Version bump to 0.1.1 [skip ci]

.github/workflows/build.yml

@@ -10,14 +10,10 @@ jobs:
         postgres: [13, 12, 11, 10, 9.6]
     steps:
     - uses: actions/checkout@v2
-    - run: |
-        sudo sh -c 'echo "deb http://apt.postgresql.org/pub/repos/apt $(lsb_release -cs)-pgdg main" > /etc/apt/sources.list.d/pgdg.list'
-        wget --quiet -O - https://www.postgresql.org/media/keys/ACCC4CF8.asc | sudo apt-key add -
-        sudo apt-get update
     - uses: ankane/setup-postgres@v1
       with:
         postgres-version: ${{ matrix.postgres }}
-    - run: sudo apt-get install postgresql-server-dev-${{ matrix.postgres }} libipc-run-perl
+    - run: sudo apt-get update && sudo apt-get install postgresql-server-dev-${{ matrix.postgres }} libipc-run-perl
     - run: make
     - run: |
         export PG_CONFIG=`which pg_config`

CHANGELOG.md

@@ -1,3 +1,8 @@
+## 0.1.1 (2021-04-25)
+
+- Added binary representation for `COPY`
+- Marked functions as `PARALLEL SAFE`
+
 ## 0.1.0 (2021-04-20)
 
 - First release

Makefile

@@ -1,11 +1,11 @@
 EXTENSION = vector
-DATA = vector--0.1.0.sql
+DATA = vector--0.1.1.sql vector--0.1.0--0.1.1.sql
 MODULE_big = vector
 OBJS = src/ivfbuild.o src/ivfflat.o src/ivfinsert.o src/ivfkmeans.o src/ivfscan.o src/ivfutils.o src/ivfvacuum.o src/vector.o
 
 TESTS = $(wildcard sql/*.sql)
 
-REGRESS = btree cast functions ivfflat_cosine ivfflat_ip ivfflat_l2 ivfflat_unlogged vector
+REGRESS = btree cast copy functions ivfflat_cosine ivfflat_ip ivfflat_l2 ivfflat_unlogged vector
 
 PG_CONFIG ?= pg_config
 PGXS := $(shell $(PG_CONFIG) --pgxs)

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 https://github.com/ankane/pgvector.git
+git clone --branch v0.1.1 https://github.com/ankane/pgvector.git
 cd pgvector
 make
 make install # may need sudo
@@ -130,6 +130,22 @@ l2_distance(vector, vector) | Euclidean distance
 vector_dims(vector) | number of dimensions
 vector_norm(vector) | Euclidean norm
 
+## Libraries
+
+Libraries that use pgvector:
+
+- [Neighbor](https://github.com/ankane/neighbor) (Ruby)
+
+## Upgrading
+
+### 0.1.1
+
+Follow the [installation instructions](#installation) and run:
+
+```sql
+ALTER EXTENSION vector UPDATE TO '0.1.1';
+```
+
 ## Thanks
 
 Thanks to:

expected/copy.out

@@ -0,0 +1,18 @@
+SET client_min_messages = warning;
+CREATE EXTENSION IF NOT EXISTS vector;
+CREATE TABLE t (val vector(3));
+INSERT INTO t (val) VALUES ('[0,0,0]'), ('[1,2,3]'), ('[1,1,1]'), (NULL);
+CREATE TABLE t2 (val vector(3));
+\copy t TO '/tmp/data.bin' WITH (FORMAT binary)
+\copy t2 FROM '/tmp/data.bin' WITH (FORMAT binary)
+SELECT * FROM t2 ORDER BY val;
+   val   
+---------
+ [0,0,0]
+ [1,1,1]
+ [1,2,3]
+ 
+(4 rows)
+
+DROP TABLE t;
+DROP TABLE t2;

sql/copy.sql

@@ -0,0 +1,15 @@
+SET client_min_messages = warning;
+CREATE EXTENSION IF NOT EXISTS vector;
+
+CREATE TABLE t (val vector(3));
+INSERT INTO t (val) VALUES ('[0,0,0]'), ('[1,2,3]'), ('[1,1,1]'), (NULL);
+
+CREATE TABLE t2 (val vector(3));
+
+\copy t TO '/tmp/data.bin' WITH (FORMAT binary)
+\copy t2 FROM '/tmp/data.bin' WITH (FORMAT binary)
+
+SELECT * FROM t2 ORDER BY val;
+
+DROP TABLE t;
+DROP TABLE t2;

src/vector.c

@@ -6,6 +6,7 @@
 #include "fmgr.h"
 #include "catalog/pg_type.h"
 #include "lib/stringinfo.h"
+#include "libpq/pqformat.h"
 #include "utils/array.h"
 #include "utils/builtins.h"
 #include "utils/lsyscache.h"
@@ -40,6 +41,21 @@ CheckExpectedDim(int32 typmod, int dim)
 				 errmsg("expected %d dimensions, not %d", typmod, dim)));
 }
 
