Debug [skip ci]

.github/workflows/build.yml

@@ -8,10 +8,6 @@ jobs:
       fail-fast: false
       matrix:
         include:
-          - postgres: 17
-            os: ubuntu-22.04
-          - postgres: 16
-            os: ubuntu-22.04
           - postgres: 15
             os: ubuntu-22.04
           - postgres: 14
@@ -20,15 +16,15 @@ jobs:
             os: ubuntu-20.04
           - postgres: 12
             os: ubuntu-20.04
+          - postgres: 11
+            os: ubuntu-18.04
     steps:
-      - uses: actions/checkout@v4
+      - uses: actions/checkout@v3
       - uses: ankane/setup-postgres@v1
         with:
           postgres-version: ${{ matrix.postgres }}
           dev-files: true
       - run: make
-        env:
-          PG_CFLAGS: -DUSE_ASSERT_CHECKING -Wall -Wextra -Werror -Wno-unused-parameter -Wno-sign-compare
       - run: |
           export PG_CONFIG=`which pg_config`
           sudo --preserve-env=PG_CONFIG make install
@@ -43,13 +39,11 @@ jobs:
     runs-on: macos-latest
     if: ${{ !startsWith(github.ref_name, 'windows') }}
     steps:
-      - uses: actions/checkout@v4
+      - uses: actions/checkout@v3
       - uses: ankane/setup-postgres@v1
         with:
           postgres-version: 14
       - run: make
-        env:
-          PG_CFLAGS: -DUSE_ASSERT_CHECKING -Wall -Wextra -Werror -Wno-unused-parameter
       - run: make install
       - run: make installcheck
       - if: ${{ failure() }}
@@ -57,58 +51,22 @@ jobs:
       - run: |
           brew install cpanm
           cpanm --notest IPC::Run
-          wget -q https://github.com/postgres/postgres/archive/refs/tags/REL_14_10.tar.gz
+          wget -q https://github.com/postgres/postgres/archive/refs/tags/REL_14_5.tar.gz
-          tar xf REL_14_10.tar.gz
+          tar xf REL_14_5.tar.gz
-      - run: make prove_installcheck PROVE_FLAGS="-I ./postgres-REL_14_10/src/test/perl" PERL5LIB="/Users/runner/perl5/lib/perl5"
+      - run: make prove_installcheck PROVE_FLAGS="-I ./postgres-REL_14_5/src/test/perl" PERL5LIB="/Users/runner/perl5/lib/perl5"
-      - run: make clean && /usr/local/opt/llvm@15/bin/scan-build --status-bugs make
-        env:
-          PG_CFLAGS: -DUSE_ASSERT_CHECKING
   windows:
     runs-on: windows-latest
     if: ${{ !startsWith(github.ref_name, 'mac') }}
     steps:
-      - uses: actions/checkout@v4
+      - uses: actions/checkout@v3
       - uses: ankane/setup-postgres@v1
         with:
           postgres-version: 14
       - run: |
           call "C:\Program Files\Microsoft Visual Studio\2022\Enterprise\VC\Auxiliary\Build\vcvars64.bat" && ^
-          cd %TEMP% && ^
           nmake /NOLOGO /F Makefile.win && ^
           nmake /NOLOGO /F Makefile.win install && ^
           nmake /NOLOGO /F Makefile.win installcheck && ^
           nmake /NOLOGO /F Makefile.win clean && ^
           nmake /NOLOGO /F Makefile.win uninstall
         shell: cmd
-  i386:
-    if: ${{ !startsWith(github.ref_name, 'mac') && !startsWith(github.ref_name, 'windows') }}
-    runs-on: ubuntu-latest
-    container:
-      image: debian:12
-      options: --platform linux/386
-    steps:
-      - run: apt-get update && apt-get install -y build-essential git libipc-run-perl postgresql-15 postgresql-server-dev-15 sudo
-      - run: service postgresql start
-      - run: |
-          git clone https://github.com/${{ github.repository }}.git pgvector
-          cd pgvector
-          git fetch origin ${{ github.ref }}
-          git reset --hard FETCH_HEAD
-          make
-          make install
-          chown -R postgres .
-          sudo -u postgres make installcheck
-          sudo -u postgres make prove_installcheck
-        env:
-          PG_CFLAGS: -DUSE_ASSERT_CHECKING -Wall -Wextra -Werror -Wno-unused-parameter -Wno-sign-compare
-  valgrind:
-    if: ${{ !startsWith(github.ref_name, 'mac') && !startsWith(github.ref_name, 'windows') }}
-    runs-on: ubuntu-latest
-    steps:
-      - uses: actions/checkout@v4
-      - uses: ankane/setup-postgres-valgrind@v1
-        with:
-          postgres-version: 16
-      - run: make
-      - run: sudo --preserve-env=PG_CONFIG make install
-      - run: make installcheck

CHANGELOG.md

@@ -1,72 +1,3 @@
-## 0.7.0 (unreleased)
-
-- Added subscript function for vectors
-
-## 0.6.2 (unreleased)
-
-- Reduced lock contention with parallel HNSW index builds
-
-## 0.6.1 (2024-03-04)
-
-- Fixed error with `ANALYZE` and vectors with different dimensions
-- Fixed error with `shared_preload_libraries`
-- Fixed vector subtraction being marked as commutative
-
-## 0.6.0 (2024-01-29)
-
-If upgrading with Postgres 12 or Docker, see [these notes](https://github.com/pgvector/pgvector#060).
-
-- Added support for parallel index builds for HNSW
-- Added validation for GUC parameters
-- Changed storage for vector from `extended` to `external`
-- Improved performance of HNSW
-- Reduced memory usage for HNSW index builds
-- Reduced WAL generation for HNSW index builds
-- Fixed error with logical replication
-- Fixed `invalid memory alloc request size` error with HNSW index builds
-- Moved Docker image to `pgvector` org
-- Added Docker tags for each supported version of Postgres
-- Dropped support for Postgres 11
-
-## 0.5.1 (2023-10-10)
-
-- Improved performance of HNSW index builds
-- Added check for MVCC-compliant snapshot for index scans
-
-## 0.5.0 (2023-08-28)
-
-- Added HNSW index type
-- Added support for parallel index builds for IVFFlat
-- Added `l1_distance` function
-- Added element-wise multiplication for vectors
-- Added `sum` aggregate
-- Improved performance of distance functions
-- Fixed out of range results for cosine distance
-- Fixed results for NULL and NaN distances for IVFFlat
-
-## 0.4.4 (2023-06-12)
-
-- Improved error message for malformed vector literal
-- Fixed segmentation fault with text input
-- Fixed consecutive delimiters with text input
-
-## 0.4.3 (2023-06-10)
-
-- Improved cost estimation
-- Improved support for spaces with text input
-- Fixed infinite and NaN values with binary input
-- Fixed infinite values with vector addition and subtraction
-- Fixed infinite values with list centers
-- Fixed compilation error when `float8` is pass by reference
-- Fixed compilation error on PowerPC
-- Fixed segmentation fault with index creation on i386
-
-## 0.4.2 (2023-05-13)
-
-- Added notice when index created with little data
-- Fixed dimensions check for some direct function calls
-- Fixed installation error with Postgres 12.0-12.2
-
 ## 0.4.1 (2023-03-21)
 
 - Improved performance of cosine distance
@@ -74,7 +5,7 @@ If upgrading with Postgres 12 or Docker, see [these notes](https://github.com/pg
 
 ## 0.4.0 (2023-01-11)
 
-If upgrading with Postgres < 13, see [this note](https://github.com/pgvector/pgvector/blob/v0.4.0/README.md#040).
+If upgrading with Postgres < 13, see [this note](https://github.com/pgvector/pgvector#040).
 
 - Changed text representation for vector elements to match `real`
 - Changed storage for vector from `plain` to `extended`
@@ -91,7 +22,7 @@ If upgrading with Postgres < 13, see [this note](https://github.com/pgvector/pgv
 
 ## 0.3.1 (2022-11-02)
 
-If upgrading from 0.2.7 or 0.3.0, [recreate](https://github.com/pgvector/pgvector/blob/v0.3.1/README.md#031) all `ivfflat` indexes after upgrading to ensure all data is indexed.
+If upgrading from 0.2.7 or 0.3.0, [recreate](https://github.com/pgvector/pgvector#031) all `ivfflat` indexes after upgrading to ensure all data is indexed.
 
 - Fixed issue with inserts silently corrupting `ivfflat` indexes (introduced in 0.2.7)
 - Fixed segmentation fault with index creation when lists > 6500

Dockerfile

@@ -1,12 +1,9 @@
-ARG PG_MAJOR=16
+FROM postgres:15
-FROM postgres:$PG_MAJOR
-ARG PG_MAJOR
 
 COPY . /tmp/pgvector
 
 RUN apt-get update && \
-		apt-mark hold locales && \
+		apt-get install -y --no-install-recommends build-essential postgresql-server-dev-15 && \
-		apt-get install -y --no-install-recommends build-essential postgresql-server-dev-$PG_MAJOR && \
 		cd /tmp/pgvector && \
 		make clean && \
 		make OPTFLAGS="" && \
@@ -14,7 +11,6 @@ RUN apt-get update && \
 		mkdir /usr/share/doc/pgvector && \
 		cp LICENSE README.md /usr/share/doc/pgvector && \
 		rm -r /tmp/pgvector && \
-		apt-get remove -y build-essential postgresql-server-dev-$PG_MAJOR && \
+		apt-get remove -y build-essential postgresql-server-dev-15 && \
 		apt-get autoremove -y && \
-		apt-mark unhold locales && \
 		rm -rf /var/lib/apt/lists/*

LICENSE

@@ -1,4 +1,4 @@
-Portions Copyright (c) 1996-2024, PostgreSQL Global Development Group
+Portions Copyright (c) 1996-2022, 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": "0.6.1",
+	"version": "0.4.1",
 	"maintainer": [
 		"Andrew Kane <andrew@ankane.org>"
 	],
@@ -12,7 +12,7 @@
 	"prereqs": {
 		"runtime": {
 			"requires": {
-				"PostgreSQL": "12.0.0"
+				"PostgreSQL": "11.0.0"
 			}
 		}
 	},
@@ -20,7 +20,7 @@
 		"vector": {
 			"file": "sql/vector.sql",
 			"docfile": "README.md",
-			"version": "0.6.1",
+			"version": "0.4.1",
 			"abstract": "Open-source vector similarity search for Postgres"
 		}
 	},

Makefile

@@ -1,30 +1,23 @@
 EXTENSION = vector
-EXTVERSION = 0.6.1
+EXTVERSION = 0.4.1
 
 MODULE_big = vector
 DATA = $(wildcard sql/*--*.sql)
-OBJS = src/hnsw.o src/hnswbuild.o src/hnswinsert.o src/hnswscan.o src/hnswutils.o src/hnswvacuum.o src/ivfbuild.o src/ivfflat.o src/ivfinsert.o src/ivfkmeans.o src/ivfscan.o src/ivfutils.o src/ivfvacuum.o src/vector.o
+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
-HEADERS = src/vector.h
 
 TESTS = $(wildcard test/sql/*.sql)
 REGRESS = $(patsubst test/sql/%.sql,%,$(TESTS))
-REGRESS_OPTS = --inputdir=test --load-extension=$(EXTENSION)
+REGRESS_OPTS = --inputdir=test --load-extension=vector
 
 OPTFLAGS = -march=native
 
 # Mac ARM doesn't support -march=native
 ifeq ($(shell uname -s), Darwin)
 	ifeq ($(shell uname -p), arm)
-		# no difference with -march=armv8.5-a
 		OPTFLAGS =
 	endif
 endif
 
-# PowerPC doesn't support -march=native
-ifneq ($(filter ppc64%, $(shell uname -m)), )
-	OPTFLAGS =
-endif
-
 # For auto-vectorization:
 # - GCC (needs -ftree-vectorize OR -O3) - https://gcc.gnu.org/projects/tree-ssa/vectorization.html
 # - Clang (could use pragma instead) - https://llvm.org/docs/Vectorizers.html
@@ -65,15 +58,7 @@ dist:
 	mkdir -p dist
 	git archive --format zip --prefix=$(EXTENSION)-$(EXTVERSION)/ --output dist/$(EXTENSION)-$(EXTVERSION).zip master
 
-# for Docker
-PG_MAJOR ?= 16
-
 .PHONY: docker
 
 docker:
-	docker build --pull --no-cache --build-arg PG_MAJOR=$(PG_MAJOR) -t pgvector/pgvector:pg$(PG_MAJOR) -t pgvector/pgvector:$(EXTVERSION)-pg$(PG_MAJOR) .
+	docker build --pull --no-cache --platform linux/amd64 -t ankane/pgvector:latest .
-
-.PHONY: docker-release
-
-docker-release:
-	docker buildx build --push --pull --no-cache --platform linux/amd64,linux/arm64 --build-arg PG_MAJOR=$(PG_MAJOR) -t pgvector/pgvector:pg$(PG_MAJOR) -t pgvector/pgvector:$(EXTVERSION)-pg$(PG_MAJOR) .

Makefile.win

@@ -1,11 +1,10 @@
 EXTENSION = vector
-EXTVERSION = 0.6.1
+EXTVERSION = 0.4.1
 
-OBJS = 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\vector.obj
+OBJS = src\ivfbuild.obj src\ivfflat.obj src\ivfinsert.obj src\ivfkmeans.obj src\ivfscan.obj src\ivfutils.obj src\ivfvacuum.obj src\vector.obj
-HEADERS = src\vector.h
 
 REGRESS = btree cast copy functions input ivfflat_cosine ivfflat_ip ivfflat_l2 ivfflat_options ivfflat_unlogged
-REGRESS_OPTS = --inputdir=test --load-extension=$(EXTENSION)
+REGRESS_OPTS = --inputdir=test --load-extension=vector
 
 # For /arch flags
 # https://learn.microsoft.com/en-us/cpp/build/reference/arch-minimum-cpu-architecture
@@ -55,8 +54,6 @@ install:
 	copy $(SHLIB) "$(PKGLIBDIR)"
 	copy $(EXTENSION).control "$(SHAREDIR)\extension"
 	copy sql\$(EXTENSION)--*.sql "$(SHAREDIR)\extension"
-	mkdir "$(INCLUDEDIR_SERVER)\extension\$(EXTENSION)"
-	for %f in ($(HEADERS)) do copy %f "$(INCLUDEDIR_SERVER)\extension\$(EXTENSION)"
 
 installcheck:
 	"$(BINDIR)\pg_regress" --bindir="$(BINDIR)" $(REGRESS_OPTS) $(REGRESS)
@@ -64,9 +61,7 @@ installcheck:
 uninstall:
 	del /f "$(PKGLIBDIR)\$(SHLIB)"
 	del /f "$(SHAREDIR)\extension\$(EXTENSION).control"
-	del /f "$(SHAREDIR)\extension\$(EXTENSION)--*.sql"
+	del /f "$(SHAREDIR)\extension\vector--*.sql"
-	del /f "$(INCLUDEDIR_SERVER)\extension\$(EXTENSION)\*.h"
-	rmdir "$(INCLUDEDIR_SERVER)\extension\$(EXTENSION)"
 
 clean:
 	del /f $(SHLIB) $(EXTENSION).lib $(EXTENSION).exp

sql/vector--0.4.1--0.4.2.sql

@@ -1,2 +0,0 @@
--- complain if script is sourced in psql, rather than via CREATE EXTENSION
-\echo Use "ALTER EXTENSION vector UPDATE TO '0.4.2'" to load this file. \quit

sql/vector--0.4.2--0.4.3.sql

@@ -1,2 +0,0 @@
--- complain if script is sourced in psql, rather than via CREATE EXTENSION
-\echo Use "ALTER EXTENSION vector UPDATE TO '0.4.3'" to load this file. \quit

sql/vector--0.4.3--0.4.4.sql

@@ -1,2 +0,0 @@
--- complain if script is sourced in psql, rather than via CREATE EXTENSION
-\echo Use "ALTER EXTENSION vector UPDATE TO '0.4.4'" to load this file. \quit

sql/vector--0.4.4--0.5.0.sql

@@ -1,43 +0,0 @@
--- complain if script is sourced in psql, rather than via CREATE EXTENSION
-\echo Use "ALTER EXTENSION vector UPDATE TO '0.5.0'" to load this file. \quit
-
-CREATE FUNCTION l1_distance(vector, vector) RETURNS float8
-	AS 'MODULE_PATHNAME' LANGUAGE C IMMUTABLE STRICT PARALLEL SAFE;
-
-CREATE FUNCTION vector_mul(vector, vector) RETURNS vector
-	AS 'MODULE_PATHNAME' LANGUAGE C IMMUTABLE STRICT PARALLEL SAFE;
-
-CREATE OPERATOR * (
-	LEFTARG = vector, RIGHTARG = vector, PROCEDURE = vector_mul,
-	COMMUTATOR = *
-);
-
-CREATE AGGREGATE sum(vector) (
-	SFUNC = vector_add,
-	STYPE = vector,
-	COMBINEFUNC = vector_add,
-	PARALLEL = SAFE
-);
-
-CREATE FUNCTION hnswhandler(internal) RETURNS index_am_handler
-	AS 'MODULE_PATHNAME' LANGUAGE C;
-
-CREATE ACCESS METHOD hnsw TYPE INDEX HANDLER hnswhandler;
-
-COMMENT ON ACCESS METHOD hnsw IS 'hnsw index access method';
-
-CREATE OPERATOR CLASS vector_l2_ops
-	FOR TYPE vector USING hnsw AS
-	OPERATOR 1 <-> (vector, vector) FOR ORDER BY float_ops,
-	FUNCTION 1 vector_l2_squared_distance(vector, vector);
-
-CREATE OPERATOR CLASS vector_ip_ops
-	FOR TYPE vector USING hnsw AS
-	OPERATOR 1 <#> (vector, vector) FOR ORDER BY float_ops,
-	FUNCTION 1 vector_negative_inner_product(vector, vector);
-
-CREATE OPERATOR CLASS vector_cosine_ops
-	FOR TYPE vector USING hnsw AS
-	OPERATOR 1 <=> (vector, vector) FOR ORDER BY float_ops,
-	FUNCTION 1 vector_negative_inner_product(vector, vector),
-	FUNCTION 2 vector_norm(vector);

sql/vector--0.5.0--0.5.1.sql

@@ -1,2 +0,0 @@
--- complain if script is sourced in psql, rather than via CREATE EXTENSION
-\echo Use "ALTER EXTENSION vector UPDATE TO '0.5.1'" to load this file. \quit

sql/vector--0.5.1--0.6.0.sql

@@ -1,5 +0,0 @@
--- complain if script is sourced in psql, rather than via CREATE EXTENSION
-\echo Use "ALTER EXTENSION vector UPDATE TO '0.6.0'" to load this file. \quit
-
--- remove this single line for Postgres < 13
-ALTER TYPE vector SET (STORAGE = external);

sql/vector--0.6.0--0.6.1.sql

@@ -1,16 +0,0 @@
--- complain if script is sourced in psql, rather than via CREATE EXTENSION
-\echo Use "ALTER EXTENSION vector UPDATE TO '0.6.1'" to load this file. \quit
-
-DROP OPERATOR - (vector, vector);
-
-CREATE OPERATOR - (
-	LEFTARG = vector, RIGHTARG = vector, PROCEDURE = vector_sub
-);
-
-ALTER OPERATOR <= (vector, vector) SET (
-	RESTRICT = scalarlesel, JOIN = scalarlejoinsel
-);
-
-ALTER OPERATOR >= (vector, vector) SET (
-	RESTRICT = scalargesel, JOIN = scalargejoinsel
-);

sql/vector--0.6.2--0.7.0.sql

@@ -1,7 +0,0 @@
--- complain if script is sourced in psql, rather than via CREATE EXTENSION
-\echo Use "ALTER EXTENSION vector UPDATE TO '0.7.0'" to load this file. \quit
-
-CREATE FUNCTION vector_subscript(internal) RETURNS internal
-	AS 'MODULE_PATHNAME' LANGUAGE C IMMUTABLE STRICT PARALLEL SAFE;
-
-ALTER TYPE vector SET (SUBSCRIPT = vector_subscript);

sql/vector.sql

@@ -20,17 +20,13 @@ CREATE FUNCTION vector_recv(internal, oid, integer) RETURNS vector
 CREATE FUNCTION vector_send(vector) RETURNS bytea
 	AS 'MODULE_PATHNAME' LANGUAGE C IMMUTABLE STRICT PARALLEL SAFE;
 
-CREATE FUNCTION vector_subscript(internal) RETURNS internal
-	AS 'MODULE_PATHNAME' LANGUAGE C IMMUTABLE STRICT PARALLEL SAFE;
-
 CREATE TYPE vector (
 	INPUT     = vector_in,
 	OUTPUT    = vector_out,
 	TYPMOD_IN = vector_typmod_in,
 	RECEIVE   = vector_recv,
 	SEND      = vector_send,
-	SUBSCRIPT = vector_subscript,
+	STORAGE   = extended
-	STORAGE   = external
 );
 
 -- functions
@@ -44,9 +40,6 @@ CREATE FUNCTION inner_product(vector, vector) RETURNS float8
 CREATE FUNCTION cosine_distance(vector, vector) RETURNS float8
 	AS 'MODULE_PATHNAME' LANGUAGE C IMMUTABLE STRICT PARALLEL SAFE;
 
-CREATE FUNCTION l1_distance(vector, vector) RETURNS float8
-	AS 'MODULE_PATHNAME' LANGUAGE C IMMUTABLE STRICT PARALLEL SAFE;
-
 CREATE FUNCTION vector_dims(vector) RETURNS integer
 	AS 'MODULE_PATHNAME' LANGUAGE C IMMUTABLE STRICT PARALLEL SAFE;
 
@@ -59,9 +52,6 @@ CREATE FUNCTION vector_add(vector, vector) RETURNS vector
 CREATE FUNCTION vector_sub(vector, vector) RETURNS vector
 	AS 'MODULE_PATHNAME' LANGUAGE C IMMUTABLE STRICT PARALLEL SAFE;
 
-CREATE FUNCTION vector_mul(vector, vector) RETURNS vector
-	AS 'MODULE_PATHNAME' LANGUAGE C IMMUTABLE STRICT PARALLEL SAFE;
-
 -- private functions
 
 CREATE FUNCTION vector_lt(vector, vector) RETURNS bool
@@ -114,13 +104,6 @@ CREATE AGGREGATE avg(vector) (
 	PARALLEL = SAFE
 );
 
-CREATE AGGREGATE sum(vector) (
-	SFUNC = vector_add,
-	STYPE = vector,
-	COMBINEFUNC = vector_add,
-	PARALLEL = SAFE
-);
-
 -- cast functions
 
 CREATE FUNCTION vector(vector, integer, boolean) RETURNS vector
@@ -184,12 +167,8 @@ CREATE OPERATOR + (
 );
 
 CREATE OPERATOR - (
-	LEFTARG = vector, RIGHTARG = vector, PROCEDURE = vector_sub
+	LEFTARG = vector, RIGHTARG = vector, PROCEDURE = vector_sub,
-);
+	COMMUTATOR = -
-
-CREATE OPERATOR * (
-	LEFTARG = vector, RIGHTARG = vector, PROCEDURE = vector_mul,
-	COMMUTATOR = *
 );
 
 CREATE OPERATOR < (
@@ -198,10 +177,11 @@ CREATE OPERATOR < (
 	RESTRICT = scalarltsel, JOIN = scalarltjoinsel
 );
 
+-- should use scalarlesel and scalarlejoinsel, but not supported in Postgres < 11
 CREATE OPERATOR <= (
 	LEFTARG = vector, RIGHTARG = vector, PROCEDURE = vector_le,
 	COMMUTATOR = >= , NEGATOR = > ,
-	RESTRICT = scalarlesel, JOIN = scalarlejoinsel
+	RESTRICT = scalarltsel, JOIN = scalarltjoinsel
 );
 
 CREATE OPERATOR = (
@@ -216,10 +196,11 @@ CREATE OPERATOR <> (
 	RESTRICT = eqsel, JOIN = eqjoinsel
 );
 
+-- should use scalargesel and scalargejoinsel, but not supported in Postgres < 11
 CREATE OPERATOR >= (
 	LEFTARG = vector, RIGHTARG = vector, PROCEDURE = vector_ge,
 	COMMUTATOR = <= , NEGATOR = < ,
-	RESTRICT = scalargesel, JOIN = scalargejoinsel
+	RESTRICT = scalargtsel, JOIN = scalargtjoinsel
 );
 
 CREATE OPERATOR > (
@@ -228,7 +209,7 @@ CREATE OPERATOR > (
 	RESTRICT = scalargtsel, JOIN = scalargtjoinsel
 );
 
--- access methods
+-- access method
 
 CREATE FUNCTION ivfflathandler(internal) RETURNS index_am_handler
 	AS 'MODULE_PATHNAME' LANGUAGE C;
@@ -237,13 +218,6 @@ CREATE ACCESS METHOD ivfflat TYPE INDEX HANDLER ivfflathandler;
 
 COMMENT ON ACCESS METHOD ivfflat IS 'ivfflat index access method';
 
-CREATE FUNCTION hnswhandler(internal) RETURNS index_am_handler
-	AS 'MODULE_PATHNAME' LANGUAGE C;
-
-CREATE ACCESS METHOD hnsw TYPE INDEX HANDLER hnswhandler;
-
-COMMENT ON ACCESS METHOD hnsw IS 'hnsw index access method';
-
 -- opclasses
 
 CREATE OPERATOR CLASS vector_ops
@@ -275,19 +249,3 @@ CREATE OPERATOR CLASS vector_cosine_ops
 	FUNCTION 2 vector_norm(vector),
 	FUNCTION 3 vector_spherical_distance(vector, vector),
 	FUNCTION 4 vector_norm(vector);
-
-CREATE OPERATOR CLASS vector_l2_ops
-	FOR TYPE vector USING hnsw AS
-	OPERATOR 1 <-> (vector, vector) FOR ORDER BY float_ops,
-	FUNCTION 1 vector_l2_squared_distance(vector, vector);
-
-CREATE OPERATOR CLASS vector_ip_ops
-	FOR TYPE vector USING hnsw AS
-	OPERATOR 1 <#> (vector, vector) FOR ORDER BY float_ops,
-	FUNCTION 1 vector_negative_inner_product(vector, vector);
-
-CREATE OPERATOR CLASS vector_cosine_ops
-	FOR TYPE vector USING hnsw AS
-	OPERATOR 1 <=> (vector, vector) FOR ORDER BY float_ops,
-	FUNCTION 1 vector_negative_inner_product(vector, vector),
-	FUNCTION 2 vector_norm(vector);

src/hnsw.c

@@ -1,249 +0,0 @@
-#include "postgres.h"
-
-#include <float.h>
-#include <math.h>
-
-#include "access/amapi.h"
-#include "access/reloptions.h"
-#include "commands/progress.h"
-#include "commands/vacuum.h"
-#include "hnsw.h"
-#include "miscadmin.h"
-#include "utils/guc.h"
-#include "utils/selfuncs.h"
-
-#if PG_VERSION_NUM < 150000
-#define MarkGUCPrefixReserved(x) EmitWarningsOnPlaceholders(x)
-#endif
-
-int			hnsw_ef_search;
-int			hnsw_lock_tranche_id;
-static relopt_kind hnsw_relopt_kind;
-
-/*
- * Assign a tranche ID for our LWLocks. This only needs to be done by one
- * backend, as the tranche ID is remembered in shared memory.
- *
- * This shared memory area is very small, so we just allocate it from the
- * "slop" that PostgreSQL reserves for small allocations like this. If
- * this grows bigger, we should use a shmem_request_hook and
- * RequestAddinShmemSpace() to pre-reserve space for this.
- */
-void
-HnswInitLockTranche(void)
-{
-	int		   *tranche_ids;
-	bool		found;
-
-	LWLockAcquire(AddinShmemInitLock, LW_EXCLUSIVE);
-	tranche_ids = ShmemInitStruct("hnsw LWLock ids",
-								  sizeof(int) * 1,
-								  &found);
-	if (!found)
-		tranche_ids[0] = LWLockNewTrancheId();
-	hnsw_lock_tranche_id = tranche_ids[0];
-	LWLockRelease(AddinShmemInitLock);
-
-	/* Per-backend registration of the tranche ID */
-	LWLockRegisterTranche(hnsw_lock_tranche_id, "HnswBuild");
-}
-
-/*
- * Initialize index options and variables
- */
-void
-HnswInit(void)
-{
-	if (!process_shared_preload_libraries_in_progress)
-		HnswInitLockTranche();
-
-	hnsw_relopt_kind = add_reloption_kind();
-	add_int_reloption(hnsw_relopt_kind, "m", "Max number of connections",
-					  HNSW_DEFAULT_M, HNSW_MIN_M, HNSW_MAX_M
-#if PG_VERSION_NUM >= 130000
-					  ,AccessExclusiveLock
-#endif
-		);
-	add_int_reloption(hnsw_relopt_kind, "ef_construction", "Size of the dynamic candidate list for construction",
-					  HNSW_DEFAULT_EF_CONSTRUCTION, HNSW_MIN_EF_CONSTRUCTION, HNSW_MAX_EF_CONSTRUCTION
-#if PG_VERSION_NUM >= 130000
-					  ,AccessExclusiveLock
-#endif
-		);
-
-	DefineCustomIntVariable("hnsw.ef_search", "Sets the size of the dynamic candidate list for search",
-							"Valid range is 1..1000.", &hnsw_ef_search,
-							HNSW_DEFAULT_EF_SEARCH, HNSW_MIN_EF_SEARCH, HNSW_MAX_EF_SEARCH, PGC_USERSET, 0, NULL, NULL, NULL);
-
-	MarkGUCPrefixReserved("hnsw");
-}
-
-/*
- * Get the name of index build phase
- */
-static char *
-hnswbuildphasename(int64 phasenum)
-{
-	switch (phasenum)
-	{
-		case PROGRESS_CREATEIDX_SUBPHASE_INITIALIZE:
-			return "initializing";
-		case PROGRESS_HNSW_PHASE_LOAD:
-			return "loading tuples";
-		default:
-			return NULL;
-	}
-}
-
-/*
- * Estimate the cost of an index scan
- */
-static void
-hnswcostestimate(PlannerInfo *root, IndexPath *path, double loop_count,
-				 Cost *indexStartupCost, Cost *indexTotalCost,
-				 Selectivity *indexSelectivity, double *indexCorrelation,
-				 double *indexPages)
-{
-	GenericCosts costs;
-	int			m;
-	int			entryLevel;
-	Relation	index;
-
-	/* Never use index without order */
-	if (path->indexorderbys == NULL)
-	{
-		*indexStartupCost = DBL_MAX;
-		*indexTotalCost = DBL_MAX;
-		*indexSelectivity = 0;
-		*indexCorrelation = 0;
-		*indexPages = 0;
-		return;
-	}
-
-	MemSet(&costs, 0, sizeof(costs));
-
-	index = index_open(path->indexinfo->indexoid, NoLock);
-	HnswGetMetaPageInfo(index, &m, NULL);
-	index_close(index, NoLock);
-
-	/* Approximate entry level */
-	entryLevel = (int) -log(1.0 / path->indexinfo->tuples) * HnswGetMl(m);
-
-	/* TODO Improve estimate of visited tuples (currently underestimates) */
-	/* Account for number of tuples (or entry level), m, and ef_search */
-	costs.numIndexTuples = (entryLevel + 2) * m;
-
-	genericcostestimate(root, path, loop_count, &costs);
-
-	/* Use total cost since most work happens before first tuple is returned */
-	*indexStartupCost = costs.indexTotalCost;
-	*indexTotalCost = costs.indexTotalCost;
-	*indexSelectivity = costs.indexSelectivity;
-	*indexCorrelation = costs.indexCorrelation;
-	*indexPages = costs.numIndexPages;
-}
-
-/*
- * Parse and validate the reloptions
- */
-static bytea *
-hnswoptions(Datum reloptions, bool validate)
-{
-	static const relopt_parse_elt tab[] = {
-		{"m", RELOPT_TYPE_INT, offsetof(HnswOptions, m)},
-		{"ef_construction", RELOPT_TYPE_INT, offsetof(HnswOptions, efConstruction)},
-	};
-
-#if PG_VERSION_NUM >= 130000
-	return (bytea *) build_reloptions(reloptions, validate,
-									  hnsw_relopt_kind,
-									  sizeof(HnswOptions),
-									  tab, lengthof(tab));
-#else
-	relopt_value *options;
-	int			numoptions;
-	HnswOptions *rdopts;
-
-	options = parseRelOptions(reloptions, validate, hnsw_relopt_kind, &numoptions);
-	rdopts = allocateReloptStruct(sizeof(HnswOptions), options, numoptions);
-	fillRelOptions((void *) rdopts, sizeof(HnswOptions), options, numoptions,
-				   validate, tab, lengthof(tab));
-
-	return (bytea *) rdopts;
-#endif
-}
-
-/*
- * Validate catalog entries for the specified operator class
- */
-static bool
-hnswvalidate(Oid opclassoid)
-{
-	return true;
-}
-
-/*
- * Define index handler
- *
- * See https://www.postgresql.org/docs/current/index-api.html
- */
-PGDLLEXPORT PG_FUNCTION_INFO_V1(hnswhandler);
-Datum
-hnswhandler(PG_FUNCTION_ARGS)
-{
-	IndexAmRoutine *amroutine = makeNode(IndexAmRoutine);
-
-	amroutine->amstrategies = 0;
-	amroutine->amsupport = 2;
-#if PG_VERSION_NUM >= 130000
-	amroutine->amoptsprocnum = 0;
-#endif
-	amroutine->amcanorder = false;
-	amroutine->amcanorderbyop = true;
-	amroutine->amcanbackward = false;	/* can change direction mid-scan */
-	amroutine->amcanunique = false;
-	amroutine->amcanmulticol = false;
-	amroutine->amoptionalkey = true;
-	amroutine->amsearcharray = false;
-	amroutine->amsearchnulls = false;
-	amroutine->amstorage = false;
-	amroutine->amclusterable = false;
-	amroutine->ampredlocks = false;
-	amroutine->amcanparallel = false;
-	amroutine->amcaninclude = false;
-#if PG_VERSION_NUM >= 130000
-	amroutine->amusemaintenanceworkmem = false; /* not used during VACUUM */
-	amroutine->amparallelvacuumoptions = VACUUM_OPTION_PARALLEL_BULKDEL;
-#endif
-	amroutine->amkeytype = InvalidOid;
-
-	/* Interface functions */
-	amroutine->ambuild = hnswbuild;
-	amroutine->ambuildempty = hnswbuildempty;
-	amroutine->aminsert = hnswinsert;
-	amroutine->ambulkdelete = hnswbulkdelete;
-	amroutine->amvacuumcleanup = hnswvacuumcleanup;
-	amroutine->amcanreturn = NULL;
-	amroutine->amcostestimate = hnswcostestimate;
-	amroutine->amoptions = hnswoptions;
-	amroutine->amproperty = NULL;	/* TODO AMPROP_DISTANCE_ORDERABLE */
-	amroutine->ambuildphasename = hnswbuildphasename;
-	amroutine->amvalidate = hnswvalidate;
-#if PG_VERSION_NUM >= 140000
-	amroutine->amadjustmembers = NULL;
-#endif
-	amroutine->ambeginscan = hnswbeginscan;
-	amroutine->amrescan = hnswrescan;
-	amroutine->amgettuple = hnswgettuple;
-	amroutine->amgetbitmap = NULL;
-	amroutine->amendscan = hnswendscan;
-	amroutine->ammarkpos = NULL;
-	amroutine->amrestrpos = NULL;
-
-	/* Interface functions to support parallel index scans */
-	amroutine->amestimateparallelscan = NULL;
-	amroutine->aminitparallelscan = NULL;
-	amroutine->amparallelrescan = NULL;
-
-	PG_RETURN_POINTER(amroutine);
-}

