Added support for adaptive beam search [skip ci]

.github/workflows/build.yml

@@ -23,7 +23,7 @@ jobs:
           - postgres: 13
             os: ubuntu-22.04
     steps:
-      - uses: actions/checkout@v6
+      - uses: actions/checkout@v5
       - uses: ankane/setup-postgres@v1
         with:
           postgres-version: ${{ matrix.postgres }}
@@ -53,7 +53,7 @@ jobs:
           - postgres: 14
             os: macos-15-intel
     steps:
-      - uses: actions/checkout@v6
+      - uses: actions/checkout@v5
       - uses: ankane/setup-postgres@v1
         with:
           postgres-version: ${{ matrix.postgres }}
@@ -72,7 +72,7 @@ jobs:
           tar xf $TAG.tar.gz
           mv postgres-$TAG postgres
         env:
-          TAG: ${{ matrix.postgres == 18 && 'REL_18_2' || 'REL_14_21' }}
+          TAG: ${{ matrix.postgres == 18 && 'REL_18_0' || 'REL_14_19' }}
       - run: make prove_installcheck PROVE_FLAGS="-I ./postgres/src/test/perl -I ./test/perl"
         env:
           PERL5LIB: /Users/runner/perl5/lib/perl5
@@ -92,12 +92,12 @@ jobs:
           - postgres: 14
             os: windows-2022
     steps:
-      - uses: actions/checkout@v6
+      - uses: actions/checkout@v5
       - uses: ankane/setup-postgres@v1
         with:
           postgres-version: ${{ matrix.postgres }}
       - run: |
-          call "C:\Program Files\Microsoft Visual Studio\${{ matrix.os == 'windows-2025' && 18 || 2022 }}\Enterprise\VC\Auxiliary\Build\vcvars64.bat" && ^
+          call "C:\Program Files\Microsoft Visual Studio\2022\Enterprise\VC\Auxiliary\Build\vcvars64.bat" && ^
           nmake /NOLOGO /F Makefile.win && ^
           nmake /NOLOGO /F Makefile.win install && ^
           nmake /NOLOGO /F Makefile.win installcheck ${{ matrix.postgres != 17 && 'PG_REGRESS=$(PGROOT)\bin\pg_regress' || '' }} && ^
@@ -133,7 +133,7 @@ jobs:
     if: ${{ !startsWith(github.ref_name, 'mac') && !startsWith(github.ref_name, 'windows') }}
     runs-on: ubuntu-latest
     steps:
-      - uses: actions/checkout@v6
+      - uses: actions/checkout@v5
       - uses: ankane/setup-postgres-valgrind@v1
         with:
           postgres-version: 18

CHANGELOG.md

@@ -1,11 +1,5 @@
-## 0.8.3 (2026-06-17)
+## 0.8.2 (unreleased)
 
-- Fixed possible index corruption with HNSW vacuuming
-- Fixed performance regression with Hamming distance and Jaccard distance with Postgres 18
-
-## 0.8.2 (2026-02-25)
-
-- Fixed buffer overflow with parallel HNSW index build - [more info](https://github.com/pgvector/pgvector/issues/959)
 - Improved `install` target on Windows
 - Fixed `Index Searches` in `EXPLAIN` output for Postgres 18
 

Dockerfile

@@ -5,7 +5,7 @@ ARG DEBIAN_CODENAME=bookworm
 FROM postgres:$PG_MAJOR-$DEBIAN_CODENAME
 ARG PG_MAJOR
 
-ADD https://github.com/pgvector/pgvector.git#v0.8.3 /tmp/pgvector
+ADD https://github.com/pgvector/pgvector.git#v0.8.1 /tmp/pgvector
 
 RUN apt-get update && \
 		apt-mark hold locales && \

LICENSE

@@ -1,4 +1,4 @@
-Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group
+Portions Copyright (c) 1996-2025, PostgreSQL Global Development Group
 
 Portions Copyright (c) 1994, The Regents of the University of California
 

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": "0.8.3",
+	"version": "0.8.1",
 	"maintainer": [
 		"Andrew Kane <andrew@ankane.org>"
 	],
@@ -20,7 +20,7 @@
 		"vector": {
 			"file": "sql/vector.sql",
 			"docfile": "README.md",
-			"version": "0.8.3",
+			"version": "0.8.1",
 			"abstract": "Open-source vector similarity search for Postgres"
 		}
 	},

Makefile

@@ -1,5 +1,5 @@
 EXTENSION = vector
-EXTVERSION = 0.8.3
+EXTVERSION = 0.8.1
 
 MODULE_big = vector
 DATA = $(wildcard sql/*--*--*.sql)
@@ -27,11 +27,6 @@ ifneq ($(filter ppc64%, $(shell uname -m)), )
 	OPTFLAGS =
 endif
 
-# RISC-V64 doesn't support -march=native
-ifeq ($(shell uname -m), riscv64)
-	OPTFLAGS =
-endif
-
 # For auto-vectorization:
 # - GCC (needs -ftree-vectorize OR -O3) - https://gcc.gnu.org/projects/tree-ssa/vectorization.html
 # - Clang (could use pragma instead) - https://llvm.org/docs/Vectorizers.html

Makefile.win

@@ -1,5 +1,5 @@
 EXTENSION = vector
-EXTVERSION = 0.8.3
+EXTVERSION = 0.8.1
 
 DATA_built = sql\$(EXTENSION)--$(EXTVERSION).sql
 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

README.md

@@ -11,8 +11,6 @@ Store your vectors with the rest of your data. Supports:
 
 Plus [ACID](https://en.wikipedia.org/wiki/ACID) compliance, point-in-time recovery, JOINs, and all of the other [great features](https://www.postgresql.org/about/) of Postgres
 
-Have a lot of vectors? Use [quantization](#scaling) to scale
-
 [![Build Status](https://github.com/pgvector/pgvector/actions/workflows/build.yml/badge.svg)](https://github.com/pgvector/pgvector/actions)
 
 ## Installation
@@ -23,7 +21,7 @@ Compile and install the extension (supports Postgres 13+)
 
 ```sh
 cd /tmp
-git clone --branch v0.8.3 https://github.com/pgvector/pgvector.git
+git clone --branch v0.8.1 https://github.com/pgvector/pgvector.git
 cd pgvector
 make
 make install # may need sudo
@@ -40,7 +38,7 @@ Ensure [C++ support in Visual Studio](https://learn.microsoft.com/en-us/cpp/buil
 ```cmd
 set "PGROOT=C:\Program Files\PostgreSQL\18"
 cd %TEMP%
-git clone --branch v0.8.3 https://github.com/pgvector/pgvector.git
+git clone --branch v0.8.1 https://github.com/pgvector/pgvector.git
 cd pgvector
 nmake /F Makefile.win
 nmake /F Makefile.win install
@@ -316,8 +314,6 @@ For a large number of workers, you may need to increase `max_parallel_workers` (
 
 The [index options](#index-options) also have a significant impact on build time (use the defaults unless seeing low recall)
 
-Use [binary quantization](#binary-quantization) for faster build times at scale
-
 ### Indexing Progress
 
 Check [indexing progress](https://www.postgresql.org/docs/current/progress-reporting.html#CREATE-INDEX-PROGRESS-REPORTING)
@@ -447,7 +443,13 @@ Exact indexes work well for conditions that match a low percentage of rows. Othe
 CREATE INDEX ON items USING hnsw (embedding vector_l2_ops);
 ```
 
-With approximate indexes, filtering is applied *after* the index is scanned. If a condition matches 10% of rows, with HNSW and the default `hnsw.ef_search` of 40, only 4 rows will match on average. For more rows, enable [iterative index scans](#iterative-index-scans), which will automatically scan more of the index when needed.
+With approximate indexes, filtering is applied *after* the index is scanned. If a condition matches 10% of rows, with HNSW and the default `hnsw.ef_search` of 40, only 4 rows will match on average. For more rows, increase `hnsw.ef_search`.
+
+```sql
+SET hnsw.ef_search = 200;
+```
+
+Starting with 0.8.0, you can enable [iterative index scans](#iterative-index-scans), which will automatically scan more of the index when needed.
 
 ```sql
 SET hnsw.iterative_scan = strict_order;
@@ -671,10 +673,6 @@ SHOW shared_buffers;
 
 Be sure to restart Postgres for changes to take effect.
 
-### Storing
-
-Use the `halfvec` type instead of `vector` for a smaller working set.
-
 ### Loading
 
 Use `COPY` for bulk loading data ([example](https://github.com/pgvector/pgvector-python/blob/master/examples/loading/example.py)).
@@ -689,8 +687,6 @@ Add any indexes *after* loading the initial data for best performance.
 
 See index build time for [HNSW](#index-build-time) and [IVFFlat](#index-build-time-1).
 
-Use [binary quantization](#binary-quantization) for smaller indexes and faster build times at scale.
-
 In production environments, create indexes concurrently to avoid blocking writes.
 
 ```sql
@@ -721,8 +717,6 @@ SELECT * FROM items ORDER BY embedding <#> '[3,1,2]' LIMIT 5;
 
 #### Approximate Search
 
-Use [binary quantization](#binary-quantization) with re-ranking to keep indexes in-memory at scale.
-
 To speed up queries with an IVFFlat index, increase the number of inverted lists (at the expense of recall).
 