+
+static inline void
+CheckDim(int dim)
+{
+	if (dim < 1)
+		ereport(ERROR,
+				(errcode(ERRCODE_DATA_EXCEPTION),
+				 errmsg("vector must have at least 1 dimension")));
+
+	if (dim > VECTOR_MAX_DIM)
+		ereport(ERROR,
+				(errcode(ERRCODE_PROGRAM_LIMIT_EXCEEDED),
+				 errmsg("vector cannot have more than %d dimensions", VECTOR_MAX_DIM)));
+}
+
 /*
  * Ensure finite elements
  */
@@ -217,6 +233,53 @@ vector_typmod_in(PG_FUNCTION_ARGS)
 	PG_RETURN_INT32(*tl);
 }
 
+/*
+ * Convert external binary representation to internal representation
+ */
+PG_FUNCTION_INFO_V1(vector_recv);
+Datum
+vector_recv(PG_FUNCTION_ARGS)
+{
+	StringInfo	buf = (StringInfo) PG_GETARG_POINTER(0);
+	int32		typmod = PG_GETARG_INT32(2);
+	Vector	   *result;
+	int16		dim;
+	int16		unused;
+	int			i;
+
+	dim = pq_getmsgint(buf, sizeof(int16));
+	unused = pq_getmsgint(buf, sizeof(int16));
+
+	CheckDim(dim);
+	CheckExpectedDim(typmod, dim);
+
+	result = InitVector(dim);
+	for (i = 0; i < dim; i++)
+		result->x[i] = pq_getmsgfloat4(buf);
+
+	PG_RETURN_POINTER(result);
+}
+
+/*
+ * Convert internal representation to the external binary representation
+ */
+PG_FUNCTION_INFO_V1(vector_send);
+Datum
+vector_send(PG_FUNCTION_ARGS)
+{
+	Vector	   *vec = PG_GETARG_VECTOR_P(0);
+	StringInfoData buf;
+	int			i;
+
+	pq_begintypsend(&buf);
+	pq_sendint(&buf, vec->dim, sizeof(int16));
+	pq_sendint(&buf, vec->unused, sizeof(int16));
+	for (i = 0; i < vec->dim; i++)
+		pq_sendfloat4(&buf, vec->x[i]);
+
+	PG_RETURN_BYTEA_P(pq_endtypsend(&buf));
+}
+
 /*
  * Convert vector to vector
  */
@@ -259,17 +322,7 @@ array_to_vector(PG_FUNCTION_ARGS)
 	deconstruct_array(array, ARR_ELEMTYPE(array), typlen, typbyval, typalign, &elemsp, &nullsp, &nelemsp);
 
 	if (typmod == -1)
-	{
-		if (nelemsp < 1)
-			ereport(ERROR,
-					(errcode(ERRCODE_DATA_EXCEPTION),
-					 errmsg("vector must have at least 1 dimension")));
-
-		if (nelemsp > VECTOR_MAX_DIM)
-			ereport(ERROR,
-					(errcode(ERRCODE_PROGRAM_LIMIT_EXCEEDED),
-					 errmsg("vector cannot have more than %d dimensions", VECTOR_MAX_DIM)));
-	}
+		CheckDim(nelemsp);
 	else
 		CheckExpectedDim(typmod, nelemsp);
 
@@ -308,11 +361,15 @@ l2_distance(PG_FUNCTION_ARGS)
 	Vector	   *a = PG_GETARG_VECTOR_P(0);
 	Vector	   *b = PG_GETARG_VECTOR_P(1);
 	double		distance = 0.0;
+	double		diff;
 
 	CheckDims(a, b);
 
 	for (int i = 0; i < a->dim; i++)
-		distance += pow(a->x[i] - b->x[i], 2);
+	{
+		diff = a->x[i] - b->x[i];
+		distance += diff * diff;
+	}
 
