Version bump to 0.1.7 [skip ci]

.dockerignore

@@ -0,0 +1,8 @@
+/.git/
+/dist/
+/results/
+/tmp_check/
+/sql/vector--?.?.?.sql
+regression.*
+*.o
+*.so

CHANGELOG.md

@@ -1,3 +1,20 @@
+## 0.1.7 (unreleased)
+
+- Added cast for `numeric[]`
+
+## 0.1.6 (2021-06-09)
+
+- Fixed segmentation fault with `COUNT`
+
+## 0.1.5 (2021-05-25)
+
+- Reduced memory usage during index creation
+
+## 0.1.4 (2021-05-09)
+
+- Fixed kmeans for inner product
+- Fixed multiple definition error with GCC 10
+
 ## 0.1.3 (2021-05-06)
 
 - Added Dockerfile

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": "0.1.3",
+	"version": "0.1.7",
 	"maintainer": [
 		"Andrew Kane <andrew@ankane.org>"
 	],
@@ -20,7 +20,7 @@
 		"vector": {
 			"file": "sql/vector.sql",
 			"docfile": "README.md",
-			"version": "0.1.3",
+			"version": "0.1.7",
 			"abstract": "Open-source vector similarity search for Postgres"
 		}
 	},

Makefile

@@ -1,5 +1,5 @@
 EXTENSION = vector
-EXTVERSION = 0.1.3
+EXTVERSION = 0.1.7
 
 MODULE_big = vector
 DATA = $(wildcard sql/*--*.sql)

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 v0.1.3 https://github.com/ankane/pgvector.git
+git clone --branch v0.1.7 https://github.com/ankane/pgvector.git
 cd pgvector
 make
 make install # may need sudo
@@ -136,7 +136,9 @@ vector_norm(vector) | Euclidean norm
 
 Libraries that use pgvector:
 
+- [pgvector-python](https://github.com/ankane/pgvector-python) (Python)
 - [Neighbor](https://github.com/ankane/neighbor) (Ruby)
+- [pgvector-rust](https://github.com/ankane/pgvector-rust) (Rust)
 
 ## Additional Installation Methods
 
@@ -153,7 +155,7 @@ This adds pgvector to the [Postgres image](https://hub.docker.com/_/postgres).
 You can also build the image manually
 
 ```sh
-git clone --branch v0.1.3 https://github.com/ankane/pgvector.git
+git clone --branch v0.1.7 https://github.com/ankane/pgvector.git
 cd pgvector
 docker build -t pgvector .
 ```
@@ -181,15 +183,14 @@ Some Postgres providers only support specific extensions. To request a new exten
 - Amazon RDS - follow the instructions on [this page](https://aws.amazon.com/rds/postgresql/faqs/)
 - Google Cloud SQL - follow the instructions on [this page](https://cloud.google.com/sql/docs/postgres/extensions#requesting-support-for-a-new-extension)
 - DigitalOcean Managed Databases - follow the instructions on [this page](https://docs.digitalocean.com/products/databases/postgresql/resources/supported-extensions/#supported-extensions)
+- Azure Database for PostgreSQL - follow the instructions on [this page](https://docs.microsoft.com/en-us/azure/postgresql/concepts-extensions#next-steps)
 
 ## Upgrading
 
-### 0.1.1
-
-Compile and install the latest version and run:
+Install the latest version and run:
 
 ```sql
-ALTER EXTENSION vector UPDATE TO '0.1.1';
+ALTER EXTENSION vector UPDATE;
 ```
 
 ## Thanks
@@ -234,7 +235,7 @@ make prove_installcheck  # TAP tests
 To run single tests:
 
 ```sh
-make installcheck REGRESS=vector                       # regression test
+make installcheck REGRESS=functions                    # regression test
 make prove_installcheck PROVE_TESTS=test/t/001_wal.pl  # TAP test
 ```
 

sql/vector--0.1.3--0.1.4.sql

@@ -0,0 +1,2 @@
+-- complain if script is sourced in psql, rather than via CREATE EXTENSION
+\echo Use "ALTER EXTENSION vector UPDATE TO '0.1.4'" to load this file. \quit

sql/vector--0.1.4--0.1.5.sql

@@ -0,0 +1,2 @@
+-- complain if script is sourced in psql, rather than via CREATE EXTENSION
+\echo Use "ALTER EXTENSION vector UPDATE TO '0.1.5'" to load this file. \quit

sql/vector--0.1.5--0.1.6.sql

@@ -0,0 +1,2 @@
+-- complain if script is sourced in psql, rather than via CREATE EXTENSION
+\echo Use "ALTER EXTENSION vector UPDATE TO '0.1.6'" to load this file. \quit

sql/vector--0.1.6--0.1.7.sql

@@ -0,0 +1,8 @@
+-- complain if script is sourced in psql, rather than via CREATE EXTENSION
+\echo Use "ALTER EXTENSION vector UPDATE TO '0.1.7'" to load this file. \quit
+
+CREATE FUNCTION array_to_vector(numeric[], integer, boolean) RETURNS vector
+	AS 'MODULE_PATHNAME' LANGUAGE C IMMUTABLE STRICT PARALLEL SAFE;
+
+CREATE CAST (numeric[] AS vector)
+	WITH FUNCTION array_to_vector(numeric[], integer, boolean) AS IMPLICIT;

sql/vector.sql

@@ -97,6 +97,9 @@ CREATE FUNCTION array_to_vector(real[], integer, boolean) RETURNS vector
 CREATE FUNCTION array_to_vector(double precision[], integer, boolean) RETURNS vector
 	AS 'MODULE_PATHNAME' LANGUAGE C IMMUTABLE STRICT PARALLEL SAFE;
 