src/hnsw.h

@@ -1,462 +0,0 @@
-#ifndef HNSW_H
-#define HNSW_H
-
-#include "postgres.h"
-
-#include "access/genam.h"
-#include "access/parallel.h"
-#include "lib/pairingheap.h"
-#include "nodes/execnodes.h"
-#include "port.h"				/* for random() */
-#include "utils/relptr.h"
-#include "utils/sampling.h"
-#include "vector.h"
-
-#if PG_VERSION_NUM < 120000
-#error "Requires PostgreSQL 12+"
-#endif
-
-#define HNSW_MAX_DIM 2000
-
-/* Support functions */
-#define HNSW_DISTANCE_PROC 1
-#define HNSW_NORM_PROC 2
-
-#define HNSW_VERSION	1
-#define HNSW_MAGIC_NUMBER 0xA953A953
-#define HNSW_PAGE_ID	0xFF90
-
-/* Preserved page numbers */
-#define HNSW_METAPAGE_BLKNO	0
-#define HNSW_HEAD_BLKNO		1	/* first element page */
-
-/* Must correspond to page numbers since page lock is used */
-#define HNSW_UPDATE_LOCK 	0
-#define HNSW_SCAN_LOCK		1
-
-/* HNSW parameters */
-#define HNSW_DEFAULT_M	16
-#define HNSW_MIN_M	2
-#define HNSW_MAX_M		100
-#define HNSW_DEFAULT_EF_CONSTRUCTION	64
-#define HNSW_MIN_EF_CONSTRUCTION	4
-#define HNSW_MAX_EF_CONSTRUCTION		1000
-#define HNSW_DEFAULT_EF_SEARCH	40
-#define HNSW_MIN_EF_SEARCH		1
-#define HNSW_MAX_EF_SEARCH		1000
-
-/* Tuple types */
-#define HNSW_ELEMENT_TUPLE_TYPE  1
-#define HNSW_NEIGHBOR_TUPLE_TYPE 2
-
-/* Make graph robust against non-HOT updates */
-#define HNSW_HEAPTIDS 10
-
-#define HNSW_UPDATE_ENTRY_GREATER 1
-#define HNSW_UPDATE_ENTRY_ALWAYS 2
-
-/* Build phases */
-/* PROGRESS_CREATEIDX_SUBPHASE_INITIALIZE is 1 */
-#define PROGRESS_HNSW_PHASE_LOAD		2
-
-#define HNSW_MAX_SIZE (BLCKSZ - MAXALIGN(SizeOfPageHeaderData) - MAXALIGN(sizeof(HnswPageOpaqueData)) - sizeof(ItemIdData))
-#define HNSW_TUPLE_ALLOC_SIZE BLCKSZ
-
-#define HNSW_ELEMENT_TUPLE_SIZE(size)	MAXALIGN(offsetof(HnswElementTupleData, data) + (size))
-#define HNSW_NEIGHBOR_TUPLE_SIZE(level, m)	MAXALIGN(offsetof(HnswNeighborTupleData, indextids) + ((level) + 2) * (m) * sizeof(ItemPointerData))
-
-#define HNSW_NEIGHBOR_ARRAY_SIZE(lm)	(offsetof(HnswNeighborArray, items) + sizeof(HnswCandidate) * (lm))
-
-#define HnswPageGetOpaque(page)	((HnswPageOpaque) PageGetSpecialPointer(page))
-#define HnswPageGetMeta(page)	((HnswMetaPageData *) PageGetContents(page))
-
-#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)
-#else
-#define RandomDouble() (((double) random()) / MAX_RANDOM_VALUE)
-#define SeedRandom(seed) srandom(seed)
-#endif
-
-#if PG_VERSION_NUM < 130000
-#define list_delete_last(list) list_truncate(list, list_length(list) - 1)
-#define list_sort(list, cmp) ((list) = list_qsort(list, cmp))
-#endif
-
-#define HnswIsElementTuple(tup) ((tup)->type == HNSW_ELEMENT_TUPLE_TYPE)
-#define HnswIsNeighborTuple(tup) ((tup)->type == HNSW_NEIGHBOR_TUPLE_TYPE)
-
-/* 2 * M connections for ground layer */
-#define HnswGetLayerM(m, layer) (layer == 0 ? (m) * 2 : (m))
-
-/* Optimal ML from paper */
-#define HnswGetMl(m) (1 / log(m))
-
-/* Ensure fits on page and in uint8 */
-#define HnswGetMaxLevel(m) Min(((BLCKSZ - MAXALIGN(SizeOfPageHeaderData) - MAXALIGN(sizeof(HnswPageOpaqueData)) - offsetof(HnswNeighborTupleData, indextids) - sizeof(ItemIdData)) / (sizeof(ItemPointerData)) / (m)) - 2, 255)
-
-#define HnswGetValue(base, element) PointerGetDatum(HnswPtrAccess(base, (element)->value))
-
-#if PG_VERSION_NUM < 140005
-#define relptr_offset(rp) ((rp).relptr_off - 1)
-#endif
-
-/* Pointer macros */
-#define HnswPtrAccess(base, hp) ((base) == NULL ? (hp).ptr : relptr_access(base, (hp).relptr))
-#define HnswPtrStore(base, hp, value) ((base) == NULL ? (void) ((hp).ptr = (value)) : (void) relptr_store(base, (hp).relptr, value))
-#define HnswPtrIsNull(base, hp) ((base) == NULL ? (hp).ptr == NULL : relptr_is_null((hp).relptr))
-#define HnswPtrEqual(base, hp1, hp2) ((base) == NULL ? (hp1).ptr == (hp2).ptr : relptr_offset((hp1).relptr) == relptr_offset((hp2).relptr))
-
-/* For code paths dedicated to each type */
-#define HnswPtrPointer(hp) (hp).ptr
-#define HnswPtrOffset(hp) relptr_offset((hp).relptr)
-
-/* Variables */
-extern int	hnsw_ef_search;
-extern int	hnsw_lock_tranche_id;
-
-typedef struct HnswElementData HnswElementData;
-typedef struct HnswNeighborArray HnswNeighborArray;
-
-#define HnswPtrDeclare(type, relptrtype, ptrtype) \
-	relptr_declare(type, relptrtype); \
-	typedef union { type *ptr; relptrtype relptr; } ptrtype;
-
-/* Pointers that can be absolute or relative */
-/* Use char for DatumPtr so works with Pointer */
-HnswPtrDeclare(HnswElementData, HnswElementRelptr, HnswElementPtr);
-HnswPtrDeclare(HnswNeighborArray, HnswNeighborArrayRelptr, HnswNeighborArrayPtr);
-HnswPtrDeclare(HnswNeighborArrayPtr, HnswNeighborsRelptr, HnswNeighborsPtr);
-HnswPtrDeclare(char, DatumRelptr, DatumPtr);
-
-struct HnswElementData
-{
-	HnswElementPtr next;
-	ItemPointerData heaptids[HNSW_HEAPTIDS];
-	uint8		heaptidsLength;
-	uint8		level;
-	uint8		deleted;
-	uint32		hash;
-	HnswNeighborsPtr neighbors;
-	BlockNumber blkno;
-	OffsetNumber offno;
-	OffsetNumber neighborOffno;
-	BlockNumber neighborPage;
-	DatumPtr	value;
-	LWLock		lock;
-};
-
-typedef HnswElementData * HnswElement;
-
-typedef struct HnswCandidate
-{
-	HnswElementPtr element;
-	float		distance;
-	bool		closer;
-}			HnswCandidate;
-
-struct HnswNeighborArray
-{
-	int			length;
-	bool		closerSet;
-	HnswCandidate items[FLEXIBLE_ARRAY_MEMBER];
-};
-
-typedef struct HnswPairingHeapNode
-{
-	pairingheap_node ph_node;
-	HnswCandidate *inner;
-}			HnswPairingHeapNode;
-
-/* HNSW index options */
-typedef struct HnswOptions
-{
-	int32		vl_len_;		/* varlena header (do not touch directly!) */
-	int			m;				/* number of connections */
-	int			efConstruction; /* size of dynamic candidate list */
-}			HnswOptions;
-
-typedef struct HnswGraph
-{
-	/* Graph state */
-	slock_t		lock;
-	HnswElementPtr head;
-	double		indtuples;
-
-	/* Entry state */
-	LWLock		entryLock;
-	LWLock		entryWaitLock;
-	HnswElementPtr entryPoint;
-
-	/* Allocations state */
-	LWLock		allocatorLock;
-	long		memoryUsed;
-	long		memoryTotal;
-
-	/* Flushed state */
-	LWLock		flushLock;
-	bool		flushed;
-}			HnswGraph;
-
-typedef struct HnswShared
-{
-	/* Immutable state */
-	Oid			heaprelid;
-	Oid			indexrelid;
-	bool		isconcurrent;
-
-	/* Worker progress */
-	ConditionVariable workersdonecv;
-
-	/* Mutex for mutable state */
-	slock_t		mutex;
-
-	/* Mutable state */
-	int			nparticipantsdone;
-	double		reltuples;
-	HnswGraph	graphData;
-}			HnswShared;
-
-#define ParallelTableScanFromHnswShared(shared) \
-	(ParallelTableScanDesc) ((char *) (shared) + BUFFERALIGN(sizeof(HnswShared)))
-
-typedef struct HnswLeader
-{
-	ParallelContext *pcxt;
-	int			nparticipanttuplesorts;
-	HnswShared *hnswshared;
-	Snapshot	snapshot;
-	char	   *hnswarea;
-}			HnswLeader;
-
-typedef struct HnswAllocator
-{
-	void	   *(*alloc) (Size size, void *state);
-	void	   *state;
-}			HnswAllocator;
-
-typedef struct HnswBuildState
-{
-	/* Info */
-	Relation	heap;
-	Relation	index;
-	IndexInfo  *indexInfo;
-	ForkNumber	forkNum;
-
-	/* Settings */
-	int			dimensions;
-	int			m;
-	int			efConstruction;
-
-	/* Statistics */
-	double		indtuples;
-	double		reltuples;
-
-	/* Support functions */
-	FmgrInfo   *procinfo;
-	FmgrInfo   *normprocinfo;
-	Oid			collation;
-
-	/* Variables */
-	HnswGraph	graphData;
-	HnswGraph  *graph;
-	double		ml;
-	int			maxLevel;
-	Vector	   *normvec;
-
-	/* Memory */
-	MemoryContext graphCtx;
-	MemoryContext tmpCtx;
-	HnswAllocator allocator;
-
-	/* Parallel builds */
-	HnswLeader *hnswleader;
-	HnswShared *hnswshared;
-	char	   *hnswarea;
-}			HnswBuildState;
-
-typedef struct HnswMetaPageData
-{
-	uint32		magicNumber;
-	uint32		version;
-	uint32		dimensions;
-	uint16		m;
-	uint16		efConstruction;
-	BlockNumber entryBlkno;
-	OffsetNumber entryOffno;
-	int16		entryLevel;
-	BlockNumber insertPage;
-}			HnswMetaPageData;
-
-typedef HnswMetaPageData * HnswMetaPage;
-
-typedef struct HnswPageOpaqueData
-{
-	BlockNumber nextblkno;
-	uint16		unused;
-	uint16		page_id;		/* for identification of HNSW indexes */
-}			HnswPageOpaqueData;
-
-typedef HnswPageOpaqueData * HnswPageOpaque;
-
-typedef struct HnswElementTupleData
-{
-	uint8		type;
-	uint8		level;
-	uint8		deleted;
-	uint8		unused;
-	ItemPointerData heaptids[HNSW_HEAPTIDS];
-	ItemPointerData neighbortid;
-	uint16		unused2;
-	Vector		data;
-}			HnswElementTupleData;
-
-typedef HnswElementTupleData * HnswElementTuple;
-
-typedef struct HnswNeighborTupleData
-{
-	uint8		type;
-	uint8		unused;
-	uint16		count;
-	ItemPointerData indextids[FLEXIBLE_ARRAY_MEMBER];
-}			HnswNeighborTupleData;
-
-typedef HnswNeighborTupleData * HnswNeighborTuple;
-
-typedef struct HnswScanOpaqueData
-{
-	bool		first;
-	List	   *w;
-	MemoryContext tmpCtx;
-
-	/* Support functions */
-	FmgrInfo   *procinfo;
-	FmgrInfo   *normprocinfo;
-	Oid			collation;
-}			HnswScanOpaqueData;
-
-typedef HnswScanOpaqueData * HnswScanOpaque;
-
-typedef struct HnswVacuumState
-{
-	/* Info */
-	Relation	index;
-	IndexBulkDeleteResult *stats;
-	IndexBulkDeleteCallback callback;
-	void	   *callback_state;
-
-	/* Settings */
-	int			m;
-	int			efConstruction;
-
-	/* Support functions */
-	FmgrInfo   *procinfo;
-	Oid			collation;
-
-	/* Variables */
-	struct tidhash_hash *deleted;
-	BufferAccessStrategy bas;
-	HnswNeighborTuple ntup;
-	HnswElementData highestPoint;
-
-	/* Memory */
-	MemoryContext tmpCtx;
-}			HnswVacuumState;
-
-/* Methods */
-int			HnswGetM(Relation index);
-int			HnswGetEfConstruction(Relation index);
-FmgrInfo   *HnswOptionalProcInfo(Relation index, uint16 procnum);
-bool		HnswNormValue(FmgrInfo *procinfo, Oid collation, Datum *value, Vector * result);
-Buffer		HnswNewBuffer(Relation index, ForkNumber forkNum);
-void		HnswInitPage(Buffer buf, Page page);
-void		HnswInit(void);
-List	   *HnswSearchLayer(char *base, Datum q, List *ep, int ef, int lc, Relation index, FmgrInfo *procinfo, Oid collation, int m, bool inserting, HnswElement skipElement);
-HnswElement HnswGetEntryPoint(Relation index);
-void		HnswGetMetaPageInfo(Relation index, int *m, HnswElement * entryPoint);
-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, FmgrInfo *procinfo, Oid collation, int m, int efConstruction, bool existing);
-HnswCandidate *HnswEntryCandidate(char *base, HnswElement em, Datum q, Relation rel, FmgrInfo *procinfo, Oid collation, 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);
-void		HnswInitNeighbors(char *base, HnswElement element, int m, HnswAllocator * alloc);
-bool		HnswInsertTupleOnDisk(Relation index, Datum value, Datum *values, bool *isnull, ItemPointer heap_tid, bool building);
-void		HnswUpdateNeighborsOnDisk(Relation index, FmgrInfo *procinfo, Oid collation, HnswElement e, int m, bool checkExisting, bool building);
-void		HnswLoadElementFromTuple(HnswElement element, HnswElementTuple etup, bool loadHeaptids, bool loadVec);
-void		HnswLoadElement(HnswElement element, float *distance, Datum *q, Relation index, FmgrInfo *procinfo, Oid collation, bool loadVec);
-void		HnswSetElementTuple(char *base, HnswElementTuple etup, HnswElement element);
-void		HnswUpdateConnection(char *base, HnswElement element, HnswCandidate * hc, int lm, int lc, int *updateIdx, Relation index, FmgrInfo *procinfo, Oid collation);
-void		HnswLoadNeighbors(HnswElement element, Relation index, int m);
-void		HnswInitLockTranche(void);
-PGDLLEXPORT void HnswParallelBuildMain(dsm_segment *seg, shm_toc *toc);
-
-/* Index access methods */
-IndexBuildResult *hnswbuild(Relation heap, Relation index, IndexInfo *indexInfo);
-void		hnswbuildempty(Relation index);
-bool		hnswinsert(Relation index, Datum *values, bool *isnull, ItemPointer heap_tid, Relation heap, IndexUniqueCheck checkUnique
-#if PG_VERSION_NUM >= 140000
-					   ,bool indexUnchanged
-#endif
-					   ,IndexInfo *indexInfo
-);
-IndexBulkDeleteResult *hnswbulkdelete(IndexVacuumInfo *info, IndexBulkDeleteResult *stats, IndexBulkDeleteCallback callback, void *callback_state);
-IndexBulkDeleteResult *hnswvacuumcleanup(IndexVacuumInfo *info, IndexBulkDeleteResult *stats);
-IndexScanDesc hnswbeginscan(Relation index, int nkeys, int norderbys);
-void		hnswrescan(IndexScanDesc scan, ScanKey keys, int nkeys, ScanKey orderbys, int norderbys);
-bool		hnswgettuple(IndexScanDesc scan, ScanDirection dir);
-void		hnswendscan(IndexScanDesc scan);
-
-static inline HnswNeighborArray *
-HnswGetNeighbors(char *base, HnswElement element, int lc)
-{
-	HnswNeighborArrayPtr *neighborList = HnswPtrAccess(base, element->neighbors);
-
-	Assert(element->level >= lc);
-
-	return HnswPtrAccess(base, neighborList[lc]);
-}
-
-/* Hash tables */
-typedef struct TidHashEntry
-{
-	ItemPointerData tid;
-	char		status;
-}			TidHashEntry;
-
-#define SH_PREFIX tidhash
-#define SH_ELEMENT_TYPE TidHashEntry
-#define SH_KEY_TYPE ItemPointerData
-#define SH_SCOPE extern
-#define SH_DECLARE
-#include "lib/simplehash.h"
-
-typedef struct PointerHashEntry
-{
-	uintptr_t	ptr;
-	char		status;
-}			PointerHashEntry;
-
-#define SH_PREFIX pointerhash
-#define SH_ELEMENT_TYPE PointerHashEntry
-#define SH_KEY_TYPE uintptr_t
-#define SH_SCOPE extern
-#define SH_DECLARE
-#include "lib/simplehash.h"
-
-typedef struct OffsetHashEntry
-{
-	Size		offset;
-	char		status;
-}			OffsetHashEntry;
-
-#define SH_PREFIX offsethash
-#define SH_ELEMENT_TYPE OffsetHashEntry
-#define SH_KEY_TYPE Size
-#define SH_SCOPE extern
-#define SH_DECLARE
-#include "lib/simplehash.h"
-
-#endif

src/hnswinsert.c

@@ -1,665 +0,0 @@
-#include "postgres.h"
-
-#include <math.h>
-
-#include "access/generic_xlog.h"
-#include "hnsw.h"
-#include "storage/bufmgr.h"
-#include "storage/lmgr.h"
-#include "utils/datum.h"
-#include "utils/memutils.h"
-
-/*
- * Get the insert page
- */
-static BlockNumber
-GetInsertPage(Relation index)
-{
-	Buffer		buf;
-	Page		page;
-	HnswMetaPage metap;
-	BlockNumber insertPage;
-
-	buf = ReadBuffer(index, HNSW_METAPAGE_BLKNO);
-	LockBuffer(buf, BUFFER_LOCK_SHARE);
-	page = BufferGetPage(buf);
-	metap = HnswPageGetMeta(page);
-
-	insertPage = metap->insertPage;
-
-	UnlockReleaseBuffer(buf);
-
-	return insertPage;
-}
-
-/*
- * Check for a free offset
- */
-static bool
-HnswFreeOffset(Relation index, Buffer buf, Page page, HnswElement element, Size ntupSize, Buffer *nbuf, Page *npage, OffsetNumber *freeOffno, OffsetNumber *freeNeighborOffno, BlockNumber *newInsertPage)
-{
-	OffsetNumber offno;
-	OffsetNumber maxoffno = PageGetMaxOffsetNumber(page);
-
-	for (offno = FirstOffsetNumber; offno <= maxoffno; offno = OffsetNumberNext(offno))
-	{
-		HnswElementTuple etup = (HnswElementTuple) PageGetItem(page, PageGetItemId(page, offno));
-
-		/* Skip neighbor tuples */
-		if (!HnswIsElementTuple(etup))
-			continue;
-
-		if (etup->deleted)
-		{
-			BlockNumber elementPage = BufferGetBlockNumber(buf);
-			BlockNumber neighborPage = ItemPointerGetBlockNumber(&etup->neighbortid);
-			OffsetNumber neighborOffno = ItemPointerGetOffsetNumber(&etup->neighbortid);
-			ItemId		itemid;
-
-			if (!BlockNumberIsValid(*newInsertPage))
-				*newInsertPage = elementPage;
-
-			if (neighborPage == elementPage)
-			{
-				*nbuf = buf;
-				*npage = page;
-			}
-			else
-			{
-				*nbuf = ReadBuffer(index, neighborPage);
-				LockBuffer(*nbuf, BUFFER_LOCK_EXCLUSIVE);
-
-				/* Skip WAL for now */
-				*npage = BufferGetPage(*nbuf);
-			}
-
-			itemid = PageGetItemId(*npage, neighborOffno);
-
-			/* Check for space on neighbor tuple page */
-			if (PageGetFreeSpace(*npage) + ItemIdGetLength(itemid) - sizeof(ItemIdData) >= ntupSize)
-			{
-				*freeOffno = offno;
-				*freeNeighborOffno = neighborOffno;
-				return true;
-			}
-			else if (*nbuf != buf)
-				UnlockReleaseBuffer(*nbuf);
-		}
-	}
-
-	return false;
-}
-
-/*
- * Add a new page
- */
-static void
-HnswInsertAppendPage(Relation index, Buffer *nbuf, Page *npage, GenericXLogState *state, Page page, bool building)
-{
-	/* Add a new page */
-	LockRelationForExtension(index, ExclusiveLock);
-	*nbuf = HnswNewBuffer(index, MAIN_FORKNUM);
-	UnlockRelationForExtension(index, ExclusiveLock);
-
-	/* Init new page */
-	if (building)
-		*npage = BufferGetPage(*nbuf);
-	else
-		*npage = GenericXLogRegisterBuffer(state, *nbuf, GENERIC_XLOG_FULL_IMAGE);
-
-	HnswInitPage(*nbuf, *npage);
-
-	/* Update previous buffer */
-	HnswPageGetOpaque(page)->nextblkno = BufferGetBlockNumber(*nbuf);
-}
-
-/*
- * Add to element and neighbor pages
- */
-static void
-AddElementOnDisk(Relation index, HnswElement e, int m, BlockNumber insertPage, BlockNumber *updatedInsertPage, bool building)
-{
-	Buffer		buf;
-	Page		page;
-	GenericXLogState *state;
-	Size		etupSize;
-	Size		ntupSize;
-	Size		combinedSize;
-	Size		maxSize;
-	Size		minCombinedSize;
-	HnswElementTuple etup;
-	BlockNumber currentPage = insertPage;
-	HnswNeighborTuple ntup;
-	Buffer		nbuf;
-	Page		npage;
-	OffsetNumber freeOffno = InvalidOffsetNumber;
-	OffsetNumber freeNeighborOffno = InvalidOffsetNumber;
-	BlockNumber newInsertPage = InvalidBlockNumber;
-	char	   *base = NULL;
-
-	/* Calculate sizes */
-	etupSize = HNSW_ELEMENT_TUPLE_SIZE(VARSIZE_ANY(HnswPtrAccess(base, e->value)));
-	ntupSize = HNSW_NEIGHBOR_TUPLE_SIZE(e->level, m);
-	combinedSize = etupSize + ntupSize + sizeof(ItemIdData);
-	maxSize = HNSW_MAX_SIZE;
-	minCombinedSize = etupSize + HNSW_NEIGHBOR_TUPLE_SIZE(0, m) + sizeof(ItemIdData);
-
-	/* Prepare element tuple */
-	etup = palloc0(etupSize);
-	HnswSetElementTuple(base, etup, e);
-
-	/* Prepare neighbor tuple */
-	ntup = palloc0(ntupSize);
-	HnswSetNeighborTuple(base, ntup, e, m);
-
-	/* Find a page (or two if needed) to insert the tuples */
-	for (;;)
-	{
-		buf = ReadBuffer(index, currentPage);
-		LockBuffer(buf, BUFFER_LOCK_EXCLUSIVE);
-
-		if (building)
-		{
-			state = NULL;
-			page = BufferGetPage(buf);
-		}
-		else
-		{
-			state = GenericXLogStart(index);
-			page = GenericXLogRegisterBuffer(state, buf, 0);
-		}
-
-		/* Keep track of first page where element at level 0 can fit */
-		if (!BlockNumberIsValid(newInsertPage) && PageGetFreeSpace(page) >= minCombinedSize)
-			newInsertPage = currentPage;
-
-		/* First, try the fastest path */
-		/* Space for both tuples on the current page */
-		/* This can split existing tuples in rare cases */
-		if (PageGetFreeSpace(page) >= combinedSize)
-		{
-			nbuf = buf;
-			npage = page;
-			break;
-		}
-
-		/* Next, try space from a deleted element */
-		if (HnswFreeOffset(index, buf, page, e, ntupSize, &nbuf, &npage, &freeOffno, &freeNeighborOffno, &newInsertPage))
-		{
-			if (nbuf != buf)
-			{
-				if (building)
-					npage = BufferGetPage(nbuf);
-				else
-					npage = GenericXLogRegisterBuffer(state, nbuf, 0);
-			}
-
-			break;
-		}
-
-		/* Finally, try space for element only if last page */
-		/* Skip if both tuples can fit on the same page */
-		if (combinedSize > maxSize && PageGetFreeSpace(page) >= etupSize && !BlockNumberIsValid(HnswPageGetOpaque(page)->nextblkno))
-		{
-			HnswInsertAppendPage(index, &nbuf, &npage, state, page, building);
-			break;
-		}
-
-		currentPage = HnswPageGetOpaque(page)->nextblkno;
-
-		if (BlockNumberIsValid(currentPage))
-		{
-			/* Move to next page */
-			if (!building)
-				GenericXLogAbort(state);
-			UnlockReleaseBuffer(buf);
-		}
-		else
-		{
-			Buffer		newbuf;
-			Page		newpage;
-
-			HnswInsertAppendPage(index, &newbuf, &newpage, state, page, building);
-
-			/* Commit */
-			if (building)
-				MarkBufferDirty(buf);
-			else
-				GenericXLogFinish(state);
-
-			/* Unlock previous buffer */
-			UnlockReleaseBuffer(buf);
-
-			/* Prepare new buffer */
-			buf = newbuf;
-			if (building)
-			{
-				state = NULL;
-				page = BufferGetPage(buf);
-			}
-			else
-			{
-				state = GenericXLogStart(index);
-				page = GenericXLogRegisterBuffer(state, buf, 0);
-			}
-
-			/* Create new page for neighbors if needed */
-			if (PageGetFreeSpace(page) < combinedSize)
-				HnswInsertAppendPage(index, &nbuf, &npage, state, page, building);
-			else
-			{
-				nbuf = buf;
-				npage = page;
-			}
-
-			break;
-		}
-	}
-
-	e->blkno = BufferGetBlockNumber(buf);
-	e->neighborPage = BufferGetBlockNumber(nbuf);
-
-	/* Added tuple to new page if newInsertPage is not set */
-	/* So can set to neighbor page instead of element page */
-	if (!BlockNumberIsValid(newInsertPage))
-		newInsertPage = e->neighborPage;
-
-	if (OffsetNumberIsValid(freeOffno))
-	{
-		e->offno = freeOffno;
-		e->neighborOffno = freeNeighborOffno;
-	}
-	else
-	{
-		e->offno = OffsetNumberNext(PageGetMaxOffsetNumber(page));
-		if (nbuf == buf)
-			e->neighborOffno = OffsetNumberNext(e->offno);
-		else
-			e->neighborOffno = FirstOffsetNumber;
-	}
-
-	ItemPointerSet(&etup->neighbortid, e->neighborPage, e->neighborOffno);
-
-	/* Add element and neighbors */
-	if (OffsetNumberIsValid(freeOffno))
-	{
-		if (!PageIndexTupleOverwrite(page, e->offno, (Item) etup, etupSize))
-			elog(ERROR, "failed to add index item to \"%s\"", RelationGetRelationName(index));
-
-		if (!PageIndexTupleOverwrite(npage, e->neighborOffno, (Item) ntup, ntupSize))
-			elog(ERROR, "failed to add index item to \"%s\"", RelationGetRelationName(index));
-	}
-	else
-	{
-		if (PageAddItem(page, (Item) etup, etupSize, InvalidOffsetNumber, false, false) != e->offno)
-			elog(ERROR, "failed to add index item to \"%s\"", RelationGetRelationName(index));
-
-		if (PageAddItem(npage, (Item) ntup, ntupSize, InvalidOffsetNumber, false, false) != e->neighborOffno)
-			elog(ERROR, "failed to add index item to \"%s\"", RelationGetRelationName(index));
-	}
-
-	/* Commit */
-	if (building)
-	{
-		MarkBufferDirty(buf);
-		if (nbuf != buf)
-			MarkBufferDirty(nbuf);
-	}
-	else
-		GenericXLogFinish(state);
-	UnlockReleaseBuffer(buf);
-	if (nbuf != buf)
-		UnlockReleaseBuffer(nbuf);
-
-	/* Update the insert page */
-	if (BlockNumberIsValid(newInsertPage) && newInsertPage != insertPage)
-		*updatedInsertPage = newInsertPage;
-}
-
-/*
- * Check if connection already exists
- */
-static bool
-ConnectionExists(HnswElement e, HnswNeighborTuple ntup, int startIdx, int lm)
-{
-	for (int i = 0; i < lm; i++)
-	{
-		ItemPointer indextid = &ntup->indextids[startIdx + i];
-
-		if (!ItemPointerIsValid(indextid))
-			break;
-
-		if (ItemPointerGetBlockNumber(indextid) == e->blkno && ItemPointerGetOffsetNumber(indextid) == e->offno)
-			return true;
-	}
-
-	return false;
-}
-
-/*
- * Update neighbors
- */
-void
-HnswUpdateNeighborsOnDisk(Relation index, FmgrInfo *procinfo, Oid collation, HnswElement e, int m, bool checkExisting, bool building)
-{
-	char	   *base = NULL;
-
-	for (int lc = e->level; lc >= 0; lc--)
-	{
-		int			lm = HnswGetLayerM(m, lc);
-		HnswNeighborArray *neighbors = HnswGetNeighbors(base, e, lc);
-
-		for (int i = 0; i < neighbors->length; i++)
-		{
-			HnswCandidate *hc = &neighbors->items[i];
-			Buffer		buf;
-			Page		page;
-			GenericXLogState *state;
-			HnswNeighborTuple ntup;
-			int			idx = -1;
-			int			startIdx;
-			HnswElement neighborElement = HnswPtrAccess(base, hc->element);
-			OffsetNumber offno = neighborElement->neighborOffno;
-
-			/* Get latest neighbors since they may have changed */
-			/* Do not lock yet since selecting neighbors can take time */
-			HnswLoadNeighbors(neighborElement, index, m);
-
-			/*
-			 * Could improve performance for vacuuming by checking neighbors
-			 * against list of elements being deleted to find index. It's
-			 * important to exclude already deleted elements for this since
-			 * they can be replaced at any time.
-			 */
-
-			/* Select neighbors */
-			HnswUpdateConnection(NULL, e, hc, lm, lc, &idx, index, procinfo, collation);
-
-			/* New element was not selected as a neighbor */
-			if (idx == -1)
-				continue;
-
-			/* Register page */
-			buf = ReadBuffer(index, neighborElement->neighborPage);
-			LockBuffer(buf, BUFFER_LOCK_EXCLUSIVE);
-			if (building)
-			{
-				state = NULL;
-				page = BufferGetPage(buf);
-			}
-			else
-			{
-				state = GenericXLogStart(index);
-				page = GenericXLogRegisterBuffer(state, buf, 0);
-			}
-
-			/* Get tuple */
-			ntup = (HnswNeighborTuple) PageGetItem(page, PageGetItemId(page, offno));
-
-			/* Calculate index for update */
-			startIdx = (neighborElement->level - lc) * m;
-
-			/* Check for existing connection */
-			if (checkExisting && ConnectionExists(e, ntup, startIdx, lm))
-				idx = -1;
-			else if (idx == -2)
-			{
-				/* Find free offset if still exists */
-				/* TODO Retry updating connections if not */
-				for (int j = 0; j < lm; j++)
-				{
-					if (!ItemPointerIsValid(&ntup->indextids[startIdx + j]))
-					{
-						idx = startIdx + j;
-						break;
-					}
-				}
-			}
-			else
-				idx += startIdx;
-
-			/* Make robust to issues */
-			if (idx >= 0 && idx < ntup->count)
-			{
-				ItemPointer indextid = &ntup->indextids[idx];
-
-				/* Update neighbor on the buffer */
-				ItemPointerSet(indextid, e->blkno, e->offno);
-
-				/* Commit */
-				if (building)
-					MarkBufferDirty(buf);
-				else
-					GenericXLogFinish(state);
-			}
-			else if (!building)
-				GenericXLogAbort(state);
-
-			UnlockReleaseBuffer(buf);
-		}
-	}
-}
-
-/*
- * Add a heap TID to an existing element
- */
-static bool
-AddDuplicateOnDisk(Relation index, HnswElement element, HnswElement dup, bool building)
-{
-	Buffer		buf;
-	Page		page;
-	GenericXLogState *state;
-	HnswElementTuple etup;
-	int			i;
-
-	/* Read page */
-	buf = ReadBuffer(index, dup->blkno);
-	LockBuffer(buf, BUFFER_LOCK_EXCLUSIVE);
-	if (building)
-	{
-		state = NULL;
-		page = BufferGetPage(buf);
-	}
-	else
-	{
-		state = GenericXLogStart(index);
-		page = GenericXLogRegisterBuffer(state, buf, 0);
-	}
-
-	/* Find space */
-	etup = (HnswElementTuple) PageGetItem(page, PageGetItemId(page, dup->offno));
-	for (i = 0; i < HNSW_HEAPTIDS; i++)
-	{
-		if (!ItemPointerIsValid(&etup->heaptids[i]))
-			break;
-	}
-
-	/* Either being deleted or we lost our chance to another backend */
-	if (i == 0 || i == HNSW_HEAPTIDS)
-	{
-		if (!building)
-			GenericXLogAbort(state);
-		UnlockReleaseBuffer(buf);
-		return false;
-	}
-
-	/* Add heap TID, modifying the tuple on the page directly */
-	etup->heaptids[i] = element->heaptids[0];
-
-	/* Commit */
-	if (building)
-		MarkBufferDirty(buf);
-	else
-		GenericXLogFinish(state);
-	UnlockReleaseBuffer(buf);
-
-	return true;
-}
-
-/*
- * Find duplicate element
- */
-static bool
-FindDuplicateOnDisk(Relation index, HnswElement element, bool building)
-{
-	char	   *base = NULL;
-	HnswNeighborArray *neighbors = HnswGetNeighbors(base, element, 0);
-	Datum		value = HnswGetValue(base, element);
-
-	for (int i = 0; i < neighbors->length; i++)
-	{
-		HnswCandidate *neighbor = &neighbors->items[i];
-		HnswElement neighborElement = HnswPtrAccess(base, neighbor->element);
-		Datum		neighborValue = HnswGetValue(base, neighborElement);
-
-		/* Exit early since ordered by distance */
-		if (!datumIsEqual(value, neighborValue, false, -1))
-			return false;
-
-		if (AddDuplicateOnDisk(index, element, neighborElement, building))
-			return true;
-	}
-
-	return false;
-}
-
-/*
- * Update graph on disk
- */
-static void
-UpdateGraphOnDisk(Relation index, FmgrInfo *procinfo, Oid collation, HnswElement element, int m, int efConstruction, HnswElement entryPoint, bool building)
-{
-	BlockNumber newInsertPage = InvalidBlockNumber;
-
-	/* Look for duplicate */
-	if (FindDuplicateOnDisk(index, element, building))
-		return;
-
-	/* Add element */
-	AddElementOnDisk(index, element, m, GetInsertPage(index), &newInsertPage, building);
-
-	/* Update insert page if needed */
-	if (BlockNumberIsValid(newInsertPage))
-		HnswUpdateMetaPage(index, 0, NULL, newInsertPage, MAIN_FORKNUM, building);
-
-	/* Update neighbors */
-	HnswUpdateNeighborsOnDisk(index, procinfo, collation, element, m, false, building);
-
-	/* Update entry point if needed */
-	if (entryPoint == NULL || element->level > entryPoint->level)
-		HnswUpdateMetaPage(index, HNSW_UPDATE_ENTRY_GREATER, element, InvalidBlockNumber, MAIN_FORKNUM, building);
-}
-
-/*
- * Insert a tuple into the index
- */
-bool
-HnswInsertTupleOnDisk(Relation index, Datum value, Datum *values, bool *isnull, ItemPointer heap_tid, bool building)
-{
-	HnswElement entryPoint;
-	HnswElement element;
-	int			m;
-	int			efConstruction = HnswGetEfConstruction(index);
-	FmgrInfo   *procinfo = index_getprocinfo(index, 1, HNSW_DISTANCE_PROC);
-	Oid			collation = index->rd_indcollation[0];
-	LOCKMODE	lockmode = ShareLock;
-	char	   *base = NULL;
-
-	/*
-	 * Get a shared lock. This allows vacuum to ensure no in-flight inserts
-	 * before repairing graph. Use a page lock so it does not interfere with
-	 * buffer lock (or reads when vacuuming).
-	 */
-	LockPage(index, HNSW_UPDATE_LOCK, lockmode);
-
-	/* Get m and entry point */
-	HnswGetMetaPageInfo(index, &m, &entryPoint);
-
-	/* Create an element */
-	element = HnswInitElement(base, heap_tid, m, HnswGetMl(m), HnswGetMaxLevel(m), NULL);
-	HnswPtrStore(base, element->value, DatumGetPointer(value));
-
-	/* Prevent concurrent inserts when likely updating entry point */
-	if (entryPoint == NULL || element->level > entryPoint->level)
-	{
-		/* Release shared lock */
-		UnlockPage(index, HNSW_UPDATE_LOCK, lockmode);
-
-		/* Get exclusive lock */
-		lockmode = ExclusiveLock;
-		LockPage(index, HNSW_UPDATE_LOCK, lockmode);
-
-		/* Get latest entry point after lock is acquired */
-		entryPoint = HnswGetEntryPoint(index);
-	}
-
-	/* Find neighbors for element */
-	HnswFindElementNeighbors(base, element, entryPoint, index, procinfo, collation, m, efConstruction, false);
-
-	/* Update graph on disk */
-	UpdateGraphOnDisk(index, procinfo, collation, element, m, efConstruction, entryPoint, building);
-
-	/* Release lock */
-	UnlockPage(index, HNSW_UPDATE_LOCK, lockmode);
-
-	return true;
-}
-
-/*
- * Insert a tuple into the index
- */
-static void
-HnswInsertTuple(Relation index, Datum *values, bool *isnull, ItemPointer heap_tid)
-{
-	Datum		value;
-	FmgrInfo   *normprocinfo;
-	Oid			collation = index->rd_indcollation[0];
-
-	/* Detoast once for all calls */
-	value = PointerGetDatum(PG_DETOAST_DATUM(values[0]));
-
-	/* Normalize if needed */
-	normprocinfo = HnswOptionalProcInfo(index, HNSW_NORM_PROC);
-	if (normprocinfo != NULL)
-	{
-		if (!HnswNormValue(normprocinfo, collation, &value, NULL))
-			return;
-	}
-
-	HnswInsertTupleOnDisk(index, value, values, isnull, heap_tid, false);
-}
-
-/*
- * Insert a tuple into the index
- */
-bool
-hnswinsert(Relation index, Datum *values, bool *isnull, ItemPointer heap_tid,
-		   Relation heap, IndexUniqueCheck checkUnique
-#if PG_VERSION_NUM >= 140000
-		   ,bool indexUnchanged
-#endif
-		   ,IndexInfo *indexInfo
-)
-{
-	MemoryContext oldCtx;
-	MemoryContext insertCtx;
-
-	/* Skip nulls */
-	if (isnull[0])
-		return false;
-
-	/* Create memory context */
-	insertCtx = AllocSetContextCreate(CurrentMemoryContext,
-									  "Hnsw insert temporary context",
-									  ALLOCSET_DEFAULT_SIZES);
-	oldCtx = MemoryContextSwitchTo(insertCtx);
-
-	/* Insert tuple */
-	HnswInsertTuple(index, values, isnull, heap_tid);
-
-	/* Delete memory context */
-	MemoryContextSwitchTo(oldCtx);
-	MemoryContextDelete(insertCtx);
-
-	return false;
-}

