Version bump to 0.2.0 [skip ci]

.github/workflows/build.yml

@@ -8,7 +8,7 @@ jobs:
       fail-fast: false
       matrix:
         os: [ubuntu-latest]
-        postgres: [13, 12, 11, 10, 9.6]
+        postgres: [14, 13, 12, 11, 10, 9.6]
         include:
         - os: macos-latest
           postgres: 13

CHANGELOG.md

@@ -1,3 +1,15 @@
+## 0.2.0 (2021-10-03)
+
+- Added support for Postgres 14
+
+## 0.1.8 (2021-09-07)
+
+- Added cast for `vector` to `real[]`
+
+## 0.1.7 (2021-06-13)
+
+- Added cast for `numeric[]` to `vector`
+
 ## 0.1.6 (2021-06-09)
 
 - Fixed segmentation fault with `COUNT`

Dockerfile

@@ -1,9 +1,9 @@
-FROM postgres
+FROM postgres:14
 
 COPY . /tmp/pgvector
 
 RUN apt-get update && \
-		apt-get install -y --no-install-recommends build-essential postgresql-server-dev-13 && \
+		apt-get install -y --no-install-recommends build-essential postgresql-server-dev-14 && \
 		cd /tmp/pgvector && \
 		make clean && \
 		make OPTFLAGS="" && \
@@ -11,6 +11,6 @@ RUN apt-get update && \
 		mkdir /usr/share/doc/pgvector && \
 		cp LICENSE README.md /usr/share/doc/pgvector && \
 		rm -r /tmp/pgvector && \
-		apt-get remove -y build-essential postgresql-server-dev-13 && \
+		apt-get remove -y build-essential postgresql-server-dev-14 && \
 		apt-get autoremove -y && \
 		rm -rf /var/lib/apt/lists/*

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": "0.1.6",
+	"version": "0.2.0",
 	"maintainer": [
 		"Andrew Kane <andrew@ankane.org>"
 	],
@@ -20,7 +20,7 @@
 		"vector": {
 			"file": "sql/vector.sql",
 			"docfile": "README.md",
-			"version": "0.1.6",
+			"version": "0.2.0",
 			"abstract": "Open-source vector similarity search for Postgres"
 		}
 	},

Makefile

@@ -1,5 +1,5 @@
 EXTENSION = vector
-EXTVERSION = 0.1.6
+EXTVERSION = 0.2.0
 
 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 v0.1.6 https://github.com/ankane/pgvector.git
+git clone --branch v0.2.0 https://github.com/ankane/pgvector.git
 cd pgvector
 make
 make install # may need sudo
@@ -106,6 +106,16 @@ SELECT ...
 COMMIT;
 ```
 
+### Partial Indexes
+
+Consider [partial indexes](https://www.postgresql.org/docs/current/indexes-partial.html) for queries with a `WHERE` clause
+
+```sql
+CREATE INDEX ON table USING ivfflat (column) WHERE (other_column = 123);
+```
+
+To index many different values of `other_column`, consider [partitioning](https://www.postgresql.org/docs/current/ddl-partitioning.html) on `other_column`.
+
 ## Reference
 
 ### Vector Type
@@ -136,8 +146,12 @@ 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.rs](https://github.com/ankane/pgvector.rs) (Rust)
+- [pgvector-node](https://github.com/ankane/pgvector-node) (Node.js)
+- [pgvector-go](https://github.com/ankane/pgvector-go) (Go)
+- [pgvector-rust](https://github.com/ankane/pgvector-rust) (Rust)
+- [pgvector-cpp](https://github.com/ankane/pgvector-cpp) (C++)
 
 ## Additional Installation Methods
 
@@ -154,7 +168,7 @@ This adds pgvector to the [Postgres image](https://hub.docker.com/_/postgres).
 You can also build the image manually
 