 	PG_RETURN_FLOAT8(sqrt(distance));
 }
@@ -328,11 +385,15 @@ vector_l2_squared_distance(PG_FUNCTION_ARGS)
 	Vector	   *a = PG_GETARG_VECTOR_P(0);
 	Vector	   *b = PG_GETARG_VECTOR_P(1);
 	double		distance = 0.0;
+	double		diff;
 
 	CheckDims(a, b);
 
 	for (int i = 0; i < a->dim; i++)
-		distance += pow(a->x[i] - b->x[i], 2);
+	{
+		diff = a->x[i] - b->x[i];
+		distance += diff * diff;
+	}
 
 	PG_RETURN_FLOAT8(distance);
 }
@@ -393,8 +454,8 @@ cosine_distance(PG_FUNCTION_ARGS)
 	for (int i = 0; i < a->dim; i++)
 	{
 		distance += a->x[i] * b->x[i];
-		norma += pow(a->x[i], 2);
-		normb += pow(b->x[i], 2);
+		norma += a->x[i] * a->x[i];
+		normb += b->x[i] * b->x[i];
 	}
 
 	PG_RETURN_FLOAT8(1 - (distance / (sqrt(norma) * sqrt(normb))));
@@ -450,7 +511,7 @@ vector_norm(PG_FUNCTION_ARGS)
 	double		norm = 0.0;
 
 	for (int i = 0; i < a->dim; i++)
-		norm += pow(a->x[i], 2);
+		norm += a->x[i] * a->x[i];
 
 	PG_RETURN_FLOAT8(sqrt(norm));
 }

vector--0.1.0--0.1.1.sql

@@ -0,0 +1,39 @@
+-- complain if script is sourced in psql, rather than via CREATE EXTENSION
+\echo Use "ALTER EXTENSION vector UPDATE TO '0.1.1'" to load this file. \quit
+
+CREATE FUNCTION vector_recv(internal, oid, integer) RETURNS vector
+	AS 'MODULE_PATHNAME' LANGUAGE C IMMUTABLE STRICT;
+
+CREATE FUNCTION vector_send(vector) RETURNS bytea
+	AS 'MODULE_PATHNAME' LANGUAGE C IMMUTABLE STRICT;
+
+ALTER TYPE vector SET ( RECEIVE = vector_recv, SEND = vector_send );
+
+-- functions
+
+ALTER FUNCTION vector_in(cstring, oid, integer) PARALLEL SAFE;
+ALTER FUNCTION vector_out(vector) PARALLEL SAFE;
+ALTER FUNCTION vector_typmod_in(cstring[]) PARALLEL SAFE;
+ALTER FUNCTION vector_recv(internal, oid, integer) PARALLEL SAFE;
+ALTER FUNCTION vector_send(vector) PARALLEL SAFE;
+ALTER FUNCTION l2_distance(vector, vector) PARALLEL SAFE;
+ALTER FUNCTION inner_product(vector, vector) PARALLEL SAFE;
+ALTER FUNCTION cosine_distance(vector, vector) PARALLEL SAFE;
+ALTER FUNCTION vector_dims(vector) PARALLEL SAFE;
+ALTER FUNCTION vector_norm(vector) PARALLEL SAFE;
+ALTER FUNCTION vector_add(vector, vector) PARALLEL SAFE;
+ALTER FUNCTION vector_sub(vector, vector) PARALLEL SAFE;
+ALTER FUNCTION vector_lt(vector, vector) PARALLEL SAFE;
+ALTER FUNCTION vector_le(vector, vector) PARALLEL SAFE;
+ALTER FUNCTION vector_eq(vector, vector) PARALLEL SAFE;
+ALTER FUNCTION vector_ne(vector, vector) PARALLEL SAFE;
+ALTER FUNCTION vector_ge(vector, vector) PARALLEL SAFE;
+ALTER FUNCTION vector_gt(vector, vector) PARALLEL SAFE;
+ALTER FUNCTION vector_cmp(vector, vector) PARALLEL SAFE;
+ALTER FUNCTION vector_l2_squared_distance(vector, vector) PARALLEL SAFE;
+ALTER FUNCTION vector_negative_inner_product(vector, vector) PARALLEL SAFE;
+ALTER FUNCTION vector_spherical_distance(vector, vector) PARALLEL SAFE;
+ALTER FUNCTION vector(vector, integer, boolean) PARALLEL SAFE;
+ALTER FUNCTION array_to_vector(integer[], integer, boolean) PARALLEL SAFE;
+ALTER FUNCTION array_to_vector(real[], integer, boolean) PARALLEL SAFE;
+ALTER FUNCTION array_to_vector(double precision[], integer, boolean) PARALLEL SAFE;

vector--0.1.1.sql

@@ -6,88 +6,96 @@
 CREATE TYPE vector;
 