 ```sql
@@ -738,20 +732,21 @@ REINDEX INDEX CONCURRENTLY index_name;
 VACUUM table_name;
 ```
 
-## Scaling
-
-For a smaller working set:
-
-1. Use the `halfvec` type instead of `vector` for tables
-2. Use [binary quantization](#binary-quantization) for indexes (with re-ranking for search)
-
-Scale vertically by increasing memory, CPU, and storage on a single instance. Use existing tools to [tune parameters](#tuning) and [monitor performance](#monitoring).
-
-Scale horizontally with [replicas](https://www.postgresql.org/docs/current/hot-standby.html), or use [Citus](https://github.com/citusdata/citus), [PgDog](https://github.com/pgdogdev/pgdog), or another approach for sharding ([example](https://github.com/pgvector/pgvector-python/blob/master/examples/citus/example.py)).
-
 ## Monitoring
 
-Use existing tools like [pg_stat_statements](https://www.postgresql.org/docs/current/pgstatstatements.html) or [PgHero](https://github.com/ankane/pghero) to monitor performance.
+Monitor performance with [pg_stat_statements](https://www.postgresql.org/docs/current/pgstatstatements.html) (be sure to add it to `shared_preload_libraries`).
+
+```sql
+CREATE EXTENSION pg_stat_statements;
+```
+
+Get the most time-consuming queries with:
+
+```sql
+SELECT query, calls, ROUND((total_plan_time + total_exec_time) / calls) AS avg_time_ms,
+    ROUND((total_plan_time + total_exec_time) / 60000) AS total_time_min
+    FROM pg_stat_statements ORDER BY total_plan_time + total_exec_time DESC LIMIT 20;
+```
 
 Monitor recall by comparing results from approximate search with exact search.
 
@@ -762,6 +757,14 @@ SELECT ...
 COMMIT;
 ```
 
+## Scaling
+
+Scale pgvector the same way you scale Postgres.
+
+Scale vertically by increasing memory, CPU, and storage on a single instance. Use existing tools to [tune parameters](#tuning) and [monitor performance](#monitoring).
+
+Scale horizontally with [replicas](https://www.postgresql.org/docs/current/hot-standby.html), or use [Citus](https://github.com/citusdata/citus) or another approach for sharding ([example](https://github.com/pgvector/pgvector-python/blob/master/examples/citus/example.py)).
+
 ## Languages
 
 Use pgvector from any language with a Postgres client. You can even generate and store vectors in one language and query them in another.
@@ -773,7 +776,6 @@ Algol | [pgvector-algol](https://github.com/pgvector/pgvector-algol)
 C | [pgvector-c](https://github.com/pgvector/pgvector-c)
 C++ | [pgvector-cpp](https://github.com/pgvector/pgvector-cpp)
 C#, F#, Visual Basic | [pgvector-dotnet](https://github.com/pgvector/pgvector-dotnet)
-COBOL | [pgvector-cobol](https://github.com/pgvector/pgvector-cobol)
 Crystal | [pgvector-crystal](https://github.com/pgvector/pgvector-crystal)
 D | [pgvector-d](https://github.com/pgvector/pgvector-d)
 Dart | [pgvector-dart](https://github.com/pgvector/pgvector-dart)
@@ -801,7 +803,6 @@ Raku | [pgvector-raku](https://github.com/pgvector/pgvector-raku)
 Ruby | [pgvector-ruby](https://github.com/pgvector/pgvector-ruby), [Neighbor](https://github.com/ankane/neighbor)
 Rust | [pgvector-rust](https://github.com/pgvector/pgvector-rust)
 Swift | [pgvector-swift](https://github.com/pgvector/pgvector-swift)
-Tcl | [pgvector-tcl](https://github.com/pgvector/pgvector-tcl)
 Zig | [pgvector-zig](https://github.com/pgvector/pgvector-zig)
 
 ## Frequently Asked Questions
@@ -875,8 +876,6 @@ No, but like other index types, you’ll likely see better performance if they d
 SELECT pg_size_pretty(pg_relation_size('index_name'));
 ```
 
-Use [half-precision indexing](#half-precision-indexing) or [binary quantization](#binary-quantization) for smaller indexes.
-
 ## Troubleshooting
 
 #### Why isn’t a query using an index?
@@ -1151,23 +1150,23 @@ This adds pgvector to the [Postgres image](https://hub.docker.com/_/postgres) (r
 
 Supported tags are:
 
-- `pg18-trixie`, `0.8.3-pg18-trixie`
+- `pg18-trixie`, `0.8.1-pg18-trixie`
-- `pg18-bookworm`, `0.8.3-pg18-bookworm`, `pg18`, `0.8.3-pg18`
+- `pg18-bookworm`, `0.8.1-pg18-bookworm`, `pg18`, `0.8.1-pg18`
-- `pg17-trixie`, `0.8.3-pg17-trixie`
+- `pg17-trixie`, `0.8.1-pg17-trixie`
-- `pg17-bookworm`, `0.8.3-pg17-bookworm`, `pg17`, `0.8.3-pg17`
+- `pg17-bookworm`, `0.8.1-pg17-bookworm`, `pg17`, `0.8.1-pg17`
-- `pg16-trixie`, `0.8.3-pg16-trixie`
+- `pg16-trixie`, `0.8.1-pg16-trixie`
-- `pg16-bookworm`, `0.8.3-pg16-bookworm`, `pg16`, `0.8.3-pg16`
+- `pg16-bookworm`, `0.8.1-pg16-bookworm`, `pg16`, `0.8.1-pg16`
-- `pg15-trixie`, `0.8.3-pg15-trixie`
+- `pg15-trixie`, `0.8.1-pg15-trixie`
-- `pg15-bookworm`, `0.8.3-pg15-bookworm`, `pg15`, `0.8.3-pg15`
+- `pg15-bookworm`, `0.8.1-pg15-bookworm`, `pg15`, `0.8.1-pg15`
-- `pg14-trixie`, `0.8.3-pg14-trixie`
+- `pg14-trixie`, `0.8.1-pg14-trixie`
-- `pg14-bookworm`, `0.8.3-pg14-bookworm`, `pg14`, `0.8.3-pg14`
+- `pg14-bookworm`, `0.8.1-pg14-bookworm`, `pg14`, `0.8.1-pg14`
-- `pg13-trixie`, `0.8.3-pg13-trixie`
+- `pg13-trixie`, `0.8.1-pg13-trixie`
-- `pg13-bookworm`, `0.8.3-pg13-bookworm`, `pg13`, `0.8.3-pg13`
+- `pg13-bookworm`, `0.8.1-pg13-bookworm`, `pg13`, `0.8.1-pg13`
 
 You can also build the image manually:
 
 ```sh
-git clone --branch v0.8.3 https://github.com/pgvector/pgvector.git
+git clone --branch v0.8.1 https://github.com/pgvector/pgvector.git
 cd pgvector
 docker build --pull --build-arg PG_MAJOR=18 -t myuser/pgvector .
 ```
@@ -1283,6 +1282,7 @@ Thanks to:
 - [k-means++: The Advantage of Careful Seeding](https://theory.stanford.edu/~sergei/papers/kMeansPP-soda.pdf)
 - [Concept Decompositions for Large Sparse Text Data using Clustering](https://www.cs.utexas.edu/users/inderjit/public_papers/concept_mlj.pdf)
 - [Efficient and Robust Approximate Nearest Neighbor Search using Hierarchical Navigable Small World Graphs](https://arxiv.org/ftp/arxiv/papers/1603/1603.09320.pdf)
+- [Distance Adaptive Beam Search for Provably Accurate Graph-Based Nearest Neighbor Search](https://arxiv.org/pdf/2505.15636)
 
 ## History
 
@@ -1329,7 +1329,7 @@ make clean && PG_CFLAGS="-DUSE_ASSERT_CHECKING" make && make install
 To enable benchmarking:
 
