Improved name [skip ci]

.github/workflows/build.yml

@@ -49,7 +49,7 @@ jobs:
           - postgres: 16
             os: macos-14
           - postgres: 14
-            os: macos-13
+            os: macos-12
     steps:
       - uses: actions/checkout@v4
       - uses: ankane/setup-postgres@v1

CHANGELOG.md

@@ -1,10 +1,6 @@
 ## 0.8.0 (unreleased)
 
-- Added support for inline filtering with HNSW
-- Added support for iterative index scans
 - Added casts for arrays to `sparsevec`
-- Improved cost estimation
-- Improved performance of HNSW inserts and on-disk index builds
 - Reduced memory usage for HNSW index scans
 - Dropped support for Postgres 12
 

Dockerfile

@@ -1,4 +1,4 @@
-ARG PG_MAJOR=17
+ARG PG_MAJOR=16
 FROM postgres:$PG_MAJOR
 ARG PG_MAJOR
 

Makefile

@@ -4,8 +4,8 @@ EXTVERSION = 0.7.4
 MODULE_big = vector
 DATA = $(wildcard sql/*--*--*.sql)
 DATA_built = sql/$(EXTENSION)--$(EXTVERSION).sql
-OBJS = src/bitutils.o src/bitvec.o src/halfutils.o src/halfvec.o 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/sparsevec.o src/vector.o
-HEADERS = src/halfvec.h src/sparsevec.h src/vector.h
+OBJS = src/bitutils.o src/bitvec.o src/halfutils.o src/halfvec.o 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/minivec.o src/ivfvacuum.o src/sparsevec.o src/vector.o
+HEADERS = src/halfvec.h src/minivec.h src/sparsevec.h src/vector.h
 
 TESTS = $(wildcard test/sql/*.sql)
 REGRESS = $(patsubst test/sql/%.sql,%,$(TESTS))
@@ -66,7 +66,7 @@ dist:
 	git archive --format zip --prefix=$(EXTENSION)-$(EXTVERSION)/ --output dist/$(EXTENSION)-$(EXTVERSION).zip master
 
 # for Docker
-PG_MAJOR ?= 17
+PG_MAJOR ?= 16
 
 .PHONY: docker
 

Makefile.win

@@ -2,10 +2,10 @@ EXTENSION = vector
 EXTVERSION = 0.7.4
 
 DATA_built = sql\$(EXTENSION)--$(EXTVERSION).sql
-OBJS = src\bitutils.obj src\bitvec.obj src\halfutils.obj src\halfvec.obj src\hnsw.obj src\hnswbuild.obj src\hnswinsert.obj src\hnswscan.obj src\hnswutils.obj src\hnswvacuum.obj src\ivfbuild.obj src\ivfflat.obj src\ivfinsert.obj src\ivfkmeans.obj src\ivfscan.obj src\ivfutils.obj src\ivfvacuum.obj src\sparsevec.obj src\vector.obj
-HEADERS = src\halfvec.h src\sparsevec.h src\vector.h
+OBJS = src\bitutils.obj src\bitvec.obj src\halfutils.obj src\halfvec.obj src\hnsw.obj src\hnswbuild.obj src\hnswinsert.obj src\hnswscan.obj src\hnswutils.obj src\hnswvacuum.obj src\ivfbuild.obj src\ivfflat.obj src\ivfinsert.obj src\ivfkmeans.obj src\ivfscan.obj src\ivfutils.obj src\ivfvacuum.obj src\minivec.obj src\sparsevec.obj src\vector.obj
+HEADERS = src\halfvec.h src\minivec.h src\sparsevec.h src\vector.h
 
-REGRESS = bit btree cast copy halfvec hnsw_bit hnsw_halfvec hnsw_sparsevec hnsw_vector ivfflat_bit ivfflat_halfvec ivfflat_vector sparsevec vector_type
+REGRESS = bit btree cast copy halfvec hnsw_bit hnsw_halfvec hnsw_sparsevec hnsw_vector ivfflat_bit ivfflat_halfvec ivfflat_vector minivec sparsevec vector_type
 REGRESS_OPTS = --inputdir=test --load-extension=$(EXTENSION)
 
 # For /arch flags

README.md

@@ -52,8 +52,6 @@ nmake /F Makefile.win
 nmake /F Makefile.win install
 ```
 
-Note: Postgres 17 is not supported yet due to an upstream issue
-
 See the [installation notes](#installation-notes---windows) if you run into issues
 
 You can also install it with [Docker](#docker) or [conda-forge](#conda-forge).
@@ -102,8 +100,6 @@ Or add a vector column to an existing table
 ALTER TABLE items ADD COLUMN embedding vector(3);
 ```
 
-Also supports [half-precision](#half-precision-vectors), [binary](#binary-vectors), and [sparse](#sparse-vectors) vectors
-
 Insert vectors
 
 ```sql
@@ -149,8 +145,6 @@ Supported distance functions are:
 - `<#>` - (negative) inner product
 - `<=>` - cosine distance
 - `<+>` - L1 distance (added in 0.7.0)
-- `<~>` - Hamming distance (binary vectors, added in 0.7.0)
-- `<%>` - Jaccard distance (binary vectors, added in 0.7.0)
 
 Get the nearest neighbors to a row
 
@@ -439,12 +433,6 @@ Create an index on one [or more](https://www.postgresql.org/docs/current/indexes
 CREATE INDEX ON items (category_id);
 ```
 
-Or a composite HNSW index for approximate search (added in 0.8.0)
-
-```sql
-CREATE INDEX ON items USING hnsw (embedding vector_l2_ops, category_id);
-```
-
 Or a [partial index](https://www.postgresql.org/docs/current/indexes-partial.html) on the vector column for approximate search
 
 ```sql
@@ -946,6 +934,37 @@ Function | Description | Added
 avg(halfvec) → halfvec | average | 0.7.0
 sum(halfvec) → halfvec | sum | 0.7.0
 
+### Minivec Type
+
+Each mini vector takes `dimensions + 8` bytes of storage. Each element is a E4M3 8-bit floating-point number, and all elements must be finite (no `NaN`). Mini vectors can have up to 16,000 dimensions.
+
+### Minivec Operators
+
+Operator | Description | Added
+--- | --- | ---
+\+ | element-wise addition | 0.8.0
+\- | element-wise subtraction | 0.8.0
+\* | element-wise multiplication | 0.8.0
+\|\| | concatenate | 0.8.0
+<-> | Euclidean distance | 0.8.0
+<#> | negative inner product | 0.8.0
+<=> | cosine distance | 0.8.0
+<+> | taxicab distance | 0.8.0
+
+### Minivec Functions
+
+Function | Description | Added
+--- | --- | ---
+binary_quantize(minivec) → bit | binary quantize | 0.8.0
+cosine_distance(minivec, minivec) → double precision | cosine distance | 0.8.0
+inner_product(minivec, minivec) → double precision | inner product | 0.8.0
+l1_distance(minivec, minivec) → double precision | taxicab distance | 0.8.0
+l2_distance(minivec, minivec) → double precision | Euclidean distance | 0.8.0
+l2_norm(minivec) → double precision | Euclidean norm | 0.8.0
+l2_normalize(minivec) → minivec | Normalize with Euclidean norm | 0.8.0
+subvector(minivec, integer, integer) → minivec | subvector | 0.8.0
+vector_dims(minivec) → integer | number of dimensions | 0.8.0
+
 ### Bit Type
 
 Each bit vector takes `dimensions / 8 + 8` bytes of storage. See the [Postgres docs](https://www.postgresql.org/docs/current/datatype-bit.html) for more info.
@@ -995,7 +1014,7 @@ l2_normalize(sparsevec) → sparsevec | Normalize with Euclidean norm | 0.7.0
 If your machine has multiple Postgres installations, specify the path to [pg_config](https://www.postgresql.org/docs/current/app-pgconfig.html) with:
 
 ```sh
-export PG_CONFIG=/Library/PostgreSQL/17/bin/pg_config
+export PG_CONFIG=/Library/PostgreSQL/16/bin/pg_config
 ```
 
 Then re-run the installation instructions (run `make clean` before `make` if needed). If `sudo` is needed for `make install`, use:
@@ -1006,11 +1025,11 @@ sudo --preserve-env=PG_CONFIG make install
 
 A few common paths on Mac are:
 
-- EDB installer - `/Library/PostgreSQL/17/bin/pg_config`
-- Homebrew (arm64) - `/opt/homebrew/opt/postgresql@17/bin/pg_config`
-- Homebrew (x86-64) - `/usr/local/opt/postgresql@17/bin/pg_config`
+- EDB installer - `/Library/PostgreSQL/16/bin/pg_config`
+- Homebrew (arm64) - `/opt/homebrew/opt/postgresql@16/bin/pg_config`
+- Homebrew (x86-64) - `/usr/local/opt/postgresql@16/bin/pg_config`
 
-Note: Replace `17` with your Postgres server version
+Note: Replace `16` with your Postgres server version
 
 ### Missing Header
 
@@ -1019,10 +1038,10 @@ If compilation fails with `fatal error: postgres.h: No such file or directory`,
 For Ubuntu and Debian, use:
 
 ```sh
-sudo apt install postgresql-server-dev-17
+sudo apt install postgresql-server-dev-16
 ```
 
-Note: Replace `17` with your Postgres server version
+Note: Replace `16` with your Postgres server version
 
 ### Missing SDK
 
@@ -1055,17 +1074,17 @@ If installation fails with `Access is denied`, re-run the installation instructi
 Get the [Docker image](https://hub.docker.com/r/pgvector/pgvector) with:
 
 ```sh
-docker pull pgvector/pgvector:pg17
+docker pull pgvector/pgvector:pg16
 ```
 
-This adds pgvector to the [Postgres image](https://hub.docker.com/_/postgres) (replace `17` with your Postgres server version, and run it the same way).
+This adds pgvector to the [Postgres image](https://hub.docker.com/_/postgres) (replace `16` with your Postgres server version, and run it the same way).
 
 You can also build the image manually:
 
 ```sh
 git clone --branch v0.7.4 https://github.com/pgvector/pgvector.git
 cd pgvector
-docker build --pull --build-arg PG_MAJOR=17 -t myuser/pgvector .
+docker build --pull --build-arg PG_MAJOR=16 -t myuser/pgvector .
 ```
 
 ### Homebrew
@@ -1076,7 +1095,7 @@ With Homebrew Postgres, you can use:
 brew install pgvector
 ```
 
-Note: This only adds it to the `postgresql@17` and `postgresql@14` formulas
+Note: This only adds it to the `postgresql@14` formula
 
 ### PGXN
 
@@ -1091,22 +1110,22 @@ pgxn install vector
 Debian and Ubuntu packages are available from the [PostgreSQL APT Repository](https://wiki.postgresql.org/wiki/Apt). Follow the [setup instructions](https://wiki.postgresql.org/wiki/Apt#Quickstart) and run:
 
 ```sh
-sudo apt install postgresql-17-pgvector
+sudo apt install postgresql-16-pgvector
 ```
 
-Note: Replace `17` with your Postgres server version
+Note: Replace `16` with your Postgres server version
 
 ### Yum
 
 RPM packages are available from the [PostgreSQL Yum Repository](https://yum.postgresql.org/). Follow the [setup instructions](https://www.postgresql.org/download/linux/redhat/) for your distribution and run:
 
 ```sh
-sudo yum install pgvector_17
+sudo yum install pgvector_16
 # or
-sudo dnf install pgvector_17
+sudo dnf install pgvector_16
 ```
 
-Note: Replace `17` with your Postgres server version
+Note: Replace `16` with your Postgres server version
 
 ### pkg
 
@@ -1195,7 +1214,6 @@ Thanks to:
 - [k-means++: The Advantage of Careful Seeding](https://theory.stanford.edu/~sergei/papers/kMeansPP-soda.pdf)
 - [Concept Decompositions for Large Sparse Text Data using Clustering](https://www.cs.utexas.edu/users/inderjit/public_papers/concept_mlj.pdf)
 - [Efficient and Robust Approximate Nearest Neighbor Search using Hierarchical Navigable Small World Graphs](https://arxiv.org/ftp/arxiv/papers/1603/1603.09320.pdf)
-- [HQANN: Efficient and Robust Similarity Search for Hybrid Queries with Structured and Unstructured Constraints](https://arxiv.org/pdf/2207.07940.pdf)
 
 ## History
 

sql/vector--0.7.4--0.8.0.sql

@@ -1,6 +1,8 @@
 -- complain if script is sourced in psql, rather than via CREATE EXTENSION
 \echo Use "ALTER EXTENSION vector UPDATE TO '0.8.0'" to load this file. \quit
 
+-- TODO minivec functions
+
 CREATE FUNCTION array_to_sparsevec(integer[], integer, boolean) RETURNS sparsevec
 	AS 'MODULE_PATHNAME' LANGUAGE C IMMUTABLE STRICT PARALLEL SAFE;
 
@@ -24,11 +26,3 @@ CREATE CAST (double precision[] AS sparsevec)
 
 CREATE CAST (numeric[] AS sparsevec)
 	WITH FUNCTION array_to_sparsevec(numeric[], integer, boolean) AS ASSIGNMENT;
-
-CREATE FUNCTION hnsw_attribute_distance(integer, integer) RETURNS float8
-	AS 'MODULE_PATHNAME', 'hnsw_int4_attribute_distance' LANGUAGE C IMMUTABLE STRICT PARALLEL SAFE;
-
-CREATE OPERATOR CLASS vector_integer_ops
-	DEFAULT FOR TYPE integer USING hnsw AS
-	OPERATOR 2 = (integer, integer),
-	FUNCTION 4 hnsw_attribute_distance(integer, integer);

sql/vector.sql

@@ -266,12 +266,18 @@ COMMENT ON ACCESS METHOD hnsw IS 'hnsw index access method';
 CREATE FUNCTION ivfflat_halfvec_support(internal) RETURNS internal
 	AS 'MODULE_PATHNAME' LANGUAGE C;
 
+CREATE FUNCTION ivfflat_minivec_support(internal) RETURNS internal
+	AS 'MODULE_PATHNAME' LANGUAGE C;
+
 CREATE FUNCTION ivfflat_bit_support(internal) RETURNS internal
 	AS 'MODULE_PATHNAME' LANGUAGE C;
 
 CREATE FUNCTION hnsw_halfvec_support(internal) RETURNS internal
 	AS 'MODULE_PATHNAME' LANGUAGE C;
 
+CREATE FUNCTION hnsw_minivec_support(internal) RETURNS internal
+	AS 'MODULE_PATHNAME' LANGUAGE C;
+
 CREATE FUNCTION hnsw_bit_support(internal) RETURNS internal
 	AS 'MODULE_PATHNAME' LANGUAGE C;
 
@@ -647,6 +653,295 @@ CREATE OPERATOR CLASS halfvec_l1_ops
 	FUNCTION 1 l1_distance(halfvec, halfvec),
 	FUNCTION 3 hnsw_halfvec_support(internal);
 