src/hnswscan.c

@@ -1,226 +0,0 @@
-#include "postgres.h"
-
-#include "access/relscan.h"
-#include "hnsw.h"
-#include "pgstat.h"
-#include "storage/bufmgr.h"
-#include "storage/lmgr.h"
-#include "utils/memutils.h"
-
-/*
- * Algorithm 5 from paper
- */
-static List *
-GetScanItems(IndexScanDesc scan, Datum q)
-{
-	HnswScanOpaque so = (HnswScanOpaque) scan->opaque;
-	Relation	index = scan->indexRelation;
-	FmgrInfo   *procinfo = so->procinfo;
-	Oid			collation = so->collation;
-	List	   *ep;
-	List	   *w;
-	int			m;
-	HnswElement entryPoint;
-	char	   *base = NULL;
-
-	/* Get m and entry point */
-	HnswGetMetaPageInfo(index, &m, &entryPoint);
-
-	if (entryPoint == NULL)
-		return NIL;
-
-	ep = list_make1(HnswEntryCandidate(base, entryPoint, q, index, procinfo, collation, false));
-
-	for (int lc = entryPoint->level; lc >= 1; lc--)
-	{
-		w = HnswSearchLayer(base, q, ep, 1, lc, index, procinfo, collation, m, false, NULL);
-		ep = w;
-	}
-
-	return HnswSearchLayer(base, q, ep, hnsw_ef_search, 0, index, procinfo, collation, m, false, NULL);
-}
-
-/*
- * Get dimensions from metapage
- */
-static int
-GetDimensions(Relation index)
-{
-	Buffer		buf;
-	Page		page;
-	HnswMetaPage metap;
-	int			dimensions;
-
-	buf = ReadBuffer(index, HNSW_METAPAGE_BLKNO);
-	LockBuffer(buf, BUFFER_LOCK_SHARE);
-	page = BufferGetPage(buf);
-	metap = HnswPageGetMeta(page);
-
-	dimensions = metap->dimensions;
-
-	UnlockReleaseBuffer(buf);
-
-	return dimensions;
-}
-
-/*
- * Get scan value
- */
-static Datum
-GetScanValue(IndexScanDesc scan)
-{
-	HnswScanOpaque so = (HnswScanOpaque) scan->opaque;
-	Datum		value;
-
-	if (scan->orderByData->sk_flags & SK_ISNULL)
-		value = PointerGetDatum(InitVector(GetDimensions(scan->indexRelation)));
-	else
-	{
-		value = scan->orderByData->sk_argument;
-
-		/* Value should not be compressed or toasted */
-		Assert(!VARATT_IS_COMPRESSED(DatumGetPointer(value)));
-		Assert(!VARATT_IS_EXTENDED(DatumGetPointer(value)));
-
-		/* Fine if normalization fails */
-		if (so->normprocinfo != NULL)
-			HnswNormValue(so->normprocinfo, so->collation, &value, NULL);
-	}
-
-	return value;
-}
-
-/*
- * Prepare for an index scan
- */
-IndexScanDesc
-hnswbeginscan(Relation index, int nkeys, int norderbys)
-{
-	IndexScanDesc scan;
-	HnswScanOpaque so;
-
-	scan = RelationGetIndexScan(index, nkeys, norderbys);
-
-	so = (HnswScanOpaque) palloc(sizeof(HnswScanOpaqueData));
-	so->first = true;
-	so->tmpCtx = AllocSetContextCreate(CurrentMemoryContext,
-									   "Hnsw scan temporary context",
-									   ALLOCSET_DEFAULT_SIZES);
-
-	/* Set support functions */
-	so->procinfo = index_getprocinfo(index, 1, HNSW_DISTANCE_PROC);
-	so->normprocinfo = HnswOptionalProcInfo(index, HNSW_NORM_PROC);
-	so->collation = index->rd_indcollation[0];
-
-	scan->opaque = so;
-
-	return scan;
-}
-
-/*
- * Start or restart an index scan
- */
-void
-hnswrescan(IndexScanDesc scan, ScanKey keys, int nkeys, ScanKey orderbys, int norderbys)
-{
-	HnswScanOpaque so = (HnswScanOpaque) scan->opaque;
-
-	so->first = true;
-	MemoryContextReset(so->tmpCtx);
-
-	if (keys && scan->numberOfKeys > 0)
-		memmove(scan->keyData, keys, scan->numberOfKeys * sizeof(ScanKeyData));
-
-	if (orderbys && scan->numberOfOrderBys > 0)
-		memmove(scan->orderByData, orderbys, scan->numberOfOrderBys * sizeof(ScanKeyData));
-}
-
-/*
- * Fetch the next tuple in the given scan
- */
-bool
-hnswgettuple(IndexScanDesc scan, ScanDirection dir)
-{
-	HnswScanOpaque so = (HnswScanOpaque) scan->opaque;
-	MemoryContext oldCtx = MemoryContextSwitchTo(so->tmpCtx);
-
-	/*
-	 * Index can be used to scan backward, but Postgres doesn't support
-	 * backward scan on operators
-	 */
-	Assert(ScanDirectionIsForward(dir));
-
-	if (so->first)
-	{
-		Datum		value;
-
-		/* Count index scan for stats */
-		pgstat_count_index_scan(scan->indexRelation);
-
-		/* Safety check */
-		if (scan->orderByData == NULL)
-			elog(ERROR, "cannot scan hnsw index without order");
-
-		/* Requires MVCC-compliant snapshot as not able to maintain a pin */
-		/* https://www.postgresql.org/docs/current/index-locking.html */
-		if (!IsMVCCSnapshot(scan->xs_snapshot))
-			elog(ERROR, "non-MVCC snapshots are not supported with hnsw");
-
-		/* Get scan value */
-		value = GetScanValue(scan);
-
-		/*
-		 * Get a shared lock. This allows vacuum to ensure no in-flight scans
-		 * before marking tuples as deleted.
-		 */
-		LockPage(scan->indexRelation, HNSW_SCAN_LOCK, ShareLock);
-
-		so->w = GetScanItems(scan, value);
-
-		/* Release shared lock */
-		UnlockPage(scan->indexRelation, HNSW_SCAN_LOCK, ShareLock);
-
-		so->first = false;
-	}
-
-	while (list_length(so->w) > 0)
-	{
-		char	   *base = NULL;
-		HnswCandidate *hc = llast(so->w);
-		HnswElement element = HnswPtrAccess(base, hc->element);
-		ItemPointer heaptid;
-
-		/* Move to next element if no valid heap TIDs */
-		if (element->heaptidsLength == 0)
-		{
-			so->w = list_delete_last(so->w);
-			continue;
-		}
-
-		heaptid = &element->heaptids[--element->heaptidsLength];
-
-		MemoryContextSwitchTo(oldCtx);
-
-		scan->xs_heaptid = *heaptid;
-		scan->xs_recheck = false;
-		scan->xs_recheckorderby = false;
-		return true;
-	}
-
-	MemoryContextSwitchTo(oldCtx);
-	return false;
-}
-
-/*
- * End a scan and release resources
- */
-void
-hnswendscan(IndexScanDesc scan)
-{
-	HnswScanOpaque so = (HnswScanOpaque) scan->opaque;
-
-	MemoryContextDelete(so->tmpCtx);
-
-	pfree(so);
-	scan->opaque = NULL;
-}

src/hnswvacuum.c

@@ -1,646 +0,0 @@
-#include "postgres.h"
-
-#include <math.h>
-
-#include "access/generic_xlog.h"
-#include "commands/vacuum.h"
-#include "hnsw.h"
-#include "storage/bufmgr.h"
-#include "storage/lmgr.h"
-#include "utils/memutils.h"
-
-/*
- * Check if deleted list contains an index TID
- */
-static bool
-DeletedContains(tidhash_hash * deleted, ItemPointer indextid)
-{
-	return tidhash_lookup(deleted, *indextid) != NULL;
-}
-
-/*
- * Remove deleted heap TIDs
- *
- * OK to remove for entry point, since always considered for searches and inserts
- */
-static void
-RemoveHeapTids(HnswVacuumState * vacuumstate)
-{
-	BlockNumber blkno = HNSW_HEAD_BLKNO;
-	HnswElement highestPoint = &vacuumstate->highestPoint;
-	Relation	index = vacuumstate->index;
-	BufferAccessStrategy bas = vacuumstate->bas;
-	HnswElement entryPoint = HnswGetEntryPoint(vacuumstate->index);
-	IndexBulkDeleteResult *stats = vacuumstate->stats;
-
-	/* Store separately since highestPoint.level is uint8 */
-	int			highestLevel = -1;
-
-	/* Initialize highest point */
-	highestPoint->blkno = InvalidBlockNumber;
-	highestPoint->offno = InvalidOffsetNumber;
-
-	while (BlockNumberIsValid(blkno))
-	{
-		Buffer		buf;
-		Page		page;
-		GenericXLogState *state;
-		OffsetNumber offno;
-		OffsetNumber maxoffno;
-		bool		updated = false;
-
-		vacuum_delay_point();
-
-		buf = ReadBufferExtended(index, MAIN_FORKNUM, blkno, RBM_NORMAL, bas);
-		LockBuffer(buf, BUFFER_LOCK_EXCLUSIVE);
-		state = GenericXLogStart(index);
-		page = GenericXLogRegisterBuffer(state, buf, 0);
-		maxoffno = PageGetMaxOffsetNumber(page);
-
-		/* Iterate over nodes */
-		for (offno = FirstOffsetNumber; offno <= maxoffno; offno = OffsetNumberNext(offno))
-		{
-			HnswElementTuple etup = (HnswElementTuple) PageGetItem(page, PageGetItemId(page, offno));
-			int			idx = 0;
-			bool		itemUpdated = false;
-
-			/* Skip neighbor tuples */
-			if (!HnswIsElementTuple(etup))
-				continue;
-
-			if (ItemPointerIsValid(&etup->heaptids[0]))
-			{
-				for (int i = 0; i < HNSW_HEAPTIDS; i++)
-				{
-					/* Stop at first unused */
-					if (!ItemPointerIsValid(&etup->heaptids[i]))
-						break;
-
-					if (vacuumstate->callback(&etup->heaptids[i], vacuumstate->callback_state))
-					{
-						itemUpdated = true;
-						stats->tuples_removed++;
-					}
-					else
-					{
-						/* Move to front of list */
-						etup->heaptids[idx++] = etup->heaptids[i];
-						stats->num_index_tuples++;
-					}
-				}
-
-				if (itemUpdated)
-				{
-					/* Mark rest as invalid */
-					for (int i = idx; i < HNSW_HEAPTIDS; i++)
-						ItemPointerSetInvalid(&etup->heaptids[i]);
-
-					updated = true;
-				}
-			}
-
-			if (!ItemPointerIsValid(&etup->heaptids[0]))
-			{
-				ItemPointerData ip;
-				bool		found;
-
-				/* Add to deleted list */
-				ItemPointerSet(&ip, blkno, offno);
-
-				tidhash_insert(vacuumstate->deleted, ip, &found);
-				Assert(!found);
-			}
-			else if (etup->level > highestLevel && !(entryPoint != NULL && blkno == entryPoint->blkno && offno == entryPoint->offno))
-			{
-				/* Keep track of highest non-entry point */
-				highestPoint->blkno = blkno;
-				highestPoint->offno = offno;
-				highestPoint->level = etup->level;
-				highestLevel = etup->level;
-			}
-		}
-
-		blkno = HnswPageGetOpaque(page)->nextblkno;
-
-		if (updated)
-			GenericXLogFinish(state);
-		else
-			GenericXLogAbort(state);
-
-		UnlockReleaseBuffer(buf);
-	}
-}
-
-/*
- * Check for deleted neighbors
- */
-static bool
-NeedsUpdated(HnswVacuumState * vacuumstate, HnswElement element)
-{
-	Relation	index = vacuumstate->index;
-	BufferAccessStrategy bas = vacuumstate->bas;
-	Buffer		buf;
-	Page		page;
-	HnswNeighborTuple ntup;
-	bool		needsUpdated = false;
-
-	buf = ReadBufferExtended(index, MAIN_FORKNUM, element->neighborPage, RBM_NORMAL, bas);
-	LockBuffer(buf, BUFFER_LOCK_SHARE);
-	page = BufferGetPage(buf);
-	ntup = (HnswNeighborTuple) PageGetItem(page, PageGetItemId(page, element->neighborOffno));
-
-	Assert(HnswIsNeighborTuple(ntup));
-
-	/* 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))
-		{
-			needsUpdated = true;
-			break;
-		}
-	}
-
-	/* Also update if layer 0 is not full */
-	/* This could indicate too many candidates being deleted during insert */
-	if (!needsUpdated)
-		needsUpdated = !ItemPointerIsValid(&ntup->indextids[ntup->count - 1]);
-
-	UnlockReleaseBuffer(buf);
-
-	return needsUpdated;
-}
-
-/*
- * Repair graph for a single element
- */
-static void
-RepairGraphElement(HnswVacuumState * vacuumstate, HnswElement element, HnswElement entryPoint)
-{
-	Relation	index = vacuumstate->index;
-	Buffer		buf;
-	Page		page;
-	GenericXLogState *state;
-	int			m = vacuumstate->m;
-	int			efConstruction = vacuumstate->efConstruction;
-	FmgrInfo   *procinfo = vacuumstate->procinfo;
-	Oid			collation = vacuumstate->collation;
-	BufferAccessStrategy bas = vacuumstate->bas;
-	HnswNeighborTuple ntup = vacuumstate->ntup;
-	Size		ntupSize = HNSW_NEIGHBOR_TUPLE_SIZE(element->level, m);
-	char	   *base = NULL;
-
-	/* Skip if element is entry point */
-	if (entryPoint != NULL && element->blkno == entryPoint->blkno && element->offno == entryPoint->offno)
-		return;
-
-	/* Init fields */
-	HnswInitNeighbors(base, element, m, NULL);
-	element->heaptidsLength = 0;
-
-	/* Find neighbors for element, skipping itself */
-	HnswFindElementNeighbors(base, element, entryPoint, index, procinfo, collation, m, efConstruction, true);
-
-	/* Zero memory for each element */
-	MemSet(ntup, 0, HNSW_TUPLE_ALLOC_SIZE);
-
-	/* Update neighbor tuple */
-	/* Do this before getting page to minimize locking */
-	HnswSetNeighborTuple(base, ntup, element, m);
-
-	/* Get neighbor page */
-	buf = ReadBufferExtended(index, MAIN_FORKNUM, element->neighborPage, RBM_NORMAL, bas);
-	LockBuffer(buf, BUFFER_LOCK_EXCLUSIVE);
-	state = GenericXLogStart(index);
-	page = GenericXLogRegisterBuffer(state, buf, 0);
-
-	/* Overwrite tuple */
-	if (!PageIndexTupleOverwrite(page, element->neighborOffno, (Item) ntup, ntupSize))
-		elog(ERROR, "failed to add index item to \"%s\"", RelationGetRelationName(index));
-
-	/* Commit */
-	GenericXLogFinish(state);
-	UnlockReleaseBuffer(buf);
-
-	/* Update neighbors */
-	HnswUpdateNeighborsOnDisk(index, procinfo, collation, element, m, true, false);
-}
-
-/*
- * Repair graph entry point
- */
-static void
-RepairGraphEntryPoint(HnswVacuumState * vacuumstate)
-{
-	Relation	index = vacuumstate->index;
-	HnswElement highestPoint = &vacuumstate->highestPoint;
-	HnswElement entryPoint;
-	MemoryContext oldCtx = MemoryContextSwitchTo(vacuumstate->tmpCtx);
-
-	if (!BlockNumberIsValid(highestPoint->blkno))
-		highestPoint = NULL;
-
-	/*
-	 * Repair graph for highest non-entry point. Highest point may be outdated
-	 * due to inserts that happen during and after RemoveHeapTids.
-	 */
-	if (highestPoint != NULL)
-	{
-		/* Get a shared lock */
-		LockPage(index, HNSW_UPDATE_LOCK, ShareLock);
-
-		/* Load element */
-		HnswLoadElement(highestPoint, NULL, NULL, index, vacuumstate->procinfo, vacuumstate->collation, true);
-
-		/* Repair if needed */
-		if (NeedsUpdated(vacuumstate, highestPoint))
-			RepairGraphElement(vacuumstate, highestPoint, HnswGetEntryPoint(index));
-
-		/* Release lock */
-		UnlockPage(index, HNSW_UPDATE_LOCK, ShareLock);
-	}
-
-	/* Prevent concurrent inserts when possibly updating entry point */
-	LockPage(index, HNSW_UPDATE_LOCK, ExclusiveLock);
-
-	/* Get latest entry point */
-	entryPoint = HnswGetEntryPoint(index);
-
-	if (entryPoint != NULL)
-	{
-		ItemPointerData epData;
-
-		ItemPointerSet(&epData, entryPoint->blkno, entryPoint->offno);
-
-		if (DeletedContains(vacuumstate->deleted, &epData))
-		{
-			/*
-			 * Replace the entry point with the highest point. If highest
-			 * point is outdated and empty, the entry point will be empty
-			 * until an element is repaired.
-			 */
-			HnswUpdateMetaPage(index, HNSW_UPDATE_ENTRY_ALWAYS, highestPoint, InvalidBlockNumber, MAIN_FORKNUM, false);
-		}
-		else
-		{
-			/*
-			 * Repair the entry point with the highest point. If highest point
-			 * is outdated, this can remove connections at higher levels in
-			 * the graph until they are repaired, but this should be fine.
-			 */
-			HnswLoadElement(entryPoint, NULL, NULL, index, vacuumstate->procinfo, vacuumstate->collation, true);
-
-			if (NeedsUpdated(vacuumstate, entryPoint))
-			{
-				/* Reset neighbors from previous update */
-				if (highestPoint != NULL)
-					HnswPtrStore((char *) NULL, highestPoint->neighbors, (HnswNeighborArrayPtr *) NULL);
-
-				RepairGraphElement(vacuumstate, entryPoint, highestPoint);
-			}
-		}
-	}
-
-	/* Release lock */
-	UnlockPage(index, HNSW_UPDATE_LOCK, ExclusiveLock);
-
-	/* Reset memory context */
-	MemoryContextSwitchTo(oldCtx);
-	MemoryContextReset(vacuumstate->tmpCtx);
-}
-
-/*
- * Repair graph for all elements
- */
-static void
-RepairGraph(HnswVacuumState * vacuumstate)
-{
-	Relation	index = vacuumstate->index;
-	BufferAccessStrategy bas = vacuumstate->bas;
-	BlockNumber blkno = HNSW_HEAD_BLKNO;
-
-	/*
-	 * Wait for inserts to complete. Inserts before this point may have
-	 * neighbors about to be deleted. Inserts after this point will not.
-	 */
-	LockPage(index, HNSW_UPDATE_LOCK, ExclusiveLock);
-	UnlockPage(index, HNSW_UPDATE_LOCK, ExclusiveLock);
-
-	/* Repair entry point first */
-	RepairGraphEntryPoint(vacuumstate);
-
-	while (BlockNumberIsValid(blkno))
-	{
-		Buffer		buf;
-		Page		page;
-		OffsetNumber offno;
-		OffsetNumber maxoffno;
-		List	   *elements = NIL;
-		ListCell   *lc2;
-		MemoryContext oldCtx;
-
-		vacuum_delay_point();
-
-		oldCtx = MemoryContextSwitchTo(vacuumstate->tmpCtx);
-
-		buf = ReadBufferExtended(index, MAIN_FORKNUM, blkno, RBM_NORMAL, bas);
-		LockBuffer(buf, BUFFER_LOCK_SHARE);
-		page = BufferGetPage(buf);
-		maxoffno = PageGetMaxOffsetNumber(page);
-
-		/* Load items into memory to minimize locking */
-		for (offno = FirstOffsetNumber; offno <= maxoffno; offno = OffsetNumberNext(offno))
-		{
-			HnswElementTuple etup = (HnswElementTuple) PageGetItem(page, PageGetItemId(page, offno));
-			HnswElement element;
-
-			/* Skip neighbor tuples */
-			if (!HnswIsElementTuple(etup))
-				continue;
-
-			/* Skip updating neighbors if being deleted */
-			if (!ItemPointerIsValid(&etup->heaptids[0]))
-				continue;
-
-			/* Create an element */
-			element = HnswInitElementFromBlock(blkno, offno);
-			HnswLoadElementFromTuple(element, etup, false, true);
-
-			elements = lappend(elements, element);
-		}
-
-		blkno = HnswPageGetOpaque(page)->nextblkno;
-
-		UnlockReleaseBuffer(buf);
-
-		/* Update neighbor pages */
-		foreach(lc2, elements)
-		{
-			HnswElement element = (HnswElement) lfirst(lc2);
-			HnswElement entryPoint;
-			LOCKMODE	lockmode = ShareLock;
-
-			/* Check if any neighbors point to deleted values */
-			if (!NeedsUpdated(vacuumstate, element))
-				continue;
-
-			/* Get a shared lock */
-			LockPage(index, HNSW_UPDATE_LOCK, lockmode);
-
-			/* Refresh entry point for each element */
-			entryPoint = HnswGetEntryPoint(index);
-
-			/* Prevent concurrent inserts when likely updating entry point */
-			if (entryPoint == NULL || element->level > entryPoint->level)
-			{
-				/* Release shared lock */
-				UnlockPage(index, HNSW_UPDATE_LOCK, lockmode);
-
-				/* Get exclusive lock */
-				lockmode = ExclusiveLock;
-				LockPage(index, HNSW_UPDATE_LOCK, lockmode);
-
-				/* Get latest entry point after lock is acquired */
-				entryPoint = HnswGetEntryPoint(index);
-			}
-
-			/* Repair connections */
-			RepairGraphElement(vacuumstate, element, entryPoint);
-
-			/*
-			 * Update metapage if needed. Should only happen if entry point
-			 * was replaced and highest point was outdated.
-			 */
-			if (entryPoint == NULL || element->level > entryPoint->level)
-				HnswUpdateMetaPage(index, HNSW_UPDATE_ENTRY_GREATER, element, InvalidBlockNumber, MAIN_FORKNUM, false);
-
-			/* Release lock */
-			UnlockPage(index, HNSW_UPDATE_LOCK, lockmode);
-		}
-
-		/* Reset memory context */
-		MemoryContextSwitchTo(oldCtx);
-		MemoryContextReset(vacuumstate->tmpCtx);
-	}
-}
-
-/*
- * Mark items as deleted
- */
-static void
-MarkDeleted(HnswVacuumState * vacuumstate)
-{
-	BlockNumber blkno = HNSW_HEAD_BLKNO;
-	BlockNumber insertPage = InvalidBlockNumber;
-	Relation	index = vacuumstate->index;
-	BufferAccessStrategy bas = vacuumstate->bas;
-
-	/*
-	 * Wait for index scans to complete. Scans before this point may contain
-	 * tuples about to be deleted. Scans after this point will not, since the
-	 * graph has been repaired.
-	 */
-	LockPage(index, HNSW_SCAN_LOCK, ExclusiveLock);
-	UnlockPage(index, HNSW_SCAN_LOCK, ExclusiveLock);
-
-	while (BlockNumberIsValid(blkno))
-	{
-		Buffer		buf;
-		Page		page;
-		GenericXLogState *state;
-		OffsetNumber offno;
-		OffsetNumber maxoffno;
-
-		vacuum_delay_point();
-
-		buf = ReadBufferExtended(index, MAIN_FORKNUM, blkno, RBM_NORMAL, bas);
-
-		/*
-		 * ambulkdelete cannot delete entries from pages that are pinned by
-		 * other backends
-		 *
-		 * https://www.postgresql.org/docs/current/index-locking.html
-		 */
-		LockBufferForCleanup(buf);
-
-		state = GenericXLogStart(index);
-		page = GenericXLogRegisterBuffer(state, buf, 0);
-		maxoffno = PageGetMaxOffsetNumber(page);
-
-		/* Update element and neighbors together */
-		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 deleted tuples */
-			if (etup->deleted)
-			{
-				/* Set to first free page */
-				if (!BlockNumberIsValid(insertPage))
-					insertPage = blkno;
-
-				continue;
-			}
-
-			/* Skip live tuples */
-			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_EXCLUSIVE);
-				npage = GenericXLogRegisterBuffer(state, nbuf, 0);
-			}
-
-			ntup = (HnswNeighborTuple) PageGetItem(npage, PageGetItemId(npage, neighborOffno));
-
-			/* Overwrite element */
-			etup->deleted = 1;
-			MemSet(&etup->data, 0, VARSIZE_ANY(&etup->data));
-
-			/* Overwrite neighbors */
-			for (int i = 0; i < ntup->count; i++)
-				ItemPointerSetInvalid(&ntup->indextids[i]);
-
-			/*
-			 * We modified the tuples in place, no need to call
-			 * PageIndexTupleOverwrite
-			 */
-
-			/* Commit */
-			GenericXLogFinish(state);
-			if (nbuf != buf)
-				UnlockReleaseBuffer(nbuf);
-
-			/* Set to first free page */
-			if (!BlockNumberIsValid(insertPage))
-				insertPage = blkno;
-
-			/* Prepare new xlog */
-			state = GenericXLogStart(index);
-			page = GenericXLogRegisterBuffer(state, buf, 0);
-		}
-
-		blkno = HnswPageGetOpaque(page)->nextblkno;
-
-		GenericXLogAbort(state);
-		UnlockReleaseBuffer(buf);
-	}
-
-	/* Update insert page last, after everything has been marked as deleted */
-	HnswUpdateMetaPage(index, 0, NULL, insertPage, MAIN_FORKNUM, false);
-}
-
-/*
- * Initialize the vacuum state
- */
-static void
-InitVacuumState(HnswVacuumState * vacuumstate, IndexVacuumInfo *info, IndexBulkDeleteResult *stats, IndexBulkDeleteCallback callback, void *callback_state)
-{
-	Relation	index = info->index;
-
-	if (stats == NULL)
-		stats = (IndexBulkDeleteResult *) palloc0(sizeof(IndexBulkDeleteResult));
-
-	vacuumstate->index = index;
-	vacuumstate->stats = stats;
-	vacuumstate->callback = callback;
-	vacuumstate->callback_state = callback_state;
-	vacuumstate->efConstruction = HnswGetEfConstruction(index);
-	vacuumstate->bas = GetAccessStrategy(BAS_BULKREAD);
-	vacuumstate->procinfo = index_getprocinfo(index, 1, HNSW_DISTANCE_PROC);
-	vacuumstate->collation = index->rd_indcollation[0];
-	vacuumstate->ntup = palloc0(HNSW_TUPLE_ALLOC_SIZE);
-	vacuumstate->tmpCtx = AllocSetContextCreate(CurrentMemoryContext,
-												"Hnsw vacuum temporary context",
-												ALLOCSET_DEFAULT_SIZES);
-
-	/* Get m from metapage */
-	HnswGetMetaPageInfo(index, &vacuumstate->m, NULL);
-
-	/* Create hash table */
-	vacuumstate->deleted = tidhash_create(CurrentMemoryContext, 256, NULL);
-}
-
-/*
- * Free resources
- */
-static void
-FreeVacuumState(HnswVacuumState * vacuumstate)
-{
-	tidhash_destroy(vacuumstate->deleted);
-	FreeAccessStrategy(vacuumstate->bas);
-	pfree(vacuumstate->ntup);
-	MemoryContextDelete(vacuumstate->tmpCtx);
-}
-
-/*
- * Bulk delete tuples from the index
- */
-IndexBulkDeleteResult *
-hnswbulkdelete(IndexVacuumInfo *info, IndexBulkDeleteResult *stats,
-			   IndexBulkDeleteCallback callback, void *callback_state)
-{
-	HnswVacuumState vacuumstate;
-
-	InitVacuumState(&vacuumstate, info, stats, callback, callback_state);
-
-	/* Pass 1: Remove heap TIDs */
-	RemoveHeapTids(&vacuumstate);
-
-	/* Pass 2: Repair graph */
-	RepairGraph(&vacuumstate);
-
-	/* Pass 3: Mark as deleted */
-	MarkDeleted(&vacuumstate);
-
-	FreeVacuumState(&vacuumstate);
-
-	return vacuumstate.stats;
-}
-
-/*
- * Clean up after a VACUUM operation
- */
-IndexBulkDeleteResult *
-hnswvacuumcleanup(IndexVacuumInfo *info, IndexBulkDeleteResult *stats)
-{
-	Relation	rel = info->index;
-
-	if (info->analyze_only)
-		return stats;
-
-	/* stats is NULL if ambulkdelete not called */
-	/* OK to return NULL if index not changed */
-	if (stats == NULL)
-		return NULL;
-
-	stats->num_pages = RelationGetNumberOfBlocks(rel);
-
-	return stats;
-}

src/ivfbuild.c

@@ -2,43 +2,42 @@
 
 #include <float.h>
 
-#include "access/table.h"
-#include "access/tableam.h"
-#include "access/parallel.h"
-#include "access/xact.h"
 #include "catalog/index.h"
-#include "catalog/pg_operator_d.h"
-#include "catalog/pg_type_d.h"
-#include "commands/progress.h"
 #include "ivfflat.h"
 #include "miscadmin.h"
-#include "optimizer/optimizer.h"
 #include "storage/bufmgr.h"
-#include "tcop/tcopprot.h"
 #include "utils/memutils.h"
 
 #if PG_VERSION_NUM >= 140000
 #include "utils/backend_progress.h"
-#else
+#elif PG_VERSION_NUM >= 120000
 #include "pgstat.h"
 #endif
 
+#if PG_VERSION_NUM >= 120000
+#include "access/tableam.h"
+#include "commands/progress.h"
+#else
+#define PROGRESS_CREATEIDX_SUBPHASE 0
+#define PROGRESS_CREATEIDX_TUPLES_TOTAL 0
+#define PROGRESS_CREATEIDX_TUPLES_DONE 0
+#endif
+
+#include "catalog/pg_operator_d.h"
+#include "catalog/pg_type_d.h"
+
 #if PG_VERSION_NUM >= 130000
 #define CALLBACK_ITEM_POINTER ItemPointer tid
 #else
 #define CALLBACK_ITEM_POINTER HeapTuple hup
 #endif
 
-#if PG_VERSION_NUM >= 140000
+#if PG_VERSION_NUM >= 120000
-#include "utils/backend_status.h"
+#define UpdateProgress(index, val) pgstat_progress_update_param(index, val)
-#include "utils/wait_event.h"
+#else
+#define UpdateProgress(index, val) ((void)val)
 #endif
 
-#define PARALLEL_KEY_IVFFLAT_SHARED		UINT64CONST(0xA000000000000001)
-#define PARALLEL_KEY_TUPLESORT			UINT64CONST(0xA000000000000002)
-#define PARALLEL_KEY_IVFFLAT_CENTERS	UINT64CONST(0xA000000000000003)
-#define PARALLEL_KEY_QUERY_TEXT			UINT64CONST(0xA000000000000004)
-
 /*
  * Add sample
  */
@@ -130,8 +129,13 @@ SampleRows(IvfflatBuildState * buildstate)
 	{
 		BlockNumber targblock = BlockSampler_Next(&buildstate->bs);
 
+#if PG_VERSION_NUM >= 120000
 		table_index_build_range_scan(buildstate->heap, buildstate->index, buildstate->indexInfo,
 									 false, true, false, targblock, 1, SampleCallback, (void *) buildstate, NULL);
+#else
+		IndexBuildHeapRangeScan(buildstate->heap, buildstate->index, buildstate->indexInfo,
+								false, true, targblock, 1, SampleCallback, (void *) buildstate, NULL);
+#endif
 	}
 }
 
@@ -143,9 +147,10 @@ AddTupleToSort(Relation index, ItemPointer tid, Datum *values, IvfflatBuildState
 {
 	double		distance;
 	double		minDistance = DBL_MAX;
-	int			closestCenter = 0;
+	int			closestCenter = -1;
 	VectorArray centers = buildstate->centers;
 	TupleTableSlot *slot = buildstate->slot;
+	int			i;
 
 	/* Detoast once for all calls */
 	Datum		value = PointerGetDatum(PG_DETOAST_DATUM(values[0]));
@@ -158,7 +163,7 @@ AddTupleToSort(Relation index, ItemPointer tid, Datum *values, IvfflatBuildState
 	}
 