+CREATE FUNCTION array_to_vector(numeric[], integer, boolean) RETURNS vector
+	AS 'MODULE_PATHNAME' LANGUAGE C IMMUTABLE STRICT PARALLEL SAFE;
+
 -- casts
 
 CREATE CAST (vector AS vector)
@@ -111,6 +114,9 @@ CREATE CAST (real[] AS vector)
 CREATE CAST (double precision[] AS vector)
 	WITH FUNCTION array_to_vector(double precision[], integer, boolean) AS IMPLICIT;
 
+CREATE CAST (numeric[] AS vector)
+	WITH FUNCTION array_to_vector(numeric[], integer, boolean) AS IMPLICIT;
+
 -- operators
 
 CREATE OPERATOR <-> (

src/ivfbuild.c

@@ -41,10 +41,13 @@ SampleCallback(Relation index, CALLBACK_ITEM_POINTER, Datum *values,
 	if (isnull[0])
 		return;
 
-	/* Normalize the value */
-	if (buildstate->normprocinfo != NULL)
+	/*
+	 * Normalize with KMEANS_NORM_PROC since spherical distance function
+	 * expects unit vectors
+	 */
+	if (buildstate->kmeansnormprocinfo != NULL)
 	{
-		if (!IvfflatNormValue(buildstate->normprocinfo, buildstate->collation, &value, buildstate->normvec))
+		if (!IvfflatNormValue(buildstate->kmeansnormprocinfo, buildstate->collation, &value, buildstate->normvec))
 			return;
 	}
 
@@ -238,6 +241,8 @@ InsertTuples(Relation index, IvfflatBuildState * buildstate, ForkNumber forkNum)
 			if (PageAddItem(page, (Item) itup, itemsz, InvalidOffsetNumber, false, false) == InvalidOffsetNumber)
 				elog(ERROR, "failed to add index item to \"%s\"", RelationGetRelationName(index));
 
+			pfree(itup);
+
 			buildstate->indtuples += 1;
 
 			GetNextTuple(buildstate->sortstate, tupdesc, slot, &itup, &list);
@@ -275,6 +280,7 @@ InitBuildState(IvfflatBuildState * buildstate, Relation heap, Relation index, In
 	/* Get support functions */
 	buildstate->procinfo = index_getprocinfo(index, 1, IVFFLAT_DISTANCE_PROC);
 	buildstate->normprocinfo = IvfflatOptionalProcInfo(index, IVFFLAT_NORM_PROC);
+	buildstate->kmeansnormprocinfo = IvfflatOptionalProcInfo(index, IVFFLAT_KMEANS_NORM_PROC);
 	buildstate->collation = index->rd_indcollation[0];
 
 	/* Create tuple description for sorting */

src/ivfflat.c

@@ -1,11 +1,14 @@
 #include "postgres.h"
 
+#include <float.h>
+
 #include "access/amapi.h"
 #include "commands/vacuum.h"
 #include "ivfflat.h"
 #include "utils/guc.h"
 #include "utils/selfuncs.h"
 
+int			ivfflat_probes;
 static relopt_kind ivfflat_relopt_kind;
 
 /*
@@ -43,18 +46,29 @@ ivfflatcostestimate(PlannerInfo *root, IndexPath *path, double loop_count,
 	int			lists;
 	double		ratio;
 	Relation	indexRel;
-
 #if PG_VERSION_NUM < 120000
 	List	   *qinfos;
-
-	qinfos = deconstruct_indexquals(path);
 #endif
 
+	/* Never use index without order */
+	if (path->indexorderbys == NULL)
+	{
+		*indexStartupCost = DBL_MAX;
+		*indexTotalCost = DBL_MAX;
+		*indexSelectivity = 0;
+		*indexCorrelation = 0;
+#if PG_VERSION_NUM >= 100000
+		*indexPages = 0;
+#endif
+		return;
+	}
+
 	MemSet(&costs, 0, sizeof(costs));
 
 #if PG_VERSION_NUM >= 120000
 	genericcostestimate(root, path, loop_count, &costs);
 #else
+	qinfos = deconstruct_indexquals(path);
 	genericcostestimate(root, path, loop_count, qinfos, &costs);
 #endif
 

src/ivfflat.h

@@ -38,7 +38,7 @@
 #endif
 
 /* Variables */
-int			ivfflat_probes;
+extern int	ivfflat_probes;
 
 typedef struct VectorArrayData
 {
@@ -81,6 +81,7 @@ typedef struct IvfflatBuildState
 	/* Support functions */
 	FmgrInfo   *procinfo;
 	FmgrInfo   *normprocinfo;
+	FmgrInfo   *kmeansnormprocinfo;
 	Oid			collation;
 