+-- minivec type
+
+CREATE TYPE minivec;
+
+CREATE FUNCTION minivec_in(cstring, oid, integer) RETURNS minivec
+	AS 'MODULE_PATHNAME' LANGUAGE C IMMUTABLE STRICT PARALLEL SAFE;
+
+CREATE FUNCTION minivec_out(minivec) RETURNS cstring
+	AS 'MODULE_PATHNAME' LANGUAGE C IMMUTABLE STRICT PARALLEL SAFE;
+
+CREATE FUNCTION minivec_typmod_in(cstring[]) RETURNS integer
+	AS 'MODULE_PATHNAME' LANGUAGE C IMMUTABLE STRICT PARALLEL SAFE;
+
+CREATE FUNCTION minivec_recv(internal, oid, integer) RETURNS minivec
+	AS 'MODULE_PATHNAME' LANGUAGE C IMMUTABLE STRICT PARALLEL SAFE;
+
+CREATE FUNCTION minivec_send(minivec) RETURNS bytea
+	AS 'MODULE_PATHNAME' LANGUAGE C IMMUTABLE STRICT PARALLEL SAFE;
+
+CREATE TYPE minivec (
+	INPUT     = minivec_in,
+	OUTPUT    = minivec_out,
+	TYPMOD_IN = minivec_typmod_in,
+	RECEIVE   = minivec_recv,
+	SEND      = minivec_send,
+	STORAGE   = external
+);
+
+-- minivec functions
+
+CREATE FUNCTION l2_distance(minivec, minivec) RETURNS float8
+	AS 'MODULE_PATHNAME', 'minivec_l2_distance' LANGUAGE C IMMUTABLE STRICT PARALLEL SAFE;
+
+CREATE FUNCTION inner_product(minivec, minivec) RETURNS float8
+	AS 'MODULE_PATHNAME', 'minivec_inner_product' LANGUAGE C IMMUTABLE STRICT PARALLEL SAFE;
+
+CREATE FUNCTION cosine_distance(minivec, minivec) RETURNS float8
+	AS 'MODULE_PATHNAME', 'minivec_cosine_distance' LANGUAGE C IMMUTABLE STRICT PARALLEL SAFE;
+
+CREATE FUNCTION l1_distance(minivec, minivec) RETURNS float8
+	AS 'MODULE_PATHNAME', 'minivec_l1_distance' LANGUAGE C IMMUTABLE STRICT PARALLEL SAFE;
+
+CREATE FUNCTION vector_dims(minivec) RETURNS integer
+	AS 'MODULE_PATHNAME', 'minivec_vector_dims' LANGUAGE C IMMUTABLE STRICT PARALLEL SAFE;
+
+CREATE FUNCTION l2_norm(minivec) RETURNS float8
+	AS 'MODULE_PATHNAME', 'minivec_l2_norm' LANGUAGE C IMMUTABLE STRICT PARALLEL SAFE;
+
+CREATE FUNCTION l2_normalize(minivec) RETURNS minivec
+	AS 'MODULE_PATHNAME', 'minivec_l2_normalize' LANGUAGE C IMMUTABLE STRICT PARALLEL SAFE;
+
+CREATE FUNCTION binary_quantize(minivec) RETURNS bit
+	AS 'MODULE_PATHNAME', 'minivec_binary_quantize' LANGUAGE C IMMUTABLE STRICT PARALLEL SAFE;
+
+CREATE FUNCTION subvector(minivec, int, int) RETURNS minivec
+	AS 'MODULE_PATHNAME', 'minivec_subvector' LANGUAGE C IMMUTABLE STRICT PARALLEL SAFE;
+
+-- minivec private functions
+
+CREATE FUNCTION minivec_add(minivec, minivec) RETURNS minivec
+	AS 'MODULE_PATHNAME' LANGUAGE C IMMUTABLE STRICT PARALLEL SAFE;
+
+CREATE FUNCTION minivec_sub(minivec, minivec) RETURNS minivec
+	AS 'MODULE_PATHNAME' LANGUAGE C IMMUTABLE STRICT PARALLEL SAFE;
+
+CREATE FUNCTION minivec_mul(minivec, minivec) RETURNS minivec
+	AS 'MODULE_PATHNAME' LANGUAGE C IMMUTABLE STRICT PARALLEL SAFE;
+
+CREATE FUNCTION minivec_concat(minivec, minivec) RETURNS minivec
+	AS 'MODULE_PATHNAME' LANGUAGE C IMMUTABLE STRICT PARALLEL SAFE;
+
+CREATE FUNCTION minivec_lt(minivec, minivec) RETURNS bool
+	AS 'MODULE_PATHNAME' LANGUAGE C IMMUTABLE STRICT PARALLEL SAFE;
+
+CREATE FUNCTION minivec_le(minivec, minivec) RETURNS bool
+	AS 'MODULE_PATHNAME' LANGUAGE C IMMUTABLE STRICT PARALLEL SAFE;
+
+CREATE FUNCTION minivec_eq(minivec, minivec) RETURNS bool
+	AS 'MODULE_PATHNAME' LANGUAGE C IMMUTABLE STRICT PARALLEL SAFE;
+
+CREATE FUNCTION minivec_ne(minivec, minivec) RETURNS bool
+	AS 'MODULE_PATHNAME' LANGUAGE C IMMUTABLE STRICT PARALLEL SAFE;
+
+CREATE FUNCTION minivec_ge(minivec, minivec) RETURNS bool
+	AS 'MODULE_PATHNAME' LANGUAGE C IMMUTABLE STRICT PARALLEL SAFE;
+
+CREATE FUNCTION minivec_gt(minivec, minivec) RETURNS bool
+	AS 'MODULE_PATHNAME' LANGUAGE C IMMUTABLE STRICT PARALLEL SAFE;
+
+CREATE FUNCTION minivec_cmp(minivec, minivec) RETURNS int4
+	AS 'MODULE_PATHNAME' LANGUAGE C IMMUTABLE STRICT PARALLEL SAFE;
+
+CREATE FUNCTION minivec_l2_squared_distance(minivec, minivec) RETURNS float8
+	AS 'MODULE_PATHNAME' LANGUAGE C IMMUTABLE STRICT PARALLEL SAFE;
+
+CREATE FUNCTION minivec_negative_inner_product(minivec, minivec) RETURNS float8
+	AS 'MODULE_PATHNAME' LANGUAGE C IMMUTABLE STRICT PARALLEL SAFE;
+
+CREATE FUNCTION minivec_spherical_distance(minivec, minivec) RETURNS float8
+	AS 'MODULE_PATHNAME' LANGUAGE C IMMUTABLE STRICT PARALLEL SAFE;
+
+-- minivec cast functions
+
+CREATE FUNCTION minivec(minivec, integer, boolean) RETURNS minivec
+	AS 'MODULE_PATHNAME' LANGUAGE C IMMUTABLE STRICT PARALLEL SAFE;
+
+CREATE FUNCTION minivec_to_vector(minivec, integer, boolean) RETURNS vector
+	AS 'MODULE_PATHNAME' LANGUAGE C IMMUTABLE STRICT PARALLEL SAFE;
+
+CREATE FUNCTION vector_to_minivec(vector, integer, boolean) RETURNS minivec
+	AS 'MODULE_PATHNAME' LANGUAGE C IMMUTABLE STRICT PARALLEL SAFE;
+
+CREATE FUNCTION array_to_minivec(integer[], integer, boolean) RETURNS minivec
+	AS 'MODULE_PATHNAME' LANGUAGE C IMMUTABLE STRICT PARALLEL SAFE;
+
+CREATE FUNCTION array_to_minivec(real[], integer, boolean) RETURNS minivec
+	AS 'MODULE_PATHNAME' LANGUAGE C IMMUTABLE STRICT PARALLEL SAFE;
+
+CREATE FUNCTION array_to_minivec(double precision[], integer, boolean) RETURNS minivec
+	AS 'MODULE_PATHNAME' LANGUAGE C IMMUTABLE STRICT PARALLEL SAFE;
+
+CREATE FUNCTION array_to_minivec(numeric[], integer, boolean) RETURNS minivec
+	AS 'MODULE_PATHNAME' LANGUAGE C IMMUTABLE STRICT PARALLEL SAFE;
+
+CREATE FUNCTION minivec_to_float4(minivec, integer, boolean) RETURNS real[]
+	AS 'MODULE_PATHNAME' LANGUAGE C IMMUTABLE STRICT PARALLEL SAFE;
+
+-- minivec casts
+
+CREATE CAST (minivec AS minivec)
+	WITH FUNCTION minivec(minivec, integer, boolean) AS IMPLICIT;
+
+CREATE CAST (minivec AS vector)
+	WITH FUNCTION minivec_to_vector(minivec, integer, boolean) AS ASSIGNMENT;
+
+CREATE CAST (vector AS minivec)
+	WITH FUNCTION vector_to_minivec(vector, integer, boolean) AS IMPLICIT;
+
+CREATE CAST (minivec AS real[])
+	WITH FUNCTION minivec_to_float4(minivec, integer, boolean) AS ASSIGNMENT;
+
+CREATE CAST (integer[] AS minivec)
+	WITH FUNCTION array_to_minivec(integer[], integer, boolean) AS ASSIGNMENT;
+
+CREATE CAST (real[] AS minivec)
+	WITH FUNCTION array_to_minivec(real[], integer, boolean) AS ASSIGNMENT;
+
+CREATE CAST (double precision[] AS minivec)
+	WITH FUNCTION array_to_minivec(double precision[], integer, boolean) AS ASSIGNMENT;
+
+CREATE CAST (numeric[] AS minivec)
+	WITH FUNCTION array_to_minivec(numeric[], integer, boolean) AS ASSIGNMENT;
+
+-- minivec operators
+
+CREATE OPERATOR <-> (
+	LEFTARG = minivec, RIGHTARG = minivec, PROCEDURE = l2_distance,
+	COMMUTATOR = '<->'
+);
+
+CREATE OPERATOR <#> (
+	LEFTARG = minivec, RIGHTARG = minivec, PROCEDURE = minivec_negative_inner_product,
+	COMMUTATOR = '<#>'
+);
+
+CREATE OPERATOR <=> (
+	LEFTARG = minivec, RIGHTARG = minivec, PROCEDURE = cosine_distance,
+	COMMUTATOR = '<=>'
+);
+
+CREATE OPERATOR <+> (
+	LEFTARG = minivec, RIGHTARG = minivec, PROCEDURE = l1_distance,
+	COMMUTATOR = '<+>'
+);
+
+CREATE OPERATOR + (
+	LEFTARG = minivec, RIGHTARG = minivec, PROCEDURE = minivec_add,
+	COMMUTATOR = +
+);
+
+CREATE OPERATOR - (
+	LEFTARG = minivec, RIGHTARG = minivec, PROCEDURE = minivec_sub
+);
+
+CREATE OPERATOR * (
+	LEFTARG = minivec, RIGHTARG = minivec, PROCEDURE = minivec_mul,
+	COMMUTATOR = *
+);
+
+CREATE OPERATOR || (
+	LEFTARG = minivec, RIGHTARG = minivec, PROCEDURE = minivec_concat
+);
+
+CREATE OPERATOR < (
+	LEFTARG = minivec, RIGHTARG = minivec, PROCEDURE = minivec_lt,
+	COMMUTATOR = > , NEGATOR = >= ,
+	RESTRICT = scalarltsel, JOIN = scalarltjoinsel
+);
+
+CREATE OPERATOR <= (
+	LEFTARG = minivec, RIGHTARG = minivec, PROCEDURE = minivec_le,
+	COMMUTATOR = >= , NEGATOR = > ,
+	RESTRICT = scalarlesel, JOIN = scalarlejoinsel
+);
+
+CREATE OPERATOR = (
+	LEFTARG = minivec, RIGHTARG = minivec, PROCEDURE = minivec_eq,
+	COMMUTATOR = = , NEGATOR = <> ,
+	RESTRICT = eqsel, JOIN = eqjoinsel
+);
+
+CREATE OPERATOR <> (
+	LEFTARG = minivec, RIGHTARG = minivec, PROCEDURE = minivec_ne,
+	COMMUTATOR = <> , NEGATOR = = ,
+	RESTRICT = eqsel, JOIN = eqjoinsel
+);
+
+CREATE OPERATOR >= (
+	LEFTARG = minivec, RIGHTARG = minivec, PROCEDURE = minivec_ge,
+	COMMUTATOR = <= , NEGATOR = < ,
+	RESTRICT = scalargesel, JOIN = scalargejoinsel
+);
+
+CREATE OPERATOR > (
+	LEFTARG = minivec, RIGHTARG = minivec, PROCEDURE = minivec_gt,
+	COMMUTATOR = < , NEGATOR = <= ,
+	RESTRICT = scalargtsel, JOIN = scalargtjoinsel
+);
+
+-- minivec op classes
+
+CREATE OPERATOR CLASS minivec_ops
+	DEFAULT FOR TYPE minivec USING btree AS
+	OPERATOR 1 < ,
+	OPERATOR 2 <= ,
+	OPERATOR 3 = ,
+	OPERATOR 4 >= ,
+	OPERATOR 5 > ,
+	FUNCTION 1 minivec_cmp(minivec, minivec);
+
+CREATE OPERATOR CLASS minivec_l2_ops
+	FOR TYPE minivec USING ivfflat AS
+	OPERATOR 1 <-> (minivec, minivec) FOR ORDER BY float_ops,
+	FUNCTION 1 minivec_l2_squared_distance(minivec, minivec),
+	FUNCTION 3 l2_distance(minivec, minivec),
+	FUNCTION 5 ivfflat_minivec_support(internal);
+
+CREATE OPERATOR CLASS minivec_ip_ops
+	FOR TYPE minivec USING ivfflat AS
+	OPERATOR 1 <#> (minivec, minivec) FOR ORDER BY float_ops,
+	FUNCTION 1 minivec_negative_inner_product(minivec, minivec),
+	FUNCTION 3 minivec_spherical_distance(minivec, minivec),
+	FUNCTION 4 l2_norm(minivec),
+	FUNCTION 5 ivfflat_minivec_support(internal);
+
+CREATE OPERATOR CLASS minivec_cosine_ops
+	FOR TYPE minivec USING ivfflat AS
+	OPERATOR 1 <=> (minivec, minivec) FOR ORDER BY float_ops,
+	FUNCTION 1 cosine_distance(minivec, minivec),
+	FUNCTION 2 l2_norm(minivec),
+	FUNCTION 3 minivec_spherical_distance(minivec, minivec),
+	FUNCTION 4 l2_norm(minivec),
+	FUNCTION 5 ivfflat_minivec_support(internal);
+
+CREATE OPERATOR CLASS minivec_l2_ops
+	FOR TYPE minivec USING hnsw AS
+	OPERATOR 1 <-> (minivec, minivec) FOR ORDER BY float_ops,
+	FUNCTION 1 minivec_l2_squared_distance(minivec, minivec),
+	FUNCTION 3 hnsw_minivec_support(internal);
+
+CREATE OPERATOR CLASS minivec_ip_ops
+	FOR TYPE minivec USING hnsw AS
+	OPERATOR 1 <#> (minivec, minivec) FOR ORDER BY float_ops,
+	FUNCTION 1 minivec_negative_inner_product(minivec, minivec),
+	FUNCTION 3 hnsw_minivec_support(internal);
+
+CREATE OPERATOR CLASS minivec_cosine_ops
+	FOR TYPE minivec USING hnsw AS
+	OPERATOR 1 <=> (minivec, minivec) FOR ORDER BY float_ops,
+	FUNCTION 1 cosine_distance(minivec, minivec),
+	FUNCTION 2 l2_norm(minivec),
+	FUNCTION 3 hnsw_minivec_support(internal);
+
+CREATE OPERATOR CLASS minivec_l1_ops
+	FOR TYPE minivec USING hnsw AS
+	OPERATOR 1 <+> (minivec, minivec) FOR ORDER BY float_ops,
+	FUNCTION 1 l1_distance(minivec, minivec),
+	FUNCTION 3 hnsw_minivec_support(internal);
+
 -- bit functions
 
 CREATE FUNCTION hamming_distance(bit, bit) RETURNS float8
@@ -916,13 +1211,3 @@ CREATE OPERATOR CLASS sparsevec_l1_ops
 	OPERATOR 1 <+> (sparsevec, sparsevec) FOR ORDER BY float_ops,
 	FUNCTION 1 l1_distance(sparsevec, sparsevec),
 	FUNCTION 3 hnsw_sparsevec_support(internal);
-
--- hnsw attributes
-
-CREATE FUNCTION hnsw_attribute_distance(integer, integer) RETURNS float8
-	AS 'MODULE_PATHNAME', 'hnsw_int4_attribute_distance' LANGUAGE C IMMUTABLE STRICT PARALLEL SAFE;
-
-CREATE OPERATOR CLASS vector_integer_ops
-	DEFAULT FOR TYPE integer USING hnsw AS
-	OPERATOR 2 = (integer, integer),
-	FUNCTION 4 hnsw_attribute_distance(integer, integer);

src/hnsw.c

@@ -12,22 +12,12 @@
 #include "utils/float.h"
 #include "utils/guc.h"
 #include "utils/selfuncs.h"
-#include "utils/spccache.h"
 
 #if PG_VERSION_NUM < 150000
 #define MarkGUCPrefixReserved(x) EmitWarningsOnPlaceholders(x)
 #endif
 
-static const struct config_enum_entry hnsw_iterative_search_options[] = {
-	{"off", HNSW_ITERATIVE_SEARCH_OFF, false},
-	{"on", HNSW_ITERATIVE_SEARCH_RELAXED, false},
-	{"strict", HNSW_ITERATIVE_SEARCH_STRICT, false},
-	{NULL, 0, false}
-};
-
 int			hnsw_ef_search;
-int			hnsw_iterative_search_max_tuples;
-int			hnsw_iterative_search;
 int			hnsw_lock_tranche_id;
 static relopt_kind hnsw_relopt_kind;
 
@@ -78,15 +68,6 @@ HnswInit(void)
 							"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);
 
-	DefineCustomEnumVariable("hnsw.iterative_search", "Sets iterative search",
-							 NULL, &hnsw_iterative_search,
-							 HNSW_ITERATIVE_SEARCH_OFF, hnsw_iterative_search_options, PGC_USERSET, 0, NULL, NULL, NULL);
-
-	/* TODO Ensure ivfflat.max_probes uses same value for "all" */
-	DefineCustomIntVariable("hnsw.iterative_search_max_tuples", "Sets the max number of candidates to visit for iterative search",
-							"-1 means all", &hnsw_iterative_search_max_tuples,
-							-1, -1, INT_MAX, PGC_USERSET, 0, NULL, NULL, NULL);
-
 	MarkGUCPrefixReserved("hnsw");
 }
 
@@ -118,9 +99,7 @@ hnswcostestimate(PlannerInfo *root, IndexPath *path, double loop_count,
 {
 	GenericCosts costs;
 	int			m;
-	double		ratio;
-	double		startupPages;
-	double		spc_seq_page_cost;
+	int			entryLevel;
 	Relation	index;
 
 	/* Never use index without order */
@@ -136,71 +115,21 @@ hnswcostestimate(PlannerInfo *root, IndexPath *path, double loop_count,
 
 	MemSet(&costs, 0, sizeof(costs));
 
-	genericcostestimate(root, path, loop_count, &costs);
-
 	index = index_open(path->indexinfo->indexoid, NoLock);
 	HnswGetMetaPageInfo(index, &m, NULL);
 	index_close(index, NoLock);
 
-	/*
-	 * HNSW cost estimation follows a formula that accounts for the total
-	 * number of tuples indexed combined with the parameters that most
-	 * influence the duration of the index scan, namely: m - the number of
-	 * tuples that are scanned in each step of the HNSW graph traversal
-	 * ef_search - which influences the total number of steps taken at layer 0
-	 *
-	 * The source of the vector data can impact how many steps it takes to
-	 * converge on the set of vectors to return to the executor. Currently, we
-	 * use a hardcoded scaling factor (HNSWScanScalingFactor) to help
-	 * influence that, but this could later become a configurable parameter
-	 * based on the cost estimations.
-	 *
-	 * The tuple estimator formula is below:
-	 *
-	 * numIndexTuples = entryLevel * m + layer0TuplesMax * layer0Selectivity
-	 *
-	 * "entryLevel * m" represents the floor of tuples we need to scan to get
-	 * to layer 0 (L0).
-	 *
-	 * "layer0TuplesMax" is the estimated total number of tuples we'd scan at
-	 * L0 if we weren't discarding already visited tuples as part of the scan.
-	 *
-	 * "layer0Selectivity" estimates the percentage of tuples that are scanned
-	 * at L0, accounting for previously visited tuples, multiplied by the
-	 * "scalingFactor" (currently hardcoded).
-	 */
-	if (path->indexinfo->tuples > 0)
-	{
-		double		scalingFactor = 0.55;
-		int			entryLevel = (int) (log(path->indexinfo->tuples) * HnswGetMl(m));
-		int			layer0TuplesMax = HnswGetLayerM(m, 0) * hnsw_ef_search;
-		double		layer0Selectivity = scalingFactor * log(path->indexinfo->tuples) / (log(m) * (1 + log(hnsw_ef_search)));
+	/* Approximate entry level */
+	entryLevel = (int) -log(1.0 / path->indexinfo->tuples) * HnswGetMl(m);
 
-		ratio = (entryLevel * m + layer0TuplesMax * layer0Selectivity) / path->indexinfo->tuples;
+	/* 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;
 
-		if (ratio > 1)
-			ratio = 1;
-	}
-	else
-		ratio = 1;
+	genericcostestimate(root, path, loop_count, &costs);
 
-	get_tablespace_page_costs(path->indexinfo->reltablespace, NULL, &spc_seq_page_cost);
-
-	/* Startup cost is cost before returning the first row */
-	costs.indexStartupCost = costs.indexTotalCost * ratio;
-
-	/* Adjust cost if needed since TOAST not included in seq scan cost */
-	startupPages = costs.numIndexPages * ratio;
-	if (startupPages > path->indexinfo->rel->pages && ratio < 0.5)
-	{
-		/* Change all page cost from random to sequential */
-		costs.indexStartupCost -= startupPages * (costs.spc_random_page_cost - spc_seq_page_cost);
-
-		/* Remove cost of extra pages */
-		costs.indexStartupCost -= (startupPages - path->indexinfo->rel->pages) * spc_seq_page_cost;
-	}
-
-	*indexStartupCost = costs.indexStartupCost;
+	/* Use total cost since most work happens before first tuple is returned */
+	*indexStartupCost = costs.indexTotalCost;
 	*indexTotalCost = costs.indexTotalCost;
 	*indexSelectivity = costs.indexSelectivity;
 	*indexCorrelation = costs.indexCorrelation;
@@ -245,13 +174,13 @@ hnswhandler(PG_FUNCTION_ARGS)
 	IndexAmRoutine *amroutine = makeNode(IndexAmRoutine);
 
 	amroutine->amstrategies = 0;
-	amroutine->amsupport = 4;
+	amroutine->amsupport = 3;
 	amroutine->amoptsprocnum = 0;
 	amroutine->amcanorder = false;
 	amroutine->amcanorderbyop = true;
 	amroutine->amcanbackward = false;	/* can change direction mid-scan */
 	amroutine->amcanunique = false;
-	amroutine->amcanmulticol = true;
+	amroutine->amcanmulticol = false;
 	amroutine->amoptionalkey = true;
 	amroutine->amsearcharray = false;
 	amroutine->amsearchnulls = false;
@@ -303,17 +232,3 @@ hnswhandler(PG_FUNCTION_ARGS)
 
 	PG_RETURN_POINTER(amroutine);
 }
-
-/*
- * Get the distance between two int4 attributes
- */
-PGDLLEXPORT PG_FUNCTION_INFO_V1(hnsw_int4_attribute_distance);
-Datum
-hnsw_int4_attribute_distance(PG_FUNCTION_ARGS)
-{
-	int32		a = PG_GETARG_INT32(0);
-	int32		b = PG_GETARG_INT32(1);
-	double		distance = ((double) a) - ((double) b);
-
-	PG_RETURN_FLOAT8(distance);
-}

src/hnsw.h

@@ -19,7 +19,6 @@
 #define HNSW_DISTANCE_PROC 1
 #define HNSW_NORM_PROC 2
 #define HNSW_TYPE_INFO_PROC 3
-#define HNSW_ATTRIBUTE_DISTANCE_PROC 4
 
 #define HNSW_VERSION	1
 #define HNSW_MAGIC_NUMBER 0xA953A953
@@ -89,9 +88,6 @@
 /* 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 HnswGetSearchCandidate(membername, ptr) pairingheap_container(HnswSearchCandidate, membername, ptr)
-#define HnswGetSearchCandidateConst(membername, ptr) pairingheap_const_container(HnswSearchCandidate, membername, ptr)
-
 #define HnswGetValue(base, element) PointerGetDatum(HnswPtrAccess(base, (element)->value))
 
 #if PG_VERSION_NUM < 140005
@@ -108,21 +104,10 @@
 #define HnswPtrPointer(hp) (hp).ptr
 #define HnswPtrOffset(hp) relptr_offset((hp).relptr)
 
-#define HnswUseIndexTuple(index) (IndexRelationGetNumberOfAttributes(index) > 1)
-
 /* Variables */
 extern int	hnsw_ef_search;
-extern int	hnsw_iterative_search;
-extern int	hnsw_iterative_search_max_tuples;
 extern int	hnsw_lock_tranche_id;
 
-typedef enum HnswIterativeSearchType
-{
-	HNSW_ITERATIVE_SEARCH_OFF,
-	HNSW_ITERATIVE_SEARCH_RELAXED,
-	HNSW_ITERATIVE_SEARCH_STRICT
-}			HnswIterativeSearchType;
-
 typedef struct HnswElementData HnswElementData;
 typedef struct HnswNeighborArray HnswNeighborArray;
 
@@ -136,7 +121,6 @@ HnswPtrDeclare(HnswElementData, HnswElementRelptr, HnswElementPtr);
 HnswPtrDeclare(HnswNeighborArray, HnswNeighborArrayRelptr, HnswNeighborArrayPtr);
 HnswPtrDeclare(HnswNeighborArrayPtr, HnswNeighborsRelptr, HnswNeighborsPtr);
 HnswPtrDeclare(char, DatumRelptr, DatumPtr);
-HnswPtrDeclare(IndexTupleData, IndexTupleRelptr, IndexTuplePtr);
 
 struct HnswElementData
 {
@@ -145,7 +129,6 @@ struct HnswElementData
 	uint8		heaptidsLength;
 	uint8		level;
 	uint8		deleted;
-	uint8		version;
 	uint32		hash;
 	HnswNeighborsPtr neighbors;
 	BlockNumber blkno;
@@ -153,7 +136,6 @@ struct HnswElementData
 	OffsetNumber neighborOffno;
 	BlockNumber neighborPage;
 	DatumPtr	value;
-	IndexTuplePtr itup;
 	LWLock		lock;
 };
 
@@ -178,8 +160,7 @@ typedef struct HnswSearchCandidate
 	pairingheap_node c_node;
 	pairingheap_node w_node;
 	HnswElementPtr element;
-	double		distance;
-	bool		matches;
+	float		distance;
 }			HnswSearchCandidate;
 
 /* HNSW index options */
@@ -204,8 +185,8 @@ typedef struct HnswGraph
 
 	/* Allocations state */
 	LWLock		allocatorLock;
-	Size		memoryUsed;
-	Size		memoryTotal;
+	long		memoryUsed;
+	long		memoryTotal;
 
 	/* Flushed state */
 	LWLock		flushLock;
@@ -256,20 +237,6 @@ typedef struct HnswTypeInfo
 	void		(*checkValue) (Pointer v);
 }			HnswTypeInfo;
 
-typedef struct HnswSupport
-{
-	FmgrInfo   *procinfo[2];
-	FmgrInfo   *normprocinfo;
-	Oid		   *collation;
-}			HnswSupport;
-
-typedef struct HnswQuery
-{
-	Datum		value;
-	IndexTuple	itup;
-	ScanKeyData *keyData;
-}			HnswQuery;
-
 typedef struct HnswBuildState
 {
 	/* Info */
@@ -289,15 +256,15 @@ typedef struct HnswBuildState
 	double		reltuples;
 
 	/* Support functions */
-	HnswSupport support;
+	FmgrInfo   *procinfo;
+	FmgrInfo   *normprocinfo;
+	Oid			collation;
 
 	/* Variables */
 	HnswGraph	graphData;
 	HnswGraph  *graph;
 	double		ml;
 	int			maxLevel;
-	bool		useIndexTuple;
-	TupleDesc	tupdesc;
 
 	/* Memory */
 	MemoryContext graphCtx;
@@ -339,10 +306,10 @@ typedef struct HnswElementTupleData
 	uint8		type;
 	uint8		level;
 	uint8		deleted;
-	uint8		version;
+	uint8		unused;
 	ItemPointerData heaptids[HNSW_HEAPTIDS];
 	ItemPointerData neighbortid;
-	uint16		unused;
+	uint16		unused2;
 	Vector		data;
 }			HnswElementTupleData;
 