 	/* Find the list that minimizes the distance */
-	for (int i = 0; i < centers->length; i++)
+	for (i = 0; i < centers->length; i++)
 	{
 		distance = DatumGetFloat8(FunctionCall2Coll(buildstate->procinfo, buildstate->collation, value, PointerGetDatum(VectorArrayGet(centers, i))));
 
@@ -253,27 +258,32 @@ GetNextTuple(Tuplesortstate *sortstate, TupleDesc tupdesc, TupleTableSlot *slot,
 static void
 InsertTuples(Relation index, IvfflatBuildState * buildstate, ForkNumber forkNum)
 {
+	Buffer		buf;
+	Page		page;
+	GenericXLogState *state;
 	int			list;
 	IndexTuple	itup = NULL;	/* silence compiler warning */
+	BlockNumber startPage;
+	BlockNumber insertPage;
+	Size		itemsz;
+	int			i;
 	int64		inserted = 0;
 
+#if PG_VERSION_NUM >= 120000
 	TupleTableSlot *slot = MakeSingleTupleTableSlot(buildstate->tupdesc, &TTSOpsMinimalTuple);
+#else
+	TupleTableSlot *slot = MakeSingleTupleTableSlot(buildstate->tupdesc);
+#endif
 	TupleDesc	tupdesc = RelationGetDescr(index);
 
-	pgstat_progress_update_param(PROGRESS_CREATEIDX_SUBPHASE, PROGRESS_IVFFLAT_PHASE_LOAD);
+	UpdateProgress(PROGRESS_CREATEIDX_SUBPHASE, PROGRESS_IVFFLAT_PHASE_LOAD);
 
-	pgstat_progress_update_param(PROGRESS_CREATEIDX_TUPLES_TOTAL, buildstate->indtuples);
+	UpdateProgress(PROGRESS_CREATEIDX_TUPLES_TOTAL, buildstate->indtuples);
 
 	GetNextTuple(buildstate->sortstate, tupdesc, slot, &itup, &list);
 
-	for (int i = 0; i < buildstate->centers->length; i++)
+	for (i = 0; i < buildstate->centers->length; i++)
 	{
-		Buffer		buf;
-		Page		page;
-		GenericXLogState *state;
-		BlockNumber startPage;
-		BlockNumber insertPage;
-
 		/* Can take a while, so ensure we can interrupt */
 		/* Needs to be called when no buffer locks are held */
 		CHECK_FOR_INTERRUPTS();
@@ -287,8 +297,7 @@ InsertTuples(Relation index, IvfflatBuildState * buildstate, ForkNumber forkNum)
 		while (list == i)
 		{
 			/* Check for free space */
-			Size		itemsz = MAXALIGN(IndexTupleSize(itup));
+			itemsz = MAXALIGN(IndexTupleSize(itup));
-
 			if (PageGetFreeSpace(page) < itemsz)
 				IvfflatAppendPage(index, &buf, &page, &state, forkNum);
 
@@ -298,7 +307,7 @@ InsertTuples(Relation index, IvfflatBuildState * buildstate, ForkNumber forkNum)
 
 			pfree(itup);
 
-			pgstat_progress_update_param(PROGRESS_CREATEIDX_TUPLES_DONE, ++inserted);
+			UpdateProgress(PROGRESS_CREATEIDX_TUPLES_DONE, ++inserted);
 
 			GetNextTuple(buildstate->sortstate, tupdesc, slot, &itup, &list);
 		}
@@ -308,7 +317,7 @@ InsertTuples(Relation index, IvfflatBuildState * buildstate, ForkNumber forkNum)
 		IvfflatCommitBuffer(buf, state);
 
 		/* Set the start and insert pages */
-		IvfflatUpdateList(index, buildstate->listInfo[i], insertPage, InvalidBlockNumber, startPage, forkNum);
+		IvfflatUpdateList(index, state, buildstate->listInfo[i], insertPage, InvalidBlockNumber, startPage, forkNum);
 	}
 }
 
@@ -342,16 +351,26 @@ InitBuildState(IvfflatBuildState * buildstate, Relation heap, Relation index, In
 	buildstate->collation = index->rd_indcollation[0];
 
 	/* Require more than one dimension for spherical k-means */
-	if (buildstate->kmeansnormprocinfo != NULL && buildstate->dimensions == 1)
+	/* Lists check for backwards compatibility */
+	/* TODO Remove lists check in 0.3.0 */
+	if (buildstate->kmeansnormprocinfo != NULL && buildstate->dimensions == 1 && buildstate->lists > 1)
 		elog(ERROR, "dimensions must be greater than one for this opclass");
 
 	/* Create tuple description for sorting */
+#if PG_VERSION_NUM >= 120000
 	buildstate->tupdesc = CreateTemplateTupleDesc(3);
+#else
+	buildstate->tupdesc = CreateTemplateTupleDesc(3, false);
+#endif
 	TupleDescInitEntry(buildstate->tupdesc, (AttrNumber) 1, "list", INT4OID, -1, 0);
 	TupleDescInitEntry(buildstate->tupdesc, (AttrNumber) 2, "tid", TIDOID, -1, 0);
 	TupleDescInitEntry(buildstate->tupdesc, (AttrNumber) 3, "vector", RelationGetDescr(index)->attrs[0].atttypid, -1, 0);
 
+#if PG_VERSION_NUM >= 120000
 	buildstate->slot = MakeSingleTupleTableSlot(buildstate->tupdesc, &TTSOpsVirtual);
+#else
+	buildstate->slot = MakeSingleTupleTableSlot(buildstate->tupdesc);
+#endif
 
 	buildstate->centers = VectorArrayInit(buildstate->lists, buildstate->dimensions);
 	buildstate->listInfo = palloc(sizeof(ListInfo) * buildstate->lists);
@@ -368,8 +387,6 @@ InitBuildState(IvfflatBuildState * buildstate, Relation heap, Relation index, In
 	buildstate->listSums = palloc0(sizeof(double) * buildstate->lists);
 	buildstate->listCounts = palloc0(sizeof(int) * buildstate->lists);
 #endif
-
-	buildstate->ivfleader = NULL;
 }
 
 /*
@@ -398,7 +415,7 @@ ComputeCenters(IvfflatBuildState * buildstate)
 {
 	int			numSamples;
 
-	pgstat_progress_update_param(PROGRESS_CREATEIDX_SUBPHASE, PROGRESS_IVFFLAT_PHASE_KMEANS);
+	UpdateProgress(PROGRESS_CREATEIDX_SUBPHASE, PROGRESS_IVFFLAT_PHASE_KMEANS);
 
 	/* Target 50 samples per list, with at least 10000 samples */
 	/* The number of samples has a large effect on index build time */
@@ -414,18 +431,8 @@ ComputeCenters(IvfflatBuildState * buildstate)
 	/* TODO Ensure within maintenance_work_mem */
 	buildstate->samples = VectorArrayInit(numSamples, buildstate->dimensions);
 	if (buildstate->heap != NULL)
-	{
 		SampleRows(buildstate);
 
-		if (buildstate->samples->length < buildstate->lists)
-		{
-			ereport(NOTICE,
-					(errmsg("ivfflat index created with little data"),
-					 errdetail("This will cause low recall."),
-					 errhint("Drop the index until the table has more data.")));
-		}
-	}
-
 	/* Calculate centers */
 	IvfflatBench("k-means", IvfflatKmeans(buildstate->index, buildstate->samples, buildstate->centers));
 
@@ -466,33 +473,33 @@ static void
 CreateListPages(Relation index, VectorArray centers, int dimensions,
 				int lists, ForkNumber forkNum, ListInfo * *listInfo)
 {
+	int			i;
 	Buffer		buf;
 	Page		page;
 	GenericXLogState *state;
-	Size		listSize;
+	OffsetNumber offno;
+	Size		itemsz;
 	IvfflatList list;
 
-	listSize = MAXALIGN(IVFFLAT_LIST_SIZE(dimensions));
+	itemsz = MAXALIGN(IVFFLAT_LIST_SIZE(dimensions));
-	list = palloc0(listSize);
+	list = palloc(itemsz);
 
 	buf = IvfflatNewBuffer(index, forkNum);
 	IvfflatInitRegisterPage(index, &buf, &page, &state);
 
-	for (int i = 0; i < lists; i++)
+	for (i = 0; i < lists; i++)
 	{
-		OffsetNumber offno;
-
 		/* Load list */
 		list->startPage = InvalidBlockNumber;
 		list->insertPage = InvalidBlockNumber;
 		memcpy(&list->center, VectorArrayGet(centers, i), VECTOR_SIZE(dimensions));
 
 		/* Ensure free space */
-		if (PageGetFreeSpace(page) < listSize)
+		if (PageGetFreeSpace(page) < itemsz)
 			IvfflatAppendPage(index, &buf, &page, &state, forkNum);
 
 		/* Add the item */
-		offno = PageAddItem(page, (Item) list, listSize, InvalidOffsetNumber, false, false);
+		offno = PageAddItem(page, (Item) list, itemsz, InvalidOffsetNumber, false, false);
 		if (offno == InvalidOffsetNumber)
 			elog(ERROR, "failed to add index item to \"%s\"", RelationGetRelationName(index));
 
@@ -506,17 +513,17 @@ CreateListPages(Relation index, VectorArray centers, int dimensions,
 	pfree(list);
 }
 
-#ifdef IVFFLAT_KMEANS_DEBUG
 /*
  * Print k-means metrics
  */
+#ifdef IVFFLAT_KMEANS_DEBUG
 static void
 PrintKmeansMetrics(IvfflatBuildState * buildstate)
 {
 	elog(INFO, "inertia: %.3e", buildstate->inertia);
 
 	/* Calculate Davies-Bouldin index */
-	if (buildstate->lists > 1 && !buildstate->ivfleader)
+	if (buildstate->lists > 1)
 	{
 		double		db = 0.0;
 
@@ -551,432 +558,43 @@ PrintKmeansMetrics(IvfflatBuildState * buildstate)
 }
 #endif
 
-/*
- * Within leader, wait for end of heap scan
- */
-static double
-ParallelHeapScan(IvfflatBuildState * buildstate)
-{
-	IvfflatShared *ivfshared = buildstate->ivfleader->ivfshared;
-	int			nparticipanttuplesorts;
-	double		reltuples;
-
-	nparticipanttuplesorts = buildstate->ivfleader->nparticipanttuplesorts;
-	for (;;)
-	{
-		SpinLockAcquire(&ivfshared->mutex);
-		if (ivfshared->nparticipantsdone == nparticipanttuplesorts)
-		{
-			buildstate->indtuples = ivfshared->indtuples;
-			reltuples = ivfshared->reltuples;
-#ifdef IVFFLAT_KMEANS_DEBUG
-			buildstate->inertia = ivfshared->inertia;
-#endif
-			SpinLockRelease(&ivfshared->mutex);
-			break;
-		}
-		SpinLockRelease(&ivfshared->mutex);
-
-		ConditionVariableSleep(&ivfshared->workersdonecv,
-							   WAIT_EVENT_PARALLEL_CREATE_INDEX_SCAN);
-	}
-
-	ConditionVariableCancelSleep();
-
-	return reltuples;
-}
-
-/*
- * Perform a worker's portion of a parallel sort
- */
-static void
-IvfflatParallelScanAndSort(IvfflatSpool * ivfspool, IvfflatShared * ivfshared, Sharedsort *sharedsort, Vector * ivfcenters, int sortmem, bool progress)
-{
-	SortCoordinate coordinate;
-	IvfflatBuildState buildstate;
-	TableScanDesc scan;
-	double		reltuples;
-	IndexInfo  *indexInfo;
-
-	/* Sort options, which must match AssignTuples */
-	AttrNumber	attNums[] = {1};
-	Oid			sortOperators[] = {Int4LessOperator};
-	Oid			sortCollations[] = {InvalidOid};
-	bool		nullsFirstFlags[] = {false};
-
-	/* Initialize local tuplesort coordination state */
-	coordinate = palloc0(sizeof(SortCoordinateData));
-	coordinate->isWorker = true;
-	coordinate->nParticipants = -1;
-	coordinate->sharedsort = sharedsort;
-
-	/* Join parallel scan */
-	indexInfo = BuildIndexInfo(ivfspool->index);
-	indexInfo->ii_Concurrent = ivfshared->isconcurrent;
-	InitBuildState(&buildstate, ivfspool->heap, ivfspool->index, indexInfo);
-	memcpy(buildstate.centers->items, ivfcenters, VECTOR_SIZE(buildstate.centers->dim) * buildstate.centers->maxlen);
-	buildstate.centers->length = buildstate.centers->maxlen;
-	ivfspool->sortstate = tuplesort_begin_heap(buildstate.tupdesc, 1, attNums, sortOperators, sortCollations, nullsFirstFlags, sortmem, coordinate, false);
-	buildstate.sortstate = ivfspool->sortstate;
-	scan = table_beginscan_parallel(ivfspool->heap,
-									ParallelTableScanFromIvfflatShared(ivfshared));
-	reltuples = table_index_build_scan(ivfspool->heap, ivfspool->index, indexInfo,
-									   true, progress, BuildCallback,
-									   (void *) &buildstate, scan);
-
-	/* Execute this worker's part of the sort */
-	tuplesort_performsort(ivfspool->sortstate);
-
-	/* Record statistics */
-	SpinLockAcquire(&ivfshared->mutex);
-	ivfshared->nparticipantsdone++;
-	ivfshared->reltuples += reltuples;
-	ivfshared->indtuples += buildstate.indtuples;
-#ifdef IVFFLAT_KMEANS_DEBUG
-	ivfshared->inertia += buildstate.inertia;
-#endif
-	SpinLockRelease(&ivfshared->mutex);
-
-	/* Log statistics */
-	if (progress)
-		ereport(DEBUG1, (errmsg("leader processed " INT64_FORMAT " tuples", (int64) reltuples)));
-	else
-		ereport(DEBUG1, (errmsg("worker processed " INT64_FORMAT " tuples", (int64) reltuples)));
-
-	/* Notify leader */
-	ConditionVariableSignal(&ivfshared->workersdonecv);
-
-	/* We can end tuplesorts immediately */
-	tuplesort_end(ivfspool->sortstate);
-
-	FreeBuildState(&buildstate);
-}
-
-/*
- * Perform work within a launched parallel process
- */
-void
-IvfflatParallelBuildMain(dsm_segment *seg, shm_toc *toc)
-{
-	char	   *sharedquery;
-	IvfflatSpool *ivfspool;
-	IvfflatShared *ivfshared;
-	Sharedsort *sharedsort;
-	Vector	   *ivfcenters;
-	Relation	heapRel;
-	Relation	indexRel;
-	LOCKMODE	heapLockmode;
-	LOCKMODE	indexLockmode;
-	int			sortmem;
-
-	/* Set debug_query_string for individual workers first */
-	sharedquery = shm_toc_lookup(toc, PARALLEL_KEY_QUERY_TEXT, true);
-	debug_query_string = sharedquery;
-
-	/* Report the query string from leader */
-	pgstat_report_activity(STATE_RUNNING, debug_query_string);
-
-	/* Look up shared state */
-	ivfshared = shm_toc_lookup(toc, PARALLEL_KEY_IVFFLAT_SHARED, false);
-
-	/* Open relations using lock modes known to be obtained by index.c */
-	if (!ivfshared->isconcurrent)
-	{
-		heapLockmode = ShareLock;
-		indexLockmode = AccessExclusiveLock;
-	}
-	else
-	{
-		heapLockmode = ShareUpdateExclusiveLock;
-		indexLockmode = RowExclusiveLock;
-	}
-
-	/* Open relations within worker */
-	heapRel = table_open(ivfshared->heaprelid, heapLockmode);
-	indexRel = index_open(ivfshared->indexrelid, indexLockmode);
-
-	/* Initialize worker's own spool */
-	ivfspool = (IvfflatSpool *) palloc0(sizeof(IvfflatSpool));
-	ivfspool->heap = heapRel;
-	ivfspool->index = indexRel;
-
-	/* Look up shared state private to tuplesort.c */
-	sharedsort = shm_toc_lookup(toc, PARALLEL_KEY_TUPLESORT, false);
-	tuplesort_attach_shared(sharedsort, seg);
-
-	ivfcenters = shm_toc_lookup(toc, PARALLEL_KEY_IVFFLAT_CENTERS, false);
-
-	/* Perform sorting */
-	sortmem = maintenance_work_mem / ivfshared->scantuplesortstates;
-	IvfflatParallelScanAndSort(ivfspool, ivfshared, sharedsort, ivfcenters, sortmem, false);
-
-	/* Close relations within worker */
-	index_close(indexRel, indexLockmode);
-	table_close(heapRel, heapLockmode);
-}
-
-/*
- * End parallel build
- */
-static void
-IvfflatEndParallel(IvfflatLeader * ivfleader)
-{
-	/* Shutdown worker processes */
-	WaitForParallelWorkersToFinish(ivfleader->pcxt);
-
-	/* Free last reference to MVCC snapshot, if one was used */
-	if (IsMVCCSnapshot(ivfleader->snapshot))
-		UnregisterSnapshot(ivfleader->snapshot);
-	DestroyParallelContext(ivfleader->pcxt);
-	ExitParallelMode();
-}
-
-/*
- * Return size of shared memory required for parallel index build
- */
-static Size
-ParallelEstimateShared(Relation heap, Snapshot snapshot)
-{
-	return add_size(BUFFERALIGN(sizeof(IvfflatShared)), table_parallelscan_estimate(heap, snapshot));
-}
-
-/*
- * Within leader, participate as a parallel worker
- */
-static void
-IvfflatLeaderParticipateAsWorker(IvfflatBuildState * buildstate)
-{
-	IvfflatLeader *ivfleader = buildstate->ivfleader;
-	IvfflatSpool *leaderworker;
-	int			sortmem;
-
-	/* Allocate memory and initialize private spool */
-	leaderworker = (IvfflatSpool *) palloc0(sizeof(IvfflatSpool));
-	leaderworker->heap = buildstate->heap;
-	leaderworker->index = buildstate->index;
-
-	/* Perform work common to all participants */
-	sortmem = maintenance_work_mem / ivfleader->nparticipanttuplesorts;
-	IvfflatParallelScanAndSort(leaderworker, ivfleader->ivfshared,
-							   ivfleader->sharedsort, ivfleader->ivfcenters,
-							   sortmem, true);
-}
-
-/*
- * Begin parallel build
- */
-static void
-IvfflatBeginParallel(IvfflatBuildState * buildstate, bool isconcurrent, int request)
-{
-	ParallelContext *pcxt;
-	int			scantuplesortstates;
-	Snapshot	snapshot;
-	Size		estivfshared;
-	Size		estsort;
-	Size		estcenters;
-	IvfflatShared *ivfshared;
-	Sharedsort *sharedsort;
-	Vector	   *ivfcenters;
-	IvfflatLeader *ivfleader = (IvfflatLeader *) palloc0(sizeof(IvfflatLeader));
-	bool		leaderparticipates = true;
-	int			querylen;
-
-#ifdef DISABLE_LEADER_PARTICIPATION
-	leaderparticipates = false;
-#endif
-
-	/* Enter parallel mode and create context */
-	EnterParallelMode();
-	Assert(request > 0);
-	pcxt = CreateParallelContext("vector", "IvfflatParallelBuildMain", request);
-
-	scantuplesortstates = leaderparticipates ? request + 1 : request;
-
-	/* Get snapshot for table scan */
-	if (!isconcurrent)
-		snapshot = SnapshotAny;
-	else
-		snapshot = RegisterSnapshot(GetTransactionSnapshot());
-
-	/* Estimate size of workspaces */
-	estivfshared = ParallelEstimateShared(buildstate->heap, snapshot);
-	shm_toc_estimate_chunk(&pcxt->estimator, estivfshared);
-	estsort = tuplesort_estimate_shared(scantuplesortstates);
-	shm_toc_estimate_chunk(&pcxt->estimator, estsort);
-	estcenters = VECTOR_SIZE(buildstate->dimensions) * buildstate->lists;
-	shm_toc_estimate_chunk(&pcxt->estimator, estcenters);
-	shm_toc_estimate_keys(&pcxt->estimator, 3);
-
-	/* Finally, estimate PARALLEL_KEY_QUERY_TEXT space */
-	if (debug_query_string)
-	{
-		querylen = strlen(debug_query_string);
-		shm_toc_estimate_chunk(&pcxt->estimator, querylen + 1);
-		shm_toc_estimate_keys(&pcxt->estimator, 1);
-	}
-	else
-		querylen = 0;			/* keep compiler quiet */
-
-	/* Everyone's had a chance to ask for space, so now create the DSM */
-	InitializeParallelDSM(pcxt);
-
-	/* If no DSM segment was available, back out (do serial build) */
-	if (pcxt->seg == NULL)
-	{
-		if (IsMVCCSnapshot(snapshot))
-			UnregisterSnapshot(snapshot);
-		DestroyParallelContext(pcxt);
-		ExitParallelMode();
-		return;
-	}
-
-	/* Store shared build state, for which we reserved space */
-	ivfshared = (IvfflatShared *) shm_toc_allocate(pcxt->toc, estivfshared);
-	/* Initialize immutable state */
-	ivfshared->heaprelid = RelationGetRelid(buildstate->heap);
-	ivfshared->indexrelid = RelationGetRelid(buildstate->index);
-	ivfshared->isconcurrent = isconcurrent;
-	ivfshared->scantuplesortstates = scantuplesortstates;
-	ConditionVariableInit(&ivfshared->workersdonecv);
-	SpinLockInit(&ivfshared->mutex);
-	/* Initialize mutable state */
-	ivfshared->nparticipantsdone = 0;
-	ivfshared->reltuples = 0;
-	ivfshared->indtuples = 0;
-#ifdef IVFFLAT_KMEANS_DEBUG
-	ivfshared->inertia = 0;
-#endif
-	table_parallelscan_initialize(buildstate->heap,
-								  ParallelTableScanFromIvfflatShared(ivfshared),
-								  snapshot);
-
-	/* Store shared tuplesort-private state, for which we reserved space */
-	sharedsort = (Sharedsort *) shm_toc_allocate(pcxt->toc, estsort);
-	tuplesort_initialize_shared(sharedsort, scantuplesortstates,
-								pcxt->seg);
-
-	ivfcenters = (Vector *) shm_toc_allocate(pcxt->toc, estcenters);
-	memcpy(ivfcenters, buildstate->centers->items, estcenters);
-
-	shm_toc_insert(pcxt->toc, PARALLEL_KEY_IVFFLAT_SHARED, ivfshared);
-	shm_toc_insert(pcxt->toc, PARALLEL_KEY_TUPLESORT, sharedsort);
-	shm_toc_insert(pcxt->toc, PARALLEL_KEY_IVFFLAT_CENTERS, ivfcenters);
-
-	/* Store query string for workers */
-	if (debug_query_string)
-	{
-		char	   *sharedquery;
-
-		sharedquery = (char *) shm_toc_allocate(pcxt->toc, querylen + 1);
-		memcpy(sharedquery, debug_query_string, querylen + 1);
-		shm_toc_insert(pcxt->toc, PARALLEL_KEY_QUERY_TEXT, sharedquery);
-	}
-
-	/* Launch workers, saving status for leader/caller */
-	LaunchParallelWorkers(pcxt);
-	ivfleader->pcxt = pcxt;
-	ivfleader->nparticipanttuplesorts = pcxt->nworkers_launched;
-	if (leaderparticipates)
-		ivfleader->nparticipanttuplesorts++;
-	ivfleader->ivfshared = ivfshared;
-	ivfleader->sharedsort = sharedsort;
-	ivfleader->snapshot = snapshot;
-	ivfleader->ivfcenters = ivfcenters;
-
-	/* If no workers were successfully launched, back out (do serial build) */
-	if (pcxt->nworkers_launched == 0)
-	{
-		IvfflatEndParallel(ivfleader);
-		return;
-	}
-
-	/* Log participants */
-	ereport(DEBUG1, (errmsg("using %d parallel workers", pcxt->nworkers_launched)));
-
-	/* Save leader state now that it's clear build will be parallel */
-	buildstate->ivfleader = ivfleader;
-
-	/* Join heap scan ourselves */
-	if (leaderparticipates)
-		IvfflatLeaderParticipateAsWorker(buildstate);
-
-	/* Wait for all launched workers */
-	WaitForParallelWorkersToAttach(pcxt);
-}
-
-/*
- * Scan table for tuples to index
- */
-static void
-AssignTuples(IvfflatBuildState * buildstate)
-{
-	int			parallel_workers = 0;
-	SortCoordinate coordinate = NULL;
-
-	/* Sort options, which must match IvfflatParallelScanAndSort */
-	AttrNumber	attNums[] = {1};
-	Oid			sortOperators[] = {Int4LessOperator};
-	Oid			sortCollations[] = {InvalidOid};
-	bool		nullsFirstFlags[] = {false};
-
-	pgstat_progress_update_param(PROGRESS_CREATEIDX_SUBPHASE, PROGRESS_IVFFLAT_PHASE_ASSIGN);
-
-	/* Calculate parallel workers */
-	if (buildstate->heap != NULL)
-		parallel_workers = plan_create_index_workers(RelationGetRelid(buildstate->heap), RelationGetRelid(buildstate->index));
-
-	/* Attempt to launch parallel worker scan when required */
-	if (parallel_workers > 0)
-		IvfflatBeginParallel(buildstate, buildstate->indexInfo->ii_Concurrent, parallel_workers);
-
-	/* Set up coordination state if at least one worker launched */
-	if (buildstate->ivfleader)
-	{
-		coordinate = (SortCoordinate) palloc0(sizeof(SortCoordinateData));
-		coordinate->isWorker = false;
-		coordinate->nParticipants = buildstate->ivfleader->nparticipanttuplesorts;
-		coordinate->sharedsort = buildstate->ivfleader->sharedsort;
-	}
-
-	/* Begin serial/leader tuplesort */
-	buildstate->sortstate = tuplesort_begin_heap(buildstate->tupdesc, 1, attNums, sortOperators, sortCollations, nullsFirstFlags, maintenance_work_mem, coordinate, false);
-
-	/* Add tuples to sort */
-	if (buildstate->heap != NULL)
-	{
-		if (buildstate->ivfleader)
-			buildstate->reltuples = ParallelHeapScan(buildstate);
-		else
-			buildstate->reltuples = table_index_build_scan(buildstate->heap, buildstate->index, buildstate->indexInfo,
-														   true, true, BuildCallback, (void *) buildstate, NULL);
-
-#ifdef IVFFLAT_KMEANS_DEBUG
-		PrintKmeansMetrics(buildstate);
-#endif
-	}
-}
-
 /*
  * Create entry pages
  */
 static void
 CreateEntryPages(IvfflatBuildState * buildstate, ForkNumber forkNum)
 {
-	/* Assign */
+	AttrNumber	attNums[] = {1};
-	IvfflatBench("assign tuples", AssignTuples(buildstate));
+	Oid			sortOperators[] = {Float8LessOperator};
+	Oid			sortCollations[] = {InvalidOid};
+	bool		nullsFirstFlags[] = {false};
+
+	UpdateProgress(PROGRESS_CREATEIDX_SUBPHASE, PROGRESS_IVFFLAT_PHASE_SORT);
+
+	buildstate->sortstate = tuplesort_begin_heap(buildstate->tupdesc, 1, attNums, sortOperators, sortCollations, nullsFirstFlags, maintenance_work_mem, NULL, false);
+
+	/* Add tuples to sort */
+	if (buildstate->heap != NULL)
+	{
+#if PG_VERSION_NUM >= 120000
+		buildstate->reltuples = table_index_build_scan(buildstate->heap, buildstate->index, buildstate->indexInfo,
+													   true, true, BuildCallback, (void *) buildstate, NULL);
+#else
+		buildstate->reltuples = IndexBuildHeapScan(buildstate->heap, buildstate->index, buildstate->indexInfo,
+												   true, BuildCallback, (void *) buildstate, NULL);
+#endif
+	}
 
 	/* Sort */
-	IvfflatBench("sort tuples", tuplesort_performsort(buildstate->sortstate));
+	tuplesort_performsort(buildstate->sortstate);
 
-	/* Load */
+#ifdef IVFFLAT_KMEANS_DEBUG
-	IvfflatBench("load tuples", InsertTuples(buildstate->index, buildstate, forkNum));
+	PrintKmeansMetrics(buildstate);
+#endif
 
-	/* End sort */
+	/* Insert */
+	InsertTuples(buildstate->index, buildstate, forkNum);
 	tuplesort_end(buildstate->sortstate);
-
-	/* End parallel build */
-	if (buildstate->ivfleader)
-		IvfflatEndParallel(buildstate->ivfleader);
 }
 
 /*
@@ -993,7 +611,7 @@ BuildIndex(Relation heap, Relation index, IndexInfo *indexInfo,
 	/* Create pages */
 	CreateMetaPage(index, buildstate->dimensions, buildstate->lists, forkNum);
 	CreateListPages(index, buildstate->centers, buildstate->dimensions, buildstate->lists, forkNum, &buildstate->listInfo);
-	CreateEntryPages(buildstate, forkNum);
+	IvfflatBench("CreateEntryPages", CreateEntryPages(buildstate, forkNum));
 
 	FreeBuildState(buildstate);
 }

src/ivfflat.c

@@ -3,16 +3,13 @@
 #include <float.h>
 
 #include "access/amapi.h"
-#include "access/reloptions.h"
-#include "commands/progress.h"
 #include "commands/vacuum.h"
 #include "ivfflat.h"
 #include "utils/guc.h"
 #include "utils/selfuncs.h"
-#include "utils/spccache.h"
 
-#if PG_VERSION_NUM < 150000
+#if PG_VERSION_NUM >= 120000
-#define MarkGUCPrefixReserved(x) EmitWarningsOnPlaceholders(x)
+#include "commands/progress.h"
 #endif
 
 int			ivfflat_probes;
@@ -22,11 +19,11 @@ static relopt_kind ivfflat_relopt_kind;
  * Initialize index options and variables
  */
 void
-IvfflatInit(void)
+_PG_init(void)
 {
 	ivfflat_relopt_kind = add_reloption_kind();
 	add_int_reloption(ivfflat_relopt_kind, "lists", "Number of inverted lists",
-					  IVFFLAT_DEFAULT_LISTS, IVFFLAT_MIN_LISTS, IVFFLAT_MAX_LISTS
+					  IVFFLAT_DEFAULT_LISTS, 1, IVFFLAT_MAX_LISTS
 #if PG_VERSION_NUM >= 130000
 					  ,AccessExclusiveLock
 #endif
@@ -34,14 +31,13 @@ IvfflatInit(void)
 
 	DefineCustomIntVariable("ivfflat.probes", "Sets the number of probes",
 							"Valid range is 1..lists.", &ivfflat_probes,
-							IVFFLAT_DEFAULT_PROBES, IVFFLAT_MIN_LISTS, IVFFLAT_MAX_LISTS, PGC_USERSET, 0, NULL, NULL, NULL);
+							1, 1, IVFFLAT_MAX_LISTS, PGC_USERSET, 0, NULL, NULL, NULL);
-
-	MarkGUCPrefixReserved("ivfflat");
 }
 
 /*
  * Get the name of index build phase
  */
+#if PG_VERSION_NUM >= 120000
 static char *
 ivfflatbuildphasename(int64 phasenum)
 {
@@ -51,14 +47,15 @@ ivfflatbuildphasename(int64 phasenum)
 			return "initializing";
 		case PROGRESS_IVFFLAT_PHASE_KMEANS:
 			return "performing k-means";
-		case PROGRESS_IVFFLAT_PHASE_ASSIGN:
+		case PROGRESS_IVFFLAT_PHASE_SORT:
-			return "assigning tuples";
+			return "sorting tuples";
 		case PROGRESS_IVFFLAT_PHASE_LOAD:
 			return "loading tuples";
 		default:
 			return NULL;
 	}
 }
+#endif
 
 /*
  * Estimate the cost of an index scan
@@ -66,14 +63,17 @@ ivfflatbuildphasename(int64 phasenum)
 static void
 ivfflatcostestimate(PlannerInfo *root, IndexPath *path, double loop_count,
 					Cost *indexStartupCost, Cost *indexTotalCost,
-					Selectivity *indexSelectivity, double *indexCorrelation,
+					Selectivity *indexSelectivity, double *indexCorrelation
-					double *indexPages)
+					,double *indexPages
+)
 {
 	GenericCosts costs;
 	int			lists;
 	double		ratio;
-	double		spc_seq_page_cost;
+	Relation	indexRel;
-	Relation	index;
+#if PG_VERSION_NUM < 120000
+	List	   *qinfos;
+#endif
 
 	/* Never use index without order */
 	if (path->indexorderbys == NULL)
@@ -88,49 +88,30 @@ ivfflatcostestimate(PlannerInfo *root, IndexPath *path, double loop_count,
 
 	MemSet(&costs, 0, sizeof(costs));
 
-	index = index_open(path->indexinfo->indexoid, NoLock);
+#if PG_VERSION_NUM >= 120000
-	IvfflatGetMetaPageInfo(index, &lists, NULL);
-	index_close(index, NoLock);
-
-	/* Get the ratio of lists that we need to visit */
-	ratio = ((double) ivfflat_probes) / lists;
-	if (ratio > 1.0)
-		ratio = 1.0;
-
-	/*
-	 * This gives us the subset of tuples to visit. This value is passed into
-	 * the generic cost estimator to determine the number of pages to visit
-	 * during the index scan.
-	 */
-	costs.numIndexTuples = path->indexinfo->tuples * ratio;
-
 	genericcostestimate(root, path, loop_count, &costs);
+#else
+	qinfos = deconstruct_indexquals(path);
+	genericcostestimate(root, path, loop_count, qinfos, &costs);
+#endif
 
-	get_tablespace_page_costs(path->indexinfo->reltablespace, NULL, &spc_seq_page_cost);
+	indexRel = index_open(path->indexinfo->indexoid, NoLock);
+	lists = IvfflatGetLists(indexRel);
+	index_close(indexRel, NoLock);
 
-	/* Adjust cost if needed since TOAST not included in seq scan cost */
+	ratio = ((double) ivfflat_probes) / lists;
-	if (costs.numIndexPages > path->indexinfo->rel->pages && ratio < 0.5)
+	if (ratio > 1)
-	{
+		ratio = 1;
-		/* Change all page cost from random to sequential */
-		costs.indexTotalCost -= costs.numIndexPages * (costs.spc_random_page_cost - spc_seq_page_cost);
 
-		/* Remove cost of extra pages */
+	// cost estimates for parallel workers applied outside of amcostestimate
-		costs.indexTotalCost -= (costs.numIndexPages - path->indexinfo->rel->pages) * spc_seq_page_cost;
+	elog(INFO, "parallel_workers = %d, parallel aware = %d", path->path.parallel_workers, path->path.parallel_aware);
-	}
-	else
-	{
-		/* Change some page cost from random to sequential */
-		costs.indexTotalCost -= 0.5 * costs.numIndexPages * (costs.spc_random_page_cost - spc_seq_page_cost);
-	}
 
-	/*
+	costs.indexTotalCost *= ratio;
-	 * If the list selectivity is lower than what is returned from the generic
+	costs.numIndexPages *= ratio;
-	 * cost estimator, use that.
-	 */
-	if (ratio < costs.indexSelectivity)
-		costs.indexSelectivity = ratio;
 
-	/* Use total cost since most work happens before first tuple is returned */
+	elog(INFO, "ivfflatcostestimate = %f", costs.indexTotalCost);
+
+	/* Startup cost and total cost are same */
 	*indexStartupCost = costs.indexTotalCost;
 	*indexTotalCost = costs.indexTotalCost;
 	*indexSelectivity = costs.indexSelectivity;
@@ -176,6 +157,25 @@ ivfflatvalidate(Oid opclassoid)
 	return true;
 }
 
