
@@ -21,7 +21,7 @@ InitCenters(Relation index, VectorArray samples, VectorArray centers, double *lo
 	double		sum;
 	double		choice;
 	Vector	   *vec;
-	double	   *weight = palloc(samples->length * sizeof(double));
+	float	   *weight = palloc(samples->length * sizeof(float));
 	int			numCenters = centers->maxlen;
 	int			numSamples = samples->length;
 
@@ -121,15 +121,18 @@ QuickCenters(Relation index, VectorArray samples, VectorArray centers)
 	FmgrInfo   *normprocinfo = IvfflatOptionalProcInfo(index, IVFFLAT_KMEANS_NORM_PROC);
 
 	/* Copy existing vectors while avoiding duplicates */
-	qsort(samples->items, samples->length, VECTOR_SIZE(samples->dim), CompareVectors);
+	if (samples->length > 0)
-	for (i = 0; i < samples->length; i++)
 	{
-		vec = VectorArrayGet(samples, i);
+		qsort(samples->items, samples->length, VECTOR_SIZE(samples->dim), CompareVectors);
-
+		for (i = 0; i < samples->length; i++)
-		if (i == 0 || CompareVectors(vec, VectorArrayGet(samples, i - 1)) != 0)
 		{
-			VectorArraySet(centers, centers->length, vec);
+			vec = VectorArrayGet(samples, i);
-			centers->length++;
+
+			if (i == 0 || CompareVectors(vec, VectorArrayGet(samples, i - 1)) != 0)
+			{
+				VectorArraySet(centers, centers->length, vec);
+				centers->length++;
+			}
 		}
 	}
 
@@ -177,11 +180,11 @@ ElkanKmeans(Relation index, VectorArray samples, VectorArray centers)
 	VectorArray newCenters;
 	int		   *centerCounts;
 	int		   *closestCenters;
-	double	   *lowerBound;
+	float	   *lowerBound;
-	double	   *upperBound;
+	float	   *upperBound;
-	double	   *s;
+	float	   *s;
-	double	   *halfcdist;
+	float	   *halfcdist;
-	double	   *newcdist;
+	float	   *newcdist;
 	int			changes;
 	double		minDistance;
 	int			closestCenter;
@@ -191,19 +194,43 @@ ElkanKmeans(Relation index, VectorArray samples, VectorArray centers)
 	double		dxcx;
 	double		dxc;
 
+	/* Calculate allocation sizes */
+	Size		samplesSize = VECTOR_ARRAY_SIZE(samples->maxlen, samples->dim);
+	Size		centersSize = VECTOR_ARRAY_SIZE(centers->maxlen, centers->dim);
+	Size		newCentersSize = VECTOR_ARRAY_SIZE(numCenters, dimensions);
+	Size		centerCountsSize = sizeof(int) * numCenters;
+	Size		closestCentersSize = sizeof(int) * numSamples;
+	Size		lowerBoundSize = sizeof(float) * numSamples * numCenters;
+	Size		upperBoundSize = sizeof(float) * numSamples;
+	Size		sSize = sizeof(float) * numCenters;
+	Size		halfcdistSize = sizeof(float) * numCenters * numCenters;
+	Size		newcdistSize = sizeof(float) * numCenters;
+
+	/* Calculate total size */
+	Size		totalSize = samplesSize + centersSize + newCentersSize + centerCountsSize + closestCentersSize + lowerBoundSize + upperBoundSize + sSize + halfcdistSize + newcdistSize;
+
+	/* Check memory requirements */
+	/* Add one to error message to ceil */
+	if (totalSize / 1024 > maintenance_work_mem)
+		ereport(ERROR,
+				(errcode(ERRCODE_PROGRAM_LIMIT_EXCEEDED),
+				 errmsg("memory required is %zu MB, maintenance_work_mem is %d MB",
+						totalSize / (1024 * 1024) + 1, maintenance_work_mem / 1024)));
+
 	/* Set support functions */
 	procinfo = index_getprocinfo(index, 1, IVFFLAT_KMEANS_DISTANCE_PROC);
 	normprocinfo = IvfflatOptionalProcInfo(index, IVFFLAT_KMEANS_NORM_PROC);
 	collation = index->rd_indcollation[0];
 