@@ -351,41 +318,24 @@ typedef HnswElementTupleData * HnswElementTuple;
 typedef struct HnswNeighborTupleData
 {
 	uint8		type;
-	uint8		version;
+	uint8		unused;
 	uint16		count;
 	ItemPointerData indextids[FLEXIBLE_ARRAY_MEMBER];
 }			HnswNeighborTupleData;
 
 typedef HnswNeighborTupleData * HnswNeighborTuple;
 
-typedef union
-{
-	struct pointerhash_hash *pointers;
-	struct offsethash_hash *offsets;
-	struct tidhash_hash *tids;
-}			visited_hash;
-
-typedef union
-{
-	HnswElement element;
-	ItemPointerData indextid;
-}			HnswUnvisited;
-
 typedef struct HnswScanOpaqueData
 {
 	const		HnswTypeInfo *typeInfo;
 	bool		first;
 	List	   *w;
-	visited_hash v;
-	pairingheap *discarded;
-	HnswQuery	q;
-	int			m;
-	int64		tuples;
-	double		previousDistance;
 	MemoryContext tmpCtx;
 
 	/* Support functions */
-	HnswSupport support;
+	FmgrInfo   *procinfo;
+	FmgrInfo   *normprocinfo;
+	Oid			collation;
 }			HnswScanOpaqueData;
 
 typedef HnswScanOpaqueData * HnswScanOpaque;
@@ -403,7 +353,8 @@ typedef struct HnswVacuumState
 	int			efConstruction;
 
 	/* Support functions */
-	HnswSupport support;
+	FmgrInfo   *procinfo;
+	Oid			collation;
 
 	/* Variables */
 	struct tidhash_hash *deleted;
@@ -419,38 +370,33 @@ typedef struct HnswVacuumState
 int			HnswGetM(Relation index);
 int			HnswGetEfConstruction(Relation index);
 FmgrInfo   *HnswOptionalProcInfo(Relation index, uint16 procnum);
-void		HnswInitSupport(HnswSupport * support, Relation index);
 Datum		HnswNormValue(const HnswTypeInfo * typeInfo, Oid collation, Datum value);
-bool		HnswCheckNorm(HnswSupport * support, Datum value);
+bool		HnswCheckNorm(FmgrInfo *procinfo, Oid collation, Datum value);
 Buffer		HnswNewBuffer(Relation index, ForkNumber forkNum);
 void		HnswInitPage(Buffer buf, Page page);
 void		HnswInit(void);
-List	   *HnswSearchLayer(char *base, HnswQuery * q, List *ep, int ef, int lc, Relation index, HnswSupport * support, int m, bool inserting, HnswElement skipElement, bool inMemory, visited_hash * v, pairingheap **discarded, bool initVisited, int64 *tuples);
+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, HnswSupport * support, int m, int efConstruction, bool existing, bool inMemory);
-HnswSearchCandidate *HnswEntryCandidate(char *base, HnswElement em, HnswQuery * q, Relation rel, HnswSupport * support, bool loadVec, bool inMemory);
+void		HnswFindElementNeighbors(char *base, HnswElement element, HnswElement entryPoint, Relation index, FmgrInfo *procinfo, Oid collation, int m, int efConstruction, bool existing);
+HnswSearchCandidate *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);
-HnswNeighborArray *HnswInitNeighborArray(int lm, HnswAllocator * allocator);
 void		HnswInitNeighbors(char *base, HnswElement element, int m, HnswAllocator * alloc);
-bool		HnswInsertTupleOnDisk(Relation index, HnswSupport * support, IndexTuple itup, ItemPointer heaptid, bool building, TupleDesc tupdesc);
-void		HnswUpdateNeighborsOnDisk(Relation index, HnswSupport * support, HnswElement e, int m, bool checkExisting, bool building);
-void		HnswLoadElementFromTuple(HnswElement element, HnswElementTuple etup, bool loadHeaptids, bool loadVec, Relation index);
-void		HnswLoadElement(HnswElement element, double *distance, bool *matches, HnswQuery * q, Relation index, HnswSupport * support, bool loadVec, double *maxDistance);
-void		HnswSetElementTuple(char *base, HnswElementTuple etup, HnswElement element, bool useIndexTuple);
-void		HnswUpdateConnection(char *base, HnswNeighborArray * neighbors, HnswElement newElement, float distance, int lm, int *updateIdx, Relation index, HnswSupport * support);
-bool		HnswFormIndexTuple(IndexTuple *out, Datum *values, bool *isnull, const HnswTypeInfo * typeInfo, HnswSupport * support, TupleDesc tupdesc);
-bool		HnswLoadNeighborTids(HnswElement element, ItemPointerData *indextids, Relation index, int m, int lm, int lc);
+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, float *maxDistance);
+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);
 const		HnswTypeInfo *HnswGetTypeInfo(Relation index);
 PGDLLEXPORT void HnswParallelBuildMain(dsm_segment *seg, shm_toc *toc);
-Size		HnswGetElementTupleSize(char *base, HnswElement element, bool useIndexTuple);
-bool		HnswIndexTupleIsEqual(IndexTuple a, IndexTuple b, TupleDesc tupdesc);
 
 /* Index access methods */
 IndexBuildResult *hnswbuild(Relation heap, Relation index, IndexInfo *indexInfo);

src/hnswbuild.c

@@ -148,7 +148,6 @@ CreateGraphPages(HnswBuildState * buildstate)
 	Page		page;
 	HnswElementPtr iter = buildstate->graph->head;
 	char	   *base = buildstate->hnswarea;
-	bool		useIndexTuple = buildstate->useIndexTuple;
 
 	/* Calculate sizes */
 	maxSize = HNSW_MAX_SIZE;
@@ -168,6 +167,7 @@ CreateGraphPages(HnswBuildState * buildstate)
 		Size		etupSize;
 		Size		ntupSize;
 		Size		combinedSize;
+		Pointer		valuePtr = HnswPtrAccess(base, element->value);
 
 		/* Update iterator */
 		iter = element->next;
@@ -176,7 +176,7 @@ CreateGraphPages(HnswBuildState * buildstate)
 		MemSet(etup, 0, HNSW_TUPLE_ALLOC_SIZE);
 
 		/* Calculate sizes */
-		etupSize = HnswGetElementTupleSize(base, element, useIndexTuple);
+		etupSize = HNSW_ELEMENT_TUPLE_SIZE(VARSIZE_ANY(valuePtr));
 		ntupSize = HNSW_NEIGHBOR_TUPLE_SIZE(element->level, buildstate->m);
 		combinedSize = etupSize + ntupSize + sizeof(ItemIdData);
 
@@ -186,7 +186,7 @@ CreateGraphPages(HnswBuildState * buildstate)
 					(errcode(ERRCODE_PROGRAM_LIMIT_EXCEEDED),
 					 errmsg("index tuple too large")));
 
-		HnswSetElementTuple(base, etup, element, useIndexTuple);
+		HnswSetElementTuple(base, etup, element);
 
 		/* Keep element and neighbors on the same page if possible */
 		if (PageGetFreeSpace(page) < etupSize || (combinedSize <= maxSize && PageGetFreeSpace(page) < combinedSize))
@@ -327,18 +327,19 @@ AddDuplicateInMemory(HnswElement element, HnswElement dup)
  * Find duplicate element
  */
 static bool
-FindDuplicateInMemory(char *base, HnswElement element, bool useIndexTuple, TupleDesc tupdesc)
+FindDuplicateInMemory(char *base, HnswElement element)
 {
 	HnswNeighborArray *neighbors = HnswGetNeighbors(base, element, 0);
-	IndexTuple	itup = HnswPtrAccess(base, element->itup);
+	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 (!HnswIndexTupleIsEqual(itup, HnswPtrAccess(base, neighborElement->itup), tupdesc))
+		if (!datumIsEqual(value, neighborValue, false, -1))
 			return false;
 
 		/* Check for space */
@@ -365,7 +366,7 @@ AddElementInMemory(char *base, HnswGraph * graph, HnswElement element)
  * Update neighbors
  */
 static void
-UpdateNeighborsInMemory(char *base, Relation index, HnswSupport * support, HnswElement e, int m)
+UpdateNeighborsInMemory(char *base, FmgrInfo *procinfo, Oid collation, HnswElement e, int m)
 {
 	for (int lc = e->level; lc >= 0; lc--)
 	{
@@ -387,7 +388,7 @@ UpdateNeighborsInMemory(char *base, Relation index, HnswSupport * support, HnswE
 			Assert(neighborElement);
 
 			LWLockAcquire(&neighborElement->lock, LW_EXCLUSIVE);
-			HnswUpdateConnection(base, HnswGetNeighbors(base, neighborElement, lc), e, hc->distance, lm, NULL, index, support);
+			HnswUpdateConnection(base, e, hc, lm, lc, NULL, NULL, procinfo, collation);
 			LWLockRelease(&neighborElement->lock);
 		}
 	}
@@ -397,20 +398,20 @@ UpdateNeighborsInMemory(char *base, Relation index, HnswSupport * support, HnswE
  * Update graph in memory
  */
 static void
-UpdateGraphInMemory(HnswSupport * support, HnswElement element, int m, int efConstruction, HnswElement entryPoint, HnswBuildState * buildstate)
+UpdateGraphInMemory(FmgrInfo *procinfo, Oid collation, HnswElement element, int m, int efConstruction, HnswElement entryPoint, HnswBuildState * buildstate)
 {
 	HnswGraph  *graph = buildstate->graph;
 	char	   *base = buildstate->hnswarea;
 
 	/* Look for duplicate */
-	if (FindDuplicateInMemory(base, element, buildstate->useIndexTuple, buildstate->tupdesc))
+	if (FindDuplicateInMemory(base, element))
 		return;
 
 	/* Add element */
 	AddElementInMemory(base, graph, element);
 
 	/* Update neighbors */
-	UpdateNeighborsInMemory(base, buildstate->index, support, element, m);
+	UpdateNeighborsInMemory(base, procinfo, collation, element, m);
 
 	/* Update entry point if needed (already have lock) */
 	if (entryPoint == NULL || element->level > entryPoint->level)
@@ -423,9 +424,9 @@ UpdateGraphInMemory(HnswSupport * support, HnswElement element, int m, int efCon
 static void
 InsertTupleInMemory(HnswBuildState * buildstate, HnswElement element)
 {
-	Relation	index = buildstate->index;
+	FmgrInfo   *procinfo = buildstate->procinfo;
+	Oid			collation = buildstate->collation;
 	HnswGraph  *graph = buildstate->graph;
-	HnswSupport *support = &buildstate->support;
 	HnswElement entryPoint;
 	LWLock	   *entryLock = &graph->entryLock;
 	LWLock	   *entryWaitLock = &graph->entryWaitLock;
@@ -457,10 +458,10 @@ InsertTupleInMemory(HnswBuildState * buildstate, HnswElement element)
 	}
 
 	/* Find neighbors for element */
-	HnswFindElementNeighbors(base, element, entryPoint, index, support, m, efConstruction, false, true);
+	HnswFindElementNeighbors(base, element, entryPoint, NULL, procinfo, collation, m, efConstruction, false);
 
 	/* Update graph in memory */
-	UpdateGraphInMemory(support, element, m, efConstruction, entryPoint, buildstate);
+	UpdateGraphInMemory(procinfo, collation, element, m, efConstruction, entryPoint, buildstate);
 
 	/* Release entry lock */
 	LWLockRelease(entryLock);
@@ -472,24 +473,33 @@ InsertTupleInMemory(HnswBuildState * buildstate, HnswElement element)
 static bool
 InsertTuple(Relation index, Datum *values, bool *isnull, ItemPointer heaptid, HnswBuildState * buildstate)
 {
+	const		HnswTypeInfo *typeInfo = buildstate->typeInfo;
 	HnswGraph  *graph = buildstate->graph;
 	HnswElement element;
 	HnswAllocator *allocator = &buildstate->allocator;
-	HnswSupport *support = &buildstate->support;
+	Size		valueSize;
+	Pointer		valuePtr;
 	LWLock	   *flushLock = &graph->flushLock;
 	char	   *base = buildstate->hnswarea;
-	TupleDesc	tupdesc = buildstate->tupdesc;
-	IndexTuple	itup;
-	Size		itupSize;
-	IndexTuple	itupShared;
-	bool		unused;
 
-	/* Form index tuple */
-	if (!HnswFormIndexTuple(&itup, values, isnull, buildstate->typeInfo, support, tupdesc))
-		return false;
+	/* Detoast once for all calls */
+	Datum		value = PointerGetDatum(PG_DETOAST_DATUM(values[0]));
 
-	/* Get tuple size */
-	itupSize = IndexTupleSize(itup);
+	/* Check value */
+	if (typeInfo->checkValue != NULL)
+		typeInfo->checkValue(DatumGetPointer(value));
+
+	/* Normalize if needed */
+	if (buildstate->normprocinfo != NULL)
+	{
+		if (!HnswCheckNorm(buildstate->normprocinfo, buildstate->collation, value))
+			return false;
+
+		value = HnswNormValue(typeInfo, buildstate->collation, value);
+	}
+
+	/* Get datum size */
+	valueSize = VARSIZE_ANY(DatumGetPointer(value));
 
 	/* Ensure graph not flushed when inserting */
 	LWLockAcquire(flushLock, LW_SHARED);
@@ -499,7 +509,7 @@ InsertTuple(Relation index, Datum *values, bool *isnull, ItemPointer heaptid, Hn
 	{
 		LWLockRelease(flushLock);
 
-		return HnswInsertTupleOnDisk(index, support, itup, heaptid, true, tupdesc);
+		return HnswInsertTupleOnDisk(index, value, values, isnull, heaptid, true);
 	}
 
 	/*
@@ -531,12 +541,12 @@ InsertTuple(Relation index, Datum *values, bool *isnull, ItemPointer heaptid, Hn
 
 		LWLockRelease(flushLock);
 
-		return HnswInsertTupleOnDisk(index, support, itup, heaptid, true, tupdesc);
+		return HnswInsertTupleOnDisk(index, value, values, isnull, heaptid, true);
 	}
 
 	/* Ok, we can proceed to allocate the element */
 	element = HnswInitElement(base, heaptid, buildstate->m, buildstate->ml, buildstate->maxLevel, allocator);
-	itupShared = HnswAlloc(allocator, itupSize);
+	valuePtr = HnswAlloc(allocator, valueSize);
 
 	/*
 	 * We have now allocated the space needed for the element, so we don't
@@ -545,10 +555,9 @@ InsertTuple(Relation index, Datum *values, bool *isnull, ItemPointer heaptid, Hn
 	 */
 	LWLockRelease(&graph->allocatorLock);
 
-	/* Copy the tuple */
-	memcpy(itupShared, itup, itupSize);
-	HnswPtrStore(base, element->itup, itupShared);
-	HnswPtrStore(base, element->value, DatumGetPointer(index_getattr(itupShared, 1, tupdesc, &unused)));
+	/* Copy the datum */
+	memcpy(valuePtr, DatumGetPointer(value), valueSize);
+	HnswPtrStore(base, element->value, valuePtr);
 
 	/* Create a lock for the element */
 	LWLockInitialize(&element->lock, hnsw_lock_tranche_id);
@@ -598,7 +607,7 @@ BuildCallback(Relation index, ItemPointer tid, Datum *values,
  * Initialize the graph
  */
 static void
-InitGraph(HnswGraph * graph, char *base, Size memoryTotal)
+InitGraph(HnswGraph * graph, char *base, long memoryTotal)
 {
 	/* Initialize the lock tranche if needed */
 	HnswInitLockTranche();
@@ -675,19 +684,6 @@ InitBuildState(HnswBuildState * buildstate, Relation heap, Relation index, Index
 				(errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
 				 errmsg("type not supported for hnsw index")));
 
-	/* TODO See if needed */
-	if (IndexRelationGetNumberOfKeyAttributes(index) > 2)
-		elog(ERROR, "index cannot have more than two columns");
-
-	if (!OidIsValid(index_getprocid(index, 1, HNSW_DISTANCE_PROC)))
-		elog(ERROR, "first column must be a vector");
-
-	for (int i = 1; i < IndexRelationGetNumberOfKeyAttributes(index); i++)
-	{
-		if (!OidIsValid(index_getprocid(index, i + 1, HNSW_ATTRIBUTE_DISTANCE_PROC)))
-			elog(ERROR, "column %d cannot be a vector", i + 1);
-	}
-
 	/* Require column to have dimensions to be indexed */
 	if (buildstate->dimensions < 0)
 		ereport(ERROR,
@@ -708,14 +704,14 @@ InitBuildState(HnswBuildState * buildstate, Relation heap, Relation index, Index
 	buildstate->indtuples = 0;
 
 	/* Get support functions */
-	HnswInitSupport(&buildstate->support, index);
+	buildstate->procinfo = index_getprocinfo(index, 1, HNSW_DISTANCE_PROC);
+	buildstate->normprocinfo = HnswOptionalProcInfo(index, HNSW_NORM_PROC);
+	buildstate->collation = index->rd_indcollation[0];
 
-	InitGraph(&buildstate->graphData, NULL, (Size) maintenance_work_mem * 1024L);
+	InitGraph(&buildstate->graphData, NULL, maintenance_work_mem * 1024L);
 	buildstate->graph = &buildstate->graphData;
 	buildstate->ml = HnswGetMl(buildstate->m);
 	buildstate->maxLevel = HnswGetMaxLevel(buildstate->m);
-	buildstate->useIndexTuple = HnswUseIndexTuple(index);
-	buildstate->tupdesc = RelationGetDescr(index);
 
 	buildstate->graphCtx = GenerationContextCreate(CurrentMemoryContext,
 												   "Hnsw build graph context",

src/hnswinsert.c

@@ -36,7 +36,7 @@ GetInsertPage(Relation index)
  * Check for a free offset
  */
 static bool
-HnswFreeOffset(Relation index, Buffer buf, Page page, HnswElement element, Size etupSize, Size ntupSize, Buffer *nbuf, Page *npage, OffsetNumber *freeOffno, OffsetNumber *freeNeighborOffno, BlockNumber *newInsertPage, uint8 *tupleVersion)
+HnswFreeOffset(Relation index, Buffer buf, Page page, HnswElement element, Size etupSize, Size ntupSize, Buffer *nbuf, Page *npage, OffsetNumber *freeOffno, OffsetNumber *freeNeighborOffno, BlockNumber *newInsertPage)
 {
 	OffsetNumber offno;
 	OffsetNumber maxoffno = PageGetMaxOffsetNumber(page);
@@ -98,7 +98,6 @@ HnswFreeOffset(Relation index, Buffer buf, Page page, HnswElement element, Size
 			{
 				*freeOffno = offno;
 				*freeNeighborOffno = neighborOffno;
-				*tupleVersion = etup->version;
 				return true;
 			}
 			else if (*nbuf != buf)
@@ -154,12 +153,10 @@ AddElementOnDisk(Relation index, HnswElement e, int m, BlockNumber insertPage, B
 	OffsetNumber freeOffno = InvalidOffsetNumber;
 	OffsetNumber freeNeighborOffno = InvalidOffsetNumber;
 	BlockNumber newInsertPage = InvalidBlockNumber;
-	uint8		tupleVersion;
 	char	   *base = NULL;
-	bool		useIndexTuple = HnswUseIndexTuple(index);
 
 	/* Calculate sizes */
-	etupSize = HnswGetElementTupleSize(base, e, useIndexTuple);
+	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;
@@ -167,7 +164,7 @@ AddElementOnDisk(Relation index, HnswElement e, int m, BlockNumber insertPage, B
 
 	/* Prepare element tuple */
 	etup = palloc0(etupSize);
-	HnswSetElementTuple(base, etup, e, useIndexTuple);
+	HnswSetElementTuple(base, etup, e);
 
 	/* Prepare neighbor tuple */
 	ntup = palloc0(ntupSize);
@@ -205,7 +202,7 @@ AddElementOnDisk(Relation index, HnswElement e, int m, BlockNumber insertPage, B
 		}
 
 		/* Next, try space from a deleted element */
-		if (HnswFreeOffset(index, buf, page, e, etupSize, ntupSize, &nbuf, &npage, &freeOffno, &freeNeighborOffno, &newInsertPage, &tupleVersion))
+		if (HnswFreeOffset(index, buf, page, e, etupSize, ntupSize, &nbuf, &npage, &freeOffno, &freeNeighborOffno, &newInsertPage))
 		{
 			if (nbuf != buf)
 			{
@@ -215,10 +212,6 @@ AddElementOnDisk(Relation index, HnswElement e, int m, BlockNumber insertPage, B
 					npage = GenericXLogRegisterBuffer(state, nbuf, 0);
 			}
 
-			/* Set tuple version */
-			etup->version = tupleVersion;
-			ntup->version = tupleVersion;
-
 			break;
 		}
 
@@ -341,110 +334,6 @@ AddElementOnDisk(Relation index, HnswElement e, int m, BlockNumber insertPage, B
 		*updatedInsertPage = newInsertPage;
 }
 