+static Size
+ivfflatestimateparallelscan()
+{
+	elog(INFO, "ivfflatestimateparallelscan");
+	return 0;
+}
+
+static void
+ivfflatinitparallelscan(void *target)
+{
+	elog(INFO, "ivfflatinitparallelscan");
+}
+
+static void
+ivfflatparallelrescan(IndexScanDesc scan)
+{
+	elog(INFO, "ivfflatparallelrescan");
+}
+
 /*
  * Define index handler
  *
@@ -203,7 +203,7 @@ ivfflathandler(PG_FUNCTION_ARGS)
 	amroutine->amstorage = false;
 	amroutine->amclusterable = false;
 	amroutine->ampredlocks = false;
-	amroutine->amcanparallel = false;
+	amroutine->amcanparallel = true;
 	amroutine->amcaninclude = false;
 #if PG_VERSION_NUM >= 130000
 	amroutine->amusemaintenanceworkmem = false; /* not used during VACUUM */
@@ -221,7 +221,9 @@ ivfflathandler(PG_FUNCTION_ARGS)
 	amroutine->amcostestimate = ivfflatcostestimate;
 	amroutine->amoptions = ivfflatoptions;
 	amroutine->amproperty = NULL;	/* TODO AMPROP_DISTANCE_ORDERABLE */
+#if PG_VERSION_NUM >= 120000
 	amroutine->ambuildphasename = ivfflatbuildphasename;
+#endif
 	amroutine->amvalidate = ivfflatvalidate;
 #if PG_VERSION_NUM >= 140000
 	amroutine->amadjustmembers = NULL;
@@ -235,9 +237,9 @@ ivfflathandler(PG_FUNCTION_ARGS)
 	amroutine->amrestrpos = NULL;
 
 	/* Interface functions to support parallel index scans */
-	amroutine->amestimateparallelscan = NULL;
+	amroutine->amestimateparallelscan = ivfflatestimateparallelscan;
-	amroutine->aminitparallelscan = NULL;
+	amroutine->aminitparallelscan = ivfflatinitparallelscan;
-	amroutine->amparallelrescan = NULL;
+	amroutine->amparallelrescan = ivfflatparallelrescan;
 
 	PG_RETURN_POINTER(amroutine);
 }

src/ivfflat.h

@@ -3,12 +3,14 @@
 
 #include "postgres.h"
 
-#include "access/genam.h"
+#if PG_VERSION_NUM < 110000
+#error "Requires PostgreSQL 11+"
+#endif
+
 #include "access/generic_xlog.h"
-#include "access/parallel.h"
+#include "access/reloptions.h"
-#include "lib/pairingheap.h"
 #include "nodes/execnodes.h"
-#include "port.h"				/* for random() */
+#include "port.h"				/* for strtof() and random() */
 #include "utils/sampling.h"
 #include "utils/tuplesort.h"
 #include "vector.h"
@@ -37,16 +39,13 @@
 #define IVFFLAT_METAPAGE_BLKNO	0
 #define IVFFLAT_HEAD_BLKNO		1	/* first list page */
 
-/* IVFFlat parameters */
 #define IVFFLAT_DEFAULT_LISTS	100
-#define IVFFLAT_MIN_LISTS		1
 #define IVFFLAT_MAX_LISTS		32768
-#define IVFFLAT_DEFAULT_PROBES	1
 
 /* Build phases */
 /* PROGRESS_CREATEIDX_SUBPHASE_INITIALIZE is 1 */
 #define PROGRESS_IVFFLAT_PHASE_KMEANS	2
-#define PROGRESS_IVFFLAT_PHASE_ASSIGN	3
+#define PROGRESS_IVFFLAT_PHASE_SORT		3
 #define PROGRESS_IVFFLAT_PHASE_LOAD		4
 
 #define IVFFLAT_LIST_SIZE(_dim)	(offsetof(IvfflatListData, center) + VECTOR_SIZE(_dim))
@@ -80,6 +79,9 @@
 /* Variables */
 extern int	ivfflat_probes;
 
+/* Exported functions */
+PGDLLEXPORT void _PG_init(void);
+
 typedef struct VectorArrayData
 {
 	int			length;
@@ -103,50 +105,6 @@ typedef struct IvfflatOptions
 	int			lists;			/* number of lists */
 }			IvfflatOptions;
 
-typedef struct IvfflatSpool
-{
-	Tuplesortstate *sortstate;
-	Relation	heap;
-	Relation	index;
-}			IvfflatSpool;
-
-typedef struct IvfflatShared
-{
-	/* Immutable state */
-	Oid			heaprelid;
-	Oid			indexrelid;
-	bool		isconcurrent;
-	int			scantuplesortstates;
-
-	/* Worker progress */
-	ConditionVariable workersdonecv;
-
-	/* Mutex for mutable state */
-	slock_t		mutex;
-
-	/* Mutable state */
-	int			nparticipantsdone;
-	double		reltuples;
-	double		indtuples;
-
-#ifdef IVFFLAT_KMEANS_DEBUG
-	double		inertia;
-#endif
-}			IvfflatShared;
-
-#define ParallelTableScanFromIvfflatShared(shared) \
-	(ParallelTableScanDesc) ((char *) (shared) + BUFFERALIGN(sizeof(IvfflatShared)))
-
-typedef struct IvfflatLeader
-{
-	ParallelContext *pcxt;
-	int			nparticipanttuplesorts;
-	IvfflatShared *ivfshared;
-	Sharedsort *sharedsort;
-	Snapshot	snapshot;
-	Vector	   *ivfcenters;
-}			IvfflatLeader;
-
 typedef struct IvfflatBuildState
 {
 	/* Info */
@@ -192,9 +150,6 @@ typedef struct IvfflatBuildState
 
 	/* Memory */
 	MemoryContext tmpCtx;
-
-	/* Parallel builds */
-	IvfflatLeader *ivfleader;
 }			IvfflatBuildState;
 
 typedef struct IvfflatMetaPageData
@@ -235,8 +190,8 @@ typedef struct IvfflatScanList
 typedef struct IvfflatScanOpaqueData
 {
 	int			probes;
-	int			dimensions;
 	bool		first;
+	Buffer		buf;
 
 	/* Sorting */
 	Tuplesortstate *sortstate;
@@ -266,18 +221,15 @@ VectorArray VectorArrayInit(int maxlen, int dimensions);
 void		VectorArrayFree(VectorArray arr);
 void		PrintVectorArray(char *msg, VectorArray arr);
 void		IvfflatKmeans(Relation index, VectorArray samples, VectorArray centers);
-FmgrInfo   *IvfflatOptionalProcInfo(Relation index, uint16 procnum);
+FmgrInfo   *IvfflatOptionalProcInfo(Relation rel, uint16 procnum);
 bool		IvfflatNormValue(FmgrInfo *procinfo, Oid collation, Datum *value, Vector * result);
 int			IvfflatGetLists(Relation index);
-void		IvfflatGetMetaPageInfo(Relation index, int *lists, int *dimensions);
+void		IvfflatUpdateList(Relation index, GenericXLogState *state, ListInfo listInfo, BlockNumber insertPage, BlockNumber originalInsertPage, BlockNumber startPage, ForkNumber forkNum);
-void		IvfflatUpdateList(Relation index, ListInfo listInfo, BlockNumber insertPage, BlockNumber originalInsertPage, BlockNumber startPage, ForkNumber forkNum);
 void		IvfflatCommitBuffer(Buffer buf, GenericXLogState *state);
 void		IvfflatAppendPage(Relation index, Buffer *buf, Page *page, GenericXLogState **state, ForkNumber forkNum);
 Buffer		IvfflatNewBuffer(Relation index, ForkNumber forkNum);
 void		IvfflatInitPage(Buffer buf, Page page);
 void		IvfflatInitRegisterPage(Relation index, Buffer *buf, Page *page, GenericXLogState **state);
-void		IvfflatInit(void);
-PGDLLEXPORT void IvfflatParallelBuildMain(dsm_segment *seg, shm_toc *toc);
 
 /* Index access methods */
 IndexBuildResult *ivfflatbuild(Relation heap, Relation index, IndexInfo *indexInfo);

src/ivfinsert.c

@@ -2,51 +2,44 @@
 
 #include <float.h>
 
-#include "access/generic_xlog.h"
 #include "ivfflat.h"
 #include "storage/bufmgr.h"
-#include "storage/lmgr.h"
 #include "utils/memutils.h"
 
 /*
  * Find the list that minimizes the distance function
  */
 static void
-FindInsertPage(Relation index, Datum *values, BlockNumber *insertPage, ListInfo * listInfo)
+FindInsertPage(Relation rel, Datum *values, BlockNumber *insertPage, ListInfo * listInfo)
 {
+	Buffer		cbuf;
+	Page		cpage;
+	IvfflatList list;
+	double		distance;
 	double		minDistance = DBL_MAX;
 	BlockNumber nextblkno = IVFFLAT_HEAD_BLKNO;
 	FmgrInfo   *procinfo;
 	Oid			collation;
+	OffsetNumber offno;
+	OffsetNumber maxoffno;
 
-	/* Avoid compiler warning */
+	procinfo = index_getprocinfo(rel, 1, IVFFLAT_DISTANCE_PROC);
-	listInfo->blkno = nextblkno;
+	collation = rel->rd_indcollation[0];
-	listInfo->offno = FirstOffsetNumber;
-
-	procinfo = index_getprocinfo(index, 1, IVFFLAT_DISTANCE_PROC);
-	collation = index->rd_indcollation[0];
 
 	/* Search all list pages */
 	while (BlockNumberIsValid(nextblkno))
 	{
-		Buffer		cbuf;
+		cbuf = ReadBuffer(rel, nextblkno);
-		Page		cpage;
-		OffsetNumber maxoffno;
-
-		cbuf = ReadBuffer(index, nextblkno);
 		LockBuffer(cbuf, BUFFER_LOCK_SHARE);
 		cpage = BufferGetPage(cbuf);
 		maxoffno = PageGetMaxOffsetNumber(cpage);
 
-		for (OffsetNumber offno = FirstOffsetNumber; offno <= maxoffno; offno = OffsetNumberNext(offno))
+		for (offno = FirstOffsetNumber; offno <= maxoffno; offno = OffsetNumberNext(offno))
 		{
-			IvfflatList list;
-			double		distance;
-
 			list = (IvfflatList) PageGetItem(cpage, PageGetItemId(cpage, offno));
 			distance = DatumGetFloat8(FunctionCall2Coll(procinfo, collation, values[0], PointerGetDatum(&list->center)));
 
-			if (distance < minDistance || !BlockNumberIsValid(*insertPage))
+			if (distance < minDistance)
 			{
 				*insertPage = list->insertPage;
 				listInfo->blkno = nextblkno;
@@ -65,7 +58,7 @@ FindInsertPage(Relation index, Datum *values, BlockNumber *insertPage, ListInfo
  * Insert a tuple into the index
  */
 static void
-InsertTuple(Relation index, Datum *values, bool *isnull, ItemPointer heap_tid, Relation heapRel)
+InsertTuple(Relation rel, Datum *values, bool *isnull, ItemPointer heap_tid, Relation heapRel)
 {
 	IndexTuple	itup;
 	Datum		value;
@@ -82,33 +75,33 @@ InsertTuple(Relation index, Datum *values, bool *isnull, ItemPointer heap_tid, R
 	value = PointerGetDatum(PG_DETOAST_DATUM(values[0]));
 
 	/* Normalize if needed */
-	normprocinfo = IvfflatOptionalProcInfo(index, IVFFLAT_NORM_PROC);
+	normprocinfo = IvfflatOptionalProcInfo(rel, IVFFLAT_NORM_PROC);
 	if (normprocinfo != NULL)
 	{
-		if (!IvfflatNormValue(normprocinfo, index->rd_indcollation[0], &value, NULL))
+		if (!IvfflatNormValue(normprocinfo, rel->rd_indcollation[0], &value, NULL))
 			return;
 	}
 
 	/* Find the insert page - sets the page and list info */
-	FindInsertPage(index, values, &insertPage, &listInfo);
+	FindInsertPage(rel, values, &insertPage, &listInfo);
 	Assert(BlockNumberIsValid(insertPage));
 	originalInsertPage = insertPage;
 
 	/* Form tuple */
-	itup = index_form_tuple(RelationGetDescr(index), &value, isnull);
+	itup = index_form_tuple(RelationGetDescr(rel), &value, isnull);
 	itup->t_tid = *heap_tid;
 
 	/* Get tuple size */
 	itemsz = MAXALIGN(IndexTupleSize(itup));
-	Assert(itemsz <= BLCKSZ - MAXALIGN(SizeOfPageHeaderData) - MAXALIGN(sizeof(IvfflatPageOpaqueData)) - sizeof(ItemIdData));
+	Assert(itemsz <= BLCKSZ - MAXALIGN(SizeOfPageHeaderData) - MAXALIGN(sizeof(IvfflatPageOpaqueData)));
 
 	/* Find a page to insert the item */
 	for (;;)
 	{
-		buf = ReadBuffer(index, insertPage);
+		buf = ReadBuffer(rel, insertPage);
 		LockBuffer(buf, BUFFER_LOCK_EXCLUSIVE);
 
-		state = GenericXLogStart(index);
+		state = GenericXLogStart(rel);
 		page = GenericXLogRegisterBuffer(state, buf, 0);
 
 		if (PageGetFreeSpace(page) >= itemsz)
@@ -124,16 +117,23 @@ InsertTuple(Relation index, Datum *values, bool *isnull, ItemPointer heap_tid, R
 		}
 		else
 		{
+			Buffer		metabuf;
 			Buffer		newbuf;
 			Page		newpage;
 
+			/*
+			 * From ReadBufferExtended: Caller is responsible for ensuring
+			 * that only one backend tries to extend a relation at the same
+			 * time!
+			 */
+			metabuf = ReadBuffer(rel, IVFFLAT_METAPAGE_BLKNO);
+			LockBuffer(metabuf, BUFFER_LOCK_EXCLUSIVE);
+
 			/* Add a new page */
-			LockRelationForExtension(index, ExclusiveLock);
+			newbuf = IvfflatNewBuffer(rel, MAIN_FORKNUM);
-			newbuf = IvfflatNewBuffer(index, MAIN_FORKNUM);
+			newpage = GenericXLogRegisterBuffer(state, newbuf, GENERIC_XLOG_FULL_IMAGE);
-			UnlockRelationForExtension(index, ExclusiveLock);
 
 			/* Init new page */
-			newpage = GenericXLogRegisterBuffer(state, newbuf, GENERIC_XLOG_FULL_IMAGE);
 			IvfflatInitPage(newbuf, newpage);
 
 			/* Update insert page */
@@ -143,13 +143,18 @@ InsertTuple(Relation index, Datum *values, bool *isnull, ItemPointer heap_tid, R
 			IvfflatPageGetOpaque(page)->nextblkno = insertPage;
 
 			/* Commit */
+			MarkBufferDirty(newbuf);
+			MarkBufferDirty(buf);
 			GenericXLogFinish(state);
 
+			/* Unlock extend relation lock as early as possible */
+			UnlockReleaseBuffer(metabuf);
+
 			/* Unlock previous buffer */
 			UnlockReleaseBuffer(buf);
 
 			/* Prepare new buffer */
-			state = GenericXLogStart(index);
+			state = GenericXLogStart(rel);
 			buf = newbuf;
 			page = GenericXLogRegisterBuffer(state, buf, 0);
 			break;
@@ -158,13 +163,13 @@ InsertTuple(Relation index, Datum *values, bool *isnull, ItemPointer heap_tid, R
 
 	/* Add to next offset */
 	if (PageAddItem(page, (Item) itup, itemsz, InvalidOffsetNumber, false, false) == InvalidOffsetNumber)
-		elog(ERROR, "failed to add index item to \"%s\"", RelationGetRelationName(index));
+		elog(ERROR, "failed to add index item to \"%s\"", RelationGetRelationName(rel));
 
 	IvfflatCommitBuffer(buf, state);
 
 	/* Update the insert page */
 	if (insertPage != originalInsertPage)
-		IvfflatUpdateList(index, listInfo, insertPage, originalInsertPage, InvalidBlockNumber, MAIN_FORKNUM);
+		IvfflatUpdateList(rel, state, listInfo, insertPage, originalInsertPage, InvalidBlockNumber, MAIN_FORKNUM);
 }
 
 /*

src/ivfkmeans.c

@@ -1,15 +1,10 @@
 #include "postgres.h"
 
 #include <float.h>
-#include <math.h>
 
 #include "ivfflat.h"
 #include "miscadmin.h"
 
-#ifdef IVFFLAT_MEMORY
-#include "utils/memutils.h"
-#endif
-
 /*
  * Initialize with kmeans++
  *
@@ -20,7 +15,12 @@ InitCenters(Relation index, VectorArray samples, VectorArray centers, float *low
 {
 	FmgrInfo   *procinfo;
 	Oid			collation;
+	int			i;
 	int64		j;
+	double		distance;
+	double		sum;
+	double		choice;
+	Vector	   *vec;
 	float	   *weight = palloc(samples->length * sizeof(float));
 	int			numCenters = centers->maxlen;
 	int			numSamples = samples->length;
@@ -33,21 +33,17 @@ InitCenters(Relation index, VectorArray samples, VectorArray centers, float *low
 	centers->length++;
 
 	for (j = 0; j < numSamples; j++)
-		weight[j] = FLT_MAX;
+		weight[j] = DBL_MAX;
 
-	for (int i = 0; i < numCenters; i++)
+	for (i = 0; i < numCenters; i++)
 	{
-		double		sum;
-		double		choice;
-
 		CHECK_FOR_INTERRUPTS();
 
 		sum = 0.0;
 
 		for (j = 0; j < numSamples; j++)
 		{
-			Vector	   *vec = VectorArrayGet(samples, j);
+			vec = VectorArrayGet(samples, j);
-			double		distance;
 
 			/* Only need to compute distance for new center */
 			/* TODO Use triangle inequality to reduce distance calculations */
@@ -91,12 +87,13 @@ InitCenters(Relation index, VectorArray samples, VectorArray centers, float *low
 static inline void
 ApplyNorm(FmgrInfo *normprocinfo, Oid collation, Vector * vec)
 {
+	int			i;
 	double		norm = DatumGetFloat8(FunctionCall1Coll(normprocinfo, collation, PointerGetDatum(vec)));
 
 	/* TODO Handle zero norm */
 	if (norm > 0)
 	{
-		for (int i = 0; i < vec->dim; i++)
+		for (i = 0; i < vec->dim; i++)
 			vec->x[i] /= norm;
 	}
 }
@@ -116,6 +113,9 @@ CompareVectors(const void *a, const void *b)
 static void
 QuickCenters(Relation index, VectorArray samples, VectorArray centers)
 {
+	int			i;
+	int			j;
+	Vector	   *vec;
 	int			dimensions = centers->dim;
 	Oid			collation = index->rd_indcollation[0];
 	FmgrInfo   *normprocinfo = IvfflatOptionalProcInfo(index, IVFFLAT_KMEANS_NORM_PROC);
@@ -124,9 +124,9 @@ QuickCenters(Relation index, VectorArray samples, VectorArray centers)
 	if (samples->length > 0)
 	{
 		qsort(samples->items, samples->length, VECTOR_SIZE(samples->dim), CompareVectors);
-		for (int i = 0; i < samples->length; i++)
+		for (i = 0; i < samples->length; i++)
 		{
-			Vector	   *vec = VectorArrayGet(samples, i);
+			vec = VectorArrayGet(samples, i);
 
 			if (i == 0 || CompareVectors(vec, VectorArrayGet(samples, i - 1)) != 0)
 			{
@@ -139,12 +139,12 @@ QuickCenters(Relation index, VectorArray samples, VectorArray centers)
 	/* Fill remaining with random data */
 	while (centers->length < centers->maxlen)
 	{
-		Vector	   *vec = VectorArrayGet(centers, centers->length);
+		vec = VectorArrayGet(centers, centers->length);
 
 		SET_VARSIZE(vec, VECTOR_SIZE(dimensions));
 		vec->dim = dimensions;
 
-		for (int j = 0; j < dimensions; j++)
+		for (j = 0; j < dimensions; j++)
 			vec->x[j] = RandomDouble();
 
 		/* Normalize if needed (only needed for random centers) */
@@ -155,23 +155,6 @@ QuickCenters(Relation index, VectorArray samples, VectorArray centers)
 	}
 }
 
-#ifdef IVFFLAT_MEMORY
-/*
- * Show memory usage
- */
-static void
-ShowMemoryUsage(Size estimatedSize)
-{
-#if PG_VERSION_NUM >= 130000
-	elog(INFO, "total memory: %zu MB",
-		 MemoryContextMemAllocated(CurrentMemoryContext, true) / (1024 * 1024));
-#else
-	MemoryContextStats(CurrentMemoryContext);
-#endif
-	elog(INFO, "estimated memory: %zu MB", estimatedSize / (1024 * 1024));
-}
-#endif
-
 /*
  * Use Elkan for performance. This requires distance function to satisfy triangle inequality.
  *
@@ -188,6 +171,7 @@ ElkanKmeans(Relation index, VectorArray samples, VectorArray centers)
 	Oid			collation;
 	Vector	   *vec;
 	Vector	   *newCenter;
+	int			iteration;
 	int64		j;
 	int64		k;
 	int			dimensions = centers->dim;
@@ -201,6 +185,14 @@ ElkanKmeans(Relation index, VectorArray samples, VectorArray centers)
 	float	   *s;
 	float	   *halfcdist;
 	float	   *newcdist;
+	int			changes;
+	double		minDistance;
+	int			closestCenter;
+	double		distance;
+	bool		rj;
+	bool		rjreset;
+	double		dxcx;
+	double		dxc;
 
 	/* Calculate allocation sizes */
 	Size		samplesSize = VECTOR_ARRAY_SIZE(samples->maxlen, samples->dim);
@@ -219,7 +211,7 @@ ElkanKmeans(Relation index, VectorArray samples, VectorArray centers)
 
 	/* Check memory requirements */
 	/* Add one to error message to ceil */
-	if (totalSize > (Size) maintenance_work_mem * 1024L)
+	if (totalSize / 1024 > maintenance_work_mem)
 		ereport(ERROR,
 				(errcode(ERRCODE_PROGRAM_LIMIT_EXCEEDED),
 				 errmsg("memory required is %zu MB, maintenance_work_mem is %d MB",
@@ -252,24 +244,20 @@ ElkanKmeans(Relation index, VectorArray samples, VectorArray centers)
 		vec->dim = dimensions;
 	}
 
-#ifdef IVFFLAT_MEMORY
-	ShowMemoryUsage(totalSize);
-#endif
-
 	/* Pick initial centers */
 	InitCenters(index, samples, centers, lowerBound);
 
 	/* Assign each x to its closest initial center c(x) = argmin d(x,c) */
 	for (j = 0; j < numSamples; j++)
 	{
-		float		minDistance = FLT_MAX;
+		minDistance = DBL_MAX;
-		int			closestCenter = 0;
+		closestCenter = -1;
 
 		/* Find closest center */
 		for (k = 0; k < numCenters; k++)
 		{
 			/* TODO Use Lemma 1 in k-means++ initialization */
-			float		distance = lowerBound[j * numCenters + k];
+			distance = lowerBound[j * numCenters + k];
 
 			if (distance < minDistance)
 			{
@@ -283,14 +271,13 @@ ElkanKmeans(Relation index, VectorArray samples, VectorArray centers)
 	}
 
 	/* Give 500 iterations to converge */
-	for (int iteration = 0; iteration < 500; iteration++)
+	for (iteration = 0; iteration < 500; iteration++)
 	{
-		int			changes = 0;
-		bool		rjreset;
-
 		/* Can take a while, so ensure we can interrupt */
 		CHECK_FOR_INTERRUPTS();
 
+		changes = 0;
+
 		/* Step 1: For all centers, compute distance */
 		for (j = 0; j < numCenters; j++)
 		{
@@ -298,8 +285,7 @@ ElkanKmeans(Relation index, VectorArray samples, VectorArray centers)
 
 			for (k = j + 1; k < numCenters; k++)
 			{
-				float		distance = 0.5 * DatumGetFloat8(FunctionCall2Coll(procinfo, collation, PointerGetDatum(vec), PointerGetDatum(VectorArrayGet(centers, k))));
+				distance = 0.5 * DatumGetFloat8(FunctionCall2Coll(procinfo, collation, PointerGetDatum(vec), PointerGetDatum(VectorArrayGet(centers, k))));
-
 				halfcdist[j * numCenters + k] = distance;
 				halfcdist[k * numCenters + j] = distance;
 			}
@@ -308,12 +294,10 @@ ElkanKmeans(Relation index, VectorArray samples, VectorArray centers)
 		/* For all centers c, compute s(c) */
 		for (j = 0; j < numCenters; j++)
 		{
-			float		minDistance = FLT_MAX;
+			minDistance = DBL_MAX;
 
 			for (k = 0; k < numCenters; k++)
 			{
-				float		distance;
-
 				if (j == k)
 					continue;
 
@@ -329,8 +313,6 @@ ElkanKmeans(Relation index, VectorArray samples, VectorArray centers)
 
 		for (j = 0; j < numSamples; j++)
 		{
-			bool		rj;
-
 			/* Step 2: Identify all points x such that u(x) <= s(c(x)) */
 			if (upperBound[j] <= s[closestCenters[j]])
 				continue;
@@ -339,8 +321,6 @@ ElkanKmeans(Relation index, VectorArray samples, VectorArray centers)
 
 			for (k = 0; k < numCenters; k++)
 			{
-				float		dxcx;
-
 				/* Step 3: For all remaining points x and centers c */
 				if (k == closestCenters[j])
 					continue;
@@ -370,7 +350,7 @@ ElkanKmeans(Relation index, VectorArray samples, VectorArray centers)
 				/* Step 3b */
 				if (dxcx > lowerBound[j * numCenters + k] || dxcx > halfcdist[closestCenters[j] * numCenters + k])
 				{
-					float		dxc = DatumGetFloat8(FunctionCall2Coll(procinfo, collation, PointerGetDatum(vec), PointerGetDatum(VectorArrayGet(centers, k))));
+					dxc = DatumGetFloat8(FunctionCall2Coll(procinfo, collation, PointerGetDatum(vec), PointerGetDatum(VectorArrayGet(centers, k))));
 
 					/* d(x,c) calculated */
 					lowerBound[j * numCenters + k] = dxc;
@@ -384,6 +364,7 @@ ElkanKmeans(Relation index, VectorArray samples, VectorArray centers)
 
 						changes++;
 					}
+
 				}
 			}
 		}
@@ -400,8 +381,6 @@ ElkanKmeans(Relation index, VectorArray samples, VectorArray centers)
 
 		for (j = 0; j < numSamples; j++)
 		{
-			int			closestCenter;
-
 			vec = VectorArrayGet(samples, j);
 			closestCenter = closestCenters[j];
 
@@ -419,14 +398,6 @@ ElkanKmeans(Relation index, VectorArray samples, VectorArray centers)
 
 			if (centerCounts[j] > 0)
 			{
-				/* Double avoids overflow, but requires more memory */
-				/* TODO Update bounds */
-				for (k = 0; k < dimensions; k++)
-				{
-					if (isinf(vec->x[k]))
-						vec->x[k] = vec->x[k] > 0 ? FLT_MAX : -FLT_MAX;
-				}
-
 				for (k = 0; k < dimensions; k++)
 					vec->x[k] /= centerCounts[j];
 			}
@@ -450,7 +421,7 @@ ElkanKmeans(Relation index, VectorArray samples, VectorArray centers)
 		{
 			for (k = 0; k < numCenters; k++)
 			{
-				float		distance = lowerBound[j * numCenters + k] - newcdist[k];
+				distance = lowerBound[j * numCenters + k] - newcdist[k];
 
 				if (distance < 0)
 					distance = 0;
@@ -466,7 +437,7 @@ ElkanKmeans(Relation index, VectorArray samples, VectorArray centers)
 
 		/* Step 7 */
 		for (j = 0; j < numCenters; j++)
-			VectorArraySet(centers, j, VectorArrayGet(newCenters, j));
+			memcpy(VectorArrayGet(centers, j), VectorArrayGet(newCenters, j), VECTOR_SIZE(dimensions));
 
 		if (changes == 0 && iteration != 0)
 			break;
@@ -489,29 +460,17 @@ static void
 CheckCenters(Relation index, VectorArray centers)
 {
 	FmgrInfo   *normprocinfo;
+	Oid			collation;
+	int			i;
+	double		norm;
 
 	if (centers->length != centers->maxlen)
 		elog(ERROR, "Not enough centers. Please report a bug.");
 
-	/* Ensure no NaN or infinite values */
-	for (int i = 0; i < centers->length; i++)
-	{
-		Vector	   *vec = VectorArrayGet(centers, i);
-
-		for (int j = 0; j < vec->dim; j++)
-		{
-			if (isnan(vec->x[j]))
-				elog(ERROR, "NaN detected. Please report a bug.");
-
-			if (isinf(vec->x[j]))
-				elog(ERROR, "Infinite value detected. Please report a bug.");
-		}
-	}
-
 	/* Ensure no duplicate centers */
 	/* Fine to sort in-place */
 	qsort(centers->items, centers->length, VECTOR_SIZE(centers->dim), CompareVectors);
-	for (int i = 1; i < centers->length; i++)
+	for (i = 1; i < centers->length; i++)
 	{
 		if (CompareVectors(VectorArrayGet(centers, i), VectorArrayGet(centers, i - 1)) == 0)
 			elog(ERROR, "Duplicate centers detected. Please report a bug.");
@@ -522,12 +481,11 @@ CheckCenters(Relation index, VectorArray centers)
 	normprocinfo = IvfflatOptionalProcInfo(index, IVFFLAT_NORM_PROC);
 	if (normprocinfo != NULL)
 	{
-		Oid			collation = index->rd_indcollation[0];
+		collation = index->rd_indcollation[0];
 
-		for (int i = 0; i < centers->length; i++)
+		for (i = 0; i < centers->length; i++)
 		{
-			double		norm = DatumGetFloat8(FunctionCall1Coll(normprocinfo, collation, PointerGetDatum(VectorArrayGet(centers, i))));
+			norm = DatumGetFloat8(FunctionCall1Coll(normprocinfo, collation, PointerGetDatum(VectorArrayGet(centers, i))));
-
 			if (norm == 0)
 				elog(ERROR, "Zero norm detected. Please report a bug.");
 		}

src/ivfscan.c

@@ -3,14 +3,14 @@
 #include <float.h>
 
 #include "access/relscan.h"
-#include "catalog/pg_operator_d.h"
-#include "catalog/pg_type_d.h"
-#include "lib/pairingheap.h"
 #include "ivfflat.h"
 #include "miscadmin.h"
 #include "pgstat.h"
 #include "storage/bufmgr.h"
 
+#include "catalog/pg_operator_d.h"
+#include "catalog/pg_type_d.h"
+
 /*
  * Compare list distances
  */
@@ -32,36 +32,36 @@ CompareLists(const pairingheap_node *a, const pairingheap_node *b, void *arg)
 static void
 GetScanLists(IndexScanDesc scan, Datum value)
 {
-	IvfflatScanOpaque so = (IvfflatScanOpaque) scan->opaque;
+	Buffer		cbuf;
+	Page		cpage;
+	IvfflatList list;
+	OffsetNumber offno;
+	OffsetNumber maxoffno;
 	BlockNumber nextblkno = IVFFLAT_HEAD_BLKNO;
 	int			listCount = 0;
+	IvfflatScanOpaque so = (IvfflatScanOpaque) scan->opaque;
+	double		distance;
+	IvfflatScanList *scanlist;
 	double		maxDistance = DBL_MAX;
 
 	/* Search all list pages */
 	while (BlockNumberIsValid(nextblkno))
 	{
-		Buffer		cbuf;
-		Page		cpage;
-		OffsetNumber maxoffno;
-
 		cbuf = ReadBuffer(scan->indexRelation, nextblkno);
 		LockBuffer(cbuf, BUFFER_LOCK_SHARE);
 		cpage = BufferGetPage(cbuf);
 
 		maxoffno = PageGetMaxOffsetNumber(cpage);
 
-		for (OffsetNumber offno = FirstOffsetNumber; offno <= maxoffno; offno = OffsetNumberNext(offno))
+		for (offno = FirstOffsetNumber; offno <= maxoffno; offno = OffsetNumberNext(offno))
 		{
-			IvfflatList list = (IvfflatList) PageGetItem(cpage, PageGetItemId(cpage, offno));
+			list = (IvfflatList) PageGetItem(cpage, PageGetItemId(cpage, offno));
-			double		distance;
 
 			/* Use procinfo from the index instead of scan key for performance */
 			distance = DatumGetFloat8(FunctionCall2Coll(so->procinfo, so->collation, PointerGetDatum(&list->center), value));
 
 			if (listCount < so->probes)
 			{
-				IvfflatScanList *scanlist;
-
 				scanlist = &so->lists[listCount];
 				scanlist->startPage = list->startPage;
 				scanlist->distance = distance;
@@ -76,8 +76,6 @@ GetScanLists(IndexScanDesc scan, Datum value)
 			}
 			else if (distance < maxDistance)
 			{
-				IvfflatScanList *scanlist;
-
 				/* Remove */
 				scanlist = (IvfflatScanList *) pairingheap_remove_first(so->listQueue);
 
@@ -104,9 +102,21 @@ static void
 GetScanItems(IndexScanDesc scan, Datum value)
 {
 	IvfflatScanOpaque so = (IvfflatScanOpaque) scan->opaque;
+	Buffer		buf;
+	Page		page;
+	IndexTuple	itup;
+	BlockNumber searchPage;
+	OffsetNumber offno;
+	OffsetNumber maxoffno;
+	Datum		datum;
+	bool		isnull;
 	TupleDesc	tupdesc = RelationGetDescr(scan->indexRelation);
-	double		tuples = 0;
+
+#if PG_VERSION_NUM >= 120000
 	TupleTableSlot *slot = MakeSingleTupleTableSlot(so->tupdesc, &TTSOpsVirtual);
+#else
+	TupleTableSlot *slot = MakeSingleTupleTableSlot(so->tupdesc);
+#endif
 
 	/*
 	 * Reuse same set of shared buffers for scan
@@ -118,28 +128,19 @@ GetScanItems(IndexScanDesc scan, Datum value)
 	/* Search closest probes lists */
 	while (!pairingheap_is_empty(so->listQueue))
 	{
-		BlockNumber searchPage = ((IvfflatScanList *) pairingheap_remove_first(so->listQueue))->startPage;
+		searchPage = ((IvfflatScanList *) pairingheap_remove_first(so->listQueue))->startPage;
 
 		/* Search all entry pages for list */
 		while (BlockNumberIsValid(searchPage))
 		{
-			Buffer		buf;
-			Page		page;
-			OffsetNumber maxoffno;
-
 			buf = ReadBufferExtended(scan->indexRelation, MAIN_FORKNUM, searchPage, RBM_NORMAL, bas);
 			LockBuffer(buf, BUFFER_LOCK_SHARE);
 			page = BufferGetPage(buf);
 			maxoffno = PageGetMaxOffsetNumber(page);
 
-			for (OffsetNumber offno = FirstOffsetNumber; offno <= maxoffno; offno = OffsetNumberNext(offno))
+			for (offno = FirstOffsetNumber; offno <= maxoffno; offno = OffsetNumberNext(offno))
 			{
-				IndexTuple	itup;
+				itup = (IndexTuple) PageGetItem(page, PageGetItemId(page, offno));
-				Datum		datum;
-				bool		isnull;
-				ItemId		itemid = PageGetItemId(page, offno);
-
-				itup = (IndexTuple) PageGetItem(page, itemid);
 				datum = index_getattr(itup, 1, tupdesc, &isnull);
 
 				/*
@@ -153,11 +154,11 @@ GetScanItems(IndexScanDesc scan, Datum value)
 				slot->tts_isnull[0] = false;
 				slot->tts_values[1] = PointerGetDatum(&itup->t_tid);
 				slot->tts_isnull[1] = false;
+				slot->tts_values[2] = Int32GetDatum((int) searchPage);
+				slot->tts_isnull[2] = false;
 				ExecStoreVirtualTuple(slot);
 
 				tuplesort_puttupleslot(so->sortstate, slot);
-
-				tuples++;
 			}
 
 			searchPage = IvfflatPageGetOpaque(page)->nextblkno;
@@ -166,14 +167,6 @@ GetScanItems(IndexScanDesc scan, Datum value)
 		}
 	}
 
-	FreeAccessStrategy(bas);
-
-	if (tuples < 100)
-		ereport(DEBUG1,
-				(errmsg("index scan found few tuples"),
-				 errdetail("Index may have been created with little data."),
-				 errhint("Recreate the index and possibly decrease lists.")));
-
 	tuplesort_performsort(so->sortstate);
 }
 
