Add pgvector extension to CloudBerry

This commit introduces the pgvector extension to CloudBerry,
enhancing the database with advanced vector processing capabilities.

The pgvector extension enables efficient storage, indexing,
and querying of vector data, opening up new possibilities for
high-performance similarity search, machine learning algorithms,
and other data analytics tasks.

src/ivfbuild.c

@@ -3,6 +3,7 @@
 #include <float.h>
 
 #include "catalog/index.h"
+#include "cdb/cdbvars.h"
 #include "ivfflat.h"
 #include "miscadmin.h"
 #include "storage/bufmgr.h"
@@ -430,7 +431,7 @@ ComputeCenters(IvfflatBuildState * buildstate)
 	/* Sample rows */
 	/* TODO Ensure within maintenance_work_mem */
 	buildstate->samples = VectorArrayInit(numSamples, buildstate->dimensions);
-	if (buildstate->heap != NULL)
+	if (buildstate->heap != NULL && Gp_role != GP_ROLE_DISPATCH)
 	{
 		SampleRows(buildstate);
 

test/expected/btree.out

@@ -1,5 +1,6 @@
 SET enable_seqscan = off;
 CREATE TABLE t (val vector(3));
+NOTICE:  Table doesn't have 'DISTRIBUTED BY' clause, and no column type is suitable for a distribution key. Creating a NULL policy entry.
 INSERT INTO t (val) VALUES ('[0,0,0]'), ('[1,2,3]'), ('[1,1,1]'), (NULL);
 CREATE INDEX ON t (val);
 SELECT * FROM t WHERE val = '[1,2,3]';

test/expected/copy.out

@@ -1,6 +1,8 @@
 CREATE TABLE t (val vector(3));
+NOTICE:  Table doesn't have 'DISTRIBUTED BY' clause, and no column type is suitable for a distribution key. Creating a NULL policy entry.
 INSERT INTO t (val) VALUES ('[0,0,0]'), ('[1,2,3]'), ('[1,1,1]'), (NULL);
 CREATE TABLE t2 (val vector(3));
+NOTICE:  Table doesn't have 'DISTRIBUTED BY' clause, and no column type is suitable for a distribution key. Creating a NULL policy entry.
 \copy t TO 'results/data.bin' WITH (FORMAT binary)
 \copy t2 FROM 'results/data.bin' WITH (FORMAT binary)
 SELECT * FROM t2 ORDER BY val;

test/expected/ivfflat_cosine.out

@@ -1,7 +1,11 @@
 SET enable_seqscan = off;
+SET optimizer = off;
 CREATE TABLE t (val vector(3));
+NOTICE:  Table doesn't have 'DISTRIBUTED BY' clause, and no column type is suitable for a distribution key. Creating a NULL policy entry.
 INSERT INTO t (val) VALUES ('[0,0,0]'), ('[1,2,3]'), ('[1,1,1]'), (NULL);
+-- start_ignore
 CREATE INDEX ON t USING ivfflat (val vector_cosine_ops) WITH (lists = 1);
+-- end_ignore
 INSERT INTO t (val) VALUES ('[1,2,4]');
 SELECT * FROM t ORDER BY val <=> '[3,3,3]';
    val   

test/expected/ivfflat_ip.out

@@ -1,7 +1,14 @@
 SET enable_seqscan = off;
+SET optimizer = off;
 CREATE TABLE t (val vector(3));
+NOTICE:  Table doesn't have 'DISTRIBUTED BY' clause, and no column type is suitable for a distribution key. Creating a NULL policy entry.
 INSERT INTO t (val) VALUES ('[0,0,0]'), ('[1,2,3]'), ('[1,1,1]'), (NULL);
+-- start_ignore
 CREATE INDEX ON t USING ivfflat (val vector_ip_ops) WITH (lists = 1);
+NOTICE:  ivfflat index created with little data  (seg1 127.0.1.1:7003 pid=424029)
+DETAIL:  This will cause low recall.
+HINT:  Drop the index until the table has more data.
+-- end_ignore
 INSERT INTO t (val) VALUES ('[1,2,4]');
 SELECT * FROM t ORDER BY val <#> '[3,3,3]';
    val   