 ```sh
-make clean && PG_CFLAGS="-DHNSW_BENCH -DIVFFLAT_BENCH" make && make install
+make clean && PG_CFLAGS="-DIVFFLAT_BENCH" make && make install
 ```
 
 To show memory usage:

sql/vector--0.8.1--0.8.2.sql

@@ -1,2 +0,0 @@
--- complain if script is sourced in psql, rather than via CREATE EXTENSION
-\echo Use "ALTER EXTENSION vector UPDATE TO '0.8.2'" to load this file. \quit

sql/vector--0.8.2--0.8.3.sql

@@ -1,2 +0,0 @@
--- complain if script is sourced in psql, rather than via CREATE EXTENSION
-\echo Use "ALTER EXTENSION vector UPDATE TO '0.8.3'" to load this file. \quit

src/bitutils.c

@@ -31,12 +31,10 @@
 #define BIT_TARGET_CLONES
 #endif
 
-/* Use built-ins when possible for Postgres < 19 for inlining */
+/* Use built-ins when possible for inlining */
-#if PG_VERSION_NUM >= 190000
+#if defined(HAVE__BUILTIN_POPCOUNT) && defined(HAVE_LONG_INT_64)
-#define popcount64(x) pg_popcount64(x)
-#elif defined(HAVE__BUILTIN_POPCOUNT) && (defined(HAVE_LONG_INT_64) || SIZEOF_LONG == 8)
 #define popcount64(x) __builtin_popcountl(x)
-#elif defined(HAVE__BUILTIN_POPCOUNT) && (defined(HAVE_LONG_LONG_INT_64) || SIZEOF_LONG_LONG == 8)
+#elif defined(HAVE__BUILTIN_POPCOUNT) && defined(HAVE_LONG_LONG_INT_64)
 #define popcount64(x) __builtin_popcountll(x)
 #elif !defined(_MSC_VER)
 /* Fails to resolve with MSVC */
@@ -171,7 +169,7 @@ BitJaccardDistanceAvx512Popcount(uint32 bytes, unsigned char *ax, unsigned char
 #endif
 
 TARGET_XSAVE static bool
-SupportsAvx512Popcount(void)
+SupportsAvx512Popcount()
 {
 	unsigned int exx[4] = {0, 0, 0, 0};
 

src/bitvec.c

@@ -2,7 +2,6 @@
 
 #include "bitutils.h"
 #include "bitvec.h"
-#include "fmgr.h"
 #include "utils/varbit.h"
 #include "vector.h"
 

src/halfutils.c

@@ -1,7 +1,5 @@
 #include "postgres.h"
 
-#include <math.h>
-
 #include "halfutils.h"
 #include "halfvec.h"
 

src/halfvec.c

@@ -13,20 +13,12 @@
 #include "port.h"				/* for strtof() */
 #include "sparsevec.h"
 #include "utils/array.h"
+#include "utils/builtins.h"
 #include "utils/float.h"
-#include "utils/fmgrprotos.h"
 #include "utils/lsyscache.h"
-#include "utils/varbit.h"
+#include "utils/numeric.h"
 #include "vector.h"
 
-#if PG_VERSION_NUM >= 160000
-#include "varatt.h"
-#endif
-
-#if PG_VERSION_NUM >= 170000
-#include "parser/scansup.h"
-#endif
-
 #define STATE_DIMS(x) (ARR_DIMS(x)[0] - 1)
 #define CreateStateDatums(dim) palloc(sizeof(Datum) * (dim + 1))
 
@@ -137,9 +129,9 @@ InitHalfVector(int dim)
 	return result;
 }
 
-#if PG_VERSION_NUM >= 170000
+/*
-#define halfvec_isspace(ch) scanner_isspace(ch)
+ * Check for whitespace, since array_isspace() is static
-#else
+ */
 static inline bool
 halfvec_isspace(char ch)
 {
@@ -152,7 +144,6 @@ halfvec_isspace(char ch)
 		return true;
 	return false;
 }
-#endif
 
 /*
  * Check state array

src/hnsw.c

@@ -1,25 +1,18 @@
 #include "postgres.h"
 
 #include <float.h>
-#include <limits.h>
 #include <math.h>
 
 #include "access/amapi.h"
-#include "access/genam.h"
 #include "access/reloptions.h"
 #include "commands/progress.h"
 #include "commands/vacuum.h"
-#include "fmgr.h"
 #include "hnsw.h"
 #include "miscadmin.h"
-#include "nodes/pg_list.h"
-#include "storage/lwlock.h"
 #include "utils/float.h"
 #include "utils/guc.h"
-#include "utils/relcache.h"
 #include "utils/selfuncs.h"
 #include "utils/spccache.h"
-#include "vector.h"
 
 #if PG_VERSION_NUM < 150000
 #define MarkGUCPrefixReserved(x) EmitWarningsOnPlaceholders(x)
@@ -267,64 +260,6 @@ FUNCTION_PREFIX PG_FUNCTION_INFO_V1(hnswhandler);
 Datum
 hnswhandler(PG_FUNCTION_ARGS)
 {
-#if PG_VERSION_NUM >= 190000
-	static const IndexAmRoutine amroutine = {
-		.type = T_IndexAmRoutine,
-		.amstrategies = 0,
-		.amsupport = 3,
-		.amoptsprocnum = 0,
-		.amcanorder = false,
-		.amcanorderbyop = true,
-		.amcanhash = false,
-		.amconsistentequality = false,
-		.amconsistentordering = false,
-		.amcanbackward = false,
-		.amcanunique = false,
-		.amcanmulticol = false,
-		.amoptionalkey = true,
-		.amsearcharray = false,
-		.amsearchnulls = false,
-		.amstorage = false,
-		.amclusterable = false,
-		.ampredlocks = false,
-		.amcanparallel = false,
-		.amcanbuildparallel = true,
-		.amcaninclude = false,
-		.amusemaintenanceworkmem = false,
-		.amsummarizing = false,
-		.amparallelvacuumoptions = VACUUM_OPTION_PARALLEL_BULKDEL,
-		.amkeytype = InvalidOid,
-
-		.ambuild = hnswbuild,
-		.ambuildempty = hnswbuildempty,
-		.aminsert = hnswinsert,
-		.aminsertcleanup = NULL,
-		.ambulkdelete = hnswbulkdelete,
-		.amvacuumcleanup = hnswvacuumcleanup,
-		.amcanreturn = NULL,
-		.amcostestimate = hnswcostestimate,
-		.amgettreeheight = NULL,
-		.amoptions = hnswoptions,
-		.amproperty = NULL,
-		.ambuildphasename = hnswbuildphasename,
-		.amvalidate = hnswvalidate,
-		.amadjustmembers = NULL,
-		.ambeginscan = hnswbeginscan,
-		.amrescan = hnswrescan,
-		.amgettuple = hnswgettuple,
-		.amgetbitmap = NULL,
-		.amendscan = hnswendscan,
-		.ammarkpos = NULL,
-		.amrestrpos = NULL,
-		.amestimateparallelscan = NULL,
-		.aminitparallelscan = NULL,
-		.amparallelrescan = NULL,
-		.amtranslatestrategy = NULL,
-		.amtranslatecmptype = NULL,
-	};
-
-	PG_RETURN_POINTER(&amroutine);
-#else
 	IndexAmRoutine *amroutine = makeNode(IndexAmRoutine);
 
 	amroutine->amstrategies = 0;
@@ -398,5 +333,4 @@ hnswhandler(PG_FUNCTION_ARGS)
 #endif
 
 	PG_RETURN_POINTER(amroutine);
-#endif
 }

src/hnsw.h

@@ -3,29 +3,15 @@
 
 #include "postgres.h"
 
-#include <math.h>
-
 #include "access/genam.h"
 #include "access/parallel.h"
 #include "lib/pairingheap.h"
 #include "nodes/execnodes.h"
 #include "port.h"				/* for random() */
-#include "storage/bufpage.h"
-#include "storage/condition_variable.h"
-#include "storage/lwlock.h"
-#include "storage/s_lock.h"
 #include "utils/relptr.h"
 #include "utils/sampling.h"
 #include "vector.h"
 
-#ifdef HNSW_BENCH
-#include "portability/instr_time.h"
-#endif
-
-#if PG_VERSION_NUM >= 190000
-typedef Pointer Item;
-#endif
-
 #define HNSW_MAX_DIM 2000
 #define HNSW_MAX_NNZ 1000
 
@@ -82,21 +68,6 @@ typedef Pointer Item;
 #define HnswPageGetOpaque(page)	((HnswPageOpaque) PageGetSpecialPointer(page))
 #define HnswPageGetMeta(page)	((HnswMetaPageData *) PageGetContents(page))
 
-#ifdef HNSW_BENCH
-#define HnswBench(name, code) \
-	do { \
-		instr_time	start; \
-		instr_time	duration; \
-		INSTR_TIME_SET_CURRENT(start); \
-		(code); \
-		INSTR_TIME_SET_CURRENT(duration); \
-		INSTR_TIME_SUBTRACT(duration, start); \
-		elog(INFO, "%s: %.3f ms", name, INSTR_TIME_GET_MILLISEC(duration)); \
-	} while (0)
-#else
-#define HnswBench(name, code) (code)
-#endif
-
 #if PG_VERSION_NUM >= 150000
 #define RandomDouble() pg_prng_double(&pg_global_prng_state)
 #define SeedRandom(seed) pg_prng_seed(&pg_global_prng_state, seed)
@@ -431,7 +402,6 @@ typedef struct HnswVacuumState
 	BufferAccessStrategy bas;
 	HnswNeighborTuple ntup;
 	HnswElementData highestPoint;
-	HnswElementData fallbackPoint;
 
 	/* Memory */
 	MemoryContext tmpCtx;