 ```sh
-git clone --branch v0.1.6 https://github.com/ankane/pgvector.git
+git clone --branch v0.2.0 https://github.com/ankane/pgvector.git
 cd pgvector
 docker build -t pgvector .
 ```

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--0.1.7--0.1.8.sql

@@ -0,0 +1,8 @@
+-- complain if script is sourced in psql, rather than via CREATE EXTENSION
+\echo Use "ALTER EXTENSION vector UPDATE TO '0.1.8'" to load this file. \quit
+
+CREATE FUNCTION vector_to_float4(vector, integer, boolean) RETURNS real[]
+	AS 'MODULE_PATHNAME' LANGUAGE C IMMUTABLE STRICT PARALLEL SAFE;
+
+CREATE CAST (vector AS real[])
+	WITH FUNCTION vector_to_float4(vector, integer, boolean) AS IMPLICIT;

sql/vector--0.1.8--0.2.0.sql

@@ -0,0 +1,2 @@
+-- complain if script is sourced in psql, rather than via CREATE EXTENSION
+\echo Use "ALTER EXTENSION vector UPDATE TO '0.2.0'" to load this file. \quit

sql/vector.sql

@@ -97,6 +97,12 @@ 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;
+
+CREATE FUNCTION vector_to_float4(vector, integer, boolean) RETURNS real[]
+	AS 'MODULE_PATHNAME' LANGUAGE C IMMUTABLE STRICT PARALLEL SAFE;
+
 -- casts
 
 CREATE CAST (vector AS vector)
@@ -111,6 +117,12 @@ 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;
+
+CREATE CAST (vector AS real[])
+	WITH FUNCTION vector_to_float4(vector, integer, boolean) AS IMPLICIT;
+
 -- operators
 
 CREATE OPERATOR <-> (

src/ivfflat.h

@@ -180,6 +180,9 @@ void		IvfflatInitPage(Relation index, Buffer *buf, Page *page, GenericXLogState
 IndexBuildResult *ivfflatbuild(Relation heap, Relation index, IndexInfo *indexInfo);
 void		ivfflatbuildempty(Relation index);
 bool		ivfflatinsert(Relation index, Datum *values, bool *isnull, ItemPointer heap_tid, Relation heap, IndexUniqueCheck checkUnique
+#if PG_VERSION_NUM >= 140000
+						  ,bool indexUnchanged
+#endif
 #if PG_VERSION_NUM >= 100000
 						  ,IndexInfo *indexInfo
 #endif

src/ivfinsert.c

@@ -128,6 +128,9 @@ InsertTuple(Relation rel, IndexTuple itup, Relation heapRel, Datum *values)
 bool
 ivfflatinsert(Relation index, Datum *values, bool *isnull, ItemPointer heap_tid,
 			  Relation heap, IndexUniqueCheck checkUnique
+#if PG_VERSION_NUM >= 140000
+			  ,bool indexUnchanged
+#endif
 #if PG_VERSION_NUM >= 100000
 			  ,IndexInfo *indexInfo
 #endif

src/vector.c

@@ -10,11 +10,16 @@
 #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"
 #endif
 
+#if PG_VERSION_NUM < 130000
+#define TYPALIGN_INT 'i'
+#endif
+
 PG_MODULE_MAGIC;
 
 /*
@@ -345,6 +350,8 @@ 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),
@@ -356,6 +363,29 @@ array_to_vector(PG_FUNCTION_ARGS)
 	PG_RETURN_POINTER(result);
 }
 
+/*
+ * Convert vector to float4[]
+ */
+PG_FUNCTION_INFO_V1(vector_to_float4);
+Datum
+vector_to_float4(PG_FUNCTION_ARGS)
+{
+	Vector	   *vec = PG_GETARG_VECTOR_P(0);
+	Datum	   *d;
+	ArrayType  *result;
+	int			i;
+
+	d = (Datum *) palloc(sizeof(Datum) * vec->dim);
+
+	for (i = 0; i < vec->dim; i++)
+		d[i] = Float4GetDatum(vec->x[i]);
+
+	/* Use TYPALIGN_INT for float4 */
+	result = construct_array(d, vec->dim, FLOAT4OID, sizeof(float4), true, TYPALIGN_INT);
+
+	PG_RETURN_POINTER(result);
+}
+
 /*
  * Get the L2 distance between vectors
  */

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  
 ---------
@@ -28,5 +34,11 @@ SELECT '{-Infinity}'::real[]::vector;
 ERROR:  infinite value not allowed in vector
 SELECT '{}'::real[]::vector;
 ERROR:  vector must have at least 1 dimension
+SELECT '[1,2,3]'::vector::real[];
+ float4  
+---------
+ {1,2,3}
+(1 row)
+
 SELECT array_agg(n)::vector FROM generate_series(1, 1025) n;
 ERROR:  vector cannot have more than 1024 dimensions

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;
@@ -9,4 +10,5 @@ SELECT '{NaN}'::real[]::vector;
 SELECT '{Infinity}'::real[]::vector;
 SELECT '{-Infinity}'::real[]::vector;
 SELECT '{}'::real[]::vector;
+SELECT '[1,2,3]'::vector::real[];
 SELECT array_agg(n)::vector FROM generate_series(1, 1025) n;

vector.control

@@ -1,4 +1,4 @@
 comment = 'vector data type and ivfflat access method'
-default_version = '0.1.6'
+default_version = '0.2.0'
 module_pathname = '$libdir/vector'
 relocatable = true