 	/* Variables */

src/ivfscan.c

@@ -251,6 +251,10 @@ ivfflatgettuple(IndexScanDesc scan, ScanDirection dir)
 	{
 		Datum		value;
 
+		/* Safety check */
+		if (scan->orderByData == NULL)
+			elog(ERROR, "cannot scan ivfflat index without order");
+
 		/* No items will match if null */
 		if (scan->orderByData->sk_flags & SK_ISNULL)
 			return false;

src/vector.c

@@ -10,6 +10,7 @@
 #include "utils/array.h"
 #include "utils/builtins.h"
 #include "utils/lsyscache.h"
+#include "utils/numeric.h"
 
 #if PG_VERSION_NUM >= 120000
 #include "utils/float.h"
@@ -244,14 +245,20 @@ vector_recv(PG_FUNCTION_ARGS)
 	int32		typmod = PG_GETARG_INT32(2);
 	Vector	   *result;
 	int16		dim;
+	int16		unused;
 	int			i;
 
 	dim = pq_getmsgint(buf, sizeof(int16));
-	pq_getmsgint(buf, sizeof(int16));	/* unused */
+	unused = pq_getmsgint(buf, sizeof(int16));
 
 	CheckDim(dim);
 	CheckExpectedDim(typmod, dim);
 
+	if (unused != 0)
+		ereport(ERROR,
+				(errcode(ERRCODE_DATA_EXCEPTION),
+				 errmsg("expected unused to be 0, not %d", unused)));
+
 	result = InitVector(dim);
 	for (i = 0; i < dim; i++)
 		result->x[i] = pq_getmsgfloat4(buf);
@@ -339,6 +346,8 @@ array_to_vector(PG_FUNCTION_ARGS)
 			result->x[i] = DatumGetFloat8(elemsp[i]);
 		else if (ARR_ELEMTYPE(array) == FLOAT4OID)
 			result->x[i] = DatumGetFloat4(elemsp[i]);
+		else if (ARR_ELEMTYPE(array) == NUMERICOID)
+			result->x[i] = DatumGetFloat4(DirectFunctionCall1(numeric_float4, NumericGetDatum(elemsp[i])));
 		else
 			ereport(ERROR,
 					(errcode(ERRCODE_DATA_EXCEPTION),

test/expected/cast.out

@@ -6,6 +6,12 @@ SELECT ARRAY[1,2,3]::vector;
  [1,2,3]
 (1 row)
 
+SELECT ARRAY[1.0,2.0,3.0]::vector;
+  array  
+---------
+ [1,2,3]
+(1 row)
+
 SELECT ARRAY[1,2,3]::float4[]::vector;
   array  
 ---------
@@ -28,3 +34,5 @@ SELECT '{-Infinity}'::real[]::vector;
 ERROR:  infinite value not allowed in vector
 SELECT '{}'::real[]::vector;
 ERROR:  vector must have at least 1 dimension
+SELECT array_agg(n)::vector FROM generate_series(1, 1025) n;
+ERROR:  vector cannot have more than 1024 dimensions

test/expected/ivfflat_l2.out

@@ -19,4 +19,10 @@ SELECT * FROM t ORDER BY val <-> (SELECT NULL::vector);
 -----
 (0 rows)
 
+SELECT COUNT(*) FROM t;
+ count 
+-------
+     5
+(1 row)
+
 DROP TABLE t;

test/sql/cast.sql

@@ -2,6 +2,7 @@ SET client_min_messages = warning;
 CREATE EXTENSION IF NOT EXISTS vector;
 
 SELECT ARRAY[1,2,3]::vector;
+SELECT ARRAY[1.0,2.0,3.0]::vector;
 SELECT ARRAY[1,2,3]::float4[]::vector;
 SELECT ARRAY[1,2,3]::float8[]::vector;
 SELECT '{NULL}'::real[]::vector;
@@ -9,3 +10,4 @@ SELECT '{NaN}'::real[]::vector;
 SELECT '{Infinity}'::real[]::vector;
 SELECT '{-Infinity}'::real[]::vector;
 SELECT '{}'::real[]::vector;
+SELECT array_agg(n)::vector FROM generate_series(1, 1025) n;

test/sql/ivfflat_l2.sql

@@ -10,5 +10,6 @@ INSERT INTO t (val) VALUES ('[1,2,4]');
 
 SELECT * FROM t ORDER BY val <-> '[3,3,3]';
 SELECT * FROM t ORDER BY val <-> (SELECT NULL::vector);
+SELECT COUNT(*) FROM t;
 
 DROP TABLE t;

vector.control

@@ -1,4 +1,4 @@
 comment = 'vector data type and ivfflat access method'
-default_version = '0.1.3'
+default_version = '0.1.7'
 module_pathname = '$libdir/vector'
 relocatable = true