@@ -454,7 +424,7 @@ void	   *HnswAlloc(HnswAllocator * allocator, Size size);
 HnswElement HnswInitElement(char *base, ItemPointer tid, int m, double ml, int maxLevel, HnswAllocator * alloc);
 HnswElement HnswInitElementFromBlock(BlockNumber blkno, OffsetNumber offno);
 void		HnswFindElementNeighbors(char *base, HnswElement element, HnswElement entryPoint, Relation index, HnswSupport * support, int m, int efConstruction, bool existing);
-HnswSearchCandidate *HnswEntryCandidate(char *base, HnswElement entryPoint, HnswQuery * q, Relation index, HnswSupport * support, bool loadVec);
+HnswSearchCandidate *HnswEntryCandidate(char *base, HnswElement em, HnswQuery * q, Relation rel, HnswSupport * support, bool loadVec);
 void		HnswUpdateMetaPage(Relation index, int updateEntry, HnswElement entryPoint, BlockNumber insertPage, ForkNumber forkNum, bool building);
 void		HnswSetNeighborTuple(char *base, HnswNeighborTuple ntup, HnswElement e, int m);
 void		HnswAddHeapTid(HnswElement element, ItemPointer heaptid);

src/hnswbuild.c

@@ -36,14 +36,11 @@
  */
 #include "postgres.h"
 
-#include <limits.h>
+#include <math.h>
 
-#include "access/genam.h"
 #include "access/parallel.h"
-#include "access/relscan.h"
 #include "access/table.h"
 #include "access/tableam.h"
-#include "access/tupdesc.h"
 #include "access/xact.h"
 #include "access/xloginsert.h"
 #include "catalog/index.h"
@@ -51,15 +48,11 @@
 #include "commands/progress.h"
 #include "hnsw.h"
 #include "miscadmin.h"
-#include "nodes/execnodes.h"
 #include "optimizer/optimizer.h"
 #include "storage/bufmgr.h"
-#include "storage/condition_variable.h"
 #include "tcop/tcopprot.h"
 #include "utils/datum.h"
 #include "utils/memutils.h"
-#include "utils/rel.h"
-#include "utils/snapmgr.h"
 
 #if PG_VERSION_NUM >= 160000
 #include "varatt.h"
@@ -80,8 +73,6 @@
 #define PARALLEL_KEY_HNSW_AREA			UINT64CONST(0xA000000000000002)
 #define PARALLEL_KEY_QUERY_TEXT			UINT64CONST(0xA000000000000003)
 
-#define HNSW_MAX_GRAPH_MEMORY (SIZE_MAX / 2)
-
 /*
  * Create the metapage
  */
@@ -494,7 +485,6 @@ InsertTuple(Relation index, Datum *values, bool *isnull, ItemPointer heaptid, Hn
 	LWLock	   *flushLock = &graph->flushLock;
 	char	   *base = buildstate->hnswarea;
 	Datum		value;
-	Size		memoryMargin;
 
 	/* Form index value */
 	if (!HnswFormIndexValue(&value, values, isnull, buildstate->typeInfo, support))
@@ -503,9 +493,6 @@ InsertTuple(Relation index, Datum *values, bool *isnull, ItemPointer heaptid, Hn
 	/* Get datum size */
 	valueSize = VARSIZE_ANY(DatumGetPointer(value));
 
-	/* In a parallel build, add a margin so allocations never fail */
-	memoryMargin = base == NULL ? 0 : 1024 * 1024;
-
 	/* Ensure graph not flushed when inserting */
 	LWLockAcquire(flushLock, LW_SHARED);
 
@@ -527,7 +514,7 @@ InsertTuple(Relation index, Datum *values, bool *isnull, ItemPointer heaptid, Hn
 	 * Check that we have enough memory available for the new element now that
 	 * we have the allocator lock, and flush pages if needed.
 	 */
-	if (graph->memoryUsed + memoryMargin >= graph->memoryTotal)
+	if (graph->memoryUsed >= graph->memoryTotal)
 	{
 		LWLockRelease(&graph->allocatorLock);
 
@@ -562,7 +549,7 @@ InsertTuple(Relation index, Datum *values, bool *isnull, ItemPointer heaptid, Hn
 
 	/* Copy the datum */
 	memcpy(valuePtr, DatumGetPointer(value), valueSize);
-	HnswPtrStore(base, element->value, (char *) valuePtr);
+	HnswPtrStore(base, element->value, valuePtr);
 
 	/* Create a lock for the element */
 	LWLockInitialize(&element->lock, hnsw_lock_tranche_id);
@@ -620,7 +607,7 @@ InitGraph(HnswGraph * graph, char *base, Size memoryTotal)
 	HnswPtrStore(base, graph->head, (HnswElement) NULL);
 	HnswPtrStore(base, graph->entryPoint, (HnswElement) NULL);
 	graph->memoryUsed = 0;
-	graph->memoryTotal = Min(memoryTotal, HNSW_MAX_GRAPH_MEMORY);
+	graph->memoryTotal = memoryTotal;
 	graph->flushed = false;
 	graph->indtuples = 0;
 	SpinLockInit(&graph->lock);
@@ -661,17 +648,9 @@ static void *
 HnswSharedMemoryAlloc(Size size, void *state)
 {
 	HnswBuildState *buildstate = (HnswBuildState *) state;
-	Size		alignedSize = MAXALIGN(size);
+	void	   *chunk = buildstate->hnswarea + buildstate->graph->memoryUsed;
-	void	   *chunk;
 
-	if (alignedSize > 1024 * 1024)
+	buildstate->graph->memoryUsed += MAXALIGN(size);
-		elog(ERROR, "hnsw allocation too large");
-
-	if (buildstate->graph->memoryUsed + alignedSize > buildstate->graph->memoryTotal)
-		elog(ERROR, "hnsw allocator out of memory");
-
-	chunk = buildstate->hnswarea + buildstate->graph->memoryUsed;
-	buildstate->graph->memoryUsed += alignedSize;
 	return chunk;
 }
 
@@ -803,11 +782,7 @@ HnswParallelScanAndInsert(Relation heapRel, Relation indexRel, HnswShared * hnsw
 	buildstate.hnswarea = hnswarea;
 	InitAllocator(&buildstate.allocator, &HnswSharedMemoryAlloc, &buildstate);
 	scan = table_beginscan_parallel(heapRel,
-									ParallelTableScanFromHnswShared(hnswshared)
+									ParallelTableScanFromHnswShared(hnswshared));
-#if PG_VERSION_NUM >= 190000
-									,SO_NONE
-#endif
-		);
 	reltuples = table_index_build_scan(heapRel, indexRel, indexInfo,
 									   true, progress, BuildCallback,
 									   (void *) &buildstate, scan);
@@ -961,8 +936,6 @@ HnswBeginParallel(HnswBuildState * buildstate, bool isconcurrent, int request)
 	if (esthnswarea > estother)
 		esthnswarea -= estother;
 
-	esthnswarea = Min(esthnswarea, HNSW_MAX_GRAPH_MEMORY);
-
 	shm_toc_estimate_chunk(&pcxt->estimator, esthnswarea);
 	shm_toc_estimate_keys(&pcxt->estimator, 2);
 
@@ -1005,7 +978,8 @@ HnswBeginParallel(HnswBuildState * buildstate, bool isconcurrent, int request)
 								  snapshot);
 
 	hnswarea = (char *) shm_toc_allocate(pcxt->toc, esthnswarea);
-	InitGraph(&hnswshared->graphData, hnswarea, esthnswarea);
+	/* Report less than allocated so never fails */
+	InitGraph(&hnswshared->graphData, hnswarea, esthnswarea - 1024 * 1024);
 
 	/*
 	 * Avoid base address for relptr for Postgres < 14.5

src/hnswinsert.c

@@ -1,15 +1,13 @@
 #include "postgres.h"
 
-#include "access/genam.h"
+#include <math.h>
+
 #include "access/generic_xlog.h"
 #include "hnsw.h"
-#include "nodes/execnodes.h"
 #include "storage/bufmgr.h"
 #include "storage/lmgr.h"
-#include "storage/lwlock.h"
 #include "utils/datum.h"
 #include "utils/memutils.h"
-#include "utils/rel.h"
 
 #if PG_VERSION_NUM >= 160000
 #include "varatt.h"
@@ -714,7 +712,7 @@ HnswInsertTupleOnDisk(Relation index, HnswSupport * support, Datum value, ItemPo
 
 	/* Create an element */
 	element = HnswInitElement(base, heaptid, m, HnswGetMl(m), HnswGetMaxLevel(m), NULL);
-	HnswPtrStore(base, element->value, (char *) DatumGetPointer(value));
+	HnswPtrStore(base, element->value, DatumGetPointer(value));
 
 	/* Prevent concurrent inserts when likely updating entry point */
 	if (entryPoint == NULL || element->level > entryPoint->level)

src/hnswscan.c

@@ -1,23 +1,12 @@
 #include "postgres.h"
 
-#include <limits.h>
-
-#include "access/genam.h"
 #include "access/relscan.h"