 CREATE FUNCTION vector_in(cstring, oid, integer) RETURNS vector
-	AS 'MODULE_PATHNAME' LANGUAGE C IMMUTABLE STRICT;
+	AS 'MODULE_PATHNAME' LANGUAGE C IMMUTABLE STRICT PARALLEL SAFE;
 
 CREATE FUNCTION vector_out(vector) RETURNS cstring
-	AS 'MODULE_PATHNAME' LANGUAGE C IMMUTABLE STRICT;
+	AS 'MODULE_PATHNAME' LANGUAGE C IMMUTABLE STRICT PARALLEL SAFE;
 
 CREATE FUNCTION vector_typmod_in(cstring[]) RETURNS integer
-	AS 'MODULE_PATHNAME' LANGUAGE C IMMUTABLE STRICT;
+	AS 'MODULE_PATHNAME' LANGUAGE C IMMUTABLE STRICT PARALLEL SAFE;
+
+CREATE FUNCTION vector_recv(internal, oid, integer) RETURNS vector
+	AS 'MODULE_PATHNAME' LANGUAGE C IMMUTABLE STRICT PARALLEL SAFE;
+
+CREATE FUNCTION vector_send(vector) RETURNS bytea
+	AS 'MODULE_PATHNAME' LANGUAGE C IMMUTABLE STRICT PARALLEL SAFE;
 
 CREATE TYPE vector (
 	INPUT     = vector_in,
 	OUTPUT    = vector_out,
-	TYPMOD_IN = vector_typmod_in
+	TYPMOD_IN = vector_typmod_in,
+	RECEIVE   = vector_recv,
+	SEND      = vector_send
 );
 
 -- functions
 
 CREATE FUNCTION l2_distance(vector, vector) RETURNS float8
-	AS 'MODULE_PATHNAME' LANGUAGE C IMMUTABLE STRICT;
+	AS 'MODULE_PATHNAME' LANGUAGE C IMMUTABLE STRICT PARALLEL SAFE;
 
 CREATE FUNCTION inner_product(vector, vector) RETURNS float8
-	AS 'MODULE_PATHNAME' LANGUAGE C IMMUTABLE STRICT;
+	AS 'MODULE_PATHNAME' LANGUAGE C IMMUTABLE STRICT PARALLEL SAFE;
 
 CREATE FUNCTION cosine_distance(vector, vector) RETURNS float8
-	AS 'MODULE_PATHNAME' LANGUAGE C IMMUTABLE STRICT;
+	AS 'MODULE_PATHNAME' LANGUAGE C IMMUTABLE STRICT PARALLEL SAFE;
 
 CREATE FUNCTION vector_dims(vector) RETURNS integer
-	AS 'MODULE_PATHNAME' LANGUAGE C IMMUTABLE STRICT;
+	AS 'MODULE_PATHNAME' LANGUAGE C IMMUTABLE STRICT PARALLEL SAFE;
 
 CREATE FUNCTION vector_norm(vector) RETURNS float8
-	AS 'MODULE_PATHNAME' LANGUAGE C IMMUTABLE STRICT;
+	AS 'MODULE_PATHNAME' LANGUAGE C IMMUTABLE STRICT PARALLEL SAFE;
 
 CREATE FUNCTION vector_add(vector, vector) RETURNS vector
-	AS 'MODULE_PATHNAME' LANGUAGE C IMMUTABLE STRICT;
+	AS 'MODULE_PATHNAME' LANGUAGE C IMMUTABLE STRICT PARALLEL SAFE;
 
 CREATE FUNCTION vector_sub(vector, vector) RETURNS vector
-	AS 'MODULE_PATHNAME' LANGUAGE C IMMUTABLE STRICT;
+	AS 'MODULE_PATHNAME' LANGUAGE C IMMUTABLE STRICT PARALLEL SAFE;
 
 -- private functions
 
 CREATE FUNCTION vector_lt(vector, vector) RETURNS bool
-	AS 'MODULE_PATHNAME' LANGUAGE C IMMUTABLE STRICT;
+	AS 'MODULE_PATHNAME' LANGUAGE C IMMUTABLE STRICT PARALLEL SAFE;
 