@@ -186,7 +179,6 @@ ivfflatbeginscan(Relation index, int nkeys, int norderbys)
 	IndexScanDesc scan;
 	IvfflatScanOpaque so;
 	int			lists;
-	int			dimensions;
 	AttrNumber	attNums[] = {1};
 	Oid			sortOperators[] = {Float8LessOperator};
 	Oid			sortCollations[] = {InvalidOid};
@@ -194,17 +186,15 @@ ivfflatbeginscan(Relation index, int nkeys, int norderbys)
 	int			probes = ivfflat_probes;
 
 	scan = RelationGetIndexScan(index, nkeys, norderbys);
-
+	lists = IvfflatGetLists(scan->indexRelation);
-	/* Get lists and dimensions from metapage */
-	IvfflatGetMetaPageInfo(index, &lists, &dimensions);
 
 	if (probes > lists)
 		probes = lists;
 
 	so = (IvfflatScanOpaque) palloc(offsetof(IvfflatScanOpaqueData, lists) + probes * sizeof(IvfflatScanList));
+	so->buf = InvalidBuffer;
 	so->first = true;
 	so->probes = probes;
-	so->dimensions = dimensions;
 
 	/* Set support functions */
 	so->procinfo = index_getprocinfo(index, 1, IVFFLAT_DISTANCE_PROC);
@@ -212,14 +202,23 @@ ivfflatbeginscan(Relation index, int nkeys, int norderbys)
 	so->collation = index->rd_indcollation[0];
 
 	/* Create tuple description for sorting */
-	so->tupdesc = CreateTemplateTupleDesc(2);
+#if PG_VERSION_NUM >= 120000
+	so->tupdesc = CreateTemplateTupleDesc(3);
+#else
+	so->tupdesc = CreateTemplateTupleDesc(3, false);
+#endif
 	TupleDescInitEntry(so->tupdesc, (AttrNumber) 1, "distance", FLOAT8OID, -1, 0);
-	TupleDescInitEntry(so->tupdesc, (AttrNumber) 2, "heaptid", TIDOID, -1, 0);
+	TupleDescInitEntry(so->tupdesc, (AttrNumber) 2, "tid", TIDOID, -1, 0);
+	TupleDescInitEntry(so->tupdesc, (AttrNumber) 3, "indexblkno", INT4OID, -1, 0);
 
 	/* Prep sort */
 	so->sortstate = tuplesort_begin_heap(so->tupdesc, 1, attNums, sortOperators, sortCollations, nullsFirstFlags, work_mem, NULL, false);
 
+#if PG_VERSION_NUM >= 120000
 	so->slot = MakeSingleTupleTableSlot(so->tupdesc, &TTSOpsMinimalTuple);
+#else
+	so->slot = MakeSingleTupleTableSlot(so->tupdesc);
+#endif
 
 	so->listQueue = pairingheap_allocate(CompareLists, scan);
 
@@ -276,24 +275,21 @@ ivfflatgettuple(IndexScanDesc scan, ScanDirection dir)
 		if (scan->orderByData == NULL)
 			elog(ERROR, "cannot scan ivfflat index without order");
 
-		/* Requires MVCC-compliant snapshot as not able to pin during sorting */
+		/* No items will match if null */
-		/* https://www.postgresql.org/docs/current/index-locking.html */
-		if (!IsMVCCSnapshot(scan->xs_snapshot))
-			elog(ERROR, "non-MVCC snapshots are not supported with ivfflat");
-
 		if (scan->orderByData->sk_flags & SK_ISNULL)
-			value = PointerGetDatum(InitVector(so->dimensions));
+			return false;
-		else
+
+		value = scan->orderByData->sk_argument;
+
+		/* Value should not be compressed or toasted */
+		Assert(!VARATT_IS_COMPRESSED(DatumGetPointer(value)));
+		Assert(!VARATT_IS_EXTENDED(DatumGetPointer(value)));
+
+		if (so->normprocinfo != NULL)
 		{
-			value = scan->orderByData->sk_argument;
+			/* No items will match if normalization fails */
-
+			if (!IvfflatNormValue(so->normprocinfo, so->collation, &value, NULL))
-			/* Value should not be compressed or toasted */
+				return false;
-			Assert(!VARATT_IS_COMPRESSED(DatumGetPointer(value)));
-			Assert(!VARATT_IS_EXTENDED(DatumGetPointer(value)));
-
-			/* Fine if normalization fails */
-			if (so->normprocinfo != NULL)
-				IvfflatNormValue(so->normprocinfo, so->collation, &value, NULL);
 		}
 
 		IvfflatBench("GetScanLists", GetScanLists(scan, value));
@@ -307,10 +303,26 @@ ivfflatgettuple(IndexScanDesc scan, ScanDirection dir)
 
 	if (tuplesort_gettupleslot(so->sortstate, true, false, so->slot, NULL))
 	{
-		ItemPointer heaptid = (ItemPointer) DatumGetPointer(slot_getattr(so->slot, 2, &so->isnull));
+		ItemPointer tid = (ItemPointer) DatumGetPointer(slot_getattr(so->slot, 2, &so->isnull));
+		BlockNumber indexblkno = DatumGetInt32(slot_getattr(so->slot, 3, &so->isnull));
+
+#if PG_VERSION_NUM >= 120000
+		scan->xs_heaptid = *tid;
+#else
+		scan->xs_ctup.t_self = *tid;
+#endif
+
+		if (BufferIsValid(so->buf))
+			ReleaseBuffer(so->buf);
+
+		/*
+		 * An index scan must maintain a pin on the index page holding the
+		 * item last returned by amgettuple
+		 *
+		 * https://www.postgresql.org/docs/current/index-locking.html
+		 */
+		so->buf = ReadBuffer(scan->indexRelation, indexblkno);
 
-		scan->xs_heaptid = *heaptid;
-		scan->xs_recheck = false;
 		scan->xs_recheckorderby = false;
 		return true;
 	}
@@ -326,6 +338,10 @@ ivfflatendscan(IndexScanDesc scan)
 {
 	IvfflatScanOpaque so = (IvfflatScanOpaque) scan->opaque;
 
+	/* Release pin */
+	if (BufferIsValid(so->buf))
+		ReleaseBuffer(so->buf);
+
 	pairingheap_free(so->listQueue);
 	tuplesort_end(so->sortstate);
 

src/ivfutils.c

@@ -1,6 +1,5 @@
 #include "postgres.h"
 
-#include "access/generic_xlog.h"
 #include "ivfflat.h"
 #include "storage/bufmgr.h"
 #include "vector.h"
@@ -36,7 +35,9 @@ VectorArrayFree(VectorArray arr)
 void
 PrintVectorArray(char *msg, VectorArray arr)
 {
-	for (int i = 0; i < arr->length; i++)
+	int			i;
+
+	for (i = 0; i < arr->length; i++)
 		PrintVector(msg, VectorArrayGet(arr, i));
 }
 
@@ -58,12 +59,12 @@ IvfflatGetLists(Relation index)
  * Get proc
  */
 FmgrInfo *
-IvfflatOptionalProcInfo(Relation index, uint16 procnum)
+IvfflatOptionalProcInfo(Relation rel, uint16 procnum)
 {
-	if (!OidIsValid(index_getprocid(index, 1, procnum)))
+	if (!OidIsValid(index_getprocid(rel, 1, procnum)))
 		return NULL;
 
-	return index_getprocinfo(index, 1, procnum);
+	return index_getprocinfo(rel, 1, procnum);
 }
 
 /*
@@ -77,16 +78,20 @@ IvfflatOptionalProcInfo(Relation index, uint16 procnum)
 bool
 IvfflatNormValue(FmgrInfo *procinfo, Oid collation, Datum *value, Vector * result)
 {
-	double		norm = DatumGetFloat8(FunctionCall1Coll(procinfo, collation, *value));
+	Vector	   *v;
+	int			i;
+	double		norm;
+
+	norm = DatumGetFloat8(FunctionCall1Coll(procinfo, collation, *value));
 
 	if (norm > 0)
 	{
-		Vector	   *v = DatumGetVector(*value);
+		v = DatumGetVector(*value);
 
 		if (result == NULL)
 			result = InitVector(v->dim);
 
-		for (int i = 0; i < v->dim; i++)
+		for (i = 0; i < v->dim; i++)
 			result->x[i] = v->x[i] / norm;
 
 		*value = PointerGetDatum(result);
@@ -137,6 +142,7 @@ IvfflatInitRegisterPage(Relation index, Buffer *buf, Page *page, GenericXLogStat
 void
 IvfflatCommitBuffer(Buffer buf, GenericXLogState *state)
 {
+	MarkBufferDirty(buf);
 	GenericXLogFinish(state);
 	UnlockReleaseBuffer(buf);
 }
@@ -160,6 +166,8 @@ IvfflatAppendPage(Relation index, Buffer *buf, Page *page, GenericXLogState **st
 	IvfflatInitPage(newbuf, newpage);
 
 	/* Commit */
+	MarkBufferDirty(*buf);
+	MarkBufferDirty(newbuf);
 	GenericXLogFinish(*state);
 
 	/* Unlock */
@@ -170,40 +178,16 @@ IvfflatAppendPage(Relation index, Buffer *buf, Page *page, GenericXLogState **st
 	*buf = newbuf;
 }
 
-/*
- * Get the metapage info
- */
-void
-IvfflatGetMetaPageInfo(Relation index, int *lists, int *dimensions)
-{
-	Buffer		buf;
-	Page		page;
-	IvfflatMetaPage metap;
-
-	buf = ReadBuffer(index, IVFFLAT_METAPAGE_BLKNO);
-	LockBuffer(buf, BUFFER_LOCK_SHARE);
-	page = BufferGetPage(buf);
-	metap = IvfflatPageGetMeta(page);
-
-	*lists = metap->lists;
-
-	if (dimensions != NULL)
-		*dimensions = metap->dimensions;
-
-	UnlockReleaseBuffer(buf);
-}
-
 /*
  * Update the start or insert page of a list
  */
 void
-IvfflatUpdateList(Relation index, ListInfo listInfo,
+IvfflatUpdateList(Relation index, GenericXLogState *state, ListInfo listInfo,
 				  BlockNumber insertPage, BlockNumber originalInsertPage,
 				  BlockNumber startPage, ForkNumber forkNum)
 {
 	Buffer		buf;
 	Page		page;
-	GenericXLogState *state;
 	IvfflatList list;
 	bool		changed = false;
 

src/ivfvacuum.c

@@ -1,6 +1,5 @@
 #include "postgres.h"
 
-#include "access/generic_xlog.h"
 #include "commands/vacuum.h"
 #include "ivfflat.h"
 #include "storage/bufmgr.h"
@@ -13,23 +12,34 @@ ivfflatbulkdelete(IndexVacuumInfo *info, IndexBulkDeleteResult *stats,
 				  IndexBulkDeleteCallback callback, void *callback_state)
 {
 	Relation	index = info->index;
-	BlockNumber blkno = IVFFLAT_HEAD_BLKNO;
+	Buffer		cbuf;
+	Page		cpage;
+	Buffer		buf;
+	Page		page;
+	IvfflatList list;
+	IndexTuple	itup;
+	ItemPointer htup;
+	OffsetNumber deletable[MaxOffsetNumber];
+	int			ndeletable;
+	BlockNumber startPages[MaxOffsetNumber];
+	BlockNumber nextblkno = IVFFLAT_HEAD_BLKNO;
+	BlockNumber searchPage;
+	BlockNumber insertPage;
+	GenericXLogState *state;
+	OffsetNumber coffno;
+	OffsetNumber cmaxoffno;
+	OffsetNumber offno;
+	OffsetNumber maxoffno;
+	ListInfo	listInfo;
 	BufferAccessStrategy bas = GetAccessStrategy(BAS_BULKREAD);
 
 	if (stats == NULL)
 		stats = (IndexBulkDeleteResult *) palloc0(sizeof(IndexBulkDeleteResult));
 
 	/* Iterate over list pages */
-	while (BlockNumberIsValid(blkno))
+	while (BlockNumberIsValid(nextblkno))
 	{
-		Buffer		cbuf;
+		cbuf = ReadBuffer(index, nextblkno);
-		Page		cpage;
-		OffsetNumber coffno;
-		OffsetNumber cmaxoffno;
-		BlockNumber startPages[MaxOffsetNumber];
-		ListInfo	listInfo;
-
-		cbuf = ReadBuffer(index, blkno);
 		LockBuffer(cbuf, BUFFER_LOCK_SHARE);
 		cpage = BufferGetPage(cbuf);
 
@@ -38,32 +48,23 @@ ivfflatbulkdelete(IndexVacuumInfo *info, IndexBulkDeleteResult *stats,
 		/* Iterate over lists */
 		for (coffno = FirstOffsetNumber; coffno <= cmaxoffno; coffno = OffsetNumberNext(coffno))
 		{
-			IvfflatList list = (IvfflatList) PageGetItem(cpage, PageGetItemId(cpage, coffno));
+			list = (IvfflatList) PageGetItem(cpage, PageGetItemId(cpage, coffno));
-
 			startPages[coffno - FirstOffsetNumber] = list->startPage;
 		}
 
-		listInfo.blkno = blkno;
+		listInfo.blkno = nextblkno;
-		blkno = IvfflatPageGetOpaque(cpage)->nextblkno;
+		nextblkno = IvfflatPageGetOpaque(cpage)->nextblkno;
 
 		UnlockReleaseBuffer(cbuf);
 
 		for (coffno = FirstOffsetNumber; coffno <= cmaxoffno; coffno = OffsetNumberNext(coffno))
 		{
-			BlockNumber searchPage = startPages[coffno - FirstOffsetNumber];
+			searchPage = startPages[coffno - FirstOffsetNumber];
-			BlockNumber insertPage = InvalidBlockNumber;
+			insertPage = InvalidBlockNumber;
 
 			/* Iterate over entry pages */
 			while (BlockNumberIsValid(searchPage))
 			{
-				Buffer		buf;
-				Page		page;
-				GenericXLogState *state;
-				OffsetNumber offno;
-				OffsetNumber maxoffno;
-				OffsetNumber deletable[MaxOffsetNumber];
-				int			ndeletable;
-
 				vacuum_delay_point();
 
 				buf = ReadBufferExtended(index, MAIN_FORKNUM, searchPage, RBM_NORMAL, bas);
@@ -85,8 +86,8 @@ ivfflatbulkdelete(IndexVacuumInfo *info, IndexBulkDeleteResult *stats,
 				/* Find deleted tuples */
 				for (offno = FirstOffsetNumber; offno <= maxoffno; offno = OffsetNumberNext(offno))
 				{
-					IndexTuple	itup = (IndexTuple) PageGetItem(page, PageGetItemId(page, offno));
+					itup = (IndexTuple) PageGetItem(page, PageGetItemId(page, offno));
-					ItemPointer htup = &(itup->t_tid);
+					htup = &(itup->t_tid);
 
 					if (callback(htup, callback_state))
 					{
@@ -108,6 +109,7 @@ ivfflatbulkdelete(IndexVacuumInfo *info, IndexBulkDeleteResult *stats,
 				{
 					/* Delete tuples */
 					PageIndexMultiDelete(page, deletable, ndeletable);
+					MarkBufferDirty(buf);
 					GenericXLogFinish(state);
 				}
 				else
@@ -125,13 +127,11 @@ ivfflatbulkdelete(IndexVacuumInfo *info, IndexBulkDeleteResult *stats,
 			if (BlockNumberIsValid(insertPage))
 			{
 				listInfo.offno = coffno;
-				IvfflatUpdateList(index, listInfo, insertPage, InvalidBlockNumber, InvalidBlockNumber, MAIN_FORKNUM);
+				IvfflatUpdateList(index, state, listInfo, insertPage, InvalidBlockNumber, InvalidBlockNumber, MAIN_FORKNUM);
 			}
 		}
 	}
 
-	FreeAccessStrategy(bas);
-
 	return stats;
 }
 

src/vector.h

@@ -1,6 +1,12 @@
 #ifndef VECTOR_H
 #define VECTOR_H
 
+#include "postgres.h"
+
+#if PG_VERSION_NUM >= 160000
+#include "varatt.h"
+#endif
+
 #define VECTOR_MAX_DIM 16000
 
 #define VECTOR_SIZE(_dim)		(offsetof(Vector, x) + sizeof(float)*(_dim))
@@ -16,8 +22,24 @@ typedef struct Vector
 	float		x[FLEXIBLE_ARRAY_MEMBER];
 }			Vector;
 
-Vector	   *InitVector(int dim);
 void		PrintVector(char *msg, Vector * vector);
 int			vector_cmp_internal(Vector * a, Vector * b);
 
+/*
+ * Allocate and initialize a new vector
+ */
+static inline Vector *
+InitVector(int dim)
+{
+	Vector	   *result;
+	int			size;
+
+	size = VECTOR_SIZE(dim);
+	result = (Vector *) palloc0(size);
+	SET_VARSIZE(result, size);
+	result->dim = dim;
+
+	return result;
+}
+
 #endif

test/expected/cast.out

@@ -29,7 +29,7 @@ SELECT ARRAY[1,2,3]::numeric[]::vector;
 (1 row)
 
 SELECT '{NULL}'::real[]::vector;
-ERROR:  array must not contain nulls
+ERROR:  array must not containing NULLs
 SELECT '{NaN}'::real[]::vector;
 ERROR:  NaN not allowed in vector
 SELECT '{Infinity}'::real[]::vector;
@@ -38,8 +38,6 @@ SELECT '{-Infinity}'::real[]::vector;
 ERROR:  infinite value not allowed in vector
 SELECT '{}'::real[]::vector;
 ERROR:  vector must have at least 1 dimension
-SELECT '{{1}}'::real[]::vector;
-ERROR:  array must be 1-D
 SELECT '[1,2,3]'::vector::real[];
  float4  
 ---------
@@ -48,8 +46,6 @@ SELECT '[1,2,3]'::vector::real[];
 
 SELECT array_agg(n)::vector FROM generate_series(1, 16001) n;
 ERROR:  vector cannot have more than 16000 dimensions
-SELECT array_to_vector(array_agg(n), 16001, false) FROM generate_series(1, 16001) n;
-ERROR:  vector cannot have more than 16000 dimensions
 -- ensure no error
 SELECT ARRAY[1,2,3] = ARRAY[1,2,3];
  ?column? 

test/expected/functions.out

@@ -4,182 +4,12 @@ SELECT '[1,2,3]'::vector + '[4,5,6]';
  [5,7,9]
 (1 row)
 
-SELECT '[3e38]'::vector + '[3e38]';
-ERROR:  value out of range: overflow
 SELECT '[1,2,3]'::vector - '[4,5,6]';
   ?column?  
 ------------
  [-3,-3,-3]
 (1 row)
 
-SELECT '[-3e38]'::vector - '[3e38]';
-ERROR:  value out of range: overflow
-SELECT '[1,2,3]'::vector * '[4,5,6]';
- ?column?  
------------
- [4,10,18]
-(1 row)
-
-SELECT '[1e37]'::vector * '[1e37]';
-ERROR:  value out of range: overflow
-SELECT '[1e-37]'::vector * '[1e-37]';
-ERROR:  value out of range: underflow
-SELECT ('[1,2,3]'::vector)[0];
- vector 
---------
-       
-(1 row)
-
-SELECT ('[1,2,3]'::vector)[1];
- vector 
---------
-      1
-(1 row)
-
-SELECT ('[1,2,3]'::vector)[2];
- vector 
---------
-      2
-(1 row)
-
-SELECT ('[1,2,3]'::vector)[3];
- vector 
---------
-      3
-(1 row)
-
-SELECT ('[1,2,3]'::vector)[4];
- vector 
---------
-       
-(1 row)
-
-SELECT ('[1,2,3]'::vector)[1:1];
- vector 
---------
- [1]
-(1 row)
-
-SELECT ('[1,2,3]'::vector)[1:2];
- vector 
---------
- [1,2]
-(1 row)
-
-SELECT ('[1,2,3]'::vector)[2:4];
- vector 
---------
- [2,3]
-(1 row)
-
-SELECT ('[1,2,3]'::vector)[-2:2];
- vector 
---------
- [1,2]
-(1 row)
-
-SELECT ('[1,2,3]'::vector)[2:1];
-ERROR:  vector must have at least 1 dimension
-SELECT ('[1,2,3]'::vector)[:];
- vector  
----------
- [1,2,3]
-(1 row)
-
-SELECT ('[1,2,3]'::vector)[:2];
- vector 
---------
- [1,2]
-(1 row)
-
-SELECT ('[1,2,3]'::vector)[2:];
- vector 
---------
- [2,3]
-(1 row)
-
-SELECT ('[1,2,3]'::vector)[:4];
- vector  
----------
- [1,2,3]
-(1 row)
-
-SELECT ('[1,2,3]'::vector)[-2:];
- vector  
----------
- [1,2,3]
-(1 row)
-
-SELECT ('[1,2,3]'::vector)[NULL];
- vector 
---------
-       
-(1 row)
-
-SELECT ('[1,2,3]'::vector)[NULL:2];
- vector 
---------
- 
-(1 row)
-
-SELECT ('[1,2,3]'::vector)[2:NULL];
- vector 
---------
- 
-(1 row)
-
-SELECT ('[1,2,3]'::vector)[1][1];
-ERROR:  vector allows only one subscript
-SELECT '[1,2,3]'::vector = '[1,2,3]';
- ?column? 
-----------
- t
-(1 row)
-
-SELECT '[1,2,3]'::vector = '[1,2]';
-ERROR:  different vector dimensions 3 and 2
-SELECT vector_cmp('[1,2,3]', '[1,2,3]');
- vector_cmp 
-------------
-          0
-(1 row)
-
-SELECT vector_cmp('[1,2,3]', '[0,0,0]');
- vector_cmp 
-------------
-          1
-(1 row)
-
-SELECT vector_cmp('[0,0,0]', '[1,2,3]');
- vector_cmp 
-------------
-         -1
-(1 row)
-
-SELECT vector_cmp('[1,2]', '[1,2,3]');
- vector_cmp 
-------------
-         -1
-(1 row)
-
-SELECT vector_cmp('[1,2,3]', '[1,2]');
- vector_cmp 
-------------
-          1
-(1 row)
-
-SELECT vector_cmp('[1,2]', '[2,3,4]');
- vector_cmp 
-------------
-         -1
-(1 row)
-
-SELECT vector_cmp('[2,3]', '[1,2,3]');
- vector_cmp 
-------------
-          1
-(1 row)
-
 SELECT vector_dims('[1,2,3]');
  vector_dims 
 -------------
@@ -204,12 +34,6 @@ SELECT vector_norm('[0,1]');
            1
 (1 row)
 
-SELECT vector_norm('[3e37,4e37]')::real;
- vector_norm 
--------------
-       5e+37
-(1 row)
-
 SELECT l2_distance('[0,0]', '[3,4]');
  l2_distance 
 -------------
@@ -224,12 +48,6 @@ SELECT l2_distance('[0,0]', '[0,1]');
 
 SELECT l2_distance('[1,2]', '[3]');
 ERROR:  different vector dimensions 2 and 1
-SELECT l2_distance('[3e38]', '[-3e38]');
- l2_distance 
--------------
-    Infinity
-(1 row)
-
 SELECT inner_product('[1,2]', '[3,4]');
  inner_product 
 ---------------
@@ -238,12 +56,6 @@ SELECT inner_product('[1,2]', '[3,4]');
 
 SELECT inner_product('[1,2]', '[3]');
 ERROR:  different vector dimensions 2 and 1
-SELECT inner_product('[3e38]', '[3e38]');
- inner_product 
----------------
-      Infinity
-(1 row)
-
 SELECT cosine_distance('[1,2]', '[2,4]');
  cosine_distance 
 -----------------
@@ -262,12 +74,6 @@ SELECT cosine_distance('[1,1]', '[1,1]');
                0
 (1 row)
 
-SELECT cosine_distance('[1,0]', '[0,2]');
- cosine_distance 
------------------
-               1
-(1 row)
-
 SELECT cosine_distance('[1,1]', '[-1,-1]');
  cosine_distance 
 -----------------
@@ -276,44 +82,6 @@ SELECT cosine_distance('[1,1]', '[-1,-1]');
 
 SELECT cosine_distance('[1,2]', '[3]');
 ERROR:  different vector dimensions 2 and 1
-SELECT cosine_distance('[1,1]', '[1.1,1.1]');
- cosine_distance 
------------------
-               0
-(1 row)
-
-SELECT cosine_distance('[1,1]', '[-1.1,-1.1]');
- cosine_distance 
------------------
-               2
-(1 row)
-
-SELECT cosine_distance('[3e38]', '[3e38]');
- cosine_distance 
------------------
-             NaN
-(1 row)
-
-SELECT l1_distance('[0,0]', '[3,4]');
- l1_distance 
--------------
-           7
-(1 row)
-
-SELECT l1_distance('[0,0]', '[0,1]');
- l1_distance 
--------------
-           1
-(1 row)
-
-SELECT l1_distance('[1,2]', '[3]');
-ERROR:  different vector dimensions 2 and 1
-SELECT l1_distance('[3e38]', '[-3e38]');
- l1_distance 
--------------
-    Infinity
-(1 row)
-
 SELECT avg(v) FROM unnest(ARRAY['[1,2,3]'::vector, '[3,5,7]']) v;
     avg    
 -----------
@@ -334,33 +102,3 @@ SELECT avg(v) FROM unnest(ARRAY[]::vector[]) v;
 
 SELECT avg(v) FROM unnest(ARRAY['[1,2]'::vector, '[3]']) v;
 ERROR:  expected 2 dimensions, not 1
-SELECT avg(v) FROM unnest(ARRAY['[3e38]'::vector, '[3e38]']) v;
-   avg   
----------
- [3e+38]
-(1 row)
-
-SELECT vector_avg(array_agg(n)) FROM generate_series(1, 16002) n;
-ERROR:  vector cannot have more than 16000 dimensions
-SELECT sum(v) FROM unnest(ARRAY['[1,2,3]'::vector, '[3,5,7]']) v;
-   sum    
-----------
- [4,7,10]
-(1 row)
-
-SELECT sum(v) FROM unnest(ARRAY['[1,2,3]'::vector, '[3,5,7]', NULL]) v;
-   sum    
-----------
- [4,7,10]
-(1 row)
-
-SELECT sum(v) FROM unnest(ARRAY[]::vector[]) v;
- sum 
------
- 
-(1 row)
-
-SELECT sum(v) FROM unnest(ARRAY['[1,2]'::vector, '[3]']) v;
-ERROR:  different vector dimensions 2 and 1
-SELECT sum(v) FROM unnest(ARRAY['[3e38]'::vector, '[3e38]']) v;
-ERROR:  value out of range: overflow

test/expected/hnsw_cosine.out

@@ -1,26 +0,0 @@
-SET enable_seqscan = off;
-CREATE TABLE t (val vector(3));
-INSERT INTO t (val) VALUES ('[0,0,0]'), ('[1,2,3]'), ('[1,1,1]'), (NULL);
-CREATE INDEX ON t USING hnsw (val vector_cosine_ops);
-INSERT INTO t (val) VALUES ('[1,2,4]');
-SELECT * FROM t ORDER BY val <=> '[3,3,3]';
-   val   
----------
- [1,1,1]
- [1,2,3]
- [1,2,4]
-(3 rows)
-
-SELECT COUNT(*) FROM (SELECT * FROM t ORDER BY val <=> '[0,0,0]') t2;
- count 
--------
-     3
-(1 row)
-
-SELECT COUNT(*) FROM (SELECT * FROM t ORDER BY val <=> (SELECT NULL::vector)) t2;
- count 
--------
-     3
-(1 row)
-
-DROP TABLE t;

test/expected/hnsw_ip.out

@@ -1,21 +0,0 @@
-SET enable_seqscan = off;
-CREATE TABLE t (val vector(3));
-INSERT INTO t (val) VALUES ('[0,0,0]'), ('[1,2,3]'), ('[1,1,1]'), (NULL);
-CREATE INDEX ON t USING hnsw (val vector_ip_ops);
-INSERT INTO t (val) VALUES ('[1,2,4]');
-SELECT * FROM t ORDER BY val <#> '[3,3,3]';
-   val   
----------
- [1,2,4]
- [1,2,3]
- [1,1,1]
- [0,0,0]
-(4 rows)
-
-SELECT COUNT(*) FROM (SELECT * FROM t ORDER BY val <#> (SELECT NULL::vector)) t2;
- count 
--------
-     4
-(1 row)
-
-DROP TABLE t;

test/expected/hnsw_l2.out

@@ -1,36 +0,0 @@
-SET enable_seqscan = off;
-CREATE TABLE t (val vector(3));
-INSERT INTO t (val) VALUES ('[0,0,0]'), ('[1,2,3]'), ('[1,1,1]'), (NULL);
-CREATE INDEX ON t USING hnsw (val vector_l2_ops);
-INSERT INTO t (val) VALUES ('[1,2,4]');
-SELECT * FROM t ORDER BY val <-> '[3,3,3]';
-   val   
----------
- [1,2,3]
- [1,2,4]
- [1,1,1]
- [0,0,0]
-(4 rows)
-
-SELECT * FROM t ORDER BY val <-> (SELECT NULL::vector);
-   val   
----------
- [0,0,0]
- [1,1,1]
- [1,2,3]
- [1,2,4]
-(4 rows)
-
-SELECT COUNT(*) FROM t;
- count 
--------
-     5
-(1 row)
-
-TRUNCATE t;
-SELECT * FROM t ORDER BY val <-> '[3,3,3]';
- val 
------
-(0 rows)
-
-DROP TABLE t;

test/expected/hnsw_options.out

@@ -1,26 +0,0 @@
-CREATE TABLE t (val vector(3));
-CREATE INDEX ON t USING hnsw (val vector_l2_ops) WITH (m = 1);
-ERROR:  value 1 out of bounds for option "m"
-DETAIL:  Valid values are between "2" and "100".
-CREATE INDEX ON t USING hnsw (val vector_l2_ops) WITH (m = 101);
-ERROR:  value 101 out of bounds for option "m"
-DETAIL:  Valid values are between "2" and "100".
-CREATE INDEX ON t USING hnsw (val vector_l2_ops) WITH (ef_construction = 3);
-ERROR:  value 3 out of bounds for option "ef_construction"
-DETAIL:  Valid values are between "4" and "1000".
-CREATE INDEX ON t USING hnsw (val vector_l2_ops) WITH (ef_construction = 1001);
-ERROR:  value 1001 out of bounds for option "ef_construction"
-DETAIL:  Valid values are between "4" and "1000".
-CREATE INDEX ON t USING hnsw (val vector_l2_ops) WITH (m = 16, ef_construction = 31);
-ERROR:  ef_construction must be greater than or equal to 2 * m
-SHOW hnsw.ef_search;
- hnsw.ef_search 
-----------------
- 40
-(1 row)
-
-SET hnsw.ef_search = 0;
-ERROR:  0 is outside the valid range for parameter "hnsw.ef_search" (1 .. 1000)
-SET hnsw.ef_search = 1001;
-ERROR:  1001 is outside the valid range for parameter "hnsw.ef_search" (1 .. 1000)
-DROP TABLE t;

test/expected/hnsw_unlogged.out

@@ -1,13 +0,0 @@
-SET enable_seqscan = off;
-CREATE UNLOGGED TABLE t (val vector(3));
-INSERT INTO t (val) VALUES ('[0,0,0]'), ('[1,2,3]'), ('[1,1,1]'), (NULL);
-CREATE INDEX ON t USING hnsw (val vector_l2_ops);
-SELECT * FROM t ORDER BY val <-> '[3,3,3]';
-   val   
----------
- [1,2,3]
- [1,1,1]
- [0,0,0]
-(3 rows)
-
-DROP TABLE t;

test/expected/input.out

@@ -4,22 +4,10 @@ SELECT '[1,2,3]'::vector;
  [1,2,3]
 (1 row)
 
-SELECT '[-1,-2,-3]'::vector;
+SELECT '[-1,2,3]'::vector;
-   vector   
+  vector  
-------------
+----------
- [-1,-2,-3]
+ [-1,2,3]
-(1 row)
-
-SELECT '[1.,2.,3.]'::vector;
- vector  
----------
- [1,2,3]
-(1 row)
-
-SELECT ' [ 1,  2 ,    3  ] '::vector;
- vector  
----------
- [1,2,3]
 (1 row)
 
 SELECT '[1.23456]'::vector;
@@ -29,7 +17,7 @@ SELECT '[1.23456]'::vector;
 (1 row)
 
 SELECT '[hello,1]'::vector;
-ERROR:  invalid input syntax for type vector: "[hello,1]"
+ERROR:  invalid input syntax for type vector: "hello"
 LINE 1: SELECT '[hello,1]'::vector;
                ^
 SELECT '[NaN,1]'::vector;
@@ -44,35 +32,13 @@ SELECT '[-Infinity,1]'::vector;
 ERROR:  infinite value not allowed in vector
 LINE 1: SELECT '[-Infinity,1]'::vector;
                ^
-SELECT '[1.5e38,-1.5e38]'::vector;
-       vector       
---------------------
- [1.5e+38,-1.5e+38]
-(1 row)
-
-SELECT '[1.5e+38,-1.5e+38]'::vector;
-       vector       
---------------------
- [1.5e+38,-1.5e+38]
-(1 row)
-
-SELECT '[1.5e-38,-1.5e-38]'::vector;
-       vector       
---------------------
- [1.5e-38,-1.5e-38]
-(1 row)
-
-SELECT '[4e38,1]'::vector;
-ERROR:  infinite value not allowed in vector
-LINE 1: SELECT '[4e38,1]'::vector;
-               ^
 SELECT '[1,2,3'::vector;
-ERROR:  malformed vector literal: "[1,2,3"
+ERROR:  malformed vector literal
 LINE 1: SELECT '[1,2,3'::vector;
                ^
 DETAIL:  Unexpected end of input.
 SELECT '[1,2,3]9'::vector;
-ERROR:  malformed vector literal: "[1,2,3]9"
+ERROR:  malformed vector literal
 LINE 1: SELECT '[1,2,3]9'::vector;
                ^
 DETAIL:  Junk after closing right brace.
@@ -81,41 +47,14 @@ ERROR:  malformed vector literal: "1,2,3"
 LINE 1: SELECT '1,2,3'::vector;
                ^
 DETAIL:  Vector contents must start with "[".
-SELECT ''::vector;
-ERROR:  malformed vector literal: ""
-LINE 1: SELECT ''::vector;
-               ^
-DETAIL:  Vector contents must start with "[".
-SELECT '['::vector;
-ERROR:  malformed vector literal: "["
-LINE 1: SELECT '['::vector;
-               ^
-DETAIL:  Unexpected end of input.
-SELECT '[,'::vector;
-ERROR:  malformed vector literal: "[,"
-LINE 1: SELECT '[,'::vector;
-               ^
-DETAIL:  Unexpected end of input.
 SELECT '[]'::vector;
 ERROR:  vector must have at least 1 dimension
 LINE 1: SELECT '[]'::vector;
                ^
 SELECT '[1,]'::vector;
-ERROR:  invalid input syntax for type vector: "[1,]"
+ERROR:  invalid input syntax for type vector: "]"
 LINE 1: SELECT '[1,]'::vector;
                ^
-SELECT '[1a]'::vector;
-ERROR:  invalid input syntax for type vector: "[1a]"
-LINE 1: SELECT '[1a]'::vector;
-               ^
-SELECT '[1,,3]'::vector;
-ERROR:  malformed vector literal: "[1,,3]"
-LINE 1: SELECT '[1,,3]'::vector;
-               ^
-SELECT '[1, ,3]'::vector;
-ERROR:  invalid input syntax for type vector: "[1, ,3]"
-LINE 1: SELECT '[1, ,3]'::vector;
-               ^
 SELECT '[1,2,3]'::vector(2);
 ERROR:  expected 2 dimensions, not 3
 SELECT unnest('{"[1,2,3]", "[4,5,6]"}'::vector[]);

test/expected/ivfflat_cosine.out

@@ -11,16 +11,9 @@ SELECT * FROM t ORDER BY val <=> '[3,3,3]';
  [1,2,4]
 (3 rows)
 
