Version bump to 0.2.7 [skip ci]

This reverts commit 94eeb83abe.

.github/workflows/build.yml

@@ -33,6 +33,6 @@ jobs:
     - if: ${{ startsWith(matrix.os, 'macos') }}
       run: |
         brew install cpanm && cpanm IPC::Run
-        wget -q https://github.com/postgres/postgres/archive/refs/tags/REL_14_1.tar.gz
-        tar xf REL_14_1.tar.gz
-        make prove_installcheck PROVE=prove PERL5LIB=postgres-REL_14_1/src/test/perl
+        wget -q https://github.com/postgres/postgres/archive/refs/tags/REL_14_4.tar.gz
+        tar xf REL_14_4.tar.gz
+        make prove_installcheck PROVE=prove PERL5LIB="postgres-REL_14_4/src/test/perl:/Users/runner/perl5/lib/perl5"

CHANGELOG.md

@@ -1,4 +1,8 @@
-## 0.2.6 (unreleased)
+## 0.2.7 (2022-07-31)
+
+- Fixed `unexpected data beyond EOF` error
+
+## 0.2.6 (2022-05-22)
 
 - Improved performance of index creation for Postgres < 12
 

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": "0.2.5",
+	"version": "0.2.7",
 	"maintainer": [
 		"Andrew Kane <andrew@ankane.org>"
 	],
@@ -20,7 +20,7 @@
 		"vector": {
 			"file": "sql/vector.sql",
 			"docfile": "README.md",
-			"version": "0.2.5",
+			"version": "0.2.7",
 			"abstract": "Open-source vector similarity search for Postgres"
 		}
 	},

Makefile

@@ -1,5 +1,5 @@
 EXTENSION = vector
-EXTVERSION = 0.2.5
+EXTVERSION = 0.2.7
 
 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 v0.2.5 https://github.com/pgvector/pgvector.git
+git clone --branch v0.2.7 https://github.com/pgvector/pgvector.git
 cd pgvector
 make
 make install # may need sudo
@@ -220,14 +220,14 @@ This adds pgvector to the [Postgres image](https://hub.docker.com/_/postgres).
 You can also build the image manually
 
 ```sh
-git clone --branch v0.2.5 https://github.com/pgvector/pgvector.git
+git clone --branch v0.2.7 https://github.com/pgvector/pgvector.git
 cd pgvector
 docker build -t pgvector .
 ```
 