 	/* Allocate space */
-	centerCounts = palloc(sizeof(int) * numCenters);
+	/* Use float instead of double to save memory */
-	closestCenters = palloc(sizeof(int) * numSamples);
+	centerCounts = palloc(centerCountsSize);
-	lowerBound = palloc(sizeof(double) * numSamples * numCenters);
+	closestCenters = palloc(closestCentersSize);
-	upperBound = palloc(sizeof(double) * numSamples);
+	lowerBound = palloc_extended(lowerBoundSize, MCXT_ALLOC_HUGE);
-	s = palloc(sizeof(double) * numCenters);
+	upperBound = palloc(upperBoundSize);
-	halfcdist = palloc(sizeof(double) * numCenters * numCenters);
+	s = palloc(sSize);
-	newcdist = palloc(sizeof(double) * numCenters);
+	halfcdist = palloc(halfcdistSize);
+	newcdist = palloc(newcdistSize);
 
 	newCenters = VectorArrayInit(numCenters, dimensions);
 	for (j = 0; j < numCenters; j++)

src/ivfscan.c

@@ -1,5 +1,7 @@
 #include "postgres.h"
 
+#include <float.h>
+
 #include "access/relscan.h"
 #include "ivfflat.h"
 #include "miscadmin.h"
@@ -17,14 +19,12 @@
  * Compare list distances
  */
 static int
-CompareLists(const void *a, const void *b)
+CompareLists(const pairingheap_node *a, const pairingheap_node *b, void *arg)
 {
-	double		diff = (((IvfflatScanList *) a)->distance - ((IvfflatScanList *) b)->distance);
+	if (((const IvfflatScanList *) a)->distance > ((const IvfflatScanList *) b)->distance)
-
-	if (diff > 0)
 		return 1;
 
-	if (diff < 0)
+	if (((const IvfflatScanList *) a)->distance < ((const IvfflatScanList *) b)->distance)
 		return -1;
 
 	return 0;
@@ -45,6 +45,8 @@ GetScanLists(IndexScanDesc scan, Datum value)
 	int			listCount = 0;
 	IvfflatScanOpaque so = (IvfflatScanOpaque) scan->opaque;
 	double		distance;
+	IvfflatScanList *scanlist;
+	double		maxDistance = DBL_MAX;
 
 	/* Search all list pages */
 	while (BlockNumberIsValid(nextblkno))
@@ -62,21 +64,39 @@ GetScanLists(IndexScanDesc scan, Datum value)
 			/* Use procinfo from the index instead of scan key for performance */
 			distance = DatumGetFloat8(FunctionCall2Coll(so->procinfo, so->collation, PointerGetDatum(&list->center), value));
 
-			so->lists[listCount].startPage = list->startPage;
+			if (listCount < so->probes)
-			so->lists[listCount].distance = distance;
+			{
-			listCount++;
+				scanlist = &so->lists[listCount];
+				scanlist->startPage = list->startPage;
+				scanlist->distance = distance;
+				listCount++;
+
+				/* Add to heap */
+				pairingheap_add(so->listQueue, &scanlist->ph_node);
+
+				/* Calculate max distance */
+				if (listCount == so->probes)
+					maxDistance = ((IvfflatScanList *) pairingheap_first(so->listQueue))->distance;
+			}
+			else if (distance < maxDistance)
+			{
+				/* Remove */
+				scanlist = (IvfflatScanList *) pairingheap_remove_first(so->listQueue);
+
+				/* Reuse */
+				scanlist->startPage = list->startPage;
+				scanlist->distance = distance;
+				pairingheap_add(so->listQueue, &scanlist->ph_node);
+
+				/* Update max distance */
+				maxDistance = ((IvfflatScanList *) pairingheap_first(so->listQueue))->distance;
+			}
 		}
 
 		nextblkno = IvfflatPageGetOpaque(cpage)->nextblkno;
 