-/*
- * Load neighbors
- */
-static HnswNeighborArray *
-HnswLoadNeighbors(HnswElement element, Relation index, int m, int lm, int lc)
-{
-	char	   *base = NULL;
-	HnswNeighborArray *neighbors = HnswInitNeighborArray(lm, NULL);
-	ItemPointerData indextids[HNSW_MAX_M * 2];
-
-	if (!HnswLoadNeighborTids(element, indextids, index, m, lm, lc))
-		return neighbors;
-
-	for (int i = 0; i < lm; i++)
-	{
-		ItemPointer indextid = &indextids[i];
-		HnswElement e;
-		HnswCandidate *hc;
-
-		if (!ItemPointerIsValid(indextid))
-			break;
-
-		e = HnswInitElementFromBlock(ItemPointerGetBlockNumber(indextid), ItemPointerGetOffsetNumber(indextid));
-		hc = &neighbors->items[neighbors->length++];
-		HnswPtrStore(base, hc->element, e);
-	}
-
-	return neighbors;
-}
-
-/*
- * Load elements for insert
- */
-static void
-LoadElementsForInsert(HnswNeighborArray * neighbors, HnswQuery * q, int *idx, Relation index, HnswSupport * support)
-{
-	char	   *base = NULL;
-
-	for (int i = 0; i < neighbors->length; i++)
-	{
-		HnswCandidate *hc = &neighbors->items[i];
-		HnswElement element = HnswPtrAccess(base, hc->element);
-		double		distance;
-		bool		matches;
-
-		HnswLoadElement(element, &distance, &matches, q, index, support, true, NULL);
-		hc->distance = distance;
-
-		/* Prune element if being deleted */
-		if (element->heaptidsLength == 0)
-		{
-			*idx = i;
-			break;
-		}
-	}
-}
-
-/*
- * Get update index
- */
-static int
-GetUpdateIndex(HnswElement element, HnswElement newElement, float distance, int m, int lm, int lc, Relation index, HnswSupport * support, MemoryContext updateCtx)
-{
-	char	   *base = NULL;
-	int			idx = -1;
-	HnswNeighborArray *neighbors;
-	MemoryContext oldCtx = MemoryContextSwitchTo(updateCtx);
-
-	/*
-	 * Get latest neighbors since they may have changed. Do not lock yet since
-	 * selecting neighbors can take time. Could use optimistic locking to
-	 * retry if another update occurs before getting exclusive lock.
-	 */
-	neighbors = HnswLoadNeighbors(element, index, m, lm, lc);
-
-	/*
-	 * 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.
-	 */
-
-	if (neighbors->length < lm)
-		idx = -2;
-	else
-	{
-		HnswQuery	q;
-
-		q.value = HnswGetValue(base, element);
-		q.itup = HnswPtrAccess(base, element->itup);
-		q.keyData = NULL;
-
-		LoadElementsForInsert(neighbors, &q, &idx, index, support);
-
-		if (idx == -1)
-			HnswUpdateConnection(base, neighbors, newElement, distance, lm, &idx, index, support);
-	}
-
-	MemoryContextSwitchTo(oldCtx);
-	MemoryContextReset(updateCtx);
-
-	return idx;
-}
-
 /*
  * Check if connection already exists
  */
@@ -465,94 +354,14 @@ ConnectionExists(HnswElement e, HnswNeighborTuple ntup, int startIdx, int lm)
 	return false;
 }
 
-/*
- * Update neighbor
- */
-static void
-UpdateNeighborOnDisk(HnswElement element, HnswElement newElement, int idx, int m, int lm, int lc, Relation index, bool checkExisting, bool building)
-{
-	Buffer		buf;
-	Page		page;
-	GenericXLogState *state;
-	HnswNeighborTuple ntup;
-	int			startIdx;
-	OffsetNumber offno = element->neighborOffno;
-
-	/* Register page */
-	buf = ReadBuffer(index, element->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 = (element->level - lc) * m;
-
-	/* Check for existing connection */
-	if (checkExisting && ConnectionExists(newElement, 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, newElement->blkno, newElement->offno);
-
-		/* Commit */
-		if (building)
-			MarkBufferDirty(buf);
-		else
-			GenericXLogFinish(state);
-	}
-	else if (!building)
-		GenericXLogAbort(state);
-
-	UnlockReleaseBuffer(buf);
-}
-
 /*
  * Update neighbors
  */
 void
-HnswUpdateNeighborsOnDisk(Relation index, HnswSupport * support, HnswElement e, int m, bool checkExisting, bool building)
+HnswUpdateNeighborsOnDisk(Relation index, FmgrInfo *procinfo, Oid collation, HnswElement e, int m, bool checkExisting, bool building)
 {
 	char	   *base = NULL;
 
-	/* Use separate memory context to improve performance for larger vectors */
-	MemoryContext updateCtx = GenerationContextCreate(CurrentMemoryContext,
-													  "Hnsw insert update context",
-#if PG_VERSION_NUM >= 150000
-													  128 * 1024, 128 * 1024,
-#endif
-													  128 * 1024);
-
 	for (int lc = e->level; lc >= 0; lc--)
 	{
 		int			lm = HnswGetLayerM(m, lc);
@@ -561,20 +370,96 @@ HnswUpdateNeighborsOnDisk(Relation index, HnswSupport * support, HnswElement e,
 		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);
-			int			idx;
+			OffsetNumber offno = neighborElement->neighborOffno;
 
-			idx = GetUpdateIndex(neighborElement, e, hc->distance, m, lm, lc, index, support, updateCtx);
+			/*
+			 * Get latest neighbors since they may have changed. Do not lock
+			 * yet since selecting neighbors can take time. Could use
+			 * optimistic locking to retry if another update occurs before
+			 * getting exclusive lock.
+			 */
+			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;
 
-			UpdateNeighborOnDisk(neighborElement, e, idx, m, lm, lc, index, checkExisting, building);
+			/* 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);
 		}
 	}
-
-	MemoryContextDelete(updateCtx);
 }
 
 /*
@@ -637,30 +522,21 @@ AddDuplicateOnDisk(Relation index, HnswElement element, HnswElement dup, bool bu
  * Find duplicate element
  */
 static bool
-FindDuplicateOnDisk(Relation index, HnswElement element, bool building, TupleDesc tupdesc)
+FindDuplicateOnDisk(Relation index, HnswElement element, bool building)
 {
 	char	   *base = NULL;
 	HnswNeighborArray *neighbors = HnswGetNeighbors(base, element, 0);
 	Datum		value = HnswGetValue(base, element);
-	IndexTuple	itup = HnswPtrAccess(base, element->itup);
 
 	for (int i = 0; i < neighbors->length; i++)
 	{
 		HnswCandidate *neighbor = &neighbors->items[i];
 		HnswElement neighborElement = HnswPtrAccess(base, neighbor->element);
+		Datum		neighborValue = HnswGetValue(base, neighborElement);
 
-		if (HnswUseIndexTuple(index))
-		{
-			/* Exit early since ordered by distance */
-			if (!HnswIndexTupleIsEqual(itup, HnswPtrAccess(base, neighborElement->itup), tupdesc))
-				return false;
-		}
-		else
-		{
-			/* Exit early since ordered by distance */
-			if (!datumIsEqual(value, HnswGetValue(base, neighborElement), false, -1))
-				return false;
-		}
+		/* Exit early since ordered by distance */
+		if (!datumIsEqual(value, neighborValue, false, -1))
+			return false;
 
 		if (AddDuplicateOnDisk(index, element, neighborElement, building))
 			return true;
@@ -673,12 +549,12 @@ FindDuplicateOnDisk(Relation index, HnswElement element, bool building, TupleDes
  * Update graph on disk
  */
 static void
-UpdateGraphOnDisk(Relation index, HnswSupport * support, HnswElement element, int m, int efConstruction, HnswElement entryPoint, bool building, TupleDesc tupdesc)
+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, tupdesc))
+	if (FindDuplicateOnDisk(index, element, building))
 		return;
 
 	/* Add element */
@@ -689,7 +565,7 @@ UpdateGraphOnDisk(Relation index, HnswSupport * support, HnswElement element, in
 		HnswUpdateMetaPage(index, 0, NULL, newInsertPage, MAIN_FORKNUM, building);
 
 	/* Update neighbors */
-	HnswUpdateNeighborsOnDisk(index, support, element, m, false, building);
+	HnswUpdateNeighborsOnDisk(index, procinfo, collation, element, m, false, building);
 
 	/* Update entry point if needed */
 	if (entryPoint == NULL || element->level > entryPoint->level)
@@ -700,15 +576,16 @@ UpdateGraphOnDisk(Relation index, HnswSupport * support, HnswElement element, in
  * Insert a tuple into the index
  */
 bool
-HnswInsertTupleOnDisk(Relation index, HnswSupport * support, IndexTuple itup, ItemPointer heaptid, bool building, TupleDesc tupdesc)
+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;
-	bool		unused;
 
 	/*
 	 * Get a shared lock. This allows vacuum to ensure no in-flight inserts
@@ -721,9 +598,8 @@ HnswInsertTupleOnDisk(Relation index, HnswSupport * support, IndexTuple itup, It
 	HnswGetMetaPageInfo(index, &m, &entryPoint);
 
 	/* Create an element */
-	element = HnswInitElement(base, heaptid, m, HnswGetMl(m), HnswGetMaxLevel(m), NULL);
-	HnswPtrStore(base, element->itup, itup);
-	HnswPtrStore(base, element->value, DatumGetPointer(index_getattr(itup, 1, tupdesc, &unused)));
+	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)
@@ -740,10 +616,10 @@ HnswInsertTupleOnDisk(Relation index, HnswSupport * support, IndexTuple itup, It
 	}
 
 	/* Find neighbors for element */
-	HnswFindElementNeighbors(base, element, entryPoint, index, support, m, efConstruction, false, false);
+	HnswFindElementNeighbors(base, element, entryPoint, index, procinfo, collation, m, efConstruction, false);
 
 	/* Update graph on disk */
-	UpdateGraphOnDisk(index, support, element, m, efConstruction, entryPoint, building, tupdesc);
+	UpdateGraphOnDisk(index, procinfo, collation, element, m, efConstruction, entryPoint, building);
 
 	/* Release lock */
 	UnlockPage(index, HNSW_UPDATE_LOCK, lockmode);
@@ -755,20 +631,31 @@ HnswInsertTupleOnDisk(Relation index, HnswSupport * support, IndexTuple itup, It
  * Insert a tuple into the index
  */
 static void
-HnswInsertTuple(Relation index, Datum *values, bool *isnull, ItemPointer heaptid)
+HnswInsertTuple(Relation index, Datum *values, bool *isnull, ItemPointer heap_tid)
 {
-	IndexTuple	itup;
+	Datum		value;
 	const		HnswTypeInfo *typeInfo = HnswGetTypeInfo(index);
-	TupleDesc	tupdesc = RelationGetDescr(index);
-	HnswSupport support;
+	FmgrInfo   *normprocinfo;
+	Oid			collation = index->rd_indcollation[0];
 
-	HnswInitSupport(&support, index);
+	/* Detoast once for all calls */
+	value = PointerGetDatum(PG_DETOAST_DATUM(values[0]));
 
-	/* Form index tuple */
-	if (!HnswFormIndexTuple(&itup, values, isnull, typeInfo, &support, tupdesc))
-		return;
+	/* Check value */
+	if (typeInfo->checkValue != NULL)
+		typeInfo->checkValue(DatumGetPointer(value));
 
-	HnswInsertTupleOnDisk(index, &support, itup, heaptid, false, tupdesc);
+	/* Normalize if needed */
+	normprocinfo = HnswOptionalProcInfo(index, HNSW_NORM_PROC);
+	if (normprocinfo != NULL)
+	{
+		if (!HnswCheckNorm(normprocinfo, collation, value))
+			return;
+
+		value = HnswNormValue(typeInfo, collation, value);
+	}
+
+	HnswInsertTupleOnDisk(index, value, values, isnull, heap_tid, false);
 }
 
 /*

src/hnswscan.c

@@ -5,78 +5,39 @@
 #include "pgstat.h"
 #include "storage/bufmgr.h"
 #include "storage/lmgr.h"
-#include "utils/float.h"
 #include "utils/memutils.h"
 
 /*
  * Algorithm 5 from paper
  */
 static List *
-GetScanItems(IndexScanDesc scan, Datum value)
+GetScanItems(IndexScanDesc scan, Datum q)
 {
 	HnswScanOpaque so = (HnswScanOpaque) scan->opaque;
 	Relation	index = scan->indexRelation;
-	HnswSupport *support = &so->support;
+	FmgrInfo   *procinfo = so->procinfo;
+	Oid			collation = so->collation;
 	List	   *ep;
 	List	   *w;
 	int			m;
 	HnswElement entryPoint;
 	char	   *base = NULL;
-	bool		inMemory = false;
-	HnswQuery  *q = &so->q;
-
-	q->value = value;
-	q->itup = NULL;
-	q->keyData = scan->keyData;
 
 	/* Get m and entry point */
 	HnswGetMetaPageInfo(index, &m, &entryPoint);
 
-	so->m = m;
-
 	if (entryPoint == NULL)
 		return NIL;
 
-	ep = list_make1(HnswEntryCandidate(base, entryPoint, q, index, support, false, inMemory));
+	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, support, m, false, NULL, inMemory, NULL, NULL, true, NULL);
+		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, support, m, false, NULL, inMemory, &so->v, hnsw_iterative_search != HNSW_ITERATIVE_SEARCH_OFF ? &so->discarded : NULL, true, &so->tuples);
-}
-
-/*
- * Resume scan at ground level with discarded candidates
- */
-static List *
-ResumeScanItems(IndexScanDesc scan)
-{
-	HnswScanOpaque so = (HnswScanOpaque) scan->opaque;
-	Relation	index = scan->indexRelation;
-	List	   *ep = NIL;
-	char	   *base = NULL;
-	int			batch_size = hnsw_ef_search;
-
-	if (pairingheap_is_empty(so->discarded))
-		return NIL;
-
-	/* Get next batch of candidates */
-	for (int i = 0; i < batch_size; i++)
-	{
-		HnswSearchCandidate *sc;
-
-		if (pairingheap_is_empty(so->discarded))
-			break;
-
-		sc = HnswGetSearchCandidate(w_node, pairingheap_remove_first(so->discarded));
-
-		ep = lappend(ep, sc);
-	}
-
-	return HnswSearchLayer(base, &so->q, ep, batch_size, 0, index, &so->support, so->m, false, NULL, false, &so->v, &so->discarded, false, &so->tuples);
+	return HnswSearchLayer(base, q, ep, hnsw_ef_search, 0, index, procinfo, collation, m, false, NULL);
 }
 
 /*
@@ -99,8 +60,8 @@ GetScanValue(IndexScanDesc scan)
 		Assert(!VARATT_IS_EXTENDED(DatumGetPointer(value)));
 
 		/* Normalize if needed */
-		if (so->support.normprocinfo != NULL)
-			value = HnswNormValue(so->typeInfo, so->support.collation[0], value);
+		if (so->normprocinfo != NULL)
+			value = HnswNormValue(so->typeInfo, so->collation, value);
 	}
 
 	return value;
@@ -120,14 +81,14 @@ hnswbeginscan(Relation index, int nkeys, int norderbys)
 	so = (HnswScanOpaque) palloc(sizeof(HnswScanOpaqueData));
 	so->typeInfo = HnswGetTypeInfo(index);
 	so->first = true;
-	so->v.tids = NULL;
-	so->discarded = NULL;
 	so->tmpCtx = AllocSetContextCreate(CurrentMemoryContext,
 									   "Hnsw scan temporary context",
 									   ALLOCSET_DEFAULT_SIZES);
 
 	/* Set support functions */
-	HnswInitSupport(&so->support, index);
+	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;
 
@@ -142,15 +103,7 @@ hnswrescan(IndexScanDesc scan, ScanKey keys, int nkeys, ScanKey orderbys, int no
 {
 	HnswScanOpaque so = (HnswScanOpaque) scan->opaque;
 
-	if (so->v.tids != NULL)
-		tidhash_reset(so->v.tids);
-
-	if (so->discarded != NULL)
-		pairingheap_reset(so->discarded);
-
 	so->first = true;
-	so->tuples = 0;
-	so->previousDistance = -get_float8_infinity();
 	MemoryContextReset(so->tmpCtx);
 
 	if (keys && scan->numberOfKeys > 0)
@@ -212,100 +165,22 @@ hnswgettuple(IndexScanDesc scan, ScanDirection dir)
 #endif
 	}
 
-	for (;;)
+	while (list_length(so->w) > 0)
 	{
 		char	   *base = NULL;
-		HnswSearchCandidate *sc;
-		HnswElement element;
+		HnswSearchCandidate *hc = llast(so->w);
+		HnswElement element = HnswPtrAccess(base, hc->element);
 		ItemPointer heaptid;
 
-		if (list_length(so->w) == 0)
-		{
-			if (hnsw_iterative_search == HNSW_ITERATIVE_SEARCH_OFF)
-				break;
-
-			/* Empty index */
-			if (so->discarded == NULL)
-				break;
-
-			/* Reached max number of additional tuples */
-			if (hnsw_iterative_search_max_tuples != -1 && so->tuples >= hnsw_iterative_search_max_tuples)
-			{
-				if (pairingheap_is_empty(so->discarded))
-					break;
-
-				/* Return remaining tuples */
-				so->w = lappend(so->w, HnswGetSearchCandidate(w_node, pairingheap_remove_first(so->discarded)));
-			}
-			/* Prevent scans from consuming too much memory */
-			else if (MemoryContextMemAllocated(so->tmpCtx, false) > (Size) work_mem * 1024L)
-			{
-				if (pairingheap_is_empty(so->discarded))
-				{
-					ereport(DEBUG1,
-							(errmsg("hnsw index scan exceeded work_mem after " INT64_FORMAT " tuples", so->tuples),
-							 errhint("Increase work_mem to scan more tuples.")));
-
-					break;
-				}
-
-				/* Return remaining tuples */
-				so->w = lappend(so->w, HnswGetSearchCandidate(w_node, pairingheap_remove_first(so->discarded)));
-			}
-			else
-			{
-				/*
-				 * Locking ensures when neighbors are read, the elements they
-				 * reference will not be deleted (and replaced) during the
-				 * iteration.
-				 *
-				 * Elements loaded into memory on previous iterations may have
-				 * been deleted (and replaced), so when reading neighbors, the
-				 * element version must be checked.
-				 */
-				LockPage(scan->indexRelation, HNSW_SCAN_LOCK, ShareLock);
-
-				so->w = ResumeScanItems(scan);
-
-				UnlockPage(scan->indexRelation, HNSW_SCAN_LOCK, ShareLock);
-
-#if defined(HNSW_MEMORY)
-				elog(INFO, "memory: %zu KB", MemoryContextMemAllocated(so->tmpCtx, false) / 1024);
-#endif
-			}
-
-			if (list_length(so->w) == 0)
-				break;
-		}
-
-		sc = llast(so->w);
-		element = HnswPtrAccess(base, sc->element);
-
 		/* Move to next element if no valid heap TIDs */
-		if (!sc->matches || element->heaptidsLength == 0)
+		if (element->heaptidsLength == 0)
 		{
 			so->w = list_delete_last(so->w);
-
-			/* Mark memory as free for next iteration */
-			if (hnsw_iterative_search != HNSW_ITERATIVE_SEARCH_OFF)
-			{
-				pfree(element);
-				pfree(sc);
-			}
-
 			continue;
 		}
 
 		heaptid = &element->heaptids[--element->heaptidsLength];
 
-		if (hnsw_iterative_search == HNSW_ITERATIVE_SEARCH_STRICT)
-		{
-			if (sc->distance < so->previousDistance)
-				continue;
-
-			so->previousDistance = sc->distance;
-		}
-
 		MemoryContextSwitchTo(oldCtx);
 
 		scan->xs_heaptid = *heaptid;

src/hnswvacuum.c

@@ -184,12 +184,13 @@ static void
 RepairGraphElement(HnswVacuumState * vacuumstate, HnswElement element, HnswElement entryPoint)
 {
 	Relation	index = vacuumstate->index;
-	HnswSupport *support = &vacuumstate->support;
 	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);
@@ -204,7 +205,7 @@ RepairGraphElement(HnswVacuumState * vacuumstate, HnswElement element, HnswEleme
 	element->heaptidsLength = 0;
 
 	/* Find neighbors for element, skipping itself */
-	HnswFindElementNeighbors(base, element, entryPoint, index, support, m, efConstruction, true, false);
+	HnswFindElementNeighbors(base, element, entryPoint, index, procinfo, collation, m, efConstruction, true);
 
 	/* Zero memory for each element */
 	MemSet(ntup, 0, HNSW_TUPLE_ALLOC_SIZE);
@@ -228,7 +229,7 @@ RepairGraphElement(HnswVacuumState * vacuumstate, HnswElement element, HnswEleme
 	UnlockReleaseBuffer(buf);
 
 	/* Update neighbors */
-	HnswUpdateNeighborsOnDisk(index, support, element, m, true, false);
+	HnswUpdateNeighborsOnDisk(index, procinfo, collation, element, m, true, false);
 }
 
 /*
@@ -238,7 +239,6 @@ static void
 RepairGraphEntryPoint(HnswVacuumState * vacuumstate)
 {
 	Relation	index = vacuumstate->index;
-	HnswSupport *support = &vacuumstate->support;
 	HnswElement highestPoint = &vacuumstate->highestPoint;
 	HnswElement entryPoint;
 	MemoryContext oldCtx = MemoryContextSwitchTo(vacuumstate->tmpCtx);
@@ -256,7 +256,7 @@ RepairGraphEntryPoint(HnswVacuumState * vacuumstate)
 		LockPage(index, HNSW_UPDATE_LOCK, ShareLock);
 
 		/* Load element */
-		HnswLoadElement(highestPoint, NULL, NULL, NULL, index, support, true, NULL);
+		HnswLoadElement(highestPoint, NULL, NULL, index, vacuumstate->procinfo, vacuumstate->collation, true, NULL);
 
 		/* Repair if needed */
 		if (NeedsUpdated(vacuumstate, highestPoint))
@@ -294,7 +294,7 @@ RepairGraphEntryPoint(HnswVacuumState * vacuumstate)
 			 * 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, NULL, index, support, true, NULL);
+			HnswLoadElement(entryPoint, NULL, NULL, index, vacuumstate->procinfo, vacuumstate->collation, true, NULL);
 
 			if (NeedsUpdated(vacuumstate, entryPoint))
 			{
@@ -370,7 +370,7 @@ RepairGraph(HnswVacuumState * vacuumstate)
 
 			/* Create an element */
 			element = HnswInitElementFromBlock(blkno, offno);
-			HnswLoadElementFromTuple(element, etup, false, true, index);
+			HnswLoadElementFromTuple(element, etup, false, true);
 
 			elements = lappend(elements, element);
 		}
@@ -440,7 +440,6 @@ MarkDeleted(HnswVacuumState * vacuumstate)
 	BlockNumber insertPage = InvalidBlockNumber;
 	Relation	index = vacuumstate->index;
 	BufferAccessStrategy bas = vacuumstate->bas;
-	bool		useIndexTuple = HnswUseIndexTuple(index);
 
 	/*
 	 * Wait for index scans to complete. Scans before this point may contain
@@ -522,27 +521,12 @@ MarkDeleted(HnswVacuumState * vacuumstate)
 