 CREATE FUNCTION vector_le(vector, vector) RETURNS bool
-	AS 'MODULE_PATHNAME' LANGUAGE C IMMUTABLE STRICT;
+	AS 'MODULE_PATHNAME' LANGUAGE C IMMUTABLE STRICT PARALLEL SAFE;
 
 CREATE FUNCTION vector_eq(vector, vector) RETURNS bool
-	AS 'MODULE_PATHNAME' LANGUAGE C IMMUTABLE STRICT;
+	AS 'MODULE_PATHNAME' LANGUAGE C IMMUTABLE STRICT PARALLEL SAFE;
 
 CREATE FUNCTION vector_ne(vector, vector) RETURNS bool
-	AS 'MODULE_PATHNAME' LANGUAGE C IMMUTABLE STRICT;
+	AS 'MODULE_PATHNAME' LANGUAGE C IMMUTABLE STRICT PARALLEL SAFE;
 
 CREATE FUNCTION vector_ge(vector, vector) RETURNS bool
-	AS 'MODULE_PATHNAME' LANGUAGE C IMMUTABLE STRICT;
+	AS 'MODULE_PATHNAME' LANGUAGE C IMMUTABLE STRICT PARALLEL SAFE;
 
 CREATE FUNCTION vector_gt(vector, vector) RETURNS bool
-	AS 'MODULE_PATHNAME' LANGUAGE C IMMUTABLE STRICT;
+	AS 'MODULE_PATHNAME' LANGUAGE C IMMUTABLE STRICT PARALLEL SAFE;
 
 CREATE FUNCTION vector_cmp(vector, vector) RETURNS int4
-	AS 'MODULE_PATHNAME' LANGUAGE C IMMUTABLE STRICT;
+	AS 'MODULE_PATHNAME' LANGUAGE C IMMUTABLE STRICT PARALLEL SAFE;
 
 CREATE FUNCTION vector_l2_squared_distance(vector, vector) RETURNS float8
-	AS 'MODULE_PATHNAME' LANGUAGE C IMMUTABLE STRICT;
+	AS 'MODULE_PATHNAME' LANGUAGE C IMMUTABLE STRICT PARALLEL SAFE;
 
 CREATE FUNCTION vector_negative_inner_product(vector, vector) RETURNS float8
-	AS 'MODULE_PATHNAME' LANGUAGE C IMMUTABLE STRICT;
+	AS 'MODULE_PATHNAME' LANGUAGE C IMMUTABLE STRICT PARALLEL SAFE;
 
 CREATE FUNCTION vector_spherical_distance(vector, vector) RETURNS float8
-	AS 'MODULE_PATHNAME' LANGUAGE C IMMUTABLE STRICT;
+	AS 'MODULE_PATHNAME' LANGUAGE C IMMUTABLE STRICT PARALLEL SAFE;
 
 -- cast functions
 
 CREATE FUNCTION vector(vector, integer, boolean) RETURNS vector
-	AS 'MODULE_PATHNAME' LANGUAGE C IMMUTABLE STRICT;
+	AS 'MODULE_PATHNAME' LANGUAGE C IMMUTABLE STRICT PARALLEL SAFE;
 
 CREATE FUNCTION array_to_vector(integer[], integer, boolean) RETURNS vector
-	AS 'MODULE_PATHNAME' LANGUAGE C IMMUTABLE STRICT;
+	AS 'MODULE_PATHNAME' LANGUAGE C IMMUTABLE STRICT PARALLEL SAFE;
 
 CREATE FUNCTION array_to_vector(real[], integer, boolean) RETURNS vector
-	AS 'MODULE_PATHNAME' LANGUAGE C IMMUTABLE STRICT;
+	AS 'MODULE_PATHNAME' LANGUAGE C IMMUTABLE STRICT PARALLEL SAFE;
 
 CREATE FUNCTION array_to_vector(double precision[], integer, boolean) RETURNS vector
-	AS 'MODULE_PATHNAME' LANGUAGE C IMMUTABLE STRICT;
+	AS 'MODULE_PATHNAME' LANGUAGE C IMMUTABLE STRICT PARALLEL SAFE;
 
 -- casts
 

vector.control

@@ -1,4 +1,4 @@
 comment = 'vector data type and ivfflat access method'
-default_version = '0.1.0'
+default_version = '0.1.1'
 module_pathname = '$libdir/vector'
 relocatable = true