 		UnlockReleaseBuffer(cbuf);
 	}
-
-	/* Sort by distance */
-	qsort(so->lists, listCount, sizeof(IvfflatScanList), CompareLists);
-
-	if (so->probes > listCount)
-		so->probes = listCount;
 }
 
 /*
@@ -94,7 +114,6 @@ GetScanItems(IndexScanDesc scan, Datum value)
 	OffsetNumber maxoffno;
 	Datum		datum;
 	bool		isnull;
-	int			i;
 	TupleDesc	tupdesc = RelationGetDescr(scan->indexRelation);
 
 #if PG_VERSION_NUM >= 120000
@@ -111,9 +130,9 @@ GetScanItems(IndexScanDesc scan, Datum value)
 	BufferAccessStrategy bas = GetAccessStrategy(BAS_BULKREAD);
 
 	/* Search closest probes lists */
-	for (i = 0; i < so->probes; i++)
+	while (!pairingheap_is_empty(so->listQueue))
 	{
-		searchPage = so->lists[i].startPage;
+		searchPage = ((IvfflatScanList *) pairingheap_remove_first(so->listQueue))->startPage;
 
 		/* Search all entry pages for list */
 		while (BlockNumberIsValid(searchPage))
@@ -137,12 +156,10 @@ GetScanItems(IndexScanDesc scan, Datum value)
 				ExecClearTuple(slot);
 				slot->tts_values[0] = FunctionCall2Coll(so->procinfo, so->collation, datum, value);
 				slot->tts_isnull[0] = false;
-				slot->tts_values[1] = Int32GetDatum((int) ItemPointerGetBlockNumberNoCheck(&itup->t_tid));
+				slot->tts_values[1] = PointerGetDatum(&itup->t_tid);
 				slot->tts_isnull[1] = false;
-				slot->tts_values[2] = Int32GetDatum((int) ItemPointerGetOffsetNumberNoCheck(&itup->t_tid));
+				slot->tts_values[2] = Int32GetDatum((int) searchPage);
 				slot->tts_isnull[2] = false;
-				slot->tts_values[3] = Int32GetDatum((int) searchPage);
-				slot->tts_isnull[3] = false;
 				ExecStoreVirtualTuple(slot);
 
 				tuplesort_puttupleslot(so->sortstate, slot);
@@ -153,6 +170,8 @@ GetScanItems(IndexScanDesc scan, Datum value)
 			UnlockReleaseBuffer(buf);
 		}
 	}
+
+	tuplesort_performsort(so->sortstate);
 }
 
 /*
@@ -168,13 +187,18 @@ ivfflatbeginscan(Relation index, int nkeys, int norderbys)
 	Oid			sortOperators[] = {Float8LessOperator};
 	Oid			sortCollations[] = {InvalidOid};
 	bool		nullsFirstFlags[] = {false};
+	int			probes = ivfflat_probes;
 
 	scan = RelationGetIndexScan(index, nkeys, norderbys);
 	lists = IvfflatGetLists(scan->indexRelation);
 
-	so = (IvfflatScanOpaque) palloc(offsetof(IvfflatScanOpaqueData, lists) + lists * sizeof(IvfflatScanList));
+	if (probes > lists)
+		probes = lists;
+
+	so = (IvfflatScanOpaque) palloc(offsetof(IvfflatScanOpaqueData, lists) + probes * sizeof(IvfflatScanList));
 	so->buf = InvalidBuffer;
 	so->first = true;
+	so->probes = probes;
 
 	/* Set support functions */
 	so->procinfo = index_getprocinfo(index, 1, IVFFLAT_DISTANCE_PROC);
@@ -183,14 +207,13 @@ ivfflatbeginscan(Relation index, int nkeys, int norderbys)
 