 			/* Overwrite element */
 			etup->deleted = 1;
-			if (useIndexTuple)
-			{
-				IndexTuple	itup = (IndexTuple) &etup->data;
-
-				MemSet(itup, 0, IndexTupleSize(itup));
-			}
-			else
-				MemSet(&etup->data, 0, VARSIZE_ANY(&etup->data));
+			MemSet(&etup->data, 0, VARSIZE_ANY(&etup->data));
 
 			/* Overwrite neighbors */
 			for (int i = 0; i < ntup->count; i++)
 				ItemPointerSetInvalid(&ntup->indextids[i]);
 
-			/* Increment version */
-			/* This is used to avoid incorrect reads for iterative scans */
-			/* Reserve some bits for future use */
-			etup->version++;
-			if (etup->version > 15)
-				etup->version = 1;
-			ntup->version = etup->version;
-
 			/*
 			 * We modified the tuples in place, no need to call
 			 * PageIndexTupleOverwrite
@@ -589,13 +573,13 @@ InitVacuumState(HnswVacuumState * vacuumstate, IndexVacuumInfo *info, IndexBulkD
 	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);
 
-	HnswInitSupport(&vacuumstate->support, index);
-
 	/* Get m from metapage */
 	HnswGetMetaPageInfo(index, &vacuumstate->m, NULL);
 

src/ivfbuild.c

@@ -138,7 +138,7 @@ SampleRows(IvfflatBuildState * buildstate)
  * Add tuple to sort
  */
 static void
-AddTupleToSort(Relation index, ItemPointer tid, Datum *values, bool *isnull, IvfflatBuildState * buildstate)
+AddTupleToSort(Relation index, ItemPointer tid, Datum *values, IvfflatBuildState * buildstate)
 {
 	double		distance;
 	double		minDistance = DBL_MAX;
@@ -184,11 +184,6 @@ AddTupleToSort(Relation index, ItemPointer tid, Datum *values, bool *isnull, Ivf
 	slot->tts_isnull[1] = false;
 	slot->tts_values[2] = value;
 	slot->tts_isnull[2] = false;
-	for (int i = 1; i < buildstate->tupdesc->natts; i++)
-	{
-		slot->tts_values[2 + i] = values[i];
-		slot->tts_isnull[2 + i] = isnull[i];
-	}
 	ExecStoreVirtualTuple(slot);
 
 	/*
@@ -220,7 +215,7 @@ BuildCallback(Relation index, ItemPointer tid, Datum *values,
 	oldCtx = MemoryContextSwitchTo(buildstate->tmpCtx);
 
 	/* Add tuple to sort */
-	AddTupleToSort(index, tid, values, isnull, buildstate);
+	AddTupleToSort(index, tid, values, buildstate);
 
 	/* Reset memory context */
 	MemoryContextSwitchTo(oldCtx);
@@ -231,20 +226,19 @@ BuildCallback(Relation index, ItemPointer tid, Datum *values,
  * Get index tuple from sort state
  */
 static inline void
-GetNextTuple(Tuplesortstate *sortstate, TupleDesc tupdesc, TupleTableSlot *slot, Datum *values, bool *isnull, IndexTuple *itup, int *list)
+GetNextTuple(Tuplesortstate *sortstate, TupleDesc tupdesc, TupleTableSlot *slot, IndexTuple *itup, int *list)
 {
+	Datum		value;
+	bool		isnull;
+
 	if (tuplesort_gettupleslot(sortstate, true, false, slot, NULL))
 	{
-		bool		unused;
-
-		*list = DatumGetInt32(slot_getattr(slot, 1, &unused));
-
-		for (int i = 0; i < tupdesc->natts; i++)
-			values[i] = slot_getattr(slot, 3 + i, &isnull[i]);
+		*list = DatumGetInt32(slot_getattr(slot, 1, &isnull));
+		value = slot_getattr(slot, 3, &isnull);
 
 		/* Form the index tuple */
-		*itup = index_form_tuple(tupdesc, values, isnull);
-		(*itup)->t_tid = *((ItemPointer) DatumGetPointer(slot_getattr(slot, 2, &unused)));
+		*itup = index_form_tuple(tupdesc, &value, &isnull);
+		(*itup)->t_tid = *((ItemPointer) DatumGetPointer(slot_getattr(slot, 2, &isnull)));
 	}
 	else
 		*list = -1;
@@ -260,16 +254,14 @@ InsertTuples(Relation index, IvfflatBuildState * buildstate, ForkNumber forkNum)
 	IndexTuple	itup = NULL;	/* silence compiler warning */
 	int64		inserted = 0;
 
-	TupleTableSlot *slot = MakeSingleTupleTableSlot(buildstate->sortdesc, &TTSOpsMinimalTuple);
-	TupleDesc	tupdesc = buildstate->tupdesc;
-	Datum	   *values = palloc(tupdesc->natts * sizeof(Datum));
-	bool	   *isnull = palloc(tupdesc->natts * sizeof(bool));
+	TupleTableSlot *slot = MakeSingleTupleTableSlot(buildstate->tupdesc, &TTSOpsMinimalTuple);
+	TupleDesc	tupdesc = RelationGetDescr(index);
 
 	pgstat_progress_update_param(PROGRESS_CREATEIDX_SUBPHASE, PROGRESS_IVFFLAT_PHASE_LOAD);
 
 	pgstat_progress_update_param(PROGRESS_CREATEIDX_TUPLES_TOTAL, buildstate->indtuples);
 
-	GetNextTuple(buildstate->sortstate, tupdesc, slot, values, isnull, &itup, &list);
+	GetNextTuple(buildstate->sortstate, tupdesc, slot, &itup, &list);
 
 	for (int i = 0; i < buildstate->centers->length; i++)
 	{
@@ -305,7 +297,7 @@ InsertTuples(Relation index, IvfflatBuildState * buildstate, ForkNumber forkNum)
 
 			pgstat_progress_update_param(PROGRESS_CREATEIDX_TUPLES_DONE, ++inserted);
 
-			GetNextTuple(buildstate->sortstate, tupdesc, slot, values, isnull, &itup, &list);
+			GetNextTuple(buildstate->sortstate, tupdesc, slot, &itup, &list);
 		}
 
 		insertPage = BufferGetBlockNumber(buf);
@@ -315,9 +307,6 @@ InsertTuples(Relation index, IvfflatBuildState * buildstate, ForkNumber forkNum)
 		/* Set the start and insert pages */
 		IvfflatUpdateList(index, buildstate->listInfo[i], insertPage, InvalidBlockNumber, startPage, forkNum);
 	}
-
-	pfree(values);
-	pfree(isnull);
 }
 
 /*
@@ -330,7 +319,6 @@ InitBuildState(IvfflatBuildState * buildstate, Relation heap, Relation index, In
 	buildstate->index = index;
 	buildstate->indexInfo = indexInfo;
 	buildstate->typeInfo = IvfflatGetTypeInfo(index);
-	buildstate->tupdesc = RelationGetDescr(index);
 
 	buildstate->lists = IvfflatGetLists(index);
 	buildstate->dimensions = TupleDescAttr(index->rd_att, 0)->atttypmod;
@@ -368,13 +356,12 @@ InitBuildState(IvfflatBuildState * buildstate, Relation heap, Relation index, In
 				 errmsg("dimensions must be greater than one for this opclass")));
 
 	/* Create tuple description for sorting */
-	buildstate->sortdesc = CreateTemplateTupleDesc(2 + buildstate->tupdesc->natts);
-	TupleDescInitEntry(buildstate->sortdesc, (AttrNumber) 1, "list", INT4OID, -1, 0);
-	TupleDescInitEntry(buildstate->sortdesc, (AttrNumber) 2, "tid", TIDOID, -1, 0);
-	for (int i = 0; i < buildstate->tupdesc->natts; i++)
-		TupleDescInitEntry(buildstate->sortdesc, (AttrNumber) (3 + i), NULL, buildstate->tupdesc->attrs[i].atttypid, -1, 0);
+	buildstate->tupdesc = CreateTemplateTupleDesc(3);
+	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);
 
-	buildstate->slot = MakeSingleTupleTableSlot(buildstate->sortdesc, &TTSOpsVirtual);
+	buildstate->slot = MakeSingleTupleTableSlot(buildstate->tupdesc, &TTSOpsVirtual);
 
 	buildstate->centers = VectorArrayInit(buildstate->lists, buildstate->dimensions, buildstate->typeInfo->itemSize(buildstate->dimensions));
 	buildstate->listInfo = palloc(sizeof(ListInfo) * buildstate->lists);
@@ -646,7 +633,7 @@ IvfflatParallelScanAndSort(IvfflatSpool * ivfspool, IvfflatShared * ivfshared, S
 	InitBuildState(&buildstate, ivfspool->heap, ivfspool->index, indexInfo);
 	memcpy(buildstate.centers->items, ivfcenters, buildstate.centers->itemsize * buildstate.centers->maxlen);
 	buildstate.centers->length = buildstate.centers->maxlen;
-	ivfspool->sortstate = InitBuildSortState(buildstate.sortdesc, sortmem, coordinate);
+	ivfspool->sortstate = InitBuildSortState(buildstate.tupdesc, sortmem, coordinate);
 	buildstate.sortstate = ivfspool->sortstate;
 	scan = table_beginscan_parallel(ivfspool->heap,
 									ParallelTableScanFromIvfflatShared(ivfshared));
@@ -963,7 +950,7 @@ AssignTuples(IvfflatBuildState * buildstate)
 	}
 
 	/* Begin serial/leader tuplesort */
-	buildstate->sortstate = InitBuildSortState(buildstate->sortdesc, maintenance_work_mem, coordinate);
+	buildstate->sortstate = InitBuildSortState(buildstate->tupdesc, maintenance_work_mem, coordinate);
 
 	/* Add tuples to sort */
 	if (buildstate->heap != NULL)

src/ivfflat.c

@@ -17,16 +17,8 @@
 #endif
 
 int			ivfflat_probes;
-int			ivfflat_iterative_search;
-int			ivfflat_iterative_search_max_probes;
 static relopt_kind ivfflat_relopt_kind;
 
-static const struct config_enum_entry ivfflat_iterative_search_options[] = {
-	{"off", IVFFLAT_ITERATIVE_SEARCH_OFF, false},
-	{"on", IVFFLAT_ITERATIVE_SEARCH_RELAXED, false},
-	{NULL, 0, false}
-};
-
 /*
  * Initialize index options and variables
  */
@@ -41,14 +33,6 @@ IvfflatInit(void)
 							"Valid range is 1..lists.", &ivfflat_probes,
 							IVFFLAT_DEFAULT_PROBES, IVFFLAT_MIN_LISTS, IVFFLAT_MAX_LISTS, PGC_USERSET, 0, NULL, NULL, NULL);
 
-	DefineCustomEnumVariable("ivfflat.iterative_search", "Sets whether to use iterative search",
-							 NULL, &ivfflat_iterative_search,
-							 IVFFLAT_ITERATIVE_SEARCH_OFF, ivfflat_iterative_search_options, PGC_USERSET, 0, NULL, NULL, NULL);
-
-	DefineCustomIntVariable("ivfflat.iterative_search_max_probes", "Sets the max number of probes for iterative search",
-							"Zero sets to the number of lists", &ivfflat_iterative_search_max_probes,
-							0, 0, IVFFLAT_MAX_LISTS, PGC_USERSET, 0, NULL, NULL, NULL);
-
 	MarkGUCPrefixReserved("ivfflat");
 }
 
@@ -85,8 +69,6 @@ ivfflatcostestimate(PlannerInfo *root, IndexPath *path, double loop_count,
 	GenericCosts costs;
 	int			lists;
 	double		ratio;
-	double		sequentialRatio = 0.5;
-	double		startupPages;
 	double		spc_seq_page_cost;
 	Relation	index;
 
@@ -103,8 +85,6 @@ ivfflatcostestimate(PlannerInfo *root, IndexPath *path, double loop_count,
 
 	MemSet(&costs, 0, sizeof(costs));
 
-	genericcostestimate(root, path, loop_count, &costs);
-
 	index = index_open(path->indexinfo->indexoid, NoLock);
 	IvfflatGetMetaPageInfo(index, &lists, NULL);
 	index_close(index, NoLock);
@@ -114,26 +94,41 @@ ivfflatcostestimate(PlannerInfo *root, IndexPath *path, double loop_count,
 	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);
+
 	get_tablespace_page_costs(path->indexinfo->reltablespace, NULL, &spc_seq_page_cost);
 
-	/* Change some page cost from random to sequential */
-	costs.indexTotalCost -= sequentialRatio * costs.numIndexPages * (costs.spc_random_page_cost - spc_seq_page_cost);
-
-	/* Startup cost is cost before returning the first row */
-	costs.indexStartupCost = costs.indexTotalCost * ratio;
-
 	/* Adjust cost if needed since TOAST not included in seq scan cost */
-	startupPages = costs.numIndexPages * ratio;
-	if (startupPages > path->indexinfo->rel->pages && ratio < 0.5)
+	if (costs.numIndexPages > path->indexinfo->rel->pages && ratio < 0.5)
 	{
-		/* Change rest of page cost from random to sequential */
-		costs.indexStartupCost -= (1 - sequentialRatio) * startupPages * (costs.spc_random_page_cost - spc_seq_page_cost);
+		/* 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 */
-		costs.indexStartupCost -= (startupPages - path->indexinfo->rel->pages) * spc_seq_page_cost;
+		costs.indexTotalCost -= (costs.numIndexPages - path->indexinfo->rel->pages) * spc_seq_page_cost;
+	}
+	else
+	{
+		/* Change some page cost from random to sequential */
+		costs.indexTotalCost -= 0.5 * costs.numIndexPages * (costs.spc_random_page_cost - spc_seq_page_cost);
 	}
 
-	*indexStartupCost = costs.indexStartupCost;
+	/*
+	 * If the list selectivity is lower than what is returned from the generic
+	 * cost estimator, use that.
+	 */
+	if (ratio < costs.indexSelectivity)
+		costs.indexSelectivity = ratio;
+
+	/* Use total cost since most work happens before first tuple is returned */
+	*indexStartupCost = costs.indexTotalCost;
 	*indexTotalCost = costs.indexTotalCost;
 	*indexSelectivity = costs.indexSelectivity;
 	*indexCorrelation = costs.indexCorrelation;

src/ivfflat.h

@@ -80,14 +80,6 @@
 
 /* Variables */
 extern int	ivfflat_probes;
-extern int	ivfflat_iterative_search;
-extern int	ivfflat_iterative_search_max_probes;
-
-typedef enum IvfflatIterativeSearchType
-{
-	IVFFLAT_ITERATIVE_SEARCH_OFF,
-	IVFFLAT_ITERATIVE_SEARCH_RELAXED
-}			IvfflatIterativeSearchType;
 
 typedef struct VectorArrayData
 {
@@ -173,7 +165,6 @@ typedef struct IvfflatBuildState
 	Relation	index;
 	IndexInfo  *indexInfo;
 	const		IvfflatTypeInfo *typeInfo;
-	TupleDesc	tupdesc;
 
 	/* Settings */
 	int			dimensions;
@@ -207,7 +198,7 @@ typedef struct IvfflatBuildState
 
 	/* Sorting */
 	Tuplesortstate *sortstate;
-	TupleDesc	sortdesc;
+	TupleDesc	tupdesc;
 	TupleTableSlot *slot;
 
 	/* Memory */
@@ -256,10 +247,8 @@ typedef struct IvfflatScanOpaqueData
 {
 	const		IvfflatTypeInfo *typeInfo;
 	int			probes;
-	int			maxProbes;
 	int			dimensions;
 	bool		first;
-	Datum		value;
 
 	/* Sorting */
 	Tuplesortstate *sortstate;
@@ -276,8 +265,6 @@ typedef struct IvfflatScanOpaqueData
 
 	/* Lists */
 	pairingheap *listQueue;
-	BlockNumber *listPages;
-	int			listIndex;
 	IvfflatScanList lists[FLEXIBLE_ARRAY_MEMBER];	/* must come last */
 }			IvfflatScanOpaqueData;
 

src/ivfinsert.c

@@ -78,8 +78,6 @@ InsertTuple(Relation index, Datum *values, bool *isnull, ItemPointer heap_tid, R
 	BlockNumber insertPage = InvalidBlockNumber;
 	ListInfo	listInfo;
 	BlockNumber originalInsertPage;
-	TupleDesc	tupdesc = RelationGetDescr(index);
-	Datum	   *newValues = palloc(tupdesc->natts * sizeof(Datum));
 
 	/* Detoast once for all calls */
 	value = PointerGetDatum(PG_DETOAST_DATUM(values[0]));
@@ -100,16 +98,12 @@ InsertTuple(Relation index, Datum *values, bool *isnull, ItemPointer heap_tid, R
 	IvfflatGetMetaPageInfo(index, NULL, NULL);
 
 	/* Find the insert page - sets the page and list info */
-	FindInsertPage(index, &value, &insertPage, &listInfo);
+	FindInsertPage(index, values, &insertPage, &listInfo);
 	Assert(BlockNumberIsValid(insertPage));
 	originalInsertPage = insertPage;
 
-	newValues[0] = value;
-	for (int i = 1; i < tupdesc->natts; i++)
-		newValues[i] = values[i];
-
 	/* Form tuple */
-	itup = index_form_tuple(tupdesc, newValues, isnull);
+	itup = index_form_tuple(RelationGetDescr(index), &value, isnull);
 	itup->t_tid = *heap_tid;
 
 	/* Get tuple size */

src/ivfscan.c

@@ -65,7 +65,7 @@ GetScanLists(IndexScanDesc scan, Datum value)
 			/* Use procinfo from the index instead of scan key for performance */
 			distance = DatumGetFloat8(so->distfunc(so->procinfo, so->collation, PointerGetDatum(&list->center), value));
 
-			if (listCount < so->maxProbes)
+			if (listCount < so->probes)
 			{
 				IvfflatScanList *scanlist;
 
@@ -78,7 +78,7 @@ GetScanLists(IndexScanDesc scan, Datum value)
 				pairingheap_add(so->listQueue, &scanlist->ph_node);
 
 				/* Calculate max distance */
-				if (listCount == so->maxProbes)
+				if (listCount == so->probes)
 					maxDistance = GetScanList(pairingheap_first(so->listQueue))->distance;
 			}
 			else if (distance < maxDistance)
@@ -102,11 +102,6 @@ GetScanLists(IndexScanDesc scan, Datum value)
 
 		UnlockReleaseBuffer(cbuf);
 	}
-
-	for (int i = listCount - 1; i >= 0; i--)
-		so->listPages[i] = GetScanList(pairingheap_remove_first(so->listQueue))->startPage;
-
-	Assert(pairingheap_is_empty(so->listQueue));
 }
 
 /*
@@ -119,14 +114,11 @@ GetScanItems(IndexScanDesc scan, Datum value)
 	TupleDesc	tupdesc = RelationGetDescr(scan->indexRelation);
 	double		tuples = 0;
 	TupleTableSlot *slot = so->vslot;
-	int			batchProbes = 0;
-
-	tuplesort_reset(so->sortstate);
 
 	/* Search closest probes lists */
-	while (so->listIndex < so->maxProbes && (++batchProbes) <= so->probes)
+	while (!pairingheap_is_empty(so->listQueue))
 	{
-		BlockNumber searchPage = so->listPages[so->listIndex++];
+		BlockNumber searchPage = GetScanList(pairingheap_remove_first(so->listQueue))->startPage;
 
 		/* Search all entry pages for list */
 		while (BlockNumberIsValid(searchPage))
@@ -174,17 +166,13 @@ GetScanItems(IndexScanDesc scan, Datum value)
 		}
 	}
 
-	if (tuples < 100 && ivfflat_iterative_search == IVFFLAT_ITERATIVE_SEARCH_OFF)
+	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);
-
-#if defined(IVFFLAT_MEMORY)
-	elog(INFO, "memory: %zu MB", MemoryContextMemAllocated(CurrentMemoryContext, true) / (1024 * 1024));
-#endif
 }
 
 /*
@@ -252,7 +240,6 @@ ivfflatbeginscan(Relation index, int nkeys, int norderbys)
 	int			lists;
 	int			dimensions;
 	int			probes = ivfflat_probes;
-	int			maxProbes;
 
 	scan = RelationGetIndexScan(index, nkeys, norderbys);
 
@@ -262,21 +249,10 @@ ivfflatbeginscan(Relation index, int nkeys, int norderbys)
 	if (probes > lists)
 		probes = lists;
 
-	if (ivfflat_iterative_search != IVFFLAT_ITERATIVE_SEARCH_OFF)
-	{
-		if (ivfflat_iterative_search_max_probes == 0)
-			maxProbes = lists;
-		else
-			maxProbes = Min(ivfflat_iterative_search_max_probes, lists);
-	}
-	else
-		maxProbes = probes;
-
-	so = (IvfflatScanOpaque) palloc(offsetof(IvfflatScanOpaqueData, lists) + maxProbes * sizeof(IvfflatScanList));
+	so = (IvfflatScanOpaque) palloc(offsetof(IvfflatScanOpaqueData, lists) + probes * sizeof(IvfflatScanList));
 	so->typeInfo = IvfflatGetTypeInfo(index);
 	so->first = true;
 	so->probes = probes;
-	so->maxProbes = maxProbes;
 	so->dimensions = dimensions;
 
 	/* Set support functions */
@@ -304,8 +280,6 @@ ivfflatbeginscan(Relation index, int nkeys, int norderbys)
 	so->bas = GetAccessStrategy(BAS_BULKREAD);
 
 	so->listQueue = pairingheap_allocate(CompareLists, scan);
-	so->listPages = palloc(maxProbes * sizeof(BlockNumber));
-	so->listIndex = 0;
 
 	scan->opaque = so;
 
@@ -320,9 +294,11 @@ ivfflatrescan(IndexScanDesc scan, ScanKey keys, int nkeys, ScanKey orderbys, int
 {
 	IvfflatScanOpaque so = (IvfflatScanOpaque) scan->opaque;
 