 #include "hnsw.h"
-#include "lib/pairingheap.h"
-#include "miscadmin.h"
-#include "nodes/pg_list.h"
 #include "pgstat.h"
+#include "storage/bufmgr.h"
 #include "storage/lmgr.h"
 #include "utils/float.h"
 #include "utils/memutils.h"
-#include "utils/relcache.h"
-#include "utils/snapmgr.h"
-
-#if PG_VERSION_NUM >= 160000
-#include "varatt.h"
-#endif
 
 /*
  * Algorithm 5 from paper
@@ -153,7 +142,7 @@ hnswbeginscan(Relation index, int nkeys, int norderbys)
 	/* Calculate max memory */
 	/* Add 256 extra bytes to fill last block when close */
 	maxMemory = (double) work_mem * hnsw_scan_mem_multiplier * 1024.0 + 256;
-	so->maxMemory = Min(maxMemory, (double) (SIZE_MAX / 2));
+	so->maxMemory = Min(maxMemory, (double) SIZE_MAX);
 
 	scan->opaque = so;
 

src/hnswutils.c

@@ -2,20 +2,18 @@
 
 #include <math.h>
 
-#include "access/genam.h"
 #include "access/generic_xlog.h"
+#include "catalog/pg_type.h"
+#include "catalog/pg_type_d.h"
 #include "common/hashfn.h"
 #include "fmgr.h"
 #include "hnsw.h"
 #include "lib/pairingheap.h"
-#include "nodes/pg_list.h"
-#include "port/atomics.h"
 #include "sparsevec.h"
 #include "storage/bufmgr.h"
 #include "utils/datum.h"
 #include "utils/memdebug.h"
 #include "utils/rel.h"
-#include "vector.h"
 
 #if PG_VERSION_NUM >= 160000
 #include "varatt.h"
@@ -262,7 +260,7 @@ HnswInitElement(char *base, ItemPointer heaptid, int m, double ml, int maxLevel,
 
 	HnswInitNeighbors(base, element, m, allocator);
 
-	HnswPtrStore(base, element->value, (char *) NULL);
+	HnswPtrStore(base, element->value, (Pointer) NULL);
 
 	return element;
 }
@@ -288,7 +286,7 @@ HnswInitElementFromBlock(BlockNumber blkno, OffsetNumber offno)
 	element->blkno = blkno;
 	element->offno = offno;
 	HnswPtrStore(base, element->neighbors, (HnswNeighborArrayPtr *) NULL);
-	HnswPtrStore(base, element->value, (char *) NULL);
+	HnswPtrStore(base, element->value, (Pointer) NULL);
 	return element;
 }
 
@@ -514,7 +512,7 @@ HnswLoadElementFromTuple(HnswElement element, HnswElementTuple etup, bool loadHe
 		char	   *base = NULL;
 		Datum		value = datumCopy(PointerGetDatum(&etup->data), false, -1);
 
-		HnswPtrStore(base, element->value, (char *) DatumGetPointer(value));
+		HnswPtrStore(base, element->value, DatumGetPointer(value));
 	}
 }
 
@@ -546,9 +544,6 @@ HnswLoadElementImpl(BlockNumber blkno, OffsetNumber offno, double *distance, Hns
 
 	Assert(HnswIsElementTuple(etup));
 
-	if (unlikely(etup->deleted))
-		elog(ERROR, "cannot load deleted element");
-
 	/* Calculate distance */
 	if (distance != NULL)
 	{
@@ -834,6 +829,7 @@ HnswSearchLayer(char *base, HnswQuery * q, List *ep, int ef, int lc, Relation in
 	HnswUnvisited *unvisited = palloc(lm * sizeof(HnswUnvisited));
 	int			unvisitedLength;
 	bool		inMemory = index == NULL;
+	double		lambda = 0.12;
 
 	if (v == NULL)
 	{
@@ -889,7 +885,7 @@ HnswSearchLayer(char *base, HnswQuery * q, List *ep, int ef, int lc, Relation in
 		HnswSearchCandidate *f = HnswGetSearchCandidate(w_node, pairingheap_first(W));
 		HnswElement cElement;
 
-		if (c->distance > f->distance)
+		if (c->distance > f->distance * (1 + lambda))
 			break;
 
 		cElement = HnswPtrAccess(base, c->element);
@@ -909,8 +905,10 @@ HnswSearchLayer(char *base, HnswQuery * q, List *ep, int ef, int lc, Relation in
 			HnswSearchCandidate *e;
 			double		eDistance;
 			bool		alwaysAdd = wlen < ef;
+			double		fDistance;
 
 			f = HnswGetSearchCandidate(w_node, pairingheap_first(W));
+			fDistance = f->distance * (1 + lambda);
 
 			if (inMemory)
 			{
@@ -925,13 +923,13 @@ HnswSearchLayer(char *base, HnswQuery * q, List *ep, int ef, int lc, Relation in
 
 				/* Avoid any allocations if not adding */
 				eElement = NULL;
-				HnswLoadElementImpl(blkno, offno, &eDistance, q, index, support, inserting, alwaysAdd || discarded != NULL ? NULL : &f->distance, &eElement);
+				HnswLoadElementImpl(blkno, offno, &eDistance, q, index, support, inserting, alwaysAdd || discarded != NULL ? NULL : &fDistance, &eElement);
 
 				if (eElement == NULL)
 					continue;
 			}
 
-			if (!(eDistance < f->distance || alwaysAdd))
+			if (eElement == NULL || !(eDistance < fDistance || alwaysAdd))
 			{
 				if (discarded != NULL)
 				{

src/hnswvacuum.c

@@ -1,14 +1,13 @@
 #include "postgres.h"
 
-#include "access/genam.h"
+#include <math.h>
+
 #include "access/generic_xlog.h"
 #include "commands/vacuum.h"
 #include "hnsw.h"
-#include "nodes/pg_list.h"
 #include "storage/bufmgr.h"
 #include "storage/lmgr.h"
 #include "utils/memutils.h"
-#include "utils/rel.h"
 
 #if PG_VERSION_NUM >= 160000
 #include "varatt.h"
@@ -37,20 +36,17 @@ RemoveHeapTids(HnswVacuumState * vacuumstate)
 {
 	BlockNumber blkno = HNSW_HEAD_BLKNO;
 	HnswElement highestPoint = &vacuumstate->highestPoint;
-	HnswElement fallbackPoint = &vacuumstate->fallbackPoint;
 	Relation	index = vacuumstate->index;
 	BufferAccessStrategy bas = vacuumstate->bas;
+	HnswElement entryPoint = HnswGetEntryPoint(vacuumstate->index);
 	IndexBulkDeleteResult *stats = vacuumstate->stats;
 
-	/* Store separately since HnswElement level is uint8 */
+	/* Store separately since highestPoint.level is uint8 */
 	int			highestLevel = -1;
-	int			fallbackLevel = -1;
 
-	/* Initialize highest point and fallback point */
+	/* Initialize highest point */
 	highestPoint->blkno = InvalidBlockNumber;
 	highestPoint->offno = InvalidOffsetNumber;
-	fallbackPoint->blkno = InvalidBlockNumber;
-	fallbackPoint->offno = InvalidOffsetNumber;
 
 	while (BlockNumberIsValid(blkno))
 	{
@@ -122,31 +118,14 @@ RemoveHeapTids(HnswVacuumState * vacuumstate)
 				tidhash_insert(vacuumstate->deleted, ip, &found);
 				Assert(!found);
 			}
-			else if (etup->level > highestLevel)
+			else if (etup->level > highestLevel && !(entryPoint != NULL && blkno == entryPoint->blkno && offno == entryPoint->offno))
 			{
-				if (BlockNumberIsValid(highestPoint->blkno))
+				/* Keep track of highest non-entry point */
-				{
-					/* Current highest point becomes fallback */
-					fallbackPoint->blkno = highestPoint->blkno;
-					fallbackPoint->offno = highestPoint->offno;
-					fallbackPoint->level = highestPoint->level;
-					fallbackLevel = highestLevel;
-				}
-
-				/* Keep track of highest point */
 				highestPoint->blkno = blkno;
 				highestPoint->offno = offno;
 				highestPoint->level = etup->level;
 				highestLevel = etup->level;
 			}
-			else if (etup->level > fallbackLevel)
-			{
-				/* Keep track of second highest point */
-				fallbackPoint->blkno = blkno;
-				fallbackPoint->offno = offno;
-				fallbackPoint->level = etup->level;
-				fallbackLevel = etup->level;
-			}
 		}
 
 		blkno = HnswPageGetOpaque(page)->nextblkno;
@@ -158,10 +137,6 @@ RemoveHeapTids(HnswVacuumState * vacuumstate)
 
 		UnlockReleaseBuffer(buf);
 	}
-
-#ifdef HNSW_MEMORY
-	elog(INFO, "memory: %zu KB", MemoryContextMemAllocated(CurrentMemoryContext, true) / 1024);
-#endif
 }
 
 /*
@@ -203,12 +178,7 @@ NeedsUpdated(HnswVacuumState * vacuumstate, HnswElement element)