 ### Homebrew
 
-On Mac with Homebrew Postgres, you can use:
+With Homebrew Postgres, you can use:
 
 ```sh
 brew install pgvector/brew/pgvector

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

sql/vector--0.2.6--0.2.7.sql

@@ -0,0 +1,2 @@
+-- complain if script is sourced in psql, rather than via CREATE EXTENSION
+\echo Use "ALTER EXTENSION vector UPDATE TO '0.2.7'" to load this file. \quit

src/ivfinsert.c

@@ -53,18 +53,6 @@ FindInsertPage(Relation rel, Datum *values, BlockNumber *insertPage, ListInfo *
 	}
 }
 
-/*
- * Prepare to insert an index tuple
- */
-static void
-LoadInsertPage(Relation index, Buffer *buf, Page *page, GenericXLogState **state, BlockNumber insertPage)
-{
-	*buf = ReadBuffer(index, insertPage);
-	LockBuffer(*buf, BUFFER_LOCK_EXCLUSIVE);
-	*state = GenericXLogStart(index);
-	*page = GenericXLogRegisterBuffer(*state, *buf, 0);
-}
-
 /*
  * Insert a tuple into the index
  */
@@ -87,11 +75,18 @@ InsertTuple(Relation rel, IndexTuple itup, Relation heapRel, Datum *values)
 	itemsz = MAXALIGN(IndexTupleSize(itup));
 	Assert(itemsz <= BLCKSZ - MAXALIGN(SizeOfPageHeaderData) - MAXALIGN(sizeof(IvfflatPageOpaqueData)));
 
-	LoadInsertPage(rel, &buf, &page, &state, insertPage);
-
 	/* Find a page to insert the item */
-	while (PageGetFreeSpace(page) < itemsz)
+	for (;;)
 	{
+		buf = ReadBuffer(rel, insertPage);
+		LockBuffer(buf, BUFFER_LOCK_EXCLUSIVE);
+
+		state = GenericXLogStart(rel);
+		page = GenericXLogRegisterBuffer(state, buf, 0);
+
+		if (PageGetFreeSpace(page) >= itemsz)
+			break;
+
 		insertPage = IvfflatPageGetOpaque(page)->nextblkno;
 
 		if (BlockNumberIsValid(insertPage))
@@ -99,15 +94,31 @@ InsertTuple(Relation rel, IndexTuple itup, Relation heapRel, Datum *values)
 			/* Move to next page */
 			GenericXLogAbort(state);
 			UnlockReleaseBuffer(buf);
-
-			LoadInsertPage(rel, &buf, &page, &state, insertPage);
 		}
 		else
 		{
 			/* Add a new page */
-			IvfflatAppendPage(rel, &buf, &page, &state, MAIN_FORKNUM);
+			Buffer		newbuf = IvfflatNewBuffer(rel, MAIN_FORKNUM);
+			Page		newpage = GenericXLogRegisterBuffer(state, buf, GENERIC_XLOG_FULL_IMAGE);
 
-			insertPage = BufferGetBlockNumber(buf);
+			insertPage = BufferGetBlockNumber(newbuf);
+
+			/* Update previous buffer */
+			IvfflatPageGetOpaque(page)->nextblkno = insertPage;
+
+			/* Init page */
+			PageInit(newpage, BufferGetPageSize(newbuf), sizeof(IvfflatPageOpaqueData));
+			IvfflatPageGetOpaque(newpage)->nextblkno = InvalidBlockNumber;
+			IvfflatPageGetOpaque(newpage)->page_id = IVFFLAT_PAGE_ID;
+
+			/* Commit */
+			MarkBufferDirty(buf);
+			MarkBufferDirty(newbuf);
+			GenericXLogFinish(state);
+
+			/* Unlock */
+			UnlockReleaseBuffer(buf);
+			UnlockReleaseBuffer(newbuf);
 		}
 	}
 

src/ivfutils.c

@@ -135,17 +135,29 @@ IvfflatCommitBuffer(Buffer buf, GenericXLogState *state)
 void
 IvfflatAppendPage(Relation index, Buffer *buf, Page *page, GenericXLogState **state, ForkNumber forkNum)
 {
-	Buffer		prevbuf = *buf;
-
 	/* Get new buffer */
-	*buf = IvfflatNewBuffer(index, forkNum);
+	Buffer		newbuf = IvfflatNewBuffer(index, forkNum);
+	Page		newpage = GenericXLogRegisterBuffer(*state, newbuf, GENERIC_XLOG_FULL_IMAGE);
 
-	/* Update and commit previous buffer */
-	IvfflatPageGetOpaque(*page)->nextblkno = BufferGetBlockNumber(*buf);
-	IvfflatCommitBuffer(prevbuf, *state);
+	/* Update the previous buffer */
+	IvfflatPageGetOpaque(*page)->nextblkno = BufferGetBlockNumber(newbuf);
 
 	/* Init new page */