+	if (!so->first)
+		tuplesort_reset(so->sortstate);
+
 	so->first = true;
 	pairingheap_reset(so->listQueue);
-	so->listIndex = 0;
 
 	if (keys && scan->numberOfKeys > 0)
 		memmove(scan->keyData, keys, scan->numberOfKeys * sizeof(ScanKeyData));
@@ -338,8 +314,6 @@ bool
 ivfflatgettuple(IndexScanDesc scan, ScanDirection dir)
 {
 	IvfflatScanOpaque so = (IvfflatScanOpaque) scan->opaque;
-	ItemPointer heaptid;
-	bool		isnull;
 
 	/*
 	 * Index can be used to scan backward, but Postgres doesn't support
@@ -367,25 +341,28 @@ ivfflatgettuple(IndexScanDesc scan, ScanDirection dir)
 		IvfflatBench("GetScanLists", GetScanLists(scan, value));
 		IvfflatBench("GetScanItems", GetScanItems(scan, value));
 		so->first = false;
-		so->value = value;
 
-		/* TODO clean up if we allocated a new value */
+#if defined(IVFFLAT_MEMORY)
+		elog(INFO, "memory: %zu MB", MemoryContextMemAllocated(CurrentMemoryContext, true) / (1024 * 1024));
+#endif
+
+		/* Clean up if we allocated a new value */
+		if (value != scan->orderByData->sk_argument)
+			pfree(DatumGetPointer(value));
 	}
 
-	while (!tuplesort_gettupleslot(so->sortstate, true, false, so->mslot, NULL))
+	if (tuplesort_gettupleslot(so->sortstate, true, false, so->mslot, NULL))
 	{
-		if (so->listIndex == so->maxProbes)
-			return false;
+		bool		isnull;
+		ItemPointer heaptid = (ItemPointer) DatumGetPointer(slot_getattr(so->mslot, 2, &isnull));
 
-		IvfflatBench("GetScanItems", GetScanItems(scan, so->value));
+		scan->xs_heaptid = *heaptid;
+		scan->xs_recheck = false;
+		scan->xs_recheckorderby = false;
+		return true;
 	}
 
-	heaptid = (ItemPointer) DatumGetPointer(slot_getattr(so->mslot, 2, &isnull));
-
-	scan->xs_heaptid = *heaptid;
-	scan->xs_recheck = false;
-	scan->xs_recheckorderby = false;
-	return true;
+	return false;
 }
 
 /*
@@ -397,7 +374,6 @@ ivfflatendscan(IndexScanDesc scan)
 	IvfflatScanOpaque so = (IvfflatScanOpaque) scan->opaque;
 
 	pairingheap_free(so->listQueue);
-	pfree(so->listPages);
 	tuplesort_end(so->sortstate);
 	FreeAccessStrategy(so->bas);
 	FreeTupleDesc(so->tupdesc);

src/ivfutils.c

@@ -7,6 +7,7 @@
 #include "halfutils.h"
 #include "halfvec.h"
 #include "ivfflat.h"
+#include "minivec.h"
 #include "storage/bufmgr.h"
 
 /*
@@ -70,6 +71,9 @@ IvfflatOptionalProcInfo(Relation index, uint16 procnum)
 Datum
 IvfflatNormValue(const IvfflatTypeInfo * typeInfo, Oid collation, Datum value)
 {
+	if (!typeInfo->normalize)
+		return value;
+
 	return DirectFunctionCall1Coll(typeInfo->normalize, collation, value);
 }
 
@@ -231,6 +235,7 @@ IvfflatUpdateList(Relation index, ListInfo listInfo,
 
 PGDLLEXPORT Datum l2_normalize(PG_FUNCTION_ARGS);
 PGDLLEXPORT Datum halfvec_l2_normalize(PG_FUNCTION_ARGS);
+PGDLLEXPORT Datum minivec_l2_normalize(PG_FUNCTION_ARGS);
 PGDLLEXPORT Datum sparsevec_l2_normalize(PG_FUNCTION_ARGS);
 
 static Size
@@ -245,6 +250,12 @@ HalfvecItemSize(int dimensions)
 	return HALFVEC_SIZE(dimensions);
 }
 
+static Size
+MinivecItemSize(int dimensions)
+{
+	return MINIVEC_SIZE(dimensions);
+}
+
 static Size
 BitItemSize(int dimensions)
 {
@@ -275,6 +286,18 @@ HalfvecUpdateCenter(Pointer v, int dimensions, float *x)
 		vec->x[k] = Float4ToHalfUnchecked(x[k]);
 }
 
+static void
+MinivecUpdateCenter(Pointer v, int dimensions, float *x)
+{
+	MiniVector *vec = (MiniVector *) v;
+
+	SET_VARSIZE(vec, MINIVEC_SIZE(dimensions));
+	vec->dim = dimensions;
+
+	for (int k = 0; k < dimensions; k++)
+		vec->x[k] = Float4ToFp8Unchecked(x[k]);
+}
+
 static void
 BitUpdateCenter(Pointer v, int dimensions, float *x)
 {
@@ -309,6 +332,15 @@ HalfvecSumCenter(Pointer v, float *x)
 		x[k] += HalfToFloat4(vec->x[k]);
 }
 
+static void
+MinivecSumCenter(Pointer v, float *x)
+{
+	MiniVector *vec = (MiniVector *) v;
+
+	for (int k = 0; k < vec->dim; k++)
+		x[k] += Fp8ToFloat4(vec->x[k]);
+}
+
 static void
 BitSumCenter(Pointer v, float *x)
 {
@@ -357,6 +389,22 @@ ivfflat_halfvec_support(PG_FUNCTION_ARGS)
 	PG_RETURN_POINTER(&typeInfo);
 };
 
+FUNCTION_PREFIX PG_FUNCTION_INFO_V1(ivfflat_minivec_support);
+Datum
+ivfflat_minivec_support(PG_FUNCTION_ARGS)
+{
+	static const IvfflatTypeInfo typeInfo = {
+		.maxDimensions = IVFFLAT_MAX_DIM * 4,
+		/* Do not normalize to maximize precision */
+		.normalize = NULL,
+		.itemSize = MinivecItemSize,
+		.updateCenter = MinivecUpdateCenter,
+		.sumCenter = MinivecSumCenter
+	};
+
+	PG_RETURN_POINTER(&typeInfo);
+};
+
 FUNCTION_PREFIX PG_FUNCTION_INFO_V1(ivfflat_bit_support);
 Datum
 ivfflat_bit_support(PG_FUNCTION_ARGS)

src/ivfvacuum.c

@@ -26,7 +26,7 @@ ivfflatbulkdelete(IndexVacuumInfo *info, IndexBulkDeleteResult *stats,
 		Page		cpage;
 		OffsetNumber coffno;
 		OffsetNumber cmaxoffno;
-		BlockNumber listPages[MaxOffsetNumber];
+		BlockNumber startPages[MaxOffsetNumber];
 		ListInfo	listInfo;
 
 		cbuf = ReadBuffer(index, blkno);
@@ -40,7 +40,7 @@ ivfflatbulkdelete(IndexVacuumInfo *info, IndexBulkDeleteResult *stats,
 		{
 			IvfflatList list = (IvfflatList) PageGetItem(cpage, PageGetItemId(cpage, coffno));
 
-			listPages[coffno - FirstOffsetNumber] = list->startPage;
+			startPages[coffno - FirstOffsetNumber] = list->startPage;
 		}
 
 		listInfo.blkno = blkno;
@@ -50,7 +50,7 @@ ivfflatbulkdelete(IndexVacuumInfo *info, IndexBulkDeleteResult *stats,
 
 		for (coffno = FirstOffsetNumber; coffno <= cmaxoffno; coffno = OffsetNumberNext(coffno))
 		{
-			BlockNumber searchPage = listPages[coffno - FirstOffsetNumber];
+			BlockNumber searchPage = startPages[coffno - FirstOffsetNumber];
 			BlockNumber insertPage = InvalidBlockNumber;
 
 			/* Iterate over entry pages */

src/minivec.h

@@ -0,0 +1,177 @@
+#ifndef MINIVEC_H
+#define MINIVEC_H
+
+#define MINIVEC_MAX_DIM 16000
+
+#define fp8 uint8
+
+#define MINIVEC_SIZE(_dim)		(offsetof(MiniVector, x) + sizeof(fp8)*(_dim))
+#define DatumGetMiniVector(x)		((MiniVector *) PG_DETOAST_DATUM(x))
+#define PG_GETARG_MINIVEC_P(x)	DatumGetMiniVector(PG_GETARG_DATUM(x))
+#define PG_RETURN_MINIVEC_P(x)	PG_RETURN_POINTER(x)
+
+typedef struct MiniVector
+{
+	int32		vl_len_;		/* varlena header (do not touch directly!) */
+	int16		dim;			/* number of dimensions */
+	int16		unused;			/* reserved for future use, always zero */
+	fp8			x[FLEXIBLE_ARRAY_MEMBER];
+}			MiniVector;
+
+MiniVector *InitMiniVector(int dim);
+
+/*
+ * Check if fp8 is NaN
+ */
+static inline bool
+Fp8IsNan(fp8 num)
+{
+	return (num & 0x7C) == 0x7C && (num & 0x7F) != 0x7C;
+}
+
+/*
+ * Check if fp8 is infinite
+ */
+static inline bool
+Fp8IsInf(fp8 num)
+{
+	return (num & 0x7F) == 0x7C;
+}
+
+/*
+ * Check if fp8 is zero
+ */
+static inline bool
+Fp8IsZero(fp8 num)
+{
+	return num == 0;
+}
+
+/*
+ * Convert a fp8 to a float4
+ */
+static inline float
+Fp8ToFloat4(fp8 num)
+{
+	/* Lookup table for non-sign bits */
+	/* Uses uint32 for correctness */
+	uint32		lookup[128] = {0, 931135488, 939524096, 943718400, 947912704, 950009856, 952107008, 954204160, 956301312, 958398464, 960495616, 962592768, 964689920, 966787072, 968884224, 970981376, 973078528, 975175680, 977272832, 979369984, 981467136, 983564288, 985661440, 987758592, 989855744, 991952896, 994050048, 996147200, 998244352, 1000341504, 1002438656, 1004535808, 1006632960, 1008730112, 1010827264, 1012924416, 1015021568, 1017118720, 1019215872, 1021313024, 1023410176, 1025507328, 1027604480, 1029701632, 1031798784, 1033895936, 1035993088, 1038090240, 1040187392, 1042284544, 1044381696, 1046478848, 1048576000, 1050673152, 1052770304, 1054867456, 1056964608, 1059061760, 1061158912, 1063256064, 1065353216, 1067450368, 1069547520, 1071644672, 1073741824, 1075838976, 1077936128, 1080033280, 1082130432, 1084227584, 1086324736, 1088421888, 1090519040, 1092616192, 1094713344, 1096810496, 1098907648, 1101004800, 1103101952, 1105199104, 1107296256, 1109393408, 1111490560, 1113587712, 1115684864, 1117782016, 1119879168, 1121976320, 1124073472, 1126170624, 1128267776, 1130364928, 1132462080, 1134559232, 1136656384, 1138753536, 1140850688, 1142947840, 1145044992, 1147142144, 1149239296, 1151336448, 1153433600, 1155530752, 1157627904, 1159725056, 1161822208, 1163919360, 1166016512, 1168113664, 1170210816, 1172307968, 1174405120, 1176502272, 1178599424, 1180696576, 1182793728, 1184890880, 1186988032, 1189085184, 1191182336, 1193279488, 1195376640, 1197473792, 2139095040, 2145386496, 2143289344, 2145386496};
+
+	union
+	{
+		float		f;
+		uint32		i;
+	}			swap;
+
+	swap.i = lookup[num & 0x7F];
+
+	return (num & 0x80) == 0x80 ? -swap.f : swap.f;
+}
+
+/*
+ * Convert a float4 to a fp8
+ */
+static inline fp8
+Float4ToFp8Unchecked(float num)
+{
+	union
+	{
+		float		f;
+		uint32		i;
+	}			swap;
+
+	uint32		bin;
+	int			exponent;
+	int			mantissa;
+	uint8		result;
+
+	swap.f = num;
+	bin = swap.i;
+	exponent = (bin & 0x7F800000) >> 23;
+	mantissa = bin & 0x007FFFFF;
+
+	/* Sign */
+	result = (bin & 0x80000000) >> 24;
+
+	if (isinf(num))
+	{
+		/* Infinite */
+		result |= 0x7C;
+	}
+	else if (isnan(num))
+	{
+		/* NaN */
+		result |= 0x7F;
+	}
+	else if (exponent > 98)
+	{
+		int			m;
+		int			gr;
+		int			s;
+
+		exponent -= 127;
+		s = mantissa & 0x000FFFFF;
+
+		/* Subnormal */
+		if (exponent < -14)
+		{
+			int			diff = -exponent - 14;
+
+			mantissa >>= diff;
+			mantissa += 1 << (23 - diff);
+			s |= mantissa & 0x000FFFFF;
+		}
+
+		m = mantissa >> 21;
+
+		/* Round */
+		gr = (mantissa >> 20) % 4;
+		if (gr == 3 || (gr == 1 && s != 0))
+			m += 1;
+
+		if (m == 4)
+		{
+			m = 0;
+			exponent += 1;
+		}
+
+		if (exponent > 15)
+		{
+			/* Infinite */
+			result |= 0x7C;
+		}
+		else
+		{
+			if (exponent >= -14)
+				result |= (exponent + 15) << 2;
+
+			result |= m;
+		}
+	}
+
+	return result;
+}
+
+/*
+ * Convert a float4 to a fp8
+ */
+static inline fp8
+Float4ToFp8(float num)
+{
+	fp8			result = Float4ToFp8Unchecked(num);
+
+	if (unlikely(Fp8IsInf(result)) && !isinf(num))
+	{
+		char	   *buf = palloc(FLOAT_SHORTEST_DECIMAL_LEN);
+
+		float_to_shortest_decimal_buf(num, buf);
+
+		ereport(ERROR,
+				(errcode(ERRCODE_NUMERIC_VALUE_OUT_OF_RANGE),
+				 errmsg("\"%s\" is out of range for type minivec", buf)));
+	}
+
+	return result;
+}
+
+#endif

src/vector.c

@@ -13,6 +13,7 @@
 #include "ivfflat.h"
 #include "lib/stringinfo.h"
 #include "libpq/pqformat.h"
+#include "minivec.h"
 #include "port.h"				/* for strtof() */
 #include "sparsevec.h"
 #include "utils/array.h"
@@ -542,6 +543,28 @@ halfvec_to_vector(PG_FUNCTION_ARGS)
 	PG_RETURN_POINTER(result);
 }
 
+/*
+ * Convert fp8 vector to vector
+ */
+FUNCTION_PREFIX PG_FUNCTION_INFO_V1(minivec_to_vector);
+Datum
+minivec_to_vector(PG_FUNCTION_ARGS)
+{
+	MiniVector *vec = PG_GETARG_MINIVEC_P(0);
+	int32		typmod = PG_GETARG_INT32(1);
+	Vector	   *result;
+
+	CheckDim(vec->dim);
+	CheckExpectedDim(typmod, vec->dim);
+
+	result = InitVector(vec->dim);
+
+	for (int i = 0; i < vec->dim; i++)
+		result->x[i] = Fp8ToFloat4(vec->x[i]);
+
+	PG_RETURN_POINTER(result);
+}
+
 VECTOR_TARGET_CLONES static float
 VectorL2SquaredDistance(int dim, float *ax, float *bx)
 {

test/expected/btree.out

@@ -38,6 +38,26 @@ SELECT * FROM t ORDER BY val;
  
 (4 rows)
 
+DROP TABLE t;
+-- minivec
+CREATE TABLE t (val minivec(3));
+INSERT INTO t (val) VALUES ('[0,0,0]'), ('[1,2,3]'), ('[1,1,1]'), (NULL);
+CREATE INDEX ON t (val);
+SELECT * FROM t WHERE val = '[1,2,3]';
+   val   
+---------
+ [1,2,3]
+(1 row)
+
+SELECT * FROM t ORDER BY val;
+   val   
+---------
+ [0,0,0]
+ [1,1,1]
+ [1,2,3]
+ 
+(4 rows)
+
 DROP TABLE t;
 -- sparsevec
 CREATE TABLE t (val sparsevec(3));

test/expected/cast.out

@@ -140,6 +140,64 @@ SELECT '{1e-8,-1e-8}'::real[]::halfvec;
  [0,-0]
 (1 row)
 
+SELECT '[1,2,3]'::vector::minivec;
+ minivec 
+---------
+ [1,2,3]
+(1 row)
+
+SELECT '[1,2,3]'::vector::minivec(3);
+ minivec 
+---------
+ [1,2,3]
+(1 row)
+
+SELECT '[1,2,3]'::vector::minivec(2);
+ERROR:  expected 2 dimensions, not 3
+SELECT '[61440]'::vector::minivec;
+ERROR:  "61440" is out of range for type minivec
+SELECT '[1e-8]'::vector::minivec;
+ minivec 
+---------
+ [0]
+(1 row)
+
+SELECT '[1,2,3]'::minivec::vector;
+ vector  
+---------
+ [1,2,3]
+(1 row)
+
+SELECT '[1,2,3]'::minivec::vector(3);
+ vector  
+---------
+ [1,2,3]
+(1 row)
+
+SELECT '[1,2,3]'::minivec::vector(2);
+ERROR:  expected 2 dimensions, not 3
+SELECT '{1,2,3}'::real[]::minivec;
+ minivec 
+---------
+ [1,2,3]
+(1 row)
+
+SELECT '{1,2,3}'::real[]::minivec(3);
+ minivec 
+---------
+ [1,2,3]
+(1 row)
+
+SELECT '{1,2,3}'::real[]::minivec(2);
+ERROR:  expected 2 dimensions, not 3
+SELECT '{61440,-61440}'::real[]::minivec;
+ERROR:  "61440" is out of range for type minivec
+SELECT '{1e-8,-1e-8}'::real[]::minivec;
+ minivec 
+---------
+ [0,-0]
+(1 row)
+
 SELECT '[0,1.5,0,3.5,0]'::vector::sparsevec;
     sparsevec    
 -----------------

test/expected/copy.out

@@ -30,6 +30,23 @@ SELECT * FROM t2 ORDER BY val;
  
 (4 rows)
 
+DROP TABLE t;
+DROP TABLE t2;
+-- minivec
+CREATE TABLE t (val minivec(3));
+INSERT INTO t (val) VALUES ('[0,0,0]'), ('[1,2,3]'), ('[1,1,1]'), (NULL);
+CREATE TABLE t2 (val minivec(3));
+\copy t TO 'results/minivec.bin' WITH (FORMAT binary)
+\copy t2 FROM 'results/minivec.bin' WITH (FORMAT binary)
+SELECT * FROM t2 ORDER BY val;
+   val   
+---------
+ [0,0,0]
+ [1,1,1]
+ [1,2,3]
+ 
+(4 rows)
+
 DROP TABLE t;
 DROP TABLE t2;
 -- sparsevec

test/expected/ivfflat_minivec.out

@@ -0,0 +1,84 @@
+SET enable_seqscan = off;
+-- L2
+CREATE TABLE t (val minivec(3));
+INSERT INTO t (val) VALUES ('[0,0,0]'), ('[1,2,3]'), ('[1,1,1]'), (NULL);
+CREATE INDEX ON t USING ivfflat (val minivec_l2_ops) WITH (lists = 1);
+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 COUNT(*) FROM (SELECT * FROM t ORDER BY val <-> (SELECT NULL::minivec)) t2;
+ count 
+-------
+     4
+(1 row)
+
+SELECT COUNT(*) FROM t;
+ count 
+-------
+     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;
+-- inner product
+CREATE TABLE t (val minivec(3));
+INSERT INTO t (val) VALUES ('[0,0,0]'), ('[1,2,3]'), ('[1,1,1]'), (NULL);
+CREATE INDEX ON t USING ivfflat (val minivec_ip_ops) WITH (lists = 1);
+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::minivec)) t2;
+ count 
+-------
+     4
+(1 row)
+
+DROP TABLE t;
+-- cosine
+CREATE TABLE t (val minivec(3));
+INSERT INTO t (val) VALUES ('[0,0,0]'), ('[1,2,3]'), ('[1,1,1]'), (NULL);
+CREATE INDEX ON t USING ivfflat (val minivec_cosine_ops) WITH (lists = 1);
+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::minivec)) t2;
+ count 
+-------
+     3
+(1 row)
+
+DROP TABLE t;