 	/* Create tuple description for sorting */
 #if PG_VERSION_NUM >= 120000
-	so->tupdesc = CreateTemplateTupleDesc(4);
+	so->tupdesc = CreateTemplateTupleDesc(3);
 #else
-	so->tupdesc = CreateTemplateTupleDesc(4, false);
+	so->tupdesc = CreateTemplateTupleDesc(3, false);
 #endif
 	TupleDescInitEntry(so->tupdesc, (AttrNumber) 1, "distance", FLOAT8OID, -1, 0);
-	TupleDescInitEntry(so->tupdesc, (AttrNumber) 2, "blkno", INT4OID, -1, 0);
+	TupleDescInitEntry(so->tupdesc, (AttrNumber) 2, "tid", TIDOID, -1, 0);
-	TupleDescInitEntry(so->tupdesc, (AttrNumber) 3, "offset", INT4OID, -1, 0);
+	TupleDescInitEntry(so->tupdesc, (AttrNumber) 3, "indexblkno", INT4OID, -1, 0);
-	TupleDescInitEntry(so->tupdesc, (AttrNumber) 4, "indexblkno", INT4OID, -1, 0);
 
 	/* Prep sort */
 #if PG_VERSION_NUM >= 110000
@@ -205,6 +228,8 @@ ivfflatbeginscan(Relation index, int nkeys, int norderbys)
 	so->slot = MakeSingleTupleTableSlot(so->tupdesc);
 #endif
 
+	so->listQueue = pairingheap_allocate(CompareLists, scan);
+
 	scan->opaque = so;
 
 	return scan;
@@ -224,7 +249,7 @@ ivfflatrescan(IndexScanDesc scan, ScanKey keys, int nkeys, ScanKey orderbys, int
 #endif
 
 	so->first = true;
-	so->probes = ivfflat_probes;
+	pairingheap_reset(so->listQueue);
 
 	if (keys && scan->numberOfKeys > 0)
 		memmove(scan->keyData, keys, scan->numberOfKeys * sizeof(ScanKeyData));
@@ -268,9 +293,8 @@ ivfflatgettuple(IndexScanDesc scan, ScanDirection dir)
 				return false;
 		}
 
-		GetScanLists(scan, value);
+		IvfflatBench("GetScanLists", GetScanLists(scan, value));
-		GetScanItems(scan, value);
+		IvfflatBench("GetScanItems", GetScanItems(scan, value));
-		tuplesort_performsort(so->sortstate);
 		so->first = false;
 
 		/* Clean up if we allocated a new value */
@@ -284,14 +308,13 @@ ivfflatgettuple(IndexScanDesc scan, ScanDirection dir)
 	if (tuplesort_gettupleslot(so->sortstate, true, so->slot, NULL))
 #endif
 	{
-		BlockNumber blkno = DatumGetInt32(slot_getattr(so->slot, 2, &so->isnull));
+		ItemPointer tid = (ItemPointer) DatumGetPointer(slot_getattr(so->slot, 2, &so->isnull));
-		OffsetNumber offset = DatumGetInt32(slot_getattr(so->slot, 3, &so->isnull));
+		BlockNumber indexblkno = DatumGetInt32(slot_getattr(so->slot, 3, &so->isnull));
-		BlockNumber indexblkno = DatumGetInt32(slot_getattr(so->slot, 4, &so->isnull));
 
 #if PG_VERSION_NUM >= 120000
-		ItemPointerSet(&scan->xs_heaptid, blkno, offset);
+		scan->xs_heaptid = *tid;
 #else
-		ItemPointerSet(&scan->xs_ctup.t_self, blkno, offset);
+		scan->xs_ctup.t_self = *tid;
 #endif
 
 		if (BufferIsValid(so->buf))
@@ -324,6 +347,7 @@ ivfflatendscan(IndexScanDesc scan)
 	if (BufferIsValid(so->buf))
 		ReleaseBuffer(so->buf);
 