-SELECT COUNT(*) FROM (SELECT * FROM t ORDER BY val <=> '[0,0,0]') t2;
+SELECT * FROM t ORDER BY val <=> (SELECT NULL::vector);
- count 
+ val 
--------
+-----
-     3
+(0 rows)
-(1 row)
-
-SELECT COUNT(*) FROM (SELECT * FROM t ORDER BY val <=> (SELECT NULL::vector)) t2;
- count 
--------
-     3
-(1 row)
 
 DROP TABLE t;

test/expected/ivfflat_ip.out

@@ -12,10 +12,9 @@ SELECT * FROM t ORDER BY val <#> '[3,3,3]';
  [0,0,0]
 (4 rows)
 
-SELECT COUNT(*) FROM (SELECT * FROM t ORDER BY val <#> (SELECT NULL::vector)) t2;
+SELECT * FROM t ORDER BY val <#> (SELECT NULL::vector);
- count 
+ val 
--------
+-----
-     4
+(0 rows)
-(1 row)
 
 DROP TABLE t;

test/expected/ivfflat_l2.out

@@ -1,7 +1,7 @@
 SET enable_seqscan = off;
 CREATE TABLE t (val vector(3));
 INSERT INTO t (val) VALUES ('[0,0,0]'), ('[1,2,3]'), ('[1,1,1]'), (NULL);
-CREATE INDEX ON t USING ivfflat (val vector_l2_ops) WITH (lists = 1);
+CREATE INDEX ON t USING ivfflat (val) WITH (lists = 1);
 INSERT INTO t (val) VALUES ('[1,2,4]');
 SELECT * FROM t ORDER BY val <-> '[3,3,3]';
    val   
@@ -13,13 +13,9 @@ SELECT * FROM t ORDER BY val <-> '[3,3,3]';
 (4 rows)
 
 SELECT * FROM t ORDER BY val <-> (SELECT NULL::vector);
-   val   
+ val 
----------
+-----
- [0,0,0]
+(0 rows)
- [1,1,1]
- [1,2,3]
- [1,2,4]
-(4 rows)
 
 SELECT COUNT(*) FROM t;
  count 
@@ -27,13 +23,4 @@ SELECT COUNT(*) FROM t;
      5
 (1 row)
 
-TRUNCATE t;
-NOTICE:  ivfflat index created with little data
-DETAIL:  This will cause low recall.
-HINT:  Drop the index until the table has more data.
-SELECT * FROM t ORDER BY val <-> '[3,3,3]';
- val 
------
-(0 rows)
-
 DROP TABLE t;

test/expected/ivfflat_options.out

@@ -1,8 +1,9 @@
+SET enable_seqscan = off;
 CREATE TABLE t (val vector(3));
-CREATE INDEX ON t USING ivfflat (val vector_l2_ops) WITH (lists = 0);
+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".
-CREATE INDEX ON t USING ivfflat (val vector_l2_ops) WITH (lists = 32769);
+CREATE INDEX ON t USING ivfflat (val) WITH (lists = 32769);
 ERROR:  value 32769 out of bounds for option "lists"
 DETAIL:  Valid values are between "1" and "32768".
 SHOW ivfflat.probes;

test/expected/ivfflat_unlogged.out

@@ -1,7 +1,7 @@
 SET enable_seqscan = off;
 CREATE UNLOGGED TABLE t (val vector(3));
 INSERT INTO t (val) VALUES ('[0,0,0]'), ('[1,2,3]'), ('[1,1,1]'), (NULL);
-CREATE INDEX ON t USING ivfflat (val vector_l2_ops) WITH (lists = 1);
+CREATE INDEX ON t USING ivfflat (val) WITH (lists = 1);
 SELECT * FROM t ORDER BY val <-> '[3,3,3]';
    val   
 ---------

test/sql/cast.sql

@@ -8,10 +8,8 @@ SELECT '{NaN}'::real[]::vector;
 SELECT '{Infinity}'::real[]::vector;
 SELECT '{-Infinity}'::real[]::vector;
 SELECT '{}'::real[]::vector;
-SELECT '{{1}}'::real[]::vector;
 SELECT '[1,2,3]'::vector::real[];
 SELECT array_agg(n)::vector FROM generate_series(1, 16001) n;
-SELECT array_to_vector(array_agg(n), 16001, false) FROM generate_series(1, 16001) n;
 
 -- ensure no error
 SELECT ARRAY[1,2,3] = ARRAY[1,2,3];

test/sql/functions.sql

@@ -1,82 +1,26 @@
 SELECT '[1,2,3]'::vector + '[4,5,6]';
-SELECT '[3e38]'::vector + '[3e38]';
 SELECT '[1,2,3]'::vector - '[4,5,6]';
-SELECT '[-3e38]'::vector - '[3e38]';
-SELECT '[1,2,3]'::vector * '[4,5,6]';
-SELECT '[1e37]'::vector * '[1e37]';
-SELECT '[1e-37]'::vector * '[1e-37]';
-
-SELECT ('[1,2,3]'::vector)[0];
-SELECT ('[1,2,3]'::vector)[1];
-SELECT ('[1,2,3]'::vector)[2];
-SELECT ('[1,2,3]'::vector)[3];
-SELECT ('[1,2,3]'::vector)[4];
-SELECT ('[1,2,3]'::vector)[1:1];
-SELECT ('[1,2,3]'::vector)[1:2];
-SELECT ('[1,2,3]'::vector)[2:4];
-SELECT ('[1,2,3]'::vector)[-2:2];
-SELECT ('[1,2,3]'::vector)[2:1];
-SELECT ('[1,2,3]'::vector)[:];
-SELECT ('[1,2,3]'::vector)[:2];
-SELECT ('[1,2,3]'::vector)[2:];
-SELECT ('[1,2,3]'::vector)[:4];
-SELECT ('[1,2,3]'::vector)[-2:];
-SELECT ('[1,2,3]'::vector)[NULL];
-SELECT ('[1,2,3]'::vector)[NULL:2];
-SELECT ('[1,2,3]'::vector)[2:NULL];
-SELECT ('[1,2,3]'::vector)[1][1];
-
-SELECT '[1,2,3]'::vector = '[1,2,3]';
-SELECT '[1,2,3]'::vector = '[1,2]';
-
-SELECT vector_cmp('[1,2,3]', '[1,2,3]');
-SELECT vector_cmp('[1,2,3]', '[0,0,0]');
-SELECT vector_cmp('[0,0,0]', '[1,2,3]');
-SELECT vector_cmp('[1,2]', '[1,2,3]');
-SELECT vector_cmp('[1,2,3]', '[1,2]');
-SELECT vector_cmp('[1,2]', '[2,3,4]');
-SELECT vector_cmp('[2,3]', '[1,2,3]');
 
 SELECT vector_dims('[1,2,3]');
 
 SELECT round(vector_norm('[1,1]')::numeric, 5);
 SELECT vector_norm('[3,4]');
 SELECT vector_norm('[0,1]');
-SELECT vector_norm('[3e37,4e37]')::real;
 
 SELECT l2_distance('[0,0]', '[3,4]');
 SELECT l2_distance('[0,0]', '[0,1]');
 SELECT l2_distance('[1,2]', '[3]');
-SELECT l2_distance('[3e38]', '[-3e38]');
 
 SELECT inner_product('[1,2]', '[3,4]');
 SELECT inner_product('[1,2]', '[3]');
-SELECT inner_product('[3e38]', '[3e38]');
 
 SELECT cosine_distance('[1,2]', '[2,4]');
 SELECT cosine_distance('[1,2]', '[0,0]');
 SELECT cosine_distance('[1,1]', '[1,1]');
-SELECT cosine_distance('[1,0]', '[0,2]');
 SELECT cosine_distance('[1,1]', '[-1,-1]');
 SELECT cosine_distance('[1,2]', '[3]');
-SELECT cosine_distance('[1,1]', '[1.1,1.1]');
-SELECT cosine_distance('[1,1]', '[-1.1,-1.1]');
-SELECT cosine_distance('[3e38]', '[3e38]');
-
-SELECT l1_distance('[0,0]', '[3,4]');
-SELECT l1_distance('[0,0]', '[0,1]');
-SELECT l1_distance('[1,2]', '[3]');
-SELECT l1_distance('[3e38]', '[-3e38]');
 
 SELECT avg(v) FROM unnest(ARRAY['[1,2,3]'::vector, '[3,5,7]']) v;
 SELECT avg(v) FROM unnest(ARRAY['[1,2,3]'::vector, '[3,5,7]', NULL]) v;
 SELECT avg(v) FROM unnest(ARRAY[]::vector[]) v;
 SELECT avg(v) FROM unnest(ARRAY['[1,2]'::vector, '[3]']) v;
-SELECT avg(v) FROM unnest(ARRAY['[3e38]'::vector, '[3e38]']) v;
-SELECT vector_avg(array_agg(n)) FROM generate_series(1, 16002) n;
-
-SELECT sum(v) FROM unnest(ARRAY['[1,2,3]'::vector, '[3,5,7]']) v;
-SELECT sum(v) FROM unnest(ARRAY['[1,2,3]'::vector, '[3,5,7]', NULL]) v;
-SELECT sum(v) FROM unnest(ARRAY[]::vector[]) v;
-SELECT sum(v) FROM unnest(ARRAY['[1,2]'::vector, '[3]']) v;
-SELECT sum(v) FROM unnest(ARRAY['[3e38]'::vector, '[3e38]']) v;

test/sql/hnsw_cosine.sql

@@ -1,13 +0,0 @@
-SET enable_seqscan = off;
-
-CREATE TABLE t (val vector(3));
-INSERT INTO t (val) VALUES ('[0,0,0]'), ('[1,2,3]'), ('[1,1,1]'), (NULL);
-CREATE INDEX ON t USING hnsw (val vector_cosine_ops);
-
-INSERT INTO t (val) VALUES ('[1,2,4]');
-
-SELECT * FROM t ORDER BY val <=> '[3,3,3]';
-SELECT COUNT(*) FROM (SELECT * FROM t ORDER BY val <=> '[0,0,0]') t2;
-SELECT COUNT(*) FROM (SELECT * FROM t ORDER BY val <=> (SELECT NULL::vector)) t2;
-
-DROP TABLE t;

test/sql/hnsw_ip.sql

@@ -1,12 +0,0 @@
-SET enable_seqscan = off;
-
-CREATE TABLE t (val vector(3));
-INSERT INTO t (val) VALUES ('[0,0,0]'), ('[1,2,3]'), ('[1,1,1]'), (NULL);
-CREATE INDEX ON t USING hnsw (val vector_ip_ops);
-
-INSERT INTO t (val) VALUES ('[1,2,4]');
-
-SELECT * FROM t ORDER BY val <#> '[3,3,3]';
-SELECT COUNT(*) FROM (SELECT * FROM t ORDER BY val <#> (SELECT NULL::vector)) t2;
-
-DROP TABLE t;

test/sql/hnsw_l2.sql

@@ -1,16 +0,0 @@
-SET enable_seqscan = off;
-
-CREATE TABLE t (val vector(3));
-INSERT INTO t (val) VALUES ('[0,0,0]'), ('[1,2,3]'), ('[1,1,1]'), (NULL);
-CREATE INDEX ON t USING hnsw (val vector_l2_ops);
-
-INSERT INTO t (val) VALUES ('[1,2,4]');
-
-SELECT * FROM t ORDER BY val <-> '[3,3,3]';
-SELECT * FROM t ORDER BY val <-> (SELECT NULL::vector);
-SELECT COUNT(*) FROM t;
-
-TRUNCATE t;
-SELECT * FROM t ORDER BY val <-> '[3,3,3]';
-
-DROP TABLE t;

test/sql/hnsw_options.sql

@@ -1,13 +0,0 @@
-CREATE TABLE t (val vector(3));
-CREATE INDEX ON t USING hnsw (val vector_l2_ops) WITH (m = 1);
-CREATE INDEX ON t USING hnsw (val vector_l2_ops) WITH (m = 101);
-CREATE INDEX ON t USING hnsw (val vector_l2_ops) WITH (ef_construction = 3);
-CREATE INDEX ON t USING hnsw (val vector_l2_ops) WITH (ef_construction = 1001);
-CREATE INDEX ON t USING hnsw (val vector_l2_ops) WITH (m = 16, ef_construction = 31);
-
-SHOW hnsw.ef_search;
-
-SET hnsw.ef_search = 0;
-SET hnsw.ef_search = 1001;
-
-DROP TABLE t;

test/sql/hnsw_unlogged.sql

@@ -1,9 +0,0 @@
-SET enable_seqscan = off;
-
-CREATE UNLOGGED TABLE t (val vector(3));
-INSERT INTO t (val) VALUES ('[0,0,0]'), ('[1,2,3]'), ('[1,1,1]'), (NULL);
-CREATE INDEX ON t USING hnsw (val vector_l2_ops);
-
-SELECT * FROM t ORDER BY val <-> '[3,3,3]';
-
-DROP TABLE t;

test/sql/input.sql

@@ -1,27 +1,15 @@
 SELECT '[1,2,3]'::vector;
-SELECT '[-1,-2,-3]'::vector;
+SELECT '[-1,2,3]'::vector;
-SELECT '[1.,2.,3.]'::vector;
-SELECT ' [ 1,  2 ,    3  ] '::vector;
 SELECT '[1.23456]'::vector;
 SELECT '[hello,1]'::vector;
 SELECT '[NaN,1]'::vector;
 SELECT '[Infinity,1]'::vector;
 SELECT '[-Infinity,1]'::vector;
-SELECT '[1.5e38,-1.5e38]'::vector;
-SELECT '[1.5e+38,-1.5e+38]'::vector;
-SELECT '[1.5e-38,-1.5e-38]'::vector;
-SELECT '[4e38,1]'::vector;
 SELECT '[1,2,3'::vector;
 SELECT '[1,2,3]9'::vector;
 SELECT '1,2,3'::vector;
-SELECT ''::vector;
-SELECT '['::vector;
-SELECT '[,'::vector;
 SELECT '[]'::vector;
 SELECT '[1,]'::vector;
-SELECT '[1a]'::vector;
-SELECT '[1,,3]'::vector;
-SELECT '[1, ,3]'::vector;
 SELECT '[1,2,3]'::vector(2);
 
 SELECT unnest('{"[1,2,3]", "[4,5,6]"}'::vector[]);

test/sql/ivfflat_cosine.sql

@@ -7,7 +7,6 @@ CREATE INDEX ON t USING ivfflat (val vector_cosine_ops) WITH (lists = 1);
 INSERT INTO t (val) VALUES ('[1,2,4]');
 
 SELECT * FROM t ORDER BY val <=> '[3,3,3]';
-SELECT COUNT(*) FROM (SELECT * FROM t ORDER BY val <=> '[0,0,0]') t2;
+SELECT * FROM t ORDER BY val <=> (SELECT NULL::vector);
-SELECT COUNT(*) FROM (SELECT * FROM t ORDER BY val <=> (SELECT NULL::vector)) t2;
 
 DROP TABLE t;

test/sql/ivfflat_ip.sql

@@ -7,6 +7,6 @@ CREATE INDEX ON t USING ivfflat (val vector_ip_ops) WITH (lists = 1);
 INSERT INTO t (val) VALUES ('[1,2,4]');
 
 SELECT * FROM t ORDER BY val <#> '[3,3,3]';
-SELECT COUNT(*) FROM (SELECT * FROM t ORDER BY val <#> (SELECT NULL::vector)) t2;
+SELECT * FROM t ORDER BY val <#> (SELECT NULL::vector);
 
 DROP TABLE t;

test/sql/ivfflat_l2.sql

@@ -2,7 +2,7 @@ SET enable_seqscan = off;
 
 CREATE TABLE t (val vector(3));
 INSERT INTO t (val) VALUES ('[0,0,0]'), ('[1,2,3]'), ('[1,1,1]'), (NULL);
-CREATE INDEX ON t USING ivfflat (val vector_l2_ops) WITH (lists = 1);
+CREATE INDEX ON t USING ivfflat (val) WITH (lists = 1);
 
 INSERT INTO t (val) VALUES ('[1,2,4]');
 
@@ -10,7 +10,4 @@ SELECT * FROM t ORDER BY val <-> '[3,3,3]';
 SELECT * FROM t ORDER BY val <-> (SELECT NULL::vector);
 SELECT COUNT(*) FROM t;
 
-TRUNCATE t;
-SELECT * FROM t ORDER BY val <-> '[3,3,3]';
-
 DROP TABLE t;

test/sql/ivfflat_options.sql

@@ -1,6 +1,8 @@
+SET enable_seqscan = off;
+
 CREATE TABLE t (val vector(3));
-CREATE INDEX ON t USING ivfflat (val vector_l2_ops) WITH (lists = 0);
+CREATE INDEX ON t USING ivfflat (val) WITH (lists = 0);
-CREATE INDEX ON t USING ivfflat (val vector_l2_ops) WITH (lists = 32769);
+CREATE INDEX ON t USING ivfflat (val) WITH (lists = 32769);
 
 SHOW ivfflat.probes;
 

test/sql/ivfflat_parallel.sql

@@ -0,0 +1,15 @@
+-- SET force_parallel_mode = on;
+SET parallel_setup_cost = 10;
+SET parallel_tuple_cost = 0.001;
+SET min_parallel_table_scan_size = 1;
+SET min_parallel_index_scan_size = 1;
+
+CREATE TABLE t (id integer, val vector(3));
+INSERT INTO t (id, val) SELECT n, ARRAY[random(), random(), random()] FROM generate_series(1,1000000) n;
+CREATE INDEX ON t USING ivfflat (val) WITH (lists = 10);
+SET ivfflat.probes = 2;
+
+EXPLAIN SELECT * FROM t ORDER BY val <-> '[0.5,0.5,0.5]' LIMIT 5;
+SELECT * FROM t ORDER BY val <-> '[0.5,0.5,0.5]' LIMIT 5;
+
+DROP TABLE t;

test/sql/ivfflat_unlogged.sql

@@ -2,7 +2,7 @@ SET enable_seqscan = off;
 
 CREATE UNLOGGED TABLE t (val vector(3));
 INSERT INTO t (val) VALUES ('[0,0,0]'), ('[1,2,3]'), ('[1,1,1]'), (NULL);
-CREATE INDEX ON t USING ivfflat (val vector_l2_ops) WITH (lists = 1);
+CREATE INDEX ON t USING ivfflat (val) WITH (lists = 1);
 
 SELECT * FROM t ORDER BY val <-> '[3,3,3]';
 

test/t/001_wal.pl

@@ -5,7 +5,7 @@ use strict;
 use warnings;
 use PostgresNode;
 use TestLib;
-use Test::More;
+use Test::More tests => 31;
 
 my $dim = 32;
 
@@ -19,13 +19,14 @@ sub test_index_replay
 
 	# Wait for replica to catch up
 	my $applname = $node_replica->name;
+
+	my $server_version_num = $node_primary->safe_psql("postgres", "SHOW server_version_num");
 	my $caughtup_query = "SELECT pg_current_wal_lsn() <= replay_lsn FROM pg_stat_replication WHERE application_name = '$applname';";
 	$node_primary->poll_query_until('postgres', $caughtup_query)
 	  or die "Timed out while waiting for replica 1 to catch up";
 
 	my @r = ();
-	for (1 .. $dim)
+	for (1 .. $dim) {
-	{
 		push(@r, rand());
 	}
 	my $sql = join(",", @r);
@@ -51,13 +52,11 @@ my $array_sql = join(",", ('random()') x $dim);
 # Initialize primary node
 $node_primary = get_new_node('primary');
 $node_primary->init(allows_streaming => 1);
-if ($dim > 32)
+if ($dim > 32) {
-{
 	# TODO use wal_keep_segments for Postgres < 13
 	$node_primary->append_conf('postgresql.conf', qq(wal_keep_size = 1GB));
 }
-if ($dim > 1500)
+if ($dim > 1500) {
-{
 	$node_primary->append_conf('postgresql.conf', qq(maintenance_work_mem = 128MB));
 }
 $node_primary->start;
@@ -68,7 +67,8 @@ $node_primary->backup($backup_name);
 
 # Create streaming replica linking to primary
 $node_replica = get_new_node('replica');
-$node_replica->init_from_backup($node_primary, $backup_name, has_streaming => 1);
+$node_replica->init_from_backup($node_primary, $backup_name,
+	has_streaming => 1);
 $node_replica->start;
 
 # Create ivfflat index on primary
@@ -77,7 +77,7 @@ $node_primary->safe_psql("postgres", "CREATE TABLE tst (i int4, v vector($dim));
 $node_primary->safe_psql("postgres",
 	"INSERT INTO tst SELECT i % 10, ARRAY[$array_sql] FROM generate_series(1, 100000) i;"
 );
-$node_primary->safe_psql("postgres", "CREATE INDEX ON tst USING ivfflat (v vector_l2_ops);");
+$node_primary->safe_psql("postgres", "CREATE INDEX ON tst USING ivfflat (v);");
 
 # Test that queries give same result
 test_index_replay('initial');
@@ -95,5 +95,3 @@ for my $i (1 .. 10)
 	);
 	test_index_replay("insert $i");
 }
-
-done_testing();

test/t/002_vacuum.pl

@@ -2,13 +2,12 @@ use strict;
 use warnings;
 use PostgresNode;
 use TestLib;
-use Test::More;
+use Test::More tests => 1;
 
 my $dim = 3;
 
 my @r = ();
-for (1 .. $dim)
+for (1 .. $dim) {
-{
 	my $v = int(rand(1000)) + 1;
 	push(@r, "i % $v");
 }
@@ -25,7 +24,7 @@ $node->safe_psql("postgres", "CREATE TABLE tst (i int4, v vector($dim));");
 $node->safe_psql("postgres",
 	"INSERT INTO tst SELECT i % 10, ARRAY[$array_sql] FROM generate_series(1, 100000) i;"
 );
-$node->safe_psql("postgres", "CREATE INDEX ON tst USING ivfflat (v vector_l2_ops);");
+$node->safe_psql("postgres", "CREATE INDEX ON tst USING ivfflat (v);");
 
 # Get size
 my $size = $node->safe_psql("postgres", "SELECT pg_total_relation_size('tst_v_idx');");
@@ -40,5 +39,3 @@ $node->safe_psql("postgres",
 # Check size
 my $new_size = $node->safe_psql("postgres", "SELECT pg_total_relation_size('tst_v_idx');");
 is($size, $new_size, "size does not change");
-
-done_testing();

test/t/003_ivfflat_build_recall.pl

@@ -1,128 +0,0 @@
-use strict;
-use warnings;
-use PostgresNode;
-use TestLib;
-use Test::More;
-
-my $node;
-my @queries = ();
-my @expected;
-my $limit = 20;
-
-sub test_recall
-{
-	my ($probes, $min, $operator) = @_;
-	my $correct = 0;
-	my $total = 0;
-
-	my $explain = $node->safe_psql("postgres", qq(
-		SET enable_seqscan = off;
-		SET ivfflat.probes = $probes;
-		EXPLAIN ANALYZE SELECT i FROM tst ORDER BY v $operator '$queries[0]' LIMIT $limit;
-	));
-	like($explain, qr/Index Scan using idx on tst/);
-
-	for my $i (0 .. $#queries)
-	{
-		my $actual = $node->safe_psql("postgres", qq(
-			SET enable_seqscan = off;
-			SET ivfflat.probes = $probes;
-			SELECT i FROM tst ORDER BY v $operator '$queries[$i]' LIMIT $limit;
-		));
-		my @actual_ids = split("\n", $actual);
-		my %actual_set = map { $_ => 1 } @actual_ids;
-
-		my @expected_ids = split("\n", $expected[$i]);
-
-		foreach (@expected_ids)
-		{
-			if (exists($actual_set{$_}))
-			{
-				$correct++;
-			}
-			$total++;
-		}
-	}
-
-	cmp_ok($correct / $total, ">=", $min, $operator);
-}
-
-# Initialize node
-$node = get_new_node('node');
-$node->init;
-$node->start;
-
-# Create table
-$node->safe_psql("postgres", "CREATE EXTENSION vector;");
-$node->safe_psql("postgres", "CREATE TABLE tst (i int4, v vector(3));");
-$node->safe_psql("postgres",
-	"INSERT INTO tst SELECT i, ARRAY[random(), random(), random()] FROM generate_series(1, 100000) i;"
-);
-
-# Generate queries
-for (1 .. 20)
-{
-	my $r1 = rand();
-	my $r2 = rand();
-	my $r3 = rand();
-	push(@queries, "[$r1,$r2,$r3]");
-}
-
-# 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];
-
-	# Get exact results
-	@expected = ();
-	foreach (@queries)
-	{
-		my $res = $node->safe_psql("postgres", "SELECT i FROM tst ORDER BY v $operator '$_' LIMIT $limit;");
-		push(@expected, $res);
-	}
-
-	# Build index serially
-	$node->safe_psql("postgres", qq(
-		SET max_parallel_maintenance_workers = 0;
-		CREATE INDEX idx ON tst USING ivfflat (v $opclass);
-	));
-
-	# Test approximate results
-	if ($operator ne "<#>")
-	{
-		# TODO Fix test (uniform random vectors all have similar inner product)
-		test_recall(1, 0.71, $operator);
-		test_recall(10, 0.95, $operator);
-	}
-	# Account for equal distances
-	test_recall(100, 0.9925, $operator);
-
-	$node->safe_psql("postgres", "DROP INDEX idx;");
-
-	# Build index in parallel
-	my ($ret, $stdout, $stderr) = $node->psql("postgres", qq(
-		SET client_min_messages = DEBUG;
-		SET min_parallel_table_scan_size = 1;
-		CREATE INDEX idx ON tst USING ivfflat (v $opclass);
-	));
-	is($ret, 0, $stderr);
-	like($stderr, qr/using \d+ parallel workers/);
-
-	# Test approximate results
-	if ($operator ne "<#>")
-	{
-		# TODO Fix test (uniform random vectors all have similar inner product)
-		test_recall(1, 0.71, $operator);
-		test_recall(10, 0.95, $operator);
-	}
-	# Account for equal distances
-	test_recall(100, 0.9925, $operator);
-
-	$node->safe_psql("postgres", "DROP INDEX idx;");
-}
-
-done_testing();

test/t/003_recall.pl

@@ -0,0 +1,88 @@
+use strict;
+use warnings;
+use PostgresNode;
+use TestLib;
+use Test::More tests => 9;
+
+my $node;
+my @queries = ();
+my @expected;
+my $limit = 20;
+
+sub test_recall
+{
+	my ($probes, $min, $operator) = @_;
+	my $correct = 0;
+	my $total = 0;
+
+	for my $i (0 .. $#queries) {
+		my $actual = $node->safe_psql("postgres", qq(
+			SET enable_seqscan = off;
+			SET ivfflat.probes = $probes;
+			SELECT i FROM tst ORDER BY v $operator '$queries[$i]' LIMIT $limit;
+		));
+		my @actual_ids = split("\n", $actual);
+		my %actual_set = map { $_ => 1 } @actual_ids;
+
+		my @expected_ids = split("\n", $expected[$i]);
+
+		foreach (@expected_ids) {
+			if (exists($actual_set{$_})) {
+				$correct++;
+			}
+			$total++;
+		}
+	}
+
+	cmp_ok($correct / $total, ">=", $min, $operator);
+}
+
+# Initialize node
+$node = get_new_node('node');
+$node->init;
+$node->start;
+
+# Create table
+$node->safe_psql("postgres", "CREATE EXTENSION vector;");
+$node->safe_psql("postgres", "CREATE TABLE tst (i int4, v vector(3));");
+$node->safe_psql("postgres",
+	"INSERT INTO tst SELECT i, ARRAY[random(), random(), random()] FROM generate_series(1, 100000) i;"
+);
+
+# Generate queries
+for (1..20) {
+	my $r1 = rand();
+	my $r2 = rand();
+	my $r3 = rand();
+	push(@queries, "[$r1,$r2,$r3]");
+}
+
+# Check each index type
+my @operators = ("<->", "<#>", "<=>");
+
+foreach (@operators) {
+	my $operator = $_;
+
+	# Get exact results
+	@expected = ();
+	foreach (@queries) {
+		my $res = $node->safe_psql("postgres", "SELECT i FROM tst ORDER BY v $operator '$_' LIMIT $limit;");
+		push(@expected, $res);
+	}
+
+	# Add index
+	my $opclass;
+	if ($operator == "<->") {
+		$opclass = "vector_l2_ops";
+	} elsif ($operator == "<#>") {
+		$opclass = "vector_ip_ops";
+	} else {
+		$opclass = "vector_cosine_ops";
+	}
+	$node->safe_psql("postgres", "CREATE INDEX ON tst USING ivfflat (v $opclass);");
+
+	# Test approximate results
+	test_recall(1, 0.75, $operator);
+	test_recall(10, 0.95, $operator);
+	test_recall(100, 1.0, $operator);
+}

test/t/004_centers.pl

@@ -2,7 +2,7 @@ use strict;
 use warnings;
 use PostgresNode;
 use TestLib;
-use Test::More;
+use Test::More tests => 3;
 
 # Initialize node
 my $node = get_new_node('node');
@@ -20,7 +20,7 @@ sub test_centers
 {
 	my ($lists, $min) = @_;
 
-	my ($ret, $stdout, $stderr) = $node->psql("postgres", "CREATE INDEX ON tst USING ivfflat (v vector_l2_ops) WITH (lists = $lists);");
+	my ($ret, $stdout, $stderr) = $node->psql("postgres", "CREATE INDEX ON tst USING ivfflat (v) WITH (lists = $lists);");
 	is($ret, 0, $stderr);
 }
 
@@ -34,5 +34,3 @@ $node->safe_psql("postgres",
 
 # Test no error for duplicate centers
 test_centers(10);
-
-done_testing();

test/t/005_query_recall.pl

@@ -2,7 +2,7 @@ use strict;
 use warnings;
 use PostgresNode;
 use TestLib;
-use Test::More;
+use Test::More tests => 60;
 
 # Initialize node
 my $node = get_new_node('node');
@@ -18,21 +18,24 @@ $node->safe_psql("postgres",
 
 # Check each index type
 my @operators = ("<->", "<#>", "<=>");
-my @opclasses = ("vector_l2_ops", "vector_ip_ops", "vector_cosine_ops");
+foreach (@operators) {
-
+	my $operator = $_;
-for my $i (0 .. $#operators)
-{
-	my $operator = $operators[$i];
-	my $opclass = $opclasses[$i];
 
 	# Add index
+	my $opclass;
+	if ($operator == "<->") {
+		$opclass = "vector_l2_ops";
+	} elsif ($operator == "<#>") {
+		$opclass = "vector_ip_ops";
+	} else {
+		$opclass = "vector_cosine_ops";
+	}
 	$node->safe_psql("postgres", "CREATE INDEX ON tst USING ivfflat (v $opclass);");
 
 	# Test 100% recall
-	for (1 .. 20)
+	for (1..20) {
-	{
+		my $i = int(rand() * 100000);
-		my $id = int(rand() * 100000);
+		my $query = $node->safe_psql("postgres", "SELECT v FROM tst WHERE i = $i;");
-		my $query = $node->safe_psql("postgres", "SELECT v FROM tst WHERE i = $id;");
 		my $res = $node->safe_psql("postgres", qq(
 			SET enable_seqscan = off;
 			SELECT v FROM tst ORDER BY v <-> '$query' LIMIT 1;
@@ -40,5 +43,3 @@ for my $i (0 .. $#operators)
 		is($res, $query);
 	}
 }
-
-done_testing();

test/t/006_lists.pl

@@ -2,7 +2,7 @@ use strict;
 use warnings;
 use PostgresNode;
 use TestLib;
-use Test::More;
+use Test::More tests => 3;
 
 # Initialize node
 my $node = get_new_node('node');
@@ -16,8 +16,8 @@ $node->safe_psql("postgres",
 	"INSERT INTO tst SELECT ARRAY[random(), random(), random()] FROM generate_series(1, 100000) i;"
 );
 
-$node->safe_psql("postgres", "CREATE INDEX lists50 ON tst USING ivfflat (v vector_l2_ops) WITH (lists = 50);");
+$node->safe_psql("postgres", "CREATE INDEX lists50 ON tst USING ivfflat (v) WITH (lists = 50);");
-$node->safe_psql("postgres", "CREATE INDEX lists100 ON tst USING ivfflat (v vector_l2_ops) WITH (lists = 100);");
+$node->safe_psql("postgres", "CREATE INDEX lists100 ON tst USING ivfflat (v) WITH (lists = 100);");
 
 # Test prefers more lists
 my $res = $node->safe_psql("postgres", "EXPLAIN SELECT v FROM tst ORDER BY v <-> '[0.5,0.5,0.5]' LIMIT 10;");
@@ -26,8 +26,6 @@ unlike($res, qr/lists50/);
 
 # Test errors with too much memory
 my ($ret, $stdout, $stderr) = $node->psql("postgres",
-	"CREATE INDEX lists10000 ON tst USING ivfflat (v vector_l2_ops) WITH (lists = 10000);"
+	"CREATE INDEX lists10000 ON tst USING ivfflat (v) WITH (lists = 10000);"
 );
 like($stderr, qr/memory required is/);
-
-done_testing();

test/t/007_inserts.pl

@@ -2,7 +2,7 @@ use strict;
 use warnings;
 use PostgresNode;
 use TestLib;
-use Test::More;
+use Test::More tests => 7;
 
 my $dim = 768;
 
@@ -19,7 +19,7 @@ $node->safe_psql("postgres", "CREATE TABLE tst (v vector($dim));");
 $node->safe_psql("postgres",
 	"INSERT INTO tst SELECT ARRAY[$array_sql] FROM generate_series(1, 10000) i;"
 );
-$node->safe_psql("postgres", "CREATE INDEX ON tst USING ivfflat (v vector_l2_ops);");
+$node->safe_psql("postgres", "CREATE INDEX ON tst USING ivfflat (v);");
 
 $node->pgbench(
 	"--no-vacuum --client=5 --transactions=100",
@@ -28,7 +28,7 @@ $node->pgbench(
 	[qr{^$}],
 	"concurrent INSERTs",
 	{
-		"007_ivfflat_inserts" => "INSERT INTO tst SELECT ARRAY[$array_sql] FROM generate_series(1, 10) i;"
+		"007_inserts" => "INSERT INTO tst SELECT ARRAY[$array_sql] FROM generate_series(1, 10) i;"
 	}
 );
 