test/expected/minivec.out

@@ -0,0 +1,590 @@
+SELECT '[1,2,3]'::minivec;
+ minivec 
+---------
+ [1,2,3]
+(1 row)
+
+SELECT '[-1,-2,-3]'::minivec;
+  minivec   
+------------
+ [-1,-2,-3]
+(1 row)
+
+SELECT '[1.,2.,3.]'::minivec;
+ minivec 
+---------
+ [1,2,3]
+(1 row)
+
+SELECT ' [ 1,  2 ,    3  ] '::minivec;
+ minivec 
+---------
+ [1,2,3]
+(1 row)
+
+SELECT '[1.23456]'::minivec;
+ minivec 
+---------
+ [1.25]
+(1 row)
+
+SELECT '[hello,1]'::minivec;
+ERROR:  invalid input syntax for type minivec: "[hello,1]"
+LINE 1: SELECT '[hello,1]'::minivec;
+               ^
+SELECT '[NaN,1]'::minivec;
+ERROR:  NaN not allowed in minivec
+LINE 1: SELECT '[NaN,1]'::minivec;
+               ^
+SELECT '[Infinity,1]'::minivec;
+ERROR:  infinite value not allowed in minivec
+LINE 1: SELECT '[Infinity,1]'::minivec;
+               ^
+SELECT '[-Infinity,1]'::minivec;
+ERROR:  infinite value not allowed in minivec
+LINE 1: SELECT '[-Infinity,1]'::minivec;
+               ^
+SELECT '[61439,-61439]'::minivec;
+    minivec     
+----------------
+ [57344,-57344]
+(1 row)
+
+SELECT '[61440,-61440]'::minivec;
+ERROR:  "61440" is out of range for type minivec
+LINE 1: SELECT '[61440,-61440]'::minivec;
+               ^
+SELECT '[1e-8,-1e-8]'::minivec;
+ minivec 
+---------
+ [0,-0]
+(1 row)
+
+SELECT '[4e38,1]'::minivec;
+ERROR:  "4e38" is out of range for type minivec
+LINE 1: SELECT '[4e38,1]'::minivec;
+               ^
+SELECT '[1e-46,1]'::minivec;
+ minivec 
+---------
+ [0,1]
+(1 row)
+
+SELECT '[1,2,3'::minivec;
+ERROR:  invalid input syntax for type minivec: "[1,2,3"
+LINE 1: SELECT '[1,2,3'::minivec;
+               ^
+SELECT '[1,2,3]9'::minivec;
+ERROR:  invalid input syntax for type minivec: "[1,2,3]9"
+LINE 1: SELECT '[1,2,3]9'::minivec;
+               ^
+DETAIL:  Junk after closing right brace.
+SELECT '1,2,3'::minivec;
+ERROR:  invalid input syntax for type minivec: "1,2,3"
+LINE 1: SELECT '1,2,3'::minivec;
+               ^
+DETAIL:  Vector contents must start with "[".
+SELECT ''::minivec;
+ERROR:  invalid input syntax for type minivec: ""
+LINE 1: SELECT ''::minivec;
+               ^
+DETAIL:  Vector contents must start with "[".
+SELECT '['::minivec;
+ERROR:  invalid input syntax for type minivec: "["
+LINE 1: SELECT '['::minivec;
+               ^
+SELECT '[ '::minivec;
+ERROR:  invalid input syntax for type minivec: "[ "
+LINE 1: SELECT '[ '::minivec;
+               ^
+SELECT '[,'::minivec;
+ERROR:  invalid input syntax for type minivec: "[,"
+LINE 1: SELECT '[,'::minivec;
+               ^
+SELECT '[]'::minivec;
+ERROR:  minivec must have at least 1 dimension
+LINE 1: SELECT '[]'::minivec;
+               ^
+SELECT '[ ]'::minivec;
+ERROR:  minivec must have at least 1 dimension
+LINE 1: SELECT '[ ]'::minivec;
+               ^
+SELECT '[,]'::minivec;
+ERROR:  invalid input syntax for type minivec: "[,]"
+LINE 1: SELECT '[,]'::minivec;
+               ^
+SELECT '[1,]'::minivec;
+ERROR:  invalid input syntax for type minivec: "[1,]"
+LINE 1: SELECT '[1,]'::minivec;
+               ^
+SELECT '[1a]'::minivec;
+ERROR:  invalid input syntax for type minivec: "[1a]"
+LINE 1: SELECT '[1a]'::minivec;
+               ^
+SELECT '[1,,3]'::minivec;
+ERROR:  invalid input syntax for type minivec: "[1,,3]"
+LINE 1: SELECT '[1,,3]'::minivec;
+               ^
+SELECT '[1, ,3]'::minivec;
+ERROR:  invalid input syntax for type minivec: "[1, ,3]"
+LINE 1: SELECT '[1, ,3]'::minivec;
+               ^
+SELECT '[1,2,3]'::minivec(3);
+ minivec 
+---------
+ [1,2,3]
+(1 row)
+
+SELECT '[1,2,3]'::minivec(2);
+ERROR:  expected 2 dimensions, not 3
+SELECT '[1,2,3]'::minivec(3, 2);
+ERROR:  invalid type modifier
+LINE 1: SELECT '[1,2,3]'::minivec(3, 2);
+                          ^
+SELECT '[1,2,3]'::minivec('a');
+ERROR:  invalid input syntax for type integer: "a"
+LINE 1: SELECT '[1,2,3]'::minivec('a');
+                          ^
+SELECT '[1,2,3]'::minivec(0);
+ERROR:  dimensions for type minivec must be at least 1
+LINE 1: SELECT '[1,2,3]'::minivec(0);
+                          ^
+SELECT '[1,2,3]'::minivec(16001);
+ERROR:  dimensions for type minivec cannot exceed 16000
+LINE 1: SELECT '[1,2,3]'::minivec(16001);
+                          ^
+SELECT unnest('{"[1,2,3]", "[4,5,6]"}'::minivec[]);
+ unnest  
+---------
+ [1,2,3]
+ [4,5,6]
+(2 rows)
+
+SELECT '{"[1,2,3]"}'::minivec(2)[];
+ERROR:  expected 2 dimensions, not 3
+SELECT '[1,2,3]'::minivec + '[4,5,6]';
+ ?column? 
+----------
+ [5,7,8]
+(1 row)
+
+SELECT '[61439]'::minivec + '[61439]';
+ERROR:  value out of range: overflow
+SELECT '[1,2]'::minivec + '[3]';
+ERROR:  different minivec dimensions 2 and 1
+SELECT '[1,2,3]'::minivec - '[4,5,6]';
+ERROR:  value out of range: overflow
+SELECT '[-61439]'::minivec - '[61439]';
+ ?column? 
+----------
+ [-0]
+(1 row)
+
+SELECT '[1,2]'::minivec - '[3]';
+ERROR:  different minivec dimensions 2 and 1
+SELECT '[1,2,3]'::minivec * '[4,5,6]';
+ ?column?  
+-----------
+ [4,10,16]
+(1 row)
+
+SELECT '[61439]'::minivec * '[61439]';
+ERROR:  value out of range: overflow
+SELECT '[1e-7]'::minivec * '[1e-7]';
+ ?column? 
+----------
+ [0]
+(1 row)
+
+SELECT '[1,2]'::minivec * '[3]';
+ERROR:  different minivec dimensions 2 and 1
+SELECT '[1,2,3]'::minivec || '[4,5]';
+  ?column?   
+-------------
+ [1,2,3,4,5]
+(1 row)
+
+SELECT array_fill(0, ARRAY[16000])::minivec || '[1]';
+ERROR:  minivec cannot have more than 16000 dimensions
+SELECT '[1,2,3]'::minivec < '[1,2,3]';
+ ?column? 
+----------
+ f
+(1 row)
+
+SELECT '[1,2,3]'::minivec < '[1,2]';
+ ?column? 
+----------
+ f
+(1 row)
+
+SELECT '[1,2,3]'::minivec <= '[1,2,3]';
+ ?column? 
+----------
+ t
+(1 row)
+
+SELECT '[1,2,3]'::minivec <= '[1,2]';
+ ?column? 
+----------
+ f
+(1 row)
+
+SELECT '[1,2,3]'::minivec = '[1,2,3]';
+ ?column? 
+----------
+ t
+(1 row)
+
+SELECT '[1,2,3]'::minivec = '[1,2]';
+ ?column? 
+----------
+ f
+(1 row)
+
+SELECT '[1,2,3]'::minivec != '[1,2,3]';
+ ?column? 
+----------
+ f
+(1 row)
+
+SELECT '[1,2,3]'::minivec != '[1,2]';
+ ?column? 
+----------
+ t
+(1 row)
+
+SELECT '[1,2,3]'::minivec >= '[1,2,3]';
+ ?column? 
+----------
+ t
+(1 row)
+
+SELECT '[1,2,3]'::minivec >= '[1,2]';
+ ?column? 
+----------
+ t
+(1 row)
+
+SELECT '[1,2,3]'::minivec > '[1,2,3]';
+ ?column? 
+----------
+ f
+(1 row)
+
+SELECT '[1,2,3]'::minivec > '[1,2]';
+ ?column? 
+----------
+ t
+(1 row)
+
+SELECT minivec_cmp('[1,2,3]', '[1,2,3]');
+ minivec_cmp 
+-------------
+           0
+(1 row)
+
+SELECT minivec_cmp('[1,2,3]', '[0,0,0]');
+ minivec_cmp 
+-------------
+           1
+(1 row)
+
+SELECT minivec_cmp('[0,0,0]', '[1,2,3]');
+ minivec_cmp 
+-------------
+          -1
+(1 row)
+
+SELECT minivec_cmp('[1,2]', '[1,2,3]');
+ minivec_cmp 
+-------------
+          -1
+(1 row)
+
+SELECT minivec_cmp('[1,2,3]', '[1,2]');
+ minivec_cmp 
+-------------
+           1
+(1 row)
+
+SELECT minivec_cmp('[1,2]', '[2,3,4]');
+ minivec_cmp 
+-------------
+          -1
+(1 row)
+
+SELECT minivec_cmp('[2,3]', '[1,2,3]');
+ minivec_cmp 
+-------------
+           1
+(1 row)
+
+SELECT vector_dims('[1,2,3]'::minivec);
+ vector_dims 
+-------------
+           3
+(1 row)
+
+SELECT round(l2_norm('[1,1]'::minivec)::numeric, 5);
+  round  
+---------
+ 1.41421
+(1 row)
+
+SELECT l2_norm('[3,4]'::minivec);
+ l2_norm 
+---------
+       5
+(1 row)
+
+SELECT l2_norm('[0,1]'::minivec);
+ l2_norm 
+---------
+       1
+(1 row)
+
+SELECT l2_norm('[0,0]'::minivec);
+ l2_norm 
+---------
+       0
+(1 row)
+
+SELECT l2_norm('[2]'::minivec);
+ l2_norm 
+---------
+       2
+(1 row)
+
+SELECT l2_distance('[0,0]'::minivec, '[3,4]');
+ l2_distance 
+-------------
+           5
+(1 row)
+
+SELECT l2_distance('[0,0]'::minivec, '[0,1]');
+ l2_distance 
+-------------
+           1
+(1 row)
+
+SELECT l2_distance('[1,2]'::minivec, '[3]');
+ERROR:  different minivec dimensions 2 and 1
+SELECT l2_distance('[1,1,1,1,1,1,1,1,1]'::minivec, '[1,1,1,1,1,1,1,4,5]');
+ l2_distance 
+-------------
+           5
+(1 row)
+
+SELECT '[0,0]'::minivec <-> '[3,4]';
+ ?column? 
+----------
+        5
+(1 row)
+
+SELECT inner_product('[1,2]'::minivec, '[3,4]');
+ inner_product 
+---------------
+            11
+(1 row)
+
+SELECT inner_product('[1,2]'::minivec, '[3]');
+ERROR:  different minivec dimensions 2 and 1
+SELECT inner_product('[448]'::minivec, '[448]');
+ inner_product 
+---------------
+        200704
+(1 row)
+
+SELECT inner_product('[1,1,1,1,1,1,1,1,1]'::minivec, '[1,2,3,4,5,6,7,8,9]');
+ inner_product 
+---------------
+            44
+(1 row)
+
+SELECT '[1,2]'::minivec <#> '[3,4]';
+ ?column? 
+----------
+      -11
+(1 row)
+
+SELECT cosine_distance('[1,2]'::minivec, '[2,4]');
+ cosine_distance 
+-----------------
+               0
+(1 row)
+
+SELECT cosine_distance('[1,2]'::minivec, '[0,0]');
+ cosine_distance 
+-----------------
+             NaN
+(1 row)
+
+SELECT cosine_distance('[1,1]'::minivec, '[1,1]');
+ cosine_distance 
+-----------------
+               0
+(1 row)
+
+SELECT cosine_distance('[1,0]'::minivec, '[0,2]');
+ cosine_distance 
+-----------------
+               1
+(1 row)
+
+SELECT cosine_distance('[1,1]'::minivec, '[-1,-1]');
+ cosine_distance 
+-----------------
+               2
+(1 row)
+
+SELECT cosine_distance('[1,2]'::minivec, '[3]');
+ERROR:  different minivec dimensions 2 and 1
+SELECT cosine_distance('[1,1]'::minivec, '[1.1,1.1]');
+ cosine_distance 
+-----------------
+               0
+(1 row)
+
+SELECT cosine_distance('[1,1]'::minivec, '[-1.1,-1.1]');
+ cosine_distance 
+-----------------
+               2
+(1 row)
+
+SELECT cosine_distance('[1,2,3,4,5,6,7,8,9]'::minivec, '[1,2,3,4,5,6,7,8,9]');
+ cosine_distance 
+-----------------
+               0
+(1 row)
+
+SELECT cosine_distance('[1,2,3,4,5,6,7,8,9]'::minivec, '[-1,-2,-3,-4,-5,-6,-7,-8,-9]');
+ cosine_distance 
+-----------------
+               2
+(1 row)
+
+SELECT '[1,2]'::minivec <=> '[2,4]';
+ ?column? 
+----------
+        0
+(1 row)
+
+SELECT l1_distance('[0,0]'::minivec, '[3,4]');
+ l1_distance 
+-------------
+           7
+(1 row)
+
+SELECT l1_distance('[0,0]'::minivec, '[0,1]');
+ l1_distance 
+-------------
+           1
+(1 row)
+
+SELECT l1_distance('[1,2]'::minivec, '[3]');
+ERROR:  different minivec dimensions 2 and 1
+SELECT l1_distance('[1,2,3,4,5,6,7,8,9]'::minivec, '[1,2,3,4,5,6,7,8,9]');
+ l1_distance 
+-------------
+           0
+(1 row)
+
+SELECT l1_distance('[1,2,3,4,5,6,7,8,9]'::minivec, '[0,3,2,5,4,7,6,9,8]');
+ l1_distance 
+-------------
+           7
+(1 row)
+
+SELECT '[0,0]'::minivec <+> '[3,4]';
+ ?column? 
+----------
+        7
+(1 row)
+
+SELECT l2_normalize('[3,4]'::minivec);
+ l2_normalize 
+--------------
+ [0.625,0.75]
+(1 row)
+
+SELECT l2_normalize('[3,0]'::minivec);
+ l2_normalize 
+--------------
+ [1,0]
+(1 row)
+
+SELECT l2_normalize('[0,0.1]'::minivec);
+ l2_normalize 
+--------------
+ [0,1]
+(1 row)
+
+SELECT l2_normalize('[0,0]'::minivec);
+ l2_normalize 
+--------------
+ [0,0]
+(1 row)
+
+SELECT l2_normalize('[448]'::minivec);
+ l2_normalize 
+--------------
+ [1]
+(1 row)
+
+SELECT binary_quantize('[1,0,-1]'::minivec);
+ binary_quantize 
+-----------------
+ 100
+(1 row)
+
+SELECT binary_quantize('[0,0.1,-0.2,-0.3,0.4,0.5,0.6,-0.7,0.8,-0.9,1]'::minivec);
+ binary_quantize 
+-----------------
+ 01001110101
+(1 row)
+
+SELECT subvector('[1,2,3,4,5]'::minivec, 1, 3);
+ subvector 
+-----------
+ [1,2,3]
+(1 row)
+
+SELECT subvector('[1,2,3,4,5]'::minivec, 3, 2);
+ subvector 
+-----------
+ [3,4]
+(1 row)
+
+SELECT subvector('[1,2,3,4,5]'::minivec, -1, 3);
+ subvector 
+-----------
+ [1]
+(1 row)
+
+SELECT subvector('[1,2,3,4,5]'::minivec, 3, 9);
+ subvector 
+-----------
+ [3,4,5]
+(1 row)
+
+SELECT subvector('[1,2,3,4,5]'::minivec, 1, 0);
+ERROR:  minivec must have at least 1 dimension
+SELECT subvector('[1,2,3,4,5]'::minivec, 3, -1);
+ERROR:  minivec must have at least 1 dimension
+SELECT subvector('[1,2,3,4,5]'::minivec, -1, 2);
+ERROR:  minivec must have at least 1 dimension
+SELECT subvector('[1,2,3,4,5]'::minivec, 2147483647, 10);
+ERROR:  minivec must have at least 1 dimension
+SELECT subvector('[1,2,3,4,5]'::minivec, 3, 2147483647);
+ subvector 
+-----------
+ [3,4,5]
+(1 row)
+
+SELECT subvector('[1,2,3,4,5]'::minivec, -2147483644, 2147483647);
+ subvector 
+-----------
+ [1,2]
+(1 row)
+

test/sql/btree.sql

@@ -22,6 +22,17 @@ SELECT * FROM t ORDER BY val;
 
 DROP TABLE t;
 
+-- minivec
+
+CREATE TABLE t (val minivec(3));
+INSERT INTO t (val) VALUES ('[0,0,0]'), ('[1,2,3]'), ('[1,1,1]'), (NULL);
+CREATE INDEX ON t (val);
+
+SELECT * FROM t WHERE val = '[1,2,3]';
+SELECT * FROM t ORDER BY val;
+
+DROP TABLE t;
+
 -- sparsevec
 
 CREATE TABLE t (val sparsevec(3));

test/sql/cast.sql

@@ -38,6 +38,22 @@ SELECT '{1,2,3}'::real[]::halfvec(2);
 SELECT '{65520,-65520}'::real[]::halfvec;
 SELECT '{1e-8,-1e-8}'::real[]::halfvec;
 
+SELECT '[1,2,3]'::vector::minivec;
+SELECT '[1,2,3]'::vector::minivec(3);
+SELECT '[1,2,3]'::vector::minivec(2);
+SELECT '[61440]'::vector::minivec;
+SELECT '[1e-8]'::vector::minivec;
+
+SELECT '[1,2,3]'::minivec::vector;
+SELECT '[1,2,3]'::minivec::vector(3);
+SELECT '[1,2,3]'::minivec::vector(2);
+
+SELECT '{1,2,3}'::real[]::minivec;
+SELECT '{1,2,3}'::real[]::minivec(3);
+SELECT '{1,2,3}'::real[]::minivec(2);
+SELECT '{61440,-61440}'::real[]::minivec;
+SELECT '{1e-8,-1e-8}'::real[]::minivec;
+
 SELECT '[0,1.5,0,3.5,0]'::vector::sparsevec;
 SELECT '[0,1.5,0,3.5,0]'::vector::sparsevec(5);
 SELECT '[0,1.5,0,3.5,0]'::vector::sparsevec(4);

test/sql/copy.sql

@@ -28,6 +28,21 @@ SELECT * FROM t2 ORDER BY val;
 DROP TABLE t;
 DROP TABLE t2;
 
+-- minivec
+
+CREATE TABLE t (val minivec(3));
+INSERT INTO t (val) VALUES ('[0,0,0]'), ('[1,2,3]'), ('[1,1,1]'), (NULL);
+
+CREATE TABLE t2 (val minivec(3));
+
+\copy t TO 'results/minivec.bin' WITH (FORMAT binary)
+\copy t2 FROM 'results/minivec.bin' WITH (FORMAT binary)
+
+SELECT * FROM t2 ORDER BY val;
+
+DROP TABLE t;
+DROP TABLE t2;
+
 -- sparsevec
 
 CREATE TABLE t (val sparsevec(3));

test/sql/ivfflat_minivec.sql

@@ -0,0 +1,45 @@
+SET enable_seqscan = off;
+
+-- L2
+
+CREATE TABLE t (val minivec(3));
+INSERT INTO t (val) VALUES ('[0,0,0]'), ('[1,2,3]'), ('[1,1,1]'), (NULL);
+CREATE INDEX ON t USING ivfflat (val minivec_l2_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::minivec)) t2;
+SELECT COUNT(*) FROM t;
+
+TRUNCATE t;
+SELECT * FROM t ORDER BY val <-> '[3,3,3]';
+
+DROP TABLE t;
+
+-- inner product
+
+CREATE TABLE t (val minivec(3));
+INSERT INTO t (val) VALUES ('[0,0,0]'), ('[1,2,3]'), ('[1,1,1]'), (NULL);
+CREATE INDEX ON t USING ivfflat (val minivec_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::minivec)) t2;
+
+DROP TABLE t;
+
+-- cosine
+
+CREATE TABLE t (val minivec(3));
+INSERT INTO t (val) VALUES ('[0,0,0]'), ('[1,2,3]'), ('[1,1,1]'), (NULL);
+CREATE INDEX ON t USING ivfflat (val minivec_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 COUNT(*) FROM (SELECT * FROM t ORDER BY val <=> (SELECT NULL::minivec)) t2;
+
+DROP TABLE t;