+	pairingheap_free(so->listQueue);
 	tuplesort_end(so->sortstate);
 
 	pfree(so);

test/expected/input.out

@@ -53,3 +53,12 @@ LINE 1: SELECT '[1,]'::vector;
                ^
 SELECT '[1,2,3]'::vector(2);
 ERROR:  expected 2 dimensions, not 3
+SELECT unnest('{"[1,2,3]", "[4,5,6]"}'::vector[]);
+ unnest  
+---------
+ [1,2,3]
+ [4,5,6]
+(2 rows)
+
+SELECT '{"[1,2,3]"}'::vector(2)[];
+ERROR:  expected 2 dimensions, not 3

test/sql/input.sql

@@ -13,3 +13,6 @@ SELECT '1,2,3'::vector;
 SELECT '[]'::vector;
 SELECT '[1,]'::vector;
 SELECT '[1,2,3]'::vector(2);
+
+SELECT unnest('{"[1,2,3]", "[4,5,6]"}'::vector[]);
+SELECT '{"[1,2,3]"}'::vector(2)[];

test/t/001_wal.pl

@@ -34,9 +34,10 @@ sub test_index_replay
 	my $r2 = rand();
 	my $r3 = rand();
 
-	my $queries = qq(SET enable_seqscan=off;
+	my $queries = qq(
-SELECT * FROM tst ORDER BY v <-> '[$r1,$r2,$r3]' LIMIT 10;
+		SET enable_seqscan = off;
-);
+		SELECT * FROM tst ORDER BY v <-> '[$r1,$r2,$r3]' LIMIT 10;
+	);
 
 	# Run test queries and compare their result
 	my $primary_result = $node_primary->safe_psql("postgres", $queries);
@@ -65,10 +66,9 @@ $node_replica->start;
 $node_primary->safe_psql("postgres", "CREATE EXTENSION vector;");
 $node_primary->safe_psql("postgres", "CREATE TABLE tst (i int4, v vector(3));");
 $node_primary->safe_psql("postgres",
-	"INSERT INTO tst SELECT i%10, ARRAY[random(), random(), random()] FROM generate_series(1,100000) i;"
+	"INSERT INTO tst SELECT i % 10, ARRAY[random(), random(), random()] FROM generate_series(1, 100000) i;"
 );
-$node_primary->safe_psql("postgres",
+$node_primary->safe_psql("postgres", "CREATE INDEX ON tst USING ivfflat (v);");
-	"CREATE INDEX ON tst USING ivfflat (v);");
 
 # Test that queries give same result
 test_index_replay('initial');
@@ -82,7 +82,7 @@ for my $i (1 .. 10)
 	test_index_replay("vacuum $i");
 	my ($start, $end) = (100001 + ($i - 1) * 10000, 100000 + $i * 10000);
 	$node_primary->safe_psql("postgres",
-		"INSERT INTO tst SELECT i%10, ARRAY[random(), random(), random()] FROM generate_series($start,$end) i;"
+		"INSERT INTO tst SELECT i % 10, ARRAY[random(), random(), random()] FROM generate_series($start, $end) i;"
 	);
 	test_index_replay("insert $i");
 }

test/t/002_vacuum.pl

@@ -13,7 +13,7 @@ $node->start;
 $node->safe_psql("postgres", "CREATE EXTENSION vector;");
 $node->safe_psql("postgres", "CREATE TABLE tst (i int4, v vector(3));");
 $node->safe_psql("postgres",
-  "INSERT INTO tst SELECT i%10, ARRAY[i%1000, i%333, i%55] FROM generate_series(1,100000) i;"
+	"INSERT INTO tst SELECT i % 10, ARRAY[i % 1000, i % 333, i % 55] FROM generate_series(1, 100000) i;"
 );
 $node->safe_psql("postgres", "CREATE INDEX ON tst USING ivfflat (v);");
 