 	/* Also update if layer 0 is not full */
 	/* This could indicate too many candidates being deleted during insert */
 	if (!needsUpdated)
-	{
-		/* Keep clang-tidy happy */
-		Assert(ntup->count > 0);
-
 		needsUpdated = !ItemPointerIsValid(&ntup->indextids[ntup->count - 1]);
-	}
 
 	UnlockReleaseBuffer(buf);
 
@@ -293,27 +263,12 @@ RepairGraphEntryPoint(HnswVacuumState * vacuumstate)
 		/* Get a shared lock */
 		LockPage(index, HNSW_UPDATE_LOCK, ShareLock);
 
-		/* Get latest entry point */
-		entryPoint = HnswGetEntryPoint(index);
-
-		/* Use fallback point if highest point is entry point */
-		if (entryPoint != NULL && entryPoint->blkno == highestPoint->blkno && entryPoint->offno == highestPoint->offno)
-		{
-			highestPoint = &vacuumstate->fallbackPoint;
-
-			if (!BlockNumberIsValid(highestPoint->blkno))
-				highestPoint = NULL;
-		}
-
-		if (highestPoint != NULL)
-		{
 		/* Load element */
 		HnswLoadElement(highestPoint, NULL, NULL, index, support, true, NULL);
 
 		/* Repair if needed */
 		if (NeedsUpdated(vacuumstate, highestPoint))
-				RepairGraphElement(vacuumstate, highestPoint, entryPoint);
+			RepairGraphElement(vacuumstate, highestPoint, HnswGetEntryPoint(index));
-		}
 
 		/* Release lock */
 		UnlockPage(index, HNSW_UPDATE_LOCK, ShareLock);
@@ -480,99 +435,6 @@ RepairGraph(HnswVacuumState * vacuumstate)
 		/* Reset memory context */
 		MemoryContextSwitchTo(oldCtx);
 		MemoryContextReset(vacuumstate->tmpCtx);
-
-#ifdef HNSW_VACUUM_PROGRESS
-		if (!BlockNumberIsValid(blkno) || (blkno - HNSW_HEAD_BLKNO) % 1000 == 0)
-		{
-			BlockNumber totalBlocks = RelationGetNumberOfBlocks(index);
-			BlockNumber currentBlocks = BlockNumberIsValid(blkno) ? blkno : totalBlocks;
-
-			elog(INFO, "hnsw vacuum progress: %.1f%%", 100.0 * currentBlocks / totalBlocks);
-		}
-#endif
-	}
-}
-
-/*
- * Confirm graph was repaired
- */
-static void
-ConfirmRepaired(HnswVacuumState * vacuumstate)
-{
-	BlockNumber blkno = HNSW_HEAD_BLKNO;
-	Relation	index = vacuumstate->index;
-	BufferAccessStrategy bas = vacuumstate->bas;
-
-	while (BlockNumberIsValid(blkno))
-	{
-		Buffer		buf;
-		Page		page;
-		OffsetNumber offno;
-		OffsetNumber maxoffno;
-
-		vacuum_delay_point();
-
-		buf = ReadBufferExtended(index, MAIN_FORKNUM, blkno, RBM_NORMAL, bas);
-		LockBuffer(buf, BUFFER_LOCK_SHARE);
-		page = BufferGetPage(buf);
-		maxoffno = PageGetMaxOffsetNumber(page);
-
-		/* Iterate over nodes */
-		for (offno = FirstOffsetNumber; offno <= maxoffno; offno = OffsetNumberNext(offno))
-		{
-			HnswElementTuple etup = (HnswElementTuple) PageGetItem(page, PageGetItemId(page, offno));
-			HnswNeighborTuple ntup;
-			Buffer		nbuf;
-			Page		npage;
-			BlockNumber neighborPage;
-			OffsetNumber neighborOffno;
-
-			/* Skip neighbor tuples */
-			if (!HnswIsElementTuple(etup))
-				continue;
-
-			/* Skip if being deleted */
-			if (!ItemPointerIsValid(&etup->heaptids[0]))
-				continue;
-
-			/* Get neighbor page */
-			neighborPage = ItemPointerGetBlockNumber(&etup->neighbortid);
-			neighborOffno = ItemPointerGetOffsetNumber(&etup->neighbortid);
-
-			if (neighborPage == blkno)
-			{
-				nbuf = buf;
-				npage = page;
-			}
-			else
-			{
-				nbuf = ReadBufferExtended(index, MAIN_FORKNUM, neighborPage, RBM_NORMAL, bas);
-				LockBuffer(nbuf, BUFFER_LOCK_SHARE);
-				npage = BufferGetPage(nbuf);
-			}
-
-			ntup = (HnswNeighborTuple) PageGetItem(npage, PageGetItemId(npage, neighborOffno));
-
-			/* Check neighbors */
-			for (int i = 0; i < ntup->count; i++)
-			{
-				ItemPointer indextid = &ntup->indextids[i];
-
-				if (!ItemPointerIsValid(indextid))
-					continue;
-
-				/* Check if in deleted list */
-				if (DeletedContains(vacuumstate->deleted, indextid))
-					elog(ERROR, "hnsw graph not repaired");
-			}
-
-			if (nbuf != buf)
-				UnlockReleaseBuffer(nbuf);
-		}
-
-		blkno = HnswPageGetOpaque(page)->nextblkno;
-
-		UnlockReleaseBuffer(buf);
 	}
 }
 
@@ -666,9 +528,8 @@ MarkDeleted(HnswVacuumState * vacuumstate)
 			ntup = (HnswNeighborTuple) PageGetItem(npage, PageGetItemId(npage, neighborOffno));
 
 			/* Overwrite element */
-			/* Use memset instead of MemSet to keep clang-tidy happy */
 			etup->deleted = 1;
-			memset(&etup->data, 0, VARSIZE_ANY(&etup->data));
+			MemSet(&etup->data, 0, VARSIZE_ANY(&etup->data));
 
 			/* Overwrite neighbors */
 			for (int i = 0; i < ntup->count; i++)
@@ -766,16 +627,13 @@ hnswbulkdelete(IndexVacuumInfo *info, IndexBulkDeleteResult *stats,
 	InitVacuumState(&vacuumstate, info, stats, callback, callback_state);
 
 	/* Pass 1: Remove heap TIDs */
-	HnswBench("RemoveHeapTids", RemoveHeapTids(&vacuumstate));
+	RemoveHeapTids(&vacuumstate);
 
 	/* Pass 2: Repair graph */
-	HnswBench("RepairGraph", RepairGraph(&vacuumstate));
+	RepairGraph(&vacuumstate);
 
-	/* Pass 3: Confirm repaired */
+	/* Pass 3: Mark as deleted */
-	HnswBench("ConfirmRepaired", ConfirmRepaired(&vacuumstate));
+	MarkDeleted(&vacuumstate);
-
-	/* Pass 4: Mark as deleted */
-	HnswBench("MarkDeleted", MarkDeleted(&vacuumstate));
 
 	FreeVacuumState(&vacuumstate);
 

src/ivfbuild.c

@@ -2,33 +2,23 @@
 
 #include <float.h>
 
-#include "access/genam.h"
-#include "access/generic_xlog.h"
-#include "access/itup.h"
-#include "access/relscan.h"
 #include "access/table.h"
 #include "access/tableam.h"
-#include "access/tupdesc.h"
 #include "access/parallel.h"
 #include "access/xact.h"
-#include "access/xloginsert.h"
+#include "bitvec.h"
 #include "catalog/index.h"
 #include "catalog/pg_operator_d.h"
 #include "catalog/pg_type_d.h"
 #include "commands/progress.h"
-#include "fmgr.h"
+#include "halfvec.h"
 #include "ivfflat.h"
 #include "miscadmin.h"
-#include "nodes/execnodes.h"
 #include "optimizer/optimizer.h"
 #include "storage/bufmgr.h"
-#include "storage/condition_variable.h"
 #include "tcop/tcopprot.h"
 #include "utils/memutils.h"
-#include "utils/rel.h"
+#include "vector.h"
-#include "utils/sampling.h"
-#include "utils/snapmgr.h"
-#include "utils/tuplesort.h"
 
 #if PG_VERSION_NUM >= 160000
 #include "varatt.h"
@@ -63,13 +53,15 @@ AddSample(Datum *values, IvfflatBuildState * buildstate)
 	Datum		value = PointerGetDatum(PG_DETOAST_DATUM(values[0]));
 
 	/*
-	 * Check with KMEANS_NORM_PROC that the value can be normalized since
+	 * Normalize with KMEANS_NORM_PROC since spherical distance function
-	 * spherical distance function expects unit vectors
+	 * expects unit vectors
 	 */
 	if (buildstate->kmeansnormprocinfo != NULL)
 	{
 		if (!IvfflatCheckNorm(buildstate->kmeansnormprocinfo, buildstate->collation, value))
 			return;
+
+		value = IvfflatNormValue(buildstate->typeInfo, buildstate->collation, value);
 	}
 
 	if (samples->length < targsamples)
@@ -80,7 +72,7 @@ AddSample(Datum *values, IvfflatBuildState * buildstate)
 	else
 	{
 		if (buildstate->rowstoskip < 0)
-			buildstate->rowstoskip = reservoir_get_next_S(&buildstate->rstate, buildstate->samplerows, targsamples);
+			buildstate->rowstoskip = reservoir_get_next_S(&buildstate->rstate, samples->length, targsamples);
 