test/expected/ivfflat_l2.out

@@ -1,7 +1,11 @@
 SET enable_seqscan = off;
+SET optimizer = off;
 CREATE TABLE t (val vector(3));
+NOTICE:  Table doesn't have 'DISTRIBUTED BY' clause, and no column type is suitable for a distribution key. Creating a NULL policy entry.
 INSERT INTO t (val) VALUES ('[0,0,0]'), ('[1,2,3]'), ('[1,1,1]'), (NULL);
+-- start_ignore
 CREATE INDEX ON t USING ivfflat (val) WITH (lists = 1);
+-- end_ignore
 INSERT INTO t (val) VALUES ('[1,2,4]');
 SELECT * FROM t ORDER BY val <-> '[3,3,3]';
    val   

test/expected/ivfflat_options.out

@@ -1,5 +1,6 @@
 SET enable_seqscan = off;
 CREATE TABLE t (val vector(3));
+NOTICE:  Table doesn't have 'DISTRIBUTED BY' clause, and no column type is suitable for a distribution key. Creating a NULL policy entry.
 CREATE INDEX ON t USING ivfflat (val) WITH (lists = 0);
 ERROR:  value 0 out of bounds for option "lists"
 DETAIL:  Valid values are between "1" and "32768".

test/expected/ivfflat_unlogged.out

@@ -1,7 +1,11 @@
 SET enable_seqscan = off;
+SET optimizer = off;
 CREATE UNLOGGED TABLE t (val vector(3));
+NOTICE:  Table doesn't have 'DISTRIBUTED BY' clause, and no column type is suitable for a distribution key. Creating a NULL policy entry.
 INSERT INTO t (val) VALUES ('[0,0,0]'), ('[1,2,3]'), ('[1,1,1]'), (NULL);
+-- start_ignore
 CREATE INDEX ON t USING ivfflat (val) WITH (lists = 1);
+-- end_ignore
 SELECT * FROM t ORDER BY val <-> '[3,3,3]';
    val   
 ---------

test/sql/ivfflat_cosine.sql

@@ -1,8 +1,11 @@
 SET enable_seqscan = off;
+SET optimizer = off;
 
 CREATE TABLE t (val vector(3));
 INSERT INTO t (val) VALUES ('[0,0,0]'), ('[1,2,3]'), ('[1,1,1]'), (NULL);
+-- start_ignore
 CREATE INDEX ON t USING ivfflat (val vector_cosine_ops) WITH (lists = 1);
+-- end_ignore
 
 INSERT INTO t (val) VALUES ('[1,2,4]');
 

test/sql/ivfflat_ip.sql

@@ -1,8 +1,11 @@
 SET enable_seqscan = off;
+SET optimizer = off;
 
 CREATE TABLE t (val vector(3));
 INSERT INTO t (val) VALUES ('[0,0,0]'), ('[1,2,3]'), ('[1,1,1]'), (NULL);
+-- start_ignore
 CREATE INDEX ON t USING ivfflat (val vector_ip_ops) WITH (lists = 1);
+-- end_ignore
 
 INSERT INTO t (val) VALUES ('[1,2,4]');
 

test/sql/ivfflat_l2.sql

@@ -1,8 +1,11 @@
 SET enable_seqscan = off;
+SET optimizer = off;
 
 CREATE TABLE t (val vector(3));
 INSERT INTO t (val) VALUES ('[0,0,0]'), ('[1,2,3]'), ('[1,1,1]'), (NULL);
+-- start_ignore
 CREATE INDEX ON t USING ivfflat (val) WITH (lists = 1);
+-- end_ignore
 
 INSERT INTO t (val) VALUES ('[1,2,4]');
 

test/sql/ivfflat_unlogged.sql

@@ -1,9 +1,11 @@
 SET enable_seqscan = off;
+SET optimizer = off;
 
 CREATE UNLOGGED TABLE t (val vector(3));
 INSERT INTO t (val) VALUES ('[0,0,0]'), ('[1,2,3]'), ('[1,1,1]'), (NULL);
+-- start_ignore
 CREATE INDEX ON t USING ivfflat (val) WITH (lists = 1);
-
+-- end_ignore
 SELECT * FROM t ORDER BY val <-> '[3,3,3]';
 
 DROP TABLE t;