@@ -24,7 +24,7 @@ my $size = $node->safe_psql("postgres", "SELECT pg_total_relation_size('tst_v_id
 $node->safe_psql("postgres", "DELETE FROM tst;");
 $node->safe_psql("postgres", "VACUUM tst;");
 $node->safe_psql("postgres",
-  "INSERT INTO tst SELECT i%10, ARRAY[i%1000, i%333, i%55] FROM generate_series(1,100000) i;"
+	"INSERT INTO tst SELECT i % 10, ARRAY[i % 1000, i % 333, i % 55] FROM generate_series(1, 100000) i;"
 );
 
 # Check size

test/t/003_recall.pl

@@ -0,0 +1,88 @@
+use strict;
+use warnings;
+use PostgresNode;
+use TestLib;
+use Test::More tests => 9;
+
+my $node;
+my @queries = ();
+my @expected;
+my $limit = 20;
+
+sub test_recall
+{
+	my ($probes, $min, $operator) = @_;
+	my $correct = 0;
+	my $total = 0;
+
+	for my $i (0 .. $#queries) {
+		my $actual = $node->safe_psql("postgres", qq(
+			SET enable_seqscan = off;
+			SET ivfflat.probes = $probes;
+			SELECT i FROM tst ORDER BY v $operator '$queries[$i]' LIMIT $limit;
+		));
+		my @actual_ids = split("\n", $actual);
+		my %actual_set = map { $_ => 1 } @actual_ids;
+
+		my @expected_ids = split("\n", $expected[$i]);
+
+		foreach (@expected_ids) {
+			if (exists($actual_set{$_})) {
+				$correct++;
+			}
+			$total++;
+		}
+	}
+
+	cmp_ok($correct / $total, ">=", $min, $operator);
+}
+
+# Initialize node
+$node = get_new_node('node');
+$node->init;
+$node->start;
+
+# Create table
+$node->safe_psql("postgres", "CREATE EXTENSION vector;");
+$node->safe_psql("postgres", "CREATE TABLE tst (i int4, v vector(3));");
+$node->safe_psql("postgres",
+	"INSERT INTO tst SELECT i, ARRAY[random(), random(), random()] FROM generate_series(1, 100000) i;"
+);
+
+# Generate queries
+for (1..20) {
+	my $r1 = rand();
+	my $r2 = rand();
+	my $r3 = rand();
+	push(@queries, "[$r1,$r2,$r3]");
+}
+
+# Check each index type
+my @operators = ("<->", "<#>", "<=>");
+
+foreach (@operators) {
+	my $operator = $_;
+
+	# Get exact results
+	@expected = ();
+	foreach (@queries) {
+		my $res = $node->safe_psql("postgres", "SELECT i FROM tst ORDER BY v $operator '$_' LIMIT $limit;");
+		push(@expected, $res);
+	}
+
+	# Add index
+	my $opclass;
+	if ($operator == "<->") {
+		$opclass = "vector_l2_ops";
+	} elsif ($operator == "<#>") {
+		$opclass = "vector_ip_ops";
+	} else {
+		$opclass = "vector_cosine_ops";
+	}
+	$node->safe_psql("postgres", "CREATE INDEX ON tst USING ivfflat (v $opclass);");
+
+	# Test approximate results
+	test_recall(1, 0.75, $operator);
+	test_recall(10, 0.95, $operator);
+	test_recall(100, 1.0, $operator);
+}