 		if (buildstate->rowstoskip <= 0)
 		{
@@ -96,9 +88,6 @@ AddSample(Datum *values, IvfflatBuildState * buildstate)
 
 		buildstate->rowstoskip -= 1;
 	}
-
-	/* Increment after reservoir_get_next_S */
-	buildstate->samplerows += 1;
 }
 
 /*
@@ -135,7 +124,6 @@ SampleRows(IvfflatBuildState * buildstate)
 	int			targsamples = buildstate->samples->maxlen;
 	BlockNumber totalblocks = RelationGetNumberOfBlocks(buildstate->heap);
 
-	buildstate->samplerows = 0;
 	buildstate->rowstoskip = -1;
 
 	BlockSampler_Init(&buildstate->bs, totalblocks, targsamples, RandomInt());
@@ -145,24 +133,8 @@ SampleRows(IvfflatBuildState * buildstate)
 	{
 		BlockNumber targblock = BlockSampler_Next(&buildstate->bs);
 
-		/* Set anyvisible to false like table_index_build_scan */
 		table_index_build_range_scan(buildstate->heap, buildstate->index, buildstate->indexInfo,
-									 false, false, false, targblock, 1, SampleCallback, (void *) buildstate, NULL);
+									 false, true, false, targblock, 1, SampleCallback, (void *) buildstate, NULL);
-	}
-
-	/* Normalize if needed */
-	if (buildstate->kmeansnormprocinfo != NULL)
-	{
-		VectorArray samples = buildstate->samples;
-
-		for (int i = 0; i < samples->length; i++)
-		{
-			Datum		value = PointerGetDatum(VectorArrayGet(samples, i));
-			Datum		normValue = IvfflatNormValue(buildstate->typeInfo, buildstate->collation, value);
-
-			VectorArraySet(samples, i, DatumGetPointer(normValue));
-			pfree(DatumGetPointer(normValue));
-		}
 	}
 }
 
@@ -393,13 +365,9 @@ InitBuildState(IvfflatBuildState * buildstate, Relation heap, Relation index, In
 	TupleDescInitEntry(buildstate->sortdesc, (AttrNumber) 1, "list", INT4OID, -1, 0);
 	TupleDescInitEntry(buildstate->sortdesc, (AttrNumber) 2, "tid", TIDOID, -1, 0);
 	TupleDescInitEntry(buildstate->sortdesc, (AttrNumber) 3, "vector", TupleDescAttr(buildstate->tupdesc, 0)->atttypid, -1, 0);
-#if PG_VERSION_NUM >= 190000
-	TupleDescFinalize(buildstate->sortdesc);
-#endif
 
 	buildstate->slot = MakeSingleTupleTableSlot(buildstate->sortdesc, &TTSOpsVirtual);
 
-	/* TODO Ensure within maintenance_work_mem */
 	buildstate->centers = VectorArrayInit(buildstate->lists, buildstate->dimensions, buildstate->typeInfo->itemSize(buildstate->dimensions));
 	buildstate->listInfo = palloc(sizeof(ListInfo) * buildstate->lists);
 
@@ -458,7 +426,7 @@ ComputeCenters(IvfflatBuildState * buildstate)
 	buildstate->samples = VectorArrayInit(numSamples, buildstate->dimensions, buildstate->centers->itemsize);
 	if (buildstate->heap != NULL)
 	{
-		IvfflatBench("sample rows", SampleRows(buildstate));
+		SampleRows(buildstate);
 
 		if (buildstate->samples->length < buildstate->lists)
 		{
@@ -673,11 +641,7 @@ IvfflatParallelScanAndSort(IvfflatSpool * ivfspool, IvfflatShared * ivfshared, S
 	ivfspool->sortstate = InitBuildSortState(buildstate.sortdesc, sortmem, coordinate);
 	buildstate.sortstate = ivfspool->sortstate;
 	scan = table_beginscan_parallel(ivfspool->heap,
-									ParallelTableScanFromIvfflatShared(ivfshared)
+									ParallelTableScanFromIvfflatShared(ivfshared));
-#if PG_VERSION_NUM >= 190000
-									,SO_NONE
-#endif
-		);
 	reltuples = table_index_build_scan(ivfspool->heap, ivfspool->index, indexInfo,
 									   true, progress, BuildCallback,
 									   (void *) &buildstate, scan);

src/ivfflat.c

@@ -3,19 +3,14 @@
 #include <float.h>
 
 #include "access/amapi.h"
-#include "access/genam.h"
 #include "access/reloptions.h"
 #include "commands/progress.h"
 #include "commands/vacuum.h"
-#include "fmgr.h"
 #include "ivfflat.h"
-#include "nodes/pg_list.h"
 #include "utils/float.h"
 #include "utils/guc.h"
-#include "utils/relcache.h"
 #include "utils/selfuncs.h"
 #include "utils/spccache.h"
-#include "vector.h"
 
 #if PG_VERSION_NUM < 150000
 #define MarkGUCPrefixReserved(x) EmitWarningsOnPlaceholders(x)
@@ -184,64 +179,6 @@ FUNCTION_PREFIX PG_FUNCTION_INFO_V1(ivfflathandler);
 Datum
 ivfflathandler(PG_FUNCTION_ARGS)
 {
-#if PG_VERSION_NUM >= 190000
-	static const IndexAmRoutine amroutine = {
-		.type = T_IndexAmRoutine,
-		.amstrategies = 0,
-		.amsupport = 5,
-		.amoptsprocnum = 0,
-		.amcanorder = false,
-		.amcanorderbyop = true,
-		.amcanhash = false,
-		.amconsistentequality = false,
-		.amconsistentordering = false,
-		.amcanbackward = false,
-		.amcanunique = false,
-		.amcanmulticol = false,
-		.amoptionalkey = true,
-		.amsearcharray = false,
-		.amsearchnulls = false,
-		.amstorage = false,
-		.amclusterable = false,
-		.ampredlocks = false,
-		.amcanparallel = false,
-		.amcanbuildparallel = true,
-		.amcaninclude = false,
-		.amusemaintenanceworkmem = false,
-		.amsummarizing = false,
-		.amparallelvacuumoptions = VACUUM_OPTION_PARALLEL_BULKDEL,
-		.amkeytype = InvalidOid,
-
-		.ambuild = ivfflatbuild,
-		.ambuildempty = ivfflatbuildempty,
-		.aminsert = ivfflatinsert,
-		.aminsertcleanup = NULL,
-		.ambulkdelete = ivfflatbulkdelete,
-		.amvacuumcleanup = ivfflatvacuumcleanup,
-		.amcanreturn = NULL,
-		.amcostestimate = ivfflatcostestimate,
-		.amgettreeheight = NULL,
-		.amoptions = ivfflatoptions,
-		.amproperty = NULL,
-		.ambuildphasename = ivfflatbuildphasename,
-		.amvalidate = ivfflatvalidate,
-		.amadjustmembers = NULL,
-		.ambeginscan = ivfflatbeginscan,
-		.amrescan = ivfflatrescan,
-		.amgettuple = ivfflatgettuple,
-		.amgetbitmap = NULL,
-		.amendscan = ivfflatendscan,
-		.ammarkpos = NULL,
-		.amrestrpos = NULL,
-		.amestimateparallelscan = NULL,
-		.aminitparallelscan = NULL,
-		.amparallelrescan = NULL,
-		.amtranslatestrategy = NULL,
-		.amtranslatecmptype = NULL,
-	};
-
-	PG_RETURN_POINTER(&amroutine);
-#else
 	IndexAmRoutine *amroutine = makeNode(IndexAmRoutine);
 
 	amroutine->amstrategies = 0;
@@ -315,5 +252,4 @@ ivfflathandler(PG_FUNCTION_ARGS)
 #endif
 
 	PG_RETURN_POINTER(amroutine);
-#endif
 }

src/ivfflat.h

@@ -9,7 +9,6 @@
 #include "lib/pairingheap.h"
 #include "nodes/execnodes.h"
 #include "port.h"				/* for random() */
-#include "storage/condition_variable.h"
 #include "utils/sampling.h"
 #include "utils/tuplesort.h"
 #include "vector.h"
@@ -26,10 +25,6 @@
 #include "portability/instr_time.h"
 #endif
 
-#if PG_VERSION_NUM >= 190000
-typedef Pointer Item;
-#endif
-
 #define IVFFLAT_MAX_DIM 2000
 
 /* Support functions */
@@ -214,8 +209,7 @@ typedef struct IvfflatBuildState
 	/* Sampling */
 	BlockSamplerData bs;
 	ReservoirStateData rstate;
-	double		samplerows;
+	int			rowstoskip;
-	double		rowstoskip;
 
 	/* Sorting */
 	Tuplesortstate *sortstate;

src/ivfinsert.c

@@ -2,16 +2,11 @@
 
 #include <float.h>
 
-#include "access/genam.h"
 #include "access/generic_xlog.h"
-#include "access/itup.h"
-#include "fmgr.h"
 #include "ivfflat.h"
-#include "nodes/execnodes.h"
 #include "storage/bufmgr.h"
 #include "storage/lmgr.h"
 #include "utils/memutils.h"
-#include "utils/rel.h"
 
 /*
  * Find the list that minimizes the distance function

src/ivfkmeans.c

@@ -1,15 +1,17 @@
 #include "postgres.h"
 