@@ -53,5 +53,3 @@ $count = $node->safe_psql("postgres", qq(
 ));
 is($count, $expected);
 is(idx_scan(), 1);
-
-done_testing();

test/t/008_aggregates.pl

@@ -1,49 +0,0 @@
-use strict;
-use warnings;
-use PostgresNode;
-use TestLib;
-use Test::More;
-
-# Initialize node
-my $node = get_new_node('node');
-$node->init;
-$node->start;
-
-# Create table
-$node->safe_psql("postgres", "CREATE EXTENSION vector;");
-$node->safe_psql("postgres", "CREATE TABLE tst (r1 real, r2 real, r3 real, v vector(3));");
-$node->safe_psql("postgres", qq(
-	INSERT INTO tst SELECT r1, r2, r3, ARRAY[r1, r2, r3] FROM (
-		SELECT random() + 1.01 AS r1, random() + 2.01 AS r2, random() + 3.01 AS r3 FROM generate_series(1, 1000000) t
-	) i;
-));
-
-sub test_aggregate
-{
-	my ($agg) = @_;
-
-	# Test value
-	my $res = $node->safe_psql("postgres", "SELECT $agg(v) FROM tst;");
-	like($res, qr/\[1\.5/);
-	like($res, qr/,2\.5/);
-	like($res, qr/,3\.5/);
-
-	# Test matches real for avg
-	# Cannot test sum since sum(real) varies between calls
-	if ($agg eq 'avg')
-	{
-		my $r1 = $node->safe_psql("postgres", "SELECT $agg(r1)::float4 FROM tst;");
-		my $r2 = $node->safe_psql("postgres", "SELECT $agg(r2)::float4 FROM tst;");
-		my $r3 = $node->safe_psql("postgres", "SELECT $agg(r3)::float4 FROM tst;");
-		is($res, "[$r1,$r2,$r3]");
-	}
-
-	# Test explain
-	my $explain = $node->safe_psql("postgres", "EXPLAIN SELECT $agg(v) FROM tst;");
-	like($explain, qr/Partial Aggregate/);
-}
-
-test_aggregate('avg');
-test_aggregate('sum');
-
-done_testing();

test/t/008_avg.pl

@@ -0,0 +1,35 @@
+use strict;
+use warnings;
+use PostgresNode;
+use TestLib;
+use Test::More tests => 5;
+
+# Initialize node
+my $node = get_new_node('node');
+$node->init;
+$node->start;
+
+# Create table
+$node->safe_psql("postgres", "CREATE EXTENSION vector;");
+$node->safe_psql("postgres", "CREATE TABLE tst (r1 real, r2 real, r3 real, v vector(3));");
+$node->safe_psql("postgres", qq(
+	INSERT INTO tst SELECT r1, r2, r3, ARRAY[r1, r2, r3] FROM (
+		SELECT random() + 1.01 AS r1, random() + 2.01 AS r2, random() + 3.01 AS r3 FROM generate_series(1, 1000000) t
+	) i;
+));
+
+# Test avg
+my $avg = $node->safe_psql("postgres", "SELECT AVG(v) FROM tst;");
+like($avg, qr/\[1\.5/);
+like($avg, qr/,2\.5/);
+like($avg, qr/,3\.5/);
+
+# Test matches real
+my $r1 = $node->safe_psql("postgres", "SELECT AVG(r1)::float4 FROM tst;");
+my $r2 = $node->safe_psql("postgres", "SELECT AVG(r2)::float4 FROM tst;");
+my $r3 = $node->safe_psql("postgres", "SELECT AVG(r3)::float4 FROM tst;");
+is($avg, "[$r1,$r2,$r3]");
+
+# Test explain
+my $explain = $node->safe_psql("postgres", "EXPLAIN SELECT AVG(v) FROM tst;");
+like($explain, qr/Partial Aggregate/);

test/t/009_storage.pl

@@ -2,7 +2,7 @@ use strict;
 use warnings;
 use PostgresNode;
 use TestLib;
-use Test::More;
+use Test::More tests => 1;
 
 my $dim = 1024;
 
@@ -30,5 +30,3 @@ my ($ret, $stdout, $stderr) = $node->psql("postgres",
 	"INSERT INTO tst SELECT array_agg(n), array_agg(n), array_agg(n) FROM generate_series(1, $dim) n"
 );
 like($stderr, qr/row is too big/);
-
-done_testing();

test/t/010_hnsw_wal.pl

@@ -1,99 +0,0 @@
-# Based on postgres/contrib/bloom/t/001_wal.pl
-
-# Test generic xlog record work for hnsw index replication.
-use strict;
-use warnings;
-use PostgresNode;
-use TestLib;
-use Test::More;
-
-my $dim = 32;
-
-my $node_primary;
-my $node_replica;
-
-# Run few queries on both primary and replica and check their results match.
-sub test_index_replay
-{
-	my ($test_name) = @_;
-
-	# Wait for replica to catch up
-	my $applname = $node_replica->name;
-	my $caughtup_query = "SELECT pg_current_wal_lsn() <= replay_lsn FROM pg_stat_replication WHERE application_name = '$applname';";
-	$node_primary->poll_query_until('postgres', $caughtup_query)
-	  or die "Timed out while waiting for replica 1 to catch up";
-
-	my @r = ();
-	for (1 .. $dim)
-	{
-		push(@r, rand());
-	}
-	my $sql = join(",", @r);
-
-	my $queries = qq(
-		SET enable_seqscan = off;
-		SELECT * FROM tst ORDER BY v <-> '[$sql]' LIMIT 10;
-	);
-
-	# Run test queries and compare their result
-	my $primary_result = $node_primary->safe_psql("postgres", $queries);
-	my $replica_result = $node_replica->safe_psql("postgres", $queries);
-
-	is($primary_result, $replica_result, "$test_name: query result matches");
-	return;
-}
-
-# Use ARRAY[random(), random(), random(), ...] over
-# SELECT array_agg(random()) FROM generate_series(1, $dim)
-# to generate different values for each row
-my $array_sql = join(",", ('random()') x $dim);
-
-# Initialize primary node
-$node_primary = get_new_node('primary');
-$node_primary->init(allows_streaming => 1);
-if ($dim > 32)
-{
-	# TODO use wal_keep_segments for Postgres < 13
-	$node_primary->append_conf('postgresql.conf', qq(wal_keep_size = 1GB));
-}
-if ($dim > 1500)
-{
-	$node_primary->append_conf('postgresql.conf', qq(maintenance_work_mem = 128MB));
-}
-$node_primary->start;
-my $backup_name = 'my_backup';
-
-# Take backup
-$node_primary->backup($backup_name);
-
-# Create streaming replica linking to primary
-$node_replica = get_new_node('replica');
-$node_replica->init_from_backup($node_primary, $backup_name, has_streaming => 1);
-$node_replica->start;
-
-# Create hnsw index on primary
-$node_primary->safe_psql("postgres", "CREATE EXTENSION vector;");
-$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[$array_sql] FROM generate_series(1, 1000) i;"
-);
-$node_primary->safe_psql("postgres", "CREATE INDEX ON tst USING hnsw (v vector_l2_ops);");
-
-# Test that queries give same result
-test_index_replay('initial');
-
-# Run 10 cycles of table modification. Run test queries after each modification.
-for my $i (1 .. 10)
-{
-	$node_primary->safe_psql("postgres", "DELETE FROM tst WHERE i = $i;");
-	test_index_replay("delete $i");
-	$node_primary->safe_psql("postgres", "VACUUM tst;");
-	test_index_replay("vacuum $i");
-	my ($start, $end) = (1001 + ($i - 1) * 100, 1000 + $i * 100);
-	$node_primary->safe_psql("postgres",
-		"INSERT INTO tst SELECT i % 10, ARRAY[$array_sql] FROM generate_series($start, $end) i;"
-	);
-	test_index_replay("insert $i");
-}
-
-done_testing();

test/t/011_hnsw_vacuum.pl

@@ -1,54 +0,0 @@
-use strict;
-use warnings;
-use PostgresNode;
-use TestLib;
-use Test::More;
-
-my $dim = 3;
-
-my @r = ();
-for (1 .. $dim)
-{
-	my $v = int(rand(1000)) + 1;
-	push(@r, "i % $v");
-}
-my $array_sql = join(", ", @r);
-
-# Initialize node
-my $node = get_new_node('node');
-$node->init;
-$node->start;
-
-# Create table and index
-$node->safe_psql("postgres", "CREATE EXTENSION vector;");
-$node->safe_psql("postgres", "CREATE TABLE tst (i int4, v vector($dim));");
-$node->safe_psql("postgres",
-	"INSERT INTO tst SELECT i, ARRAY[$array_sql] FROM generate_series(1, 10000) i;"
-);
-$node->safe_psql("postgres", "CREATE INDEX ON tst USING hnsw (v vector_l2_ops);");
-
-# Get size
-my $size = $node->safe_psql("postgres", "SELECT pg_total_relation_size('tst_v_idx');");
-
-# Delete all, vacuum, and insert same data
-$node->safe_psql("postgres", "DELETE FROM tst;");
-$node->safe_psql("postgres", "VACUUM tst;");
-$node->safe_psql("postgres",
-	"INSERT INTO tst SELECT i, ARRAY[$array_sql] FROM generate_series(1, 10000) i;"
-);
-
-# Check size
-# May increase some due to different levels
-my $new_size = $node->safe_psql("postgres", "SELECT pg_total_relation_size('tst_v_idx');");
-cmp_ok($new_size, "<=", $size * 1.02, "size does not increase too much");
-
-# Delete all but one
-$node->safe_psql("postgres", "DELETE FROM tst WHERE i != 123;");
-$node->safe_psql("postgres", "VACUUM tst;");
-my $res = $node->safe_psql("postgres", qq(
-	SET enable_seqscan = off;
-	SELECT i FROM tst ORDER BY v <-> '[0,0,0]' LIMIT 10;
-));
-is($res, 123);
-
-done_testing();

test/t/012_hnsw_build_recall.pl

@@ -1,128 +0,0 @@
-use strict;
-use warnings;
-use PostgresNode;
-use TestLib;
-use Test::More;
-
-my $node;
-my @queries = ();
-my @expected;
-my $limit = 20;
-
-sub test_recall
-{
-	my ($min, $operator) = @_;
-	my $correct = 0;
-	my $total = 0;
-
-	my $explain = $node->safe_psql("postgres", qq(
-		SET enable_seqscan = off;
-		EXPLAIN ANALYZE SELECT i FROM tst ORDER BY v $operator '$queries[0]' LIMIT $limit;
-	));
-	like($explain, qr/Index Scan/);
-
-	for my $i (0 .. $#queries)
-	{
-		my $actual = $node->safe_psql("postgres", qq(
-			SET enable_seqscan = off;
-			SELECT i FROM tst ORDER BY v $operator '$queries[$i]' LIMIT $limit;
-		));
-		my @actual_ids = split("\n", $actual);
-		my %actual_set = map { $_ => 1 } @actual_ids;
-
-		my @expected_ids = split("\n", $expected[$i]);
-
-		foreach (@expected_ids)
-		{
-			if (exists($actual_set{$_}))
-			{
-				$correct++;
-			}
-			$total++;
-		}
-	}
-
-	cmp_ok($correct / $total, ">=", $min, $operator);
-}
-
-# Initialize node
-$node = get_new_node('node');
-$node->init;
-$node->start;
-
-# Create table
-$node->safe_psql("postgres", "CREATE EXTENSION vector;");
-$node->safe_psql("postgres", "CREATE TABLE tst (i int4, v vector(3));");
-$node->safe_psql("postgres",
-	"INSERT INTO tst SELECT i, ARRAY[random(), random(), random()] FROM generate_series(1, 10000) i;"
-);
-
-# Generate queries
-for (1 .. 20)
-{
-	my $r1 = rand();
-	my $r2 = rand();
-	my $r3 = rand();
-	push(@queries, "[$r1,$r2,$r3]");
-}
-
-# 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];
-
-	# Get exact results
-	@expected = ();
-	foreach (@queries)
-	{
-		my $res = $node->safe_psql("postgres", "SELECT i FROM tst ORDER BY v $operator '$_' LIMIT $limit;");
-		push(@expected, $res);
-	}
-
-	# Build index serially
-	$node->safe_psql("postgres", qq(
-		SET max_parallel_maintenance_workers = 0;
-		CREATE INDEX idx ON tst USING hnsw (v $opclass);
-	));
-
-	# Test approximate results
-	my $min = $operator eq "<#>" ? 0.80 : 0.99;
-	test_recall($min, $operator);
-
-	$node->safe_psql("postgres", "DROP INDEX idx;");
-
-	# Build index in parallel in memory
-	my ($ret, $stdout, $stderr) = $node->psql("postgres", qq(
-		SET client_min_messages = DEBUG;
-		SET min_parallel_table_scan_size = 1;
-		CREATE INDEX idx ON tst USING hnsw (v $opclass);
-	));
-	is($ret, 0, $stderr);
-	like($stderr, qr/using \d+ parallel workers/);
-
-	# Test approximate results
-	test_recall($min, $operator);
-
-	$node->safe_psql("postgres", "DROP INDEX idx;");
-
-	# Build index in parallel on disk
-	# Set parallel_workers on table to use workers with low maintenance_work_mem
-	($ret, $stdout, $stderr) = $node->psql("postgres", qq(
-		ALTER TABLE tst SET (parallel_workers = 2);
-		SET client_min_messages = DEBUG;
-		SET maintenance_work_mem = '4MB';
-		CREATE INDEX idx ON tst USING hnsw (v $opclass);
-		ALTER TABLE tst RESET (parallel_workers);
-	));
-	is($ret, 0, $stderr);
-	like($stderr, qr/using \d+ parallel workers/);
-	like($stderr, qr/hnsw graph no longer fits into maintenance_work_mem/);
-
-	$node->safe_psql("postgres", "DROP INDEX idx;");
-}
-
-done_testing();

test/t/013_hnsw_insert_recall.pl

@@ -1,108 +0,0 @@
-use strict;
-use warnings;
-use PostgresNode;
-use TestLib;
-use Test::More;
-
-my $node;
-my @queries = ();
-my @expected;
-my $limit = 20;
-
-sub test_recall
-{
-	my ($min, $operator) = @_;
-	my $correct = 0;
-	my $total = 0;
-
-	my $explain = $node->safe_psql("postgres", qq(
-		SET enable_seqscan = off;
-		EXPLAIN ANALYZE SELECT i FROM tst ORDER BY v $operator '$queries[0]' LIMIT $limit;
-	));
-	like($explain, qr/Index Scan/);
-
-	for my $i (0 .. $#queries)
-	{
-		my $actual = $node->safe_psql("postgres", qq(
-			SET enable_seqscan = off;
-			SELECT i FROM tst ORDER BY v $operator '$queries[$i]' LIMIT $limit;
-		));
-		my @actual_ids = split("\n", $actual);
-		my %actual_set = map { $_ => 1 } @actual_ids;
-
-		my @expected_ids = split("\n", $expected[$i]);
-
-		foreach (@expected_ids)
-		{
-			if (exists($actual_set{$_}))
-			{
-				$correct++;
-			}
-			$total++;
-		}
-	}
-
-	cmp_ok($correct / $total, ">=", $min, $operator);
-}
-
-# Initialize node
-$node = get_new_node('node');
-$node->init;
-$node->start;
-
-# Create table
-$node->safe_psql("postgres", "CREATE EXTENSION vector;");
-$node->safe_psql("postgres", "CREATE TABLE tst (i serial, v vector(3));");
-
-# Generate queries
-for (1 .. 20)
-{
-	my $r1 = rand();
-	my $r2 = rand();
-	my $r3 = rand();
-	push(@queries, "[$r1,$r2,$r3]");
-}
-
-# 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 idx ON tst USING hnsw (v $opclass);");
-
-	# Use concurrent inserts
-	$node->pgbench(
-		"--no-vacuum --client=10 --transactions=1000",
-		0,
-		[qr{actually processed}],
-		[qr{^$}],
-		"concurrent INSERTs",
-		{
-			"013_hnsw_insert_recall_$opclass" => "INSERT INTO tst (v) VALUES (ARRAY[random(), random(), random()]);"
-		}
-	);
-
-	# Get exact results
-	@expected = ();
-	foreach (@queries)
-	{
-		my $res = $node->safe_psql("postgres", qq(
-			SET enable_indexscan = off;
-			SELECT i FROM tst ORDER BY v $operator '$_' LIMIT $limit;
-		));
-		push(@expected, $res);
-	}
-
-	my $min = $operator eq "<#>" ? 0.80 : 0.99;
-	test_recall($min, $operator);
-
-	$node->safe_psql("postgres", "DROP INDEX idx;");
-	$node->safe_psql("postgres", "TRUNCATE tst;");
-}
-
-done_testing();

test/t/014_hnsw_inserts.pl

@@ -1,74 +0,0 @@
-use strict;
-use warnings;
-use PostgresNode;
-use TestLib;
-use Test::More;
-
-# Ensures elements and neighbors on both same and different pages
-my $dim = 1900;
-
-my $array_sql = join(",", ('random()') x $dim);
-
-# Initialize node
-my $node = get_new_node('node');
-$node->init;
-$node->start;
-
-# Create table and index
-$node->safe_psql("postgres", "CREATE EXTENSION vector;");
-$node->safe_psql("postgres", "CREATE TABLE tst (v vector($dim));");
-$node->safe_psql("postgres", "CREATE INDEX ON tst USING hnsw (v vector_l2_ops);");
-
-sub idx_scan
-{
-	# Stats do not update instantaneously
-	# https://www.postgresql.org/docs/current/monitoring-stats.html#MONITORING-STATS-VIEWS
-	sleep(1);
-	$node->safe_psql("postgres", "SELECT idx_scan FROM pg_stat_user_indexes WHERE indexrelid = 'tst_v_idx'::regclass;");
-}
-
-for my $i (1 .. 20)
-{
-	$node->pgbench(
-		"--no-vacuum --client=10 --transactions=1",
-		0,
-		[qr{actually processed}],
-		[qr{^$}],
-		"concurrent INSERTs",
-		{
-			"014_hnsw_inserts_$i" => "INSERT INTO tst VALUES (ARRAY[$array_sql]);"
-		}
-	);
-
-	my $count = $node->safe_psql("postgres", qq(
-		SET enable_seqscan = off;
-		SELECT COUNT(*) FROM (SELECT v FROM tst ORDER BY v <-> (SELECT v FROM tst LIMIT 1)) t;
-	));
-	is($count, 10);
-
-	$node->safe_psql("postgres", "TRUNCATE tst;");
-}
-
-$node->pgbench(
-	"--no-vacuum --client=20 --transactions=5",
-	0,
-	[qr{actually processed}],
-	[qr{^$}],
-	"concurrent INSERTs",
-	{
-		"014_hnsw_inserts" => "INSERT INTO tst SELECT ARRAY[$array_sql] FROM generate_series(1, 10) i;"
-	}
-);
-
-my $count = $node->safe_psql("postgres", qq(
-	SET enable_seqscan = off;
-	SET hnsw.ef_search = 1000;
-	SELECT COUNT(*) FROM (SELECT v FROM tst ORDER BY v <-> (SELECT v FROM tst LIMIT 1)) t;
-));
-# Elements may lose all incoming connections with the HNSW algorithm
-# Vacuuming can fix this if one of the elements neighbors is deleted
-cmp_ok($count, ">=", 997);
-
-is(idx_scan(), 21);
-
-done_testing();

test/t/015_hnsw_duplicates.pl

@@ -1,58 +0,0 @@
-use strict;
-use warnings;
-use PostgresNode;
-use TestLib;
-use Test::More;
-
-# Initialize node
-my $node = get_new_node('node');
-$node->init;
-$node->start;
-
-# Create table
-$node->safe_psql("postgres", "CREATE EXTENSION vector;");
-$node->safe_psql("postgres", "CREATE TABLE tst (v vector(3));");
-
-sub insert_vectors
-{
-	for my $i (1 .. 20)
-	{
-		$node->safe_psql("postgres", "INSERT INTO tst VALUES ('[1,1,1]');");
-	}
-}
-
-sub test_duplicates
-{
-	my $res = $node->safe_psql("postgres", qq(
-		SET enable_seqscan = off;
-		SET hnsw.ef_search = 1;
-		SELECT COUNT(*) FROM (SELECT * FROM tst ORDER BY v <-> '[1,1,1]') t;
-	));
-	is($res, 10);
-}
-
-# Test duplicates with build
-insert_vectors();
-$node->safe_psql("postgres", "CREATE INDEX idx ON tst USING hnsw (v vector_l2_ops);");
-test_duplicates();
-
-# Reset
-$node->safe_psql("postgres", "TRUNCATE tst;");
-
-# Test duplicates with inserts
-insert_vectors();
-test_duplicates();
-
-# Test fallback path for inserts
-$node->pgbench(
-	"--no-vacuum --client=5 --transactions=100",
-	0,
-	[qr{actually processed}],
-	[qr{^$}],
-	"concurrent INSERTs",
-	{
-		"015_hnsw_duplicates" => "INSERT INTO tst VALUES ('[1,1,1]');"
-	}
-);
-
-done_testing();

test/t/016_hnsw_vacuum_recall.pl

@@ -1,97 +0,0 @@
-use strict;
-use warnings;
-use PostgresNode;
-use TestLib;
-use Test::More;
-
-my $node;
-my @queries = ();
-my @expected;
-my $limit = 20;
-
-sub test_recall
-{
-	my ($min, $ef_search, $test_name) = @_;
-	my $correct = 0;
-	my $total = 0;
-
-	my $explain = $node->safe_psql("postgres", qq(
-		SET enable_seqscan = off;
-		SET hnsw.ef_search = $ef_search;
-		EXPLAIN ANALYZE SELECT i FROM tst ORDER BY v <-> '$queries[0]' LIMIT $limit;
-	));
-	like($explain, qr/Index Scan/);
-
-	for my $i (0 .. $#queries)
-	{
-		my $actual = $node->safe_psql("postgres", qq(
-			SET enable_seqscan = off;
-			SET hnsw.ef_search = $ef_search;
-			SELECT i FROM tst ORDER BY v <-> '$queries[$i]' LIMIT $limit;
-		));
-		my @actual_ids = split("\n", $actual);
-		my %actual_set = map { $_ => 1 } @actual_ids;
-
-		my @expected_ids = split("\n", $expected[$i]);
-
-		foreach (@expected_ids)
-		{
-			if (exists($actual_set{$_}))
-			{
-				$correct++;
-			}
-			$total++;
-		}
-	}
-
-	cmp_ok($correct / $total, ">=", $min, $test_name);
-}
-
-# Initialize node
-$node = get_new_node('node');
-$node->init;
-$node->start;
-
-# Create table
-$node->safe_psql("postgres", "CREATE EXTENSION vector;");
-$node->safe_psql("postgres", "CREATE TABLE tst (i int4, v vector(3));");
-$node->safe_psql("postgres", "ALTER TABLE tst SET (autovacuum_enabled = false);");
-$node->safe_psql("postgres",
-	"INSERT INTO tst SELECT i, ARRAY[random(), random(), random()] FROM generate_series(1, 10000) i;"
-);
-
-# Add index
-$node->safe_psql("postgres", "CREATE INDEX ON tst USING hnsw (v vector_l2_ops) WITH (m = 4, ef_construction = 8);");
-
-# Delete data
-$node->safe_psql("postgres", "DELETE FROM tst WHERE i > 2500;");
-
-# Generate queries
-for (1 .. 20)
-{
-	my $r1 = rand();
-	my $r2 = rand();
-	my $r3 = rand();
-	push(@queries, "[$r1,$r2,$r3]");
-}
-
-# Get exact results
-@expected = ();
-foreach (@queries)
-{
-	my $res = $node->safe_psql("postgres", qq(
-		SET enable_indexscan = off;
-		SELECT i FROM tst ORDER BY v <-> '$_' LIMIT $limit;
-	));
-	push(@expected, $res);
-}
-
-test_recall(0.20, $limit, "before vacuum");
-test_recall(0.95, 100, "before vacuum");
-
-# TODO Test concurrent inserts with vacuum
-$node->safe_psql("postgres", "VACUUM tst;");
-
-test_recall(0.95, $limit, "after vacuum");
-
-done_testing();

test/t/017_ivfflat_insert_recall.pl

@@ -1,117 +0,0 @@
-use strict;
-use warnings;
-use PostgresNode;
-use TestLib;
-use Test::More;
-
-my $node;
-my @queries = ();
-my @expected;
-my $limit = 20;
-
-sub test_recall
-{
-	my ($probes, $min, $operator) = @_;
-	my $correct = 0;
-	my $total = 0;
-
-	my $explain = $node->safe_psql("postgres", qq(
-		SET enable_seqscan = off;
-		SET ivfflat.probes = $probes;
-		EXPLAIN ANALYZE SELECT i FROM tst ORDER BY v $operator '$queries[0]' LIMIT $limit;
-	));
-	like($explain, qr/Index Scan using idx on tst/);
-
-	for my $i (0 .. $#queries)
-	{
-		my $actual = $node->safe_psql("postgres", qq(
-			SET enable_seqscan = off;
-			SET ivfflat.probes = $probes;
-			SELECT i FROM tst ORDER BY v $operator '$queries[$i]' LIMIT $limit;
-		));
-		my @actual_ids = split("\n", $actual);
-		my %actual_set = map { $_ => 1 } @actual_ids;
-
-		my @expected_ids = split("\n", $expected[$i]);
-
-		foreach (@expected_ids)
-		{
-			if (exists($actual_set{$_}))
-			{
-				$correct++;
-			}
-			$total++;
-		}
-	}
-
-	cmp_ok($correct / $total, ">=", $min, $operator);
-}
-
-# Initialize node
-$node = get_new_node('node');
-$node->init;
-$node->start;
-
-# Create table
-$node->safe_psql("postgres", "CREATE EXTENSION vector;");
-$node->safe_psql("postgres", "CREATE TABLE tst (i serial, v vector(3));");
-
-# Generate queries
-for (1 .. 20)
-{
-	my $r1 = rand();
-	my $r2 = rand();
-	my $r3 = rand();
-	push(@queries, "[$r1,$r2,$r3]");
-}
-
-# 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 idx ON tst USING ivfflat (v $opclass);");
-
-	# Use concurrent inserts
-	$node->pgbench(
-		"--no-vacuum --client=10 --transactions=1000",
-		0,
-		[qr{actually processed}],
-		[qr{^$}],
-		"concurrent INSERTs",
-		{
-			"017_ivfflat_insert_recall_$opclass" => "INSERT INTO tst (v) SELECT ARRAY[random(), random(), random()] FROM generate_series(1, 10) i;"
-		}
-	);
-
-	# Get exact results
-	@expected = ();
-	foreach (@queries)
-	{
-		my $res = $node->safe_psql("postgres", qq(
-			SET enable_indexscan = off;
-			SELECT i FROM tst ORDER BY v $operator '$_' LIMIT $limit;
-		));
-		push(@expected, $res);
-	}
-
-	# Test approximate results
-	if ($operator ne "<#>")
-	{
-		# TODO Fix test (uniform random vectors all have similar inner product)
-		test_recall(1, 0.71, $operator);
-		test_recall(10, 0.95, $operator);
-	}
-	# Account for equal distances
-	test_recall(100, 0.9925, $operator);
-
-	$node->safe_psql("postgres", "DROP INDEX idx;");
-	$node->safe_psql("postgres", "TRUNCATE tst;");
-}
-
-done_testing();

test/t/018_hnsw_filtering.pl

@@ -1,114 +0,0 @@
-use strict;
-use warnings;
-use PostgresNode;
-use TestLib;
-use Test::More;
-
-my $dim = 3;
-my $nc = 50;
-my $limit = 20;
-
-my $array_sql = join(",", ('random()') x $dim);
-
-# Initialize node
-my $node = get_new_node('node');
-$node->init;
-$node->start;
-
-# Create table and index
-$node->safe_psql("postgres", "CREATE EXTENSION vector;");
-$node->safe_psql("postgres", "CREATE TABLE tst (i int4, v vector($dim), c int4, t text);");
-$node->safe_psql("postgres",
-	"INSERT INTO tst SELECT i, ARRAY[$array_sql], i % $nc, 'test ' || i FROM generate_series(1, 10000) i;"
-);
-$node->safe_psql("postgres", "CREATE INDEX idx ON tst USING hnsw (v vector_l2_ops);");
-$node->safe_psql("postgres", "ANALYZE tst;");
-
-# Generate query
-my @r = ();
-for (1 .. $dim)
-{
-	push(@r, rand());
-}
-my $query = "[" . join(",", @r) . "]";
-my $c = int(rand() * $nc);
-
-# Test attribute filtering
-my $explain = $node->safe_psql("postgres", qq(
-	EXPLAIN ANALYZE SELECT i FROM tst WHERE c = $c ORDER BY v <-> '$query' LIMIT $limit;
-));
-# TODO Do not use index
-like($explain, qr/Index Scan using idx/);
-
-# Test attribute filtering with few rows removed
-$explain = $node->safe_psql("postgres", qq(
-	EXPLAIN ANALYZE SELECT i FROM tst WHERE c != $c ORDER BY v <-> '$query' LIMIT $limit;
-));
-like($explain, qr/Index Scan using idx/);
-
-# Test attribute filtering with few rows removed comparison
-$explain = $node->safe_psql("postgres", qq(
-	EXPLAIN ANALYZE SELECT i FROM tst WHERE c >= 1 ORDER BY v <-> '$query' LIMIT $limit;
-));
-like($explain, qr/Index Scan using idx/);
-
-# Test attribute filtering with many rows removed comparison
-$explain = $node->safe_psql("postgres", qq(
-	EXPLAIN ANALYZE SELECT i FROM tst WHERE c < 1 ORDER BY v <-> '$query' LIMIT $limit;
-));
-# TODO Do not use index
-like($explain, qr/Index Scan using idx/);
-
-# Test attribute filtering with few rows removed like
-$explain = $node->safe_psql("postgres", qq(
-	EXPLAIN ANALYZE SELECT i FROM tst WHERE t LIKE '%%test%%' ORDER BY v <-> '$query' LIMIT $limit;
-));
-like($explain, qr/Index Scan using idx/);
-
-# Test attribute filtering with many rows removed like
-$explain = $node->safe_psql("postgres", qq(
-	EXPLAIN ANALYZE SELECT i FROM tst WHERE t LIKE '%%other%%' ORDER BY v <-> '$query' LIMIT $limit;
-));
-like($explain, qr/Seq Scan/);
-
-# Test distance filtering
-$explain = $node->safe_psql("postgres", qq(
-	EXPLAIN ANALYZE SELECT i FROM tst WHERE v <-> '$query' < 1 ORDER BY v <-> '$query' LIMIT $limit;
-));
-like($explain, qr/Index Scan using idx/);
-
-# Test distance filtering greater than distance
-$explain = $node->safe_psql("postgres", qq(
-	EXPLAIN ANALYZE SELECT i FROM tst WHERE v <-> '$query' > 1 ORDER BY v <-> '$query' LIMIT $limit;
-));
-# TODO Do not use index
-like($explain, qr/Index Scan using idx/);
-
-# Test distance filtering without order
-$explain = $node->safe_psql("postgres", qq(
-	EXPLAIN ANALYZE SELECT i FROM tst WHERE v <-> '$query' < 1;
-));
-like($explain, qr/Seq Scan/);
-
-# Test distance filtering without limit
-$explain = $node->safe_psql("postgres", qq(
-	EXPLAIN ANALYZE SELECT i FROM tst WHERE v <-> '$query' < 1 ORDER BY v <-> '$query';
-));
-like($explain, qr/Seq Scan/);
-
-# Test attribute index
-$node->safe_psql("postgres", "CREATE INDEX attribute_idx ON tst (c);");
-$explain = $node->safe_psql("postgres", qq(
-	EXPLAIN ANALYZE SELECT i FROM tst WHERE c = $c ORDER BY v <-> '$query' LIMIT $limit;
-));
-# TODO Use attribute index
-like($explain, qr/Index Scan using idx/);
-
-# Test partial index
-$node->safe_psql("postgres", "CREATE INDEX partial_idx ON tst USING hnsw (v vector_l2_ops) WHERE (c = $c);");
-$explain = $node->safe_psql("postgres", qq(
-	EXPLAIN ANALYZE SELECT i FROM tst WHERE c = $c ORDER BY v <-> '$query' LIMIT $limit;
-));
-like($explain, qr/Index Scan using partial_idx/);
-
-done_testing();

test/t/019_ivfflat_filtering.pl

@@ -1,116 +0,0 @@
-use strict;
-use warnings;
-use PostgresNode;
-use TestLib;
-use Test::More;
-
-my $dim = 3;
-my $nc = 50;
-my $limit = 20;
-
-my $array_sql = join(",", ('random()') x $dim);
-
-# Initialize node
-my $node = get_new_node('node');
-$node->init;
-$node->start;
-
-# Create table and index
-$node->safe_psql("postgres", "CREATE EXTENSION vector;");
-$node->safe_psql("postgres", "CREATE TABLE tst (i int4, v vector($dim), c int4, t text);");
-$node->safe_psql("postgres",
-	"INSERT INTO tst SELECT i, ARRAY[$array_sql], i % $nc, 'test ' || i FROM generate_series(1, 10000) i;"
-);
-$node->safe_psql("postgres", "CREATE INDEX idx ON tst USING ivfflat (v vector_l2_ops) WITH (lists = 100);");
-$node->safe_psql("postgres", "ANALYZE tst;");
-
-# Generate query
-my @r = ();
-for (1 .. $dim)
-{
-	push(@r, rand());
-}
-my $query = "[" . join(",", @r) . "]";
-my $c = int(rand() * $nc);
-
-# Test attribute filtering
-my $explain = $node->safe_psql("postgres", qq(
-	EXPLAIN ANALYZE SELECT i FROM tst WHERE c = $c ORDER BY v <-> '$query' LIMIT $limit;
-));
-# TODO Do not use index
-like($explain, qr/Index Scan using idx/);
-
-# Test attribute filtering with few rows removed
-$explain = $node->safe_psql("postgres", qq(
-	EXPLAIN ANALYZE SELECT i FROM tst WHERE c != $c ORDER BY v <-> '$query' LIMIT $limit;
-));
-like($explain, qr/Index Scan using idx/);
-
-# Test attribute filtering with few rows removed comparison
-$explain = $node->safe_psql("postgres", qq(
-	EXPLAIN ANALYZE SELECT i FROM tst WHERE c >= 1 ORDER BY v <-> '$query' LIMIT $limit;
-));
-like($explain, qr/Index Scan using idx/);
-
-# Test attribute filtering with many rows removed comparison
-$explain = $node->safe_psql("postgres", qq(
-	EXPLAIN ANALYZE SELECT i FROM tst WHERE c < 1 ORDER BY v <-> '$query' LIMIT $limit;
-));
-# TODO Do not use index
-like($explain, qr/Index Scan using idx/);
-
-# Test attribute filtering with few rows removed like
-$explain = $node->safe_psql("postgres", qq(
-	EXPLAIN ANALYZE SELECT i FROM tst WHERE t LIKE '%%test%%' ORDER BY v <-> '$query' LIMIT $limit;
-));
-like($explain, qr/Index Scan using idx/);
-
-# Test attribute filtering with many rows removed like
-$explain = $node->safe_psql("postgres", qq(
-	EXPLAIN ANALYZE SELECT i FROM tst WHERE t LIKE '%%other%%' ORDER BY v <-> '$query' LIMIT $limit;
-));
-like($explain, qr/Seq Scan/);
-
-# Test distance filtering
-$explain = $node->safe_psql("postgres", qq(
-	EXPLAIN ANALYZE SELECT i FROM tst WHERE v <-> '$query' < 1 ORDER BY v <-> '$query' LIMIT $limit;
-));
-like($explain, qr/Index Scan using idx/);
-
-# Test distance filtering greater than distance
-$explain = $node->safe_psql("postgres", qq(
-	EXPLAIN ANALYZE SELECT i FROM tst WHERE v <-> '$query' > 1 ORDER BY v <-> '$query' LIMIT $limit;
-));
-# TODO Do not use index
-like($explain, qr/Index Scan using idx/);
-
-# Test distance filtering without order
-$explain = $node->safe_psql("postgres", qq(
-	EXPLAIN ANALYZE SELECT i FROM tst WHERE v <-> '$query' < 1;
-));
-like($explain, qr/Seq Scan/);
-
-# Test distance filtering without limit
-$explain = $node->safe_psql("postgres", qq(
-	EXPLAIN ANALYZE SELECT i FROM tst WHERE v <-> '$query' < 1 ORDER BY v <-> '$query';
-));
-# TODO Do not use index
-like($explain, qr/Index Scan using idx/);
-
-# Test attribute index
-$node->safe_psql("postgres", "CREATE INDEX attribute_idx ON tst (c);");
-$explain = $node->safe_psql("postgres", qq(
-	EXPLAIN ANALYZE SELECT i FROM tst WHERE c = $c ORDER BY v <-> '$query' LIMIT $limit;
-));
-# TODO Use attribute index
-like($explain, qr/Index Scan using idx/);
-
-# Test partial index
-$node->safe_psql("postgres", "CREATE INDEX partial_idx ON tst USING ivfflat (v vector_l2_ops) WITH (lists = 5) WHERE (c = $c);");
-$explain = $node->safe_psql("postgres", qq(
-	EXPLAIN ANALYZE SELECT i FROM tst WHERE c = $c ORDER BY v <-> '$query' LIMIT $limit;
-));
-# TODO Use partial index
-like($explain, qr/Index Scan using idx/);
-
-done_testing();

vector.control

@@ -1,4 +1,4 @@
-comment = 'vector data type and ivfflat and hnsw access methods'
+comment = 'vector data type and ivfflat access method'
-default_version = '0.6.1'
+default_version = '0.4.1'
 module_pathname = '$libdir/vector'
 relocatable = true