test/t/004_centers.pl

@@ -0,0 +1,36 @@
+use strict;
+use warnings;
+use PostgresNode;
+use TestLib;
+use Test::More tests => 3;
+
+# Initialize node
+my $node = get_new_node('node');
+$node->init;
+$node->start;
+
+# Create table
+$node->safe_psql("postgres", "CREATE EXTENSION vector;");
+$node->safe_psql("postgres", "CREATE TABLE tst (i int4, v vector(3));");
+$node->safe_psql("postgres",
+	"INSERT INTO tst SELECT i, '[1,2,3]' FROM generate_series(1, 10) i;"
+);
+
+sub test_centers
+{
+	my ($lists, $min) = @_;
+
+	my ($ret, $stdout, $stderr) = $node->psql("postgres", "CREATE INDEX ON tst USING ivfflat (v) WITH (lists = $lists);");
+	is($ret, 0, $stderr);
+}
+
+# Test no error for duplicate centers
+test_centers(5);
+test_centers(10);
+
+$node->safe_psql("postgres",
+	"INSERT INTO tst SELECT i, '[4,5,6]' FROM generate_series(1, 10) i;"
+);
+
+# Test no error for duplicate centers
+test_centers(10);

test/t/005_query_recall.pl

@@ -0,0 +1,45 @@
+use strict;
+use warnings;
+use PostgresNode;
+use TestLib;
+use Test::More tests => 60;
+
+# Initialize node
+my $node = get_new_node('node');
+$node->init;
+$node->start;
+
+# Create table
+$node->safe_psql("postgres", "CREATE EXTENSION vector;");
+$node->safe_psql("postgres", "CREATE TABLE tst (i int4 primary key, v vector(3));");
+$node->safe_psql("postgres",
+	"INSERT INTO tst SELECT i, ARRAY[random(), random(), random()] FROM generate_series(1, 100000) i;"
+);
+
+# Check each index type
+my @operators = ("<->", "<#>", "<=>");
+foreach (@operators) {
+	my $operator = $_;
+
+	# Add index
+	my $opclass;
+	if ($operator == "<->") {
+		$opclass = "vector_l2_ops";
+	} elsif ($operator == "<#>") {
+		$opclass = "vector_ip_ops";
+	} else {
+		$opclass = "vector_cosine_ops";
+	}
+	$node->safe_psql("postgres", "CREATE INDEX ON tst USING ivfflat (v $opclass);");
+
+	# Test 100% recall
+	for (1..20) {
+		my $i = int(rand() * 100000);
+		my $query = $node->safe_psql("postgres", "SELECT v FROM tst WHERE i = $i;");
+		my $res = $node->safe_psql("postgres", qq(
+			SET enable_seqscan = off;
+			SELECT v FROM tst ORDER BY v <-> '$query' LIMIT 1;
+		));
+		is($res, $query);
+	}
+}

test/t/006_lists.pl

@@ -0,0 +1,31 @@
+use strict;
+use warnings;
+use PostgresNode;
+use TestLib;
+use Test::More tests => 3;
+
+# Initialize node
+my $node = get_new_node('node');
+$node->init;
+$node->start;
+
+# Create table
+$node->safe_psql("postgres", "CREATE EXTENSION vector;");
+$node->safe_psql("postgres", "CREATE TABLE tst (v vector(3));");
+$node->safe_psql("postgres",
+	"INSERT INTO tst SELECT ARRAY[random(), random(), random()] FROM generate_series(1, 100000) i;"
+);
+
+$node->safe_psql("postgres", "CREATE INDEX lists50 ON tst USING ivfflat (v) WITH (lists = 50);");
+$node->safe_psql("postgres", "CREATE INDEX lists100 ON tst USING ivfflat (v) WITH (lists = 100);");
+
+# Test prefers more lists
+my $res = $node->safe_psql("postgres", "EXPLAIN SELECT v FROM tst ORDER BY v <-> '[0.5,0.5,0.5]' LIMIT 10;");
+like($res, qr/lists100/);
+unlike($res, qr/lists50/);
+
+# Test errors with too much memory
+my ($ret, $stdout, $stderr) = $node->psql("postgres",
+	"CREATE INDEX lists10000 ON tst USING ivfflat (v) WITH (lists = 10000);"
+);
+like($stderr, qr/memory required is/);

vector.control

@@ -1,4 +1,4 @@
 comment = 'vector data type and ivfflat access method'
-default_version = '0.2.4'
+default_version = '0.2.6'
 module_pathname = '$libdir/vector'
 relocatable = true