 #include <float.h>
-#include <limits.h>
 #include <math.h>
 
-#include "access/genam.h"
+#include "bitvec.h"
-#include "fmgr.h"
+#include "halfutils.h"
+#include "halfvec.h"
 #include "ivfflat.h"
 #include "miscadmin.h"
+#include "utils/builtins.h"
+#include "utils/datum.h"
 #include "utils/memutils.h"
-#include "utils/relcache.h"
+#include "vector.h"
 
 #if PG_VERSION_NUM >= 160000
 #include "varatt.h"

src/ivfscan.c

@@ -2,26 +2,15 @@
 
 #include <float.h>
 
-#include "access/genam.h"
-#include "access/itup.h"
 #include "access/relscan.h"
-#include "access/tupdesc.h"
 #include "catalog/pg_operator_d.h"
 #include "catalog/pg_type_d.h"
-#include "fmgr.h"
 #include "lib/pairingheap.h"
 #include "ivfflat.h"
 #include "miscadmin.h"
 #include "pgstat.h"
 #include "storage/bufmgr.h"
 #include "utils/memutils.h"
-#include "utils/rel.h"
-#include "utils/snapmgr.h"
-#include "utils/tuplesort.h"
-
-#if PG_VERSION_NUM >= 160000
-#include "varatt.h"
-#endif
 
 #define GetScanList(ptr) pairingheap_container(IvfflatScanList, ph_node, ptr)
 #define GetScanListConst(ptr) pairingheap_const_container(IvfflatScanList, ph_node, ptr)
@@ -298,9 +287,6 @@ ivfflatbeginscan(Relation index, int nkeys, int norderbys)
 	so->tupdesc = CreateTemplateTupleDesc(2);
 	TupleDescInitEntry(so->tupdesc, (AttrNumber) 1, "distance", FLOAT8OID, -1, 0);
 	TupleDescInitEntry(so->tupdesc, (AttrNumber) 2, "heaptid", TIDOID, -1, 0);
-#if PG_VERSION_NUM >= 190000
-	TupleDescFinalize(so->tupdesc);
-#endif
 
 	/* Prep sort */
 	so->sortstate = InitScanSortState(so->tupdesc);

src/ivfutils.c

@@ -1,19 +1,13 @@
 #include "postgres.h"
 
-#include "access/genam.h"
 #include "access/generic_xlog.h"
+#include "bitvec.h"
+#include "catalog/pg_type.h"
 #include "fmgr.h"
 #include "halfutils.h"
 #include "halfvec.h"
 #include "ivfflat.h"
 #include "storage/bufmgr.h"
-#include "utils/relcache.h"
-#include "utils/varbit.h"
-#include "vector.h"
-
-#if PG_VERSION_NUM >= 160000
-#include "varatt.h"
-#endif
 
 /*
  * Allocate a vector array

src/ivfvacuum.c

@@ -1,12 +1,9 @@
 #include "postgres.h"
 
-#include "access/genam.h"
 #include "access/generic_xlog.h"
-#include "access/itup.h"
 #include "commands/vacuum.h"
 #include "ivfflat.h"
 #include "storage/bufmgr.h"
-#include "utils/relcache.h"
 
 #if PG_VERSION_NUM >= 180000
 #define vacuum_delay_point() vacuum_delay_point(false)

src/sparsevec.c

@@ -5,27 +5,18 @@
 
 #include "catalog/pg_type.h"
 #include "common/shortest_dec.h"
+#include "common/string.h"
 #include "fmgr.h"
 #include "halfutils.h"
 #include "halfvec.h"
-#include "lib/stringinfo.h"
 #include "libpq/pqformat.h"
 #include "sparsevec.h"
 #include "utils/array.h"
 #include "utils/builtins.h"
 #include "utils/float.h"
-#include "utils/fmgrprotos.h"
 #include "utils/lsyscache.h"
 #include "vector.h"
 
-#if PG_VERSION_NUM >= 160000
-#include "varatt.h"
-#endif
-
-#if PG_VERSION_NUM >= 170000
-#include "parser/scansup.h"
-#endif
-
 typedef struct SparseInputElement
 {
 	int32		index;
@@ -159,9 +150,9 @@ InitSparseVector(int dim, int nnz)
 	return result;
 }
 
-#if PG_VERSION_NUM >= 170000
+/*
-#define sparsevec_isspace(ch) scanner_isspace(ch)
+ * Check for whitespace, since array_isspace() is static
-#else
+ */
 static inline bool
 sparsevec_isspace(char ch)
 {
@@ -174,7 +165,6 @@ sparsevec_isspace(char ch)
 		return true;
 	return false;
 }
-#endif
 
 /*
  * Compare indices
@@ -182,10 +172,10 @@ sparsevec_isspace(char ch)
 static int
 CompareIndices(const void *a, const void *b)
 {
-	if (((const SparseInputElement *) a)->index < ((const SparseInputElement *) b)->index)
+	if (((SparseInputElement *) a)->index < ((SparseInputElement *) b)->index)
 		return -1;
 
-	if (((const SparseInputElement *) a)->index > ((const SparseInputElement *) b)->index)
+	if (((SparseInputElement *) a)->index > ((SparseInputElement *) b)->index)
 		return 1;
 
 	return 0;

src/vector.c

@@ -16,20 +16,16 @@
 #include "port.h"				/* for strtof() */
 #include "sparsevec.h"
 #include "utils/array.h"
+#include "utils/builtins.h"
 #include "utils/float.h"
-#include "utils/fmgrprotos.h"
 #include "utils/lsyscache.h"
-#include "utils/varbit.h"
+#include "utils/numeric.h"
 #include "vector.h"
 
 #if PG_VERSION_NUM >= 160000
 #include "varatt.h"
 #endif
 
-#if PG_VERSION_NUM >= 170000
-#include "parser/scansup.h"
-#endif
-
 #define STATE_DIMS(x) (ARR_DIMS(x)[0] - 1)
 #define CreateStateDatums(dim) palloc(sizeof(Datum) * (dim + 1))
 
@@ -40,7 +36,7 @@
 #endif
 
 #if PG_VERSION_NUM >= 180000
-PG_MODULE_MAGIC_EXT(.name = "vector", .version = "0.8.3");
+PG_MODULE_MAGIC_EXT(.name = "vector",.version = "0.8.1");
 #else
 PG_MODULE_MAGIC;
 #endif
@@ -133,9 +129,9 @@ InitVector(int dim)
 	return result;
 }
 
-#if PG_VERSION_NUM >= 170000
+/*
-#define vector_isspace(ch) scanner_isspace(ch)
+ * Check for whitespace, since array_isspace() is static
-#else
+ */
 static inline bool
 vector_isspace(char ch)
 {
@@ -148,7 +144,6 @@ vector_isspace(char ch)
 		return true;
 	return false;
 }
-#endif
 
 /*
  * Check state array

test/t/005_ivfflat_query_recall.pl

@@ -16,8 +16,17 @@ $node->safe_psql("postgres",
 	"INSERT INTO tst SELECT i, ARRAY[random(), random(), random()] FROM generate_series(1, 100000) i;"
 );
 
+# Check each index type
+my @operators = ("<->", "<#>", "<=>");
+my @opclasses = ("vector_l2_ops", "vector_ip_ops", "vector_cosine_ops");
+
+for my $i (0 .. $#operators)
+{
+	my $operator = $operators[$i];
+	my $opclass = $opclasses[$i];
+
 	# Add index
-$node->safe_psql("postgres", "CREATE INDEX ON tst USING ivfflat (v vector_l2_ops);");
+	$node->safe_psql("postgres", "CREATE INDEX ON tst USING ivfflat (v $opclass);");
 
 	# Test 100% recall
 	for (1 .. 20)
@@ -30,5 +39,6 @@ for (1 .. 20)
 		));
 		is($res, $query);
 	}
+}
 
 done_testing();

test/t/045_hnsw_low_memory_build.pl

@@ -1,29 +0,0 @@
-use strict;
-use warnings FATAL => 'all';
-use PostgreSQL::Test::Cluster;
-use PostgreSQL::Test::Utils;
-use Test::More;
-
-# Initialize node
-my $node = PostgreSQL::Test::Cluster->new('node');
-$node->init;
-$node->start;
-
-# Create table
-$node->safe_psql("postgres", "CREATE EXTENSION vector;");
-$node->safe_psql("postgres", "CREATE TABLE tst (v vector(3));");
-$node->safe_psql("postgres",
-	"INSERT INTO tst SELECT ARRAY[random(), random(), random()] FROM generate_series(1, 1000) i;"
-);
-
-my ($ret, $stdout, $stderr) = $node->psql("postgres", qq(
-	SET client_min_messages = DEBUG;
-	SET maintenance_work_mem = '3073kB';
-	ALTER TABLE tst SET (parallel_workers = 1);
-	CREATE INDEX ON tst USING hnsw (v vector_l2_ops);
-));
-is($ret, 0, $stderr);
-like($stderr, qr/using \d+ parallel workers/);
-like($stderr, qr/hnsw graph no longer fits into maintenance_work_mem after 0 tuples/);
-
-done_testing();

vector.control

@@ -1,4 +1,4 @@
 comment = 'vector data type and ivfflat and hnsw access methods'
-default_version = '0.8.3'
+default_version = '0.8.1'
 module_pathname = '$libdir/vector'
 relocatable = true