-	IvfflatInitPage(index, buf, page, state);
+	PageInit(newpage, BufferGetPageSize(newbuf), sizeof(IvfflatPageOpaqueData));
+	IvfflatPageGetOpaque(newpage)->nextblkno = InvalidBlockNumber;
+	IvfflatPageGetOpaque(newpage)->page_id = IVFFLAT_PAGE_ID;
+
+	/* Commit */
+	MarkBufferDirty(*buf);
+	MarkBufferDirty(newbuf);
+	GenericXLogFinish(*state);
+
+	/* Unlock */
+	UnlockReleaseBuffer(*buf);
+
+	*state = GenericXLogStart(index);
+	*page = GenericXLogRegisterBuffer(*state, newbuf, GENERIC_XLOG_FULL_IMAGE);
+	*buf = newbuf;
 }
 
 /*

test/t/001_wal.pl

@@ -7,6 +7,8 @@ use PostgresNode;
 use TestLib;
 use Test::More tests => 31;
 
+my $dim = 32;
+
 my $node_primary;
 my $node_replica;
 
@@ -30,13 +32,15 @@ sub test_index_replay
 	$node_primary->poll_query_until('postgres', $caughtup_query)
 	  or die "Timed out while waiting for replica 1 to catch up";
 
-	my $r1 = rand();
-	my $r2 = rand();
-	my $r3 = rand();
+	my @r = ();
+	for (1 .. $dim) {
+		push(@r, rand());
+	}
+	my $sql = join(",", @r);
 
 	my $queries = qq(
 		SET enable_seqscan = off;
-		SELECT * FROM tst ORDER BY v <-> '[$r1,$r2,$r3]' LIMIT 10;
+		SELECT * FROM tst ORDER BY v <-> '[$sql]' LIMIT 10;
 	);
 
 	# Run test queries and compare their result
@@ -50,6 +54,10 @@ sub test_index_replay
 # Initialize primary node
 $node_primary = get_new_node('primary');
 $node_primary->init(allows_streaming => 1);
+if ($dim > 32) {
+	# TODO use wal_keep_segments for Postgres < 13
+	$node_primary->append_conf('postgresql.conf', qq(wal_keep_size = 1GB));
+}
 $node_primary->start;
 my $backup_name = 'my_backup';
 
@@ -64,9 +72,9 @@ $node_replica->start;
 
 # Create ivfflat index on primary
 $node_primary->safe_psql("postgres", "CREATE EXTENSION vector;");
-$node_primary->safe_psql("postgres", "CREATE TABLE tst (i int4, v vector(3));");
+$node_primary->safe_psql("postgres", "CREATE TABLE tst (i int4, v vector($dim));");
 $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, (SELECT array_agg(random()) FROM generate_series(1, $dim)) FROM generate_series(1, 100000) i;"
 );
 $node_primary->safe_psql("postgres", "CREATE INDEX ON tst USING ivfflat (v);");
 
@@ -82,7 +90,7 @@ 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, (SELECT array_agg(random()) FROM generate_series(1, $dim)) FROM generate_series($start, $end) i;"
 	);
 	test_index_replay("insert $i");
 }

test/t/007_inserts.pl

@@ -0,0 +1,33 @@
+use strict;
+use warnings;
+use PostgresNode;
+use TestLib;
+use Test::More tests => 3;
+
+# Initialize node
+my $node = get_new_node('node');
+$node->init;
+$node->start;
+
+# Create table and index
+$node->safe_psql("postgres", "CREATE EXTENSION vector;");
+$node->safe_psql("postgres", "CREATE TABLE tst (v vector(768));");
+$node->safe_psql("postgres",
+	"INSERT INTO tst SELECT (SELECT array_agg(random()) FROM generate_series(1, 768)) FROM generate_series(1, 10000) i;"
+);
+$node->safe_psql("postgres", "CREATE INDEX ON tst USING ivfflat (v);");
+
+$node->pgbench(
+	"--no-vacuum --client=5 --transactions=100",
+	0,
+	[qr{actually processed}],
+	[qr{^$}],
+	"concurrent INSERTs",
+	{
+		"007_concurrent" => q(
+			BEGIN;
+			INSERT INTO tst SELECT (SELECT array_agg(random()) FROM generate_series(1, 768)) FROM generate_series(1, 10) i;
+			COMMIT;
+		),
+	}
+);

vector.control

@@ -1,4 +1,4 @@
 comment = 'vector data type and ivfflat access method'
-default_version = '0.2.5'
+default_version = '0.2.7'
 module_pathname = '$libdir/vector'
 relocatable = true