Version bump to 0.7.2 [skip ci]

Currently pgvector does not create any WAL records for unlogged tables

Postgres assumes INIT_FORK of unlogged tables is persistent and uses it
to reset the table index to its default empty state after a server
crash.

This patch makes INIT_FORK of unlogged table WAL-tracked, which ensures
an unlogged table is usable after a crash-restart

CHANGELOG.md

@@ -1,3 +1,7 @@
+## 0.7.2 (2024-06-11)
+
+- Fixed initialization fork for indexes on unlogged tables
+
 ## 0.7.1 (2024-06-03)
 
 - Improved performance of on-disk HNSW index builds

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": "0.7.1",
+	"version": "0.7.2",
 	"maintainer": [
 		"Andrew Kane <andrew@ankane.org>"
 	],
@@ -20,7 +20,7 @@
 		"vector": {
 			"file": "sql/vector.sql",
 			"docfile": "README.md",
-			"version": "0.7.1",
+			"version": "0.7.2",
 			"abstract": "Open-source vector similarity search for Postgres"
 		}
 	},

Makefile

@@ -1,5 +1,5 @@
 EXTENSION = vector
-EXTVERSION = 0.7.1
+EXTVERSION = 0.7.2
 
 MODULE_big = vector
 DATA = $(wildcard sql/*--*.sql)

Makefile.win

@@ -1,5 +1,5 @@
 EXTENSION = vector
-EXTVERSION = 0.7.1
+EXTVERSION = 0.7.2
 
 OBJS = src\bitutils.obj src\bitvec.obj src\halfutils.obj src\halfvec.obj src\hnsw.obj src\hnswbuild.obj src\hnswinsert.obj src\hnswscan.obj src\hnswutils.obj src\hnswvacuum.obj src\ivfbuild.obj src\ivfflat.obj src\ivfinsert.obj src\ivfkmeans.obj src\ivfscan.obj src\ivfutils.obj src\ivfvacuum.obj src\sparsevec.obj src\vector.obj
 HEADERS = src\halfvec.h src\sparsevec.h src\vector.h

README.md

@@ -21,7 +21,7 @@ Compile and install the extension (supports Postgres 12+)
 
 ```sh
 cd /tmp
-git clone --branch v0.7.1 https://github.com/pgvector/pgvector.git
+git clone --branch v0.7.2 https://github.com/pgvector/pgvector.git
 cd pgvector
 make
 make install # may need sudo
@@ -46,7 +46,7 @@ Then use `nmake` to build:
 ```cmd
 set "PGROOT=C:\Program Files\PostgreSQL\16"
 cd %TEMP%
-git clone --branch v0.7.1 https://github.com/pgvector/pgvector.git
+git clone --branch v0.7.2 https://github.com/pgvector/pgvector.git
 cd pgvector
 nmake /F Makefile.win
 nmake /F Makefile.win install
@@ -1051,7 +1051,7 @@ This adds pgvector to the [Postgres image](https://hub.docker.com/_/postgres) (r
 You can also build the image manually:
 
 ```sh
-git clone --branch v0.7.1 https://github.com/pgvector/pgvector.git
+git clone --branch v0.7.2 https://github.com/pgvector/pgvector.git
 cd pgvector
 docker build --pull --build-arg PG_MAJOR=16 -t myuser/pgvector .
 ```

sql/vector--0.7.1--0.7.2.sql

@@ -0,0 +1,2 @@
+-- complain if script is sourced in psql, rather than via CREATE EXTENSION
+\echo Use "ALTER EXTENSION vector UPDATE TO '0.7.2'" to load this file. \quit

src/hnswbuild.c

@@ -1121,8 +1121,8 @@ BuildIndex(Relation heap, Relation index, IndexInfo *indexInfo,
 
 	BuildGraph(buildstate, forkNum);
 
-	if (RelationNeedsWAL(index))
-		log_newpage_range(index, forkNum, 0, RelationGetNumberOfBlocks(index), true);
+	if (RelationNeedsWAL(index) || forkNum == INIT_FORKNUM)
+		log_newpage_range(index, forkNum, 0, RelationGetNumberOfBlocksInFork(index, forkNum), true);
 
 	FreeBuildState(buildstate);
 }

src/hnswutils.c

@@ -300,6 +300,9 @@ HnswGetMetaPageInfo(Relation index, int *m, HnswElement * entryPoint)
 	page = BufferGetPage(buf);
 	metap = HnswPageGetMeta(page);
 
+	if (unlikely(metap->magicNumber != HNSW_MAGIC_NUMBER))
+		elog(ERROR, "hnsw index is not valid");
+
 	if (m != NULL)
 		*m = metap->m;
 
@@ -796,15 +799,16 @@ HnswSearchLayer(char *base, Datum q, List *ep, int ef, int lc, Relation index, F
 			{
 				float		eDistance;
 				HnswElement eElement = HnswPtrAccess(base, e->element);
+				bool		alwaysAdd = wlen < ef;
 
 				f = ((HnswPairingHeapNode *) pairingheap_first(W))->inner;
 
 				if (index == NULL)
 					eDistance = GetCandidateDistance(base, e, q, procinfo, collation);
 				else
-					HnswLoadElement(eElement, &eDistance, &q, index, procinfo, collation, inserting, wlen >= ef ? &f->distance : NULL);
+					HnswLoadElement(eElement, &eDistance, &q, index, procinfo, collation, inserting, alwaysAdd ? NULL : &f->distance);
 
-				if (eDistance < f->distance || wlen < ef)
+				if (eDistance < f->distance || alwaysAdd)
 				{
 					HnswCandidate *ec;
 

src/ivfbuild.c

@@ -1006,6 +1006,10 @@ BuildIndex(Relation heap, Relation index, IndexInfo *indexInfo,
 	CreateListPages(index, buildstate->centers, buildstate->dimensions, buildstate->lists, forkNum, &buildstate->listInfo);
 	CreateEntryPages(buildstate, forkNum);
 
+	/* Write WAL for initialization fork since GenericXLog functions do not */
+	if (forkNum == INIT_FORKNUM)
+		log_newpage_range(index, forkNum, 0, RelationGetNumberOfBlocksInFork(index, forkNum), true);
+
 	FreeBuildState(buildstate);
 }
 

src/ivfinsert.c

@@ -94,6 +94,9 @@ InsertTuple(Relation index, Datum *values, bool *isnull, ItemPointer heap_tid, R
 		value = IvfflatNormValue(typeInfo, collation, value);
 	}
 
+	/* Ensure index is valid */
+	IvfflatGetMetaPageInfo(index, NULL, NULL);
+
 	/* Find the insert page - sets the page and list info */
 	FindInsertPage(index, values, &insertPage, &listInfo);
 	Assert(BlockNumberIsValid(insertPage));

src/ivfutils.c

@@ -170,7 +170,11 @@ IvfflatGetMetaPageInfo(Relation index, int *lists, int *dimensions)
 	page = BufferGetPage(buf);
 	metap = IvfflatPageGetMeta(page);
 
-	*lists = metap->lists;
+	if (unlikely(metap->magicNumber != IVFFLAT_MAGIC_NUMBER))
+		elog(ERROR, "ivfflat index is not valid");
+
+	if (lists != NULL)
+		*lists = metap->lists;
 
 	if (dimensions != NULL)
 		*dimensions = metap->dimensions;

vector.control

@@ -1,4 +1,4 @@
 comment = 'vector data type and ivfflat and hnsw access methods'
-default_version = '0.7.1'
+default_version = '0.7.2'
 module_pathname = '$libdir/vector'
 relocatable = true