test/sql/minivec.sql

@@ -0,0 +1,134 @@
+SELECT '[1,2,3]'::minivec;
+SELECT '[-1,-2,-3]'::minivec;
+SELECT '[1.,2.,3.]'::minivec;
+SELECT ' [ 1,  2 ,    3  ] '::minivec;
+SELECT '[1.23456]'::minivec;
+SELECT '[hello,1]'::minivec;
+SELECT '[NaN,1]'::minivec;
+SELECT '[Infinity,1]'::minivec;
+SELECT '[-Infinity,1]'::minivec;
+SELECT '[61439,-61439]'::minivec;
+SELECT '[61440,-61440]'::minivec;
+SELECT '[1e-8,-1e-8]'::minivec;
+SELECT '[4e38,1]'::minivec;
+SELECT '[1e-46,1]'::minivec;
+SELECT '[1,2,3'::minivec;
+SELECT '[1,2,3]9'::minivec;
+SELECT '1,2,3'::minivec;
+SELECT ''::minivec;
+SELECT '['::minivec;
+SELECT '[ '::minivec;
+SELECT '[,'::minivec;
+SELECT '[]'::minivec;
+SELECT '[ ]'::minivec;
+SELECT '[,]'::minivec;
+SELECT '[1,]'::minivec;
+SELECT '[1a]'::minivec;
+SELECT '[1,,3]'::minivec;
+SELECT '[1, ,3]'::minivec;
+
+SELECT '[1,2,3]'::minivec(3);
+SELECT '[1,2,3]'::minivec(2);
+SELECT '[1,2,3]'::minivec(3, 2);
+SELECT '[1,2,3]'::minivec('a');
+SELECT '[1,2,3]'::minivec(0);
+SELECT '[1,2,3]'::minivec(16001);
+
+SELECT unnest('{"[1,2,3]", "[4,5,6]"}'::minivec[]);
+SELECT '{"[1,2,3]"}'::minivec(2)[];
+
+SELECT '[1,2,3]'::minivec + '[4,5,6]';
+SELECT '[61439]'::minivec + '[61439]';
+SELECT '[1,2]'::minivec + '[3]';
+
+SELECT '[1,2,3]'::minivec - '[4,5,6]';
+SELECT '[-61439]'::minivec - '[61439]';
+SELECT '[1,2]'::minivec - '[3]';
+
+SELECT '[1,2,3]'::minivec * '[4,5,6]';
+SELECT '[61439]'::minivec * '[61439]';
+SELECT '[1e-7]'::minivec * '[1e-7]';
+SELECT '[1,2]'::minivec * '[3]';
+
+SELECT '[1,2,3]'::minivec || '[4,5]';
+SELECT array_fill(0, ARRAY[16000])::minivec || '[1]';
+
+SELECT '[1,2,3]'::minivec < '[1,2,3]';
+SELECT '[1,2,3]'::minivec < '[1,2]';
+SELECT '[1,2,3]'::minivec <= '[1,2,3]';
+SELECT '[1,2,3]'::minivec <= '[1,2]';
+SELECT '[1,2,3]'::minivec = '[1,2,3]';
+SELECT '[1,2,3]'::minivec = '[1,2]';
+SELECT '[1,2,3]'::minivec != '[1,2,3]';
+SELECT '[1,2,3]'::minivec != '[1,2]';
+SELECT '[1,2,3]'::minivec >= '[1,2,3]';
+SELECT '[1,2,3]'::minivec >= '[1,2]';
+SELECT '[1,2,3]'::minivec > '[1,2,3]';
+SELECT '[1,2,3]'::minivec > '[1,2]';
+
+SELECT minivec_cmp('[1,2,3]', '[1,2,3]');
+SELECT minivec_cmp('[1,2,3]', '[0,0,0]');
+SELECT minivec_cmp('[0,0,0]', '[1,2,3]');
+SELECT minivec_cmp('[1,2]', '[1,2,3]');
+SELECT minivec_cmp('[1,2,3]', '[1,2]');
+SELECT minivec_cmp('[1,2]', '[2,3,4]');
+SELECT minivec_cmp('[2,3]', '[1,2,3]');
+
+SELECT vector_dims('[1,2,3]'::minivec);
+
+SELECT round(l2_norm('[1,1]'::minivec)::numeric, 5);
+SELECT l2_norm('[3,4]'::minivec);
+SELECT l2_norm('[0,1]'::minivec);
+SELECT l2_norm('[0,0]'::minivec);
+SELECT l2_norm('[2]'::minivec);
+
+SELECT l2_distance('[0,0]'::minivec, '[3,4]');
+SELECT l2_distance('[0,0]'::minivec, '[0,1]');
+SELECT l2_distance('[1,2]'::minivec, '[3]');
+SELECT l2_distance('[1,1,1,1,1,1,1,1,1]'::minivec, '[1,1,1,1,1,1,1,4,5]');
+SELECT '[0,0]'::minivec <-> '[3,4]';
+
+SELECT inner_product('[1,2]'::minivec, '[3,4]');
+SELECT inner_product('[1,2]'::minivec, '[3]');
+SELECT inner_product('[448]'::minivec, '[448]');
+SELECT inner_product('[1,1,1,1,1,1,1,1,1]'::minivec, '[1,2,3,4,5,6,7,8,9]');
+SELECT '[1,2]'::minivec <#> '[3,4]';
+
+SELECT cosine_distance('[1,2]'::minivec, '[2,4]');
+SELECT cosine_distance('[1,2]'::minivec, '[0,0]');
+SELECT cosine_distance('[1,1]'::minivec, '[1,1]');
+SELECT cosine_distance('[1,0]'::minivec, '[0,2]');
+SELECT cosine_distance('[1,1]'::minivec, '[-1,-1]');
+SELECT cosine_distance('[1,2]'::minivec, '[3]');
+SELECT cosine_distance('[1,1]'::minivec, '[1.1,1.1]');
+SELECT cosine_distance('[1,1]'::minivec, '[-1.1,-1.1]');
+SELECT cosine_distance('[1,2,3,4,5,6,7,8,9]'::minivec, '[1,2,3,4,5,6,7,8,9]');
+SELECT cosine_distance('[1,2,3,4,5,6,7,8,9]'::minivec, '[-1,-2,-3,-4,-5,-6,-7,-8,-9]');
+SELECT '[1,2]'::minivec <=> '[2,4]';
+
+SELECT l1_distance('[0,0]'::minivec, '[3,4]');
+SELECT l1_distance('[0,0]'::minivec, '[0,1]');
+SELECT l1_distance('[1,2]'::minivec, '[3]');
+SELECT l1_distance('[1,2,3,4,5,6,7,8,9]'::minivec, '[1,2,3,4,5,6,7,8,9]');
+SELECT l1_distance('[1,2,3,4,5,6,7,8,9]'::minivec, '[0,3,2,5,4,7,6,9,8]');
+SELECT '[0,0]'::minivec <+> '[3,4]';
+
+SELECT l2_normalize('[3,4]'::minivec);
+SELECT l2_normalize('[3,0]'::minivec);
+SELECT l2_normalize('[0,0.1]'::minivec);
+SELECT l2_normalize('[0,0]'::minivec);
+SELECT l2_normalize('[448]'::minivec);
+
+SELECT binary_quantize('[1,0,-1]'::minivec);
+SELECT binary_quantize('[0,0.1,-0.2,-0.3,0.4,0.5,0.6,-0.7,0.8,-0.9,1]'::minivec);
+
+SELECT subvector('[1,2,3,4,5]'::minivec, 1, 3);
+SELECT subvector('[1,2,3,4,5]'::minivec, 3, 2);
+SELECT subvector('[1,2,3,4,5]'::minivec, -1, 3);
+SELECT subvector('[1,2,3,4,5]'::minivec, 3, 9);
+SELECT subvector('[1,2,3,4,5]'::minivec, 1, 0);
+SELECT subvector('[1,2,3,4,5]'::minivec, 3, -1);
+SELECT subvector('[1,2,3,4,5]'::minivec, -1, 2);
+SELECT subvector('[1,2,3,4,5]'::minivec, 2147483647, 10);
+SELECT subvector('[1,2,3,4,5]'::minivec, 3, 2147483647);
+SELECT subvector('[1,2,3,4,5]'::minivec, -2147483644, 2147483647);

test/t/002_ivfflat_vacuum.pl

@@ -6,7 +6,13 @@ use Test::More;
 
 my $dim = 3;
 
-my $array_sql = join(",", ('random()') x $dim);
+my @r = ();
+for (1 .. $dim)
+{
+	my $v = int(rand(1000)) + 1;
+	push(@r, "i % $v");
+}
+my $array_sql = join(", ", @r);
 
 # Initialize node
 my $node = PostgreSQL::Test::Cluster->new('node');
@@ -17,20 +23,19 @@ $node->start;
 $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, 100000) i;"
+	"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);");
 
 # Get size
 my $size = $node->safe_psql("postgres", "SELECT pg_total_relation_size('tst_v_idx');");
 
-# Store values
-$node->safe_psql("postgres", "CREATE TABLE tmp AS SELECT * FROM tst;");
-
 # 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 * FROM tmp;");
+$node->safe_psql("postgres",
+	"INSERT INTO tst SELECT i % 10, ARRAY[$array_sql] FROM generate_series(1, 100000) i;"
+);
 
 # Check size
 my $new_size = $node->safe_psql("postgres", "SELECT pg_total_relation_size('tst_v_idx');");

test/t/009_ivfflat_filtering.pl

@@ -94,7 +94,8 @@ like($explain, qr/Seq Scan/);
 $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/);
+# 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);");
@@ -109,6 +110,7 @@ $node->safe_psql("postgres", "CREATE INDEX partial_idx ON tst USING ivfflat (v v
 $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/);
+# TODO Use partial index
+like($explain, qr/Index Scan using idx/);
 
 done_testing();

test/t/017_hnsw_filtering.pl

@@ -18,13 +18,9 @@ $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", "CREATE TABLE cat (i int4 PRIMARY KEY, t text, b boolean);");
 $node->safe_psql("postgres",
 	"INSERT INTO tst SELECT i, ARRAY[$array_sql], i % $nc, 'test ' || i FROM generate_series(1, 10000) i;"
 );
-$node->safe_psql("postgres",
-	"INSERT INTO cat SELECT i, 'cat ' || i, i % 5 = 0 FROM generate_series(1, $nc) i;"
-);
 $node->safe_psql("postgres", "CREATE INDEX idx ON tst USING hnsw (v vector_l2_ops);");
 $node->safe_psql("postgres", "ANALYZE tst;");
 
@@ -41,7 +37,8 @@ my $c = int(rand() * $nc);
 my $explain = $node->safe_psql("postgres", qq(
 	EXPLAIN ANALYZE SELECT i FROM tst WHERE c = $c ORDER BY v <-> '$query' LIMIT $limit;
 ));
-like($explain, qr/Seq Scan/);
+# TODO Do not use index
+like($explain, qr/Index Scan using idx/);
 
 # Test attribute filtering with few rows removed
 $explain = $node->safe_psql("postgres", qq(
@@ -59,7 +56,8 @@ like($explain, qr/Index Scan using idx/);
 $explain = $node->safe_psql("postgres", qq(
 	EXPLAIN ANALYZE SELECT i FROM tst WHERE c < 1 ORDER BY v <-> '$query' LIMIT $limit;
 ));
-like($explain, qr/Seq Scan/);
+# 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(
@@ -98,25 +96,13 @@ $explain = $node->safe_psql("postgres", qq(
 ));
 like($explain, qr/Seq Scan/);
 
-# Test join
-$explain = $node->safe_psql("postgres", qq(
-	EXPLAIN ANALYZE SELECT cat.t FROM cat INNER JOIN tst ON cat.i = tst.c ORDER BY v <-> '$query' LIMIT $limit;
-));
-like($explain, qr/Index Scan using idx/);
-
-# Test join with attribute filtering
-$explain = $node->safe_psql("postgres", qq(
-	EXPLAIN ANALYZE SELECT cat.t FROM cat INNER JOIN tst ON cat.i = tst.c WHERE cat.b = 't' ORDER BY v <-> '$query' LIMIT $limit;
-));
-like($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 $limit;
 ));
-# Use attribute index
-like($explain, qr/Bitmap Index Scan on attribute_idx/);
+# 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);");

test/t/033_comparison.pl

@@ -40,6 +40,10 @@ for (1 .. 50)
 	$actual = $node->safe_psql("postgres", "SELECT halfvec_cmp(v::halfvec, '$query'::real[]::halfvec) FROM tst");
 	is($expected, $actual);
 
+	# Test minivec
+	$actual = $node->safe_psql("postgres", "SELECT minivec_cmp(v::minivec, '$query'::real[]::minivec) FROM tst");
+	is($expected, $actual);
+
 	# Test sparsevec
 	$actual = $node->safe_psql("postgres", "SELECT sparsevec_cmp(v::vector::sparsevec, '$query'::real[]::vector::sparsevec) FROM tst");
 	is($expected, $actual);

test/t/034_distance_functions.pl

@@ -45,6 +45,10 @@ for my $function (@functions)
 		my $actual = $node->safe_psql("postgres", "SELECT $function(v::halfvec, '$query'::vector::halfvec) FROM tst");
 		is($expected, $actual, "halfvec $function");
 
+		# Test minivec
+		$actual = $node->safe_psql("postgres", "SELECT $function(v::minivec, '$query'::vector::minivec) FROM tst");
+		is($expected, $actual, "minivec $function");
+
 		# Test sparsevec
 		$actual = $node->safe_psql("postgres", "SELECT $function(v::sparsevec, '$query'::vector::sparsevec) FROM tst");
 		is($expected, $actual, "sparsevec $function");

test/t/037_inputs.pl

@@ -12,8 +12,8 @@ $node->start;
 # Create extension
 $node->safe_psql("postgres", "CREATE EXTENSION vector;");
 
-my @types = ("vector", "halfvec", "sparsevec");
-my @inputs = ("[1.23,4.56,7.89]", "[1.23,4.56,7.89]", "{1:1.23,2:4.56,3:7.89}/3");
+my @types = ("vector", "halfvec", "minivec", "sparsevec");
+my @inputs = ("[1.23,4.56,7.89]", "[1.23,4.56,7.89]", "[1.23,4.56,7.89]", "{1:1.23,2:4.56,3:7.89}/3");
 my @subs = (" ", " ", ",", ":", "-", "1", "9", "\0", "2147483648", "-2147483649");
 
 for my $i (0 .. $#types)

test/t/039_hnsw_cost.pl

@@ -1,60 +0,0 @@
-use strict;
-use warnings FATAL => 'all';
-use PostgreSQL::Test::Cluster;
-use PostgreSQL::Test::Utils;
-use Test::More;
-
-my @dims = (384, 1536);
-my $limit = 10;
-
-# Initialize node
-my $node = PostgreSQL::Test::Cluster->new('node');
-$node->init;
-$node->start;
-
-$node->safe_psql("postgres", "CREATE EXTENSION vector;");
-
-for my $dim (@dims)
-{
-	my $array_sql = join(",", ('random()') x $dim);
-
-	# Create table and index
-	$node->safe_psql("postgres", "CREATE TABLE tst (i int4, v vector($dim));");
-	$node->safe_psql("postgres",
-		"INSERT INTO tst SELECT i, ARRAY[$array_sql] FROM generate_series(1, 2000) 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 $explain = $node->safe_psql("postgres", qq(
-		EXPLAIN ANALYZE SELECT i FROM tst ORDER BY v <-> '$query' LIMIT $limit;
-	));
-	like($explain, qr/Index Scan using idx/);
-
-	# 3x the rows are needed for distance filters
-	# since the planner uses DEFAULT_INEQ_SEL for the selectivity (should be 1)
-	# Recreate index for performance
-	$node->safe_psql("postgres", "DROP INDEX idx;");
-	$node->safe_psql("postgres",
-		"INSERT INTO tst SELECT i, ARRAY[$array_sql] FROM generate_series(2001, 6000) i;"
-	);
-	$node->safe_psql("postgres", "CREATE INDEX idx ON tst USING hnsw (v vector_l2_ops);");
-	$node->safe_psql("postgres", "ANALYZE tst;");
-
-	$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/);
-
-	$node->safe_psql("postgres", "DROP TABLE tst;");
-}
-
-done_testing();

test/t/039_hnsw_minivec_build_recall.pl

@@ -0,0 +1,132 @@
+use strict;
+use warnings FATAL => 'all';
+use PostgreSQL::Test::Cluster;
+use PostgreSQL::Test::Utils;
+use Test::More;
+
+my $node;
+my @queries = ();
+my @expected;
+my $limit = 20;
+my $dim = 10;
+my $array_sql = join(",", ('2 * random() * random()') x $dim);
+
+sub test_recall
+{
+	my ($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 = PostgreSQL::Test::Cluster->new('node');
+$node->init;
+$node->start;
+
+# Create table
+$node->safe_psql("postgres", "CREATE EXTENSION vector;");
+$node->safe_psql("postgres", "CREATE TABLE tst (i int4, v minivec($dim));");
+$node->safe_psql("postgres",
+	"INSERT INTO tst SELECT i, ARRAY[$array_sql] FROM generate_series(1, 10000) i;"
+);
+
+# Generate queries
+for (1 .. 20)
+{
+	my @r = ();
+	for (1 .. $dim)
+	{
+		push(@r, rand());
+	}
+	push(@queries, "[" . join(",", @r) . "]");
+}
+
+# Check each index type
+my @operators = ("<->", "<#>", "<=>", "<+>");
+my @opclasses = ("minivec_l2_ops", "minivec_ip_ops", "minivec_cosine_ops", "minivec_l1_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 $min = 0.98;
+	test_recall($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/040_hnsw_minivec_insert_recall.pl

@@ -0,0 +1,113 @@
+use strict;
+use warnings FATAL => 'all';
+use PostgreSQL::Test::Cluster;
+use PostgreSQL::Test::Utils;
+use Test::More;
+
+my $node;
+my @queries = ();
+my @expected;
+my $limit = 20;
+my $dim = 10;
+my $array_sql = join(",", ('2 * random() * random()') x $dim);
+
+sub test_recall
+{
+	my ($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 = PostgreSQL::Test::Cluster->new('node');
+$node->init;
+$node->start;
+
+# Create table
+$node->safe_psql("postgres", "CREATE EXTENSION vector;");
+$node->safe_psql("postgres", "CREATE TABLE tst (i serial, v minivec($dim));");
+
+# Generate queries
+for (1 .. 20)
+{
+	my @r = ();
+	for (1 .. $dim)
+	{
+		push(@r, rand());
+	}
+	push(@queries, "[" . join(",", @r) . "]");
+}
+
+# Check each index type
+my @operators = ("<->", "<#>", "<=>", "<+>");
+my @opclasses = ("minivec_l2_ops", "minivec_ip_ops", "minivec_cosine_ops", "minivec_l1_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",
+		{
+			"040_hnsw_minivec_insert_recall_$opclass" => "INSERT INTO tst (v) VALUES (ARRAY[$array_sql]);"
+		}
+	);
+
+	# 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
+	my $min = 0.98;
+	test_recall($min, $operator);
+
+	$node->safe_psql("postgres", "DROP INDEX idx;");
+	$node->safe_psql("postgres", "TRUNCATE tst;");
+}
+
+done_testing();

test/t/040_ivfflat_cost.pl

@@ -1,50 +0,0 @@
-use strict;
-use warnings FATAL => 'all';
-use PostgreSQL::Test::Cluster;
-use PostgreSQL::Test::Utils;
-use Test::More;
-
-my @dims = (384, 1536);
-my $limit = 10;
-
-# Initialize node
-my $node = PostgreSQL::Test::Cluster->new('node');
-$node->init;
-$node->start;
-
-$node->safe_psql("postgres", "CREATE EXTENSION vector;");
-
-for my $dim (@dims)
-{
-	my $array_sql = join(",", ('random()') x $dim);
-
-	# Create table and index
-	$node->safe_psql("postgres", "CREATE TABLE tst (i int4, v vector($dim));");
-	$node->safe_psql("postgres",
-		"INSERT INTO tst SELECT i, ARRAY[$array_sql] FROM generate_series(1, 5000) i;"
-	);
-	$node->safe_psql("postgres", "CREATE INDEX idx ON tst USING ivfflat (v vector_l2_ops) WITH (lists = 5);");
-	$node->safe_psql("postgres", "ANALYZE tst;");
-
-	# Generate query
-	my @r = ();
-	for (1 .. $dim)
-	{
-		push(@r, rand());
-	}
-	my $query = "[" . join(",", @r) . "]";
-
-	my $explain = $node->safe_psql("postgres", qq(
-		EXPLAIN ANALYZE SELECT i FROM tst ORDER BY v <-> '$query' LIMIT $limit;
-	));
-	like($explain, qr/Index Scan using idx/);
-
-	$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/);
-
-	$node->safe_psql("postgres", "DROP TABLE tst;");
-}
-
-done_testing();

test/t/041_hnsw_minivec_vacuum_recall.pl

@@ -0,0 +1,97 @@
+use strict;
+use warnings FATAL => 'all';
+use PostgreSQL::Test::Cluster;
+use PostgreSQL::Test::Utils;
+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 = PostgreSQL::Test::Cluster->new('node');
+$node->init;
+$node->start;
+
+# Create table
+$node->safe_psql("postgres", "CREATE EXTENSION vector;");
+$node->safe_psql("postgres", "CREATE TABLE tst (i int4, v minivec(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
+$node->safe_psql("postgres", "CREATE INDEX ON tst USING hnsw (v minivec_l2_ops) WITH (m = 4, ef_construction = 8);");
+
+# Delete data
+$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(0.18, $limit, "before vacuum");
+test_recall(0.93, 100, "before vacuum");
+
+# TODO Test concurrent inserts with vacuum
+$node->safe_psql("postgres", "VACUUM tst;");
+
+test_recall(0.95, $limit, "after vacuum");
+
+done_testing();

test/t/041_ivfflat_iterative_search.pl

@@ -1,54 +0,0 @@
-use strict;
-use warnings FATAL => 'all';
-use PostgreSQL::Test::Cluster;
-use PostgreSQL::Test::Utils;
-use Test::More;
-
-my $dim = 3;
-my $array_sql = join(",", ('random()') x $dim);
-
-# Initialize node
-my $node = PostgreSQL::Test::Cluster->new('node');
-$node->init;
-$node->start;
-
-# Create table
-$node->safe_psql("postgres", "CREATE EXTENSION vector;");
-$node->safe_psql("postgres", "CREATE TABLE tst (i int4 PRIMARY KEY, v vector($dim));");
-$node->safe_psql("postgres",
-	"INSERT INTO tst SELECT i, ARRAY[$array_sql] FROM generate_series(1, 100000) i;"
-);
-$node->safe_psql("postgres", "CREATE INDEX ON tst USING ivfflat (v vector_l2_ops);");
-
-my $count = $node->safe_psql("postgres", qq(
-	SET enable_seqscan = off;
-	SET ivfflat.probes = 10;
-	SET ivfflat.iterative_search = on;
-	SELECT COUNT(*) FROM (SELECT v FROM tst WHERE i % 10000 = 0 ORDER BY v <-> (SELECT v FROM tst LIMIT 1) LIMIT 11) t;
-));
-is($count, 10);
-
-foreach ((30, 50, 70))
-{
-	my $max_probes = $_;
-	my $expected = $max_probes / 10;
-	my $sum = 0;
-
-	for my $i (1 .. 20)
-	{
-		$count = $node->safe_psql("postgres", qq(
-			SET enable_seqscan = off;
-			SET ivfflat.probes = 10;
-			SET ivfflat.iterative_search = on;
-			SET ivfflat.iterative_search_max_probes = $max_probes;
-			SELECT COUNT(*) FROM (SELECT v FROM tst WHERE i % 10000 = 0 ORDER BY v <-> (SELECT v FROM tst WHERE i = $i) LIMIT 11) t;
-		));
-		$sum += $count;
-	}
-
-	my $avg = $sum / 20;
-	cmp_ok($avg, '>', $expected - 2);
-	cmp_ok($avg, '<', $expected + 2);
-}
-
-done_testing();

test/t/042_hnsw_minivec_duplicates.pl

@@ -0,0 +1,58 @@
+use strict;
+use warnings FATAL => 'all';
+use PostgreSQL::Test::Cluster;
+use PostgreSQL::Test::Utils;
+use Test::More;
+
+# Initialize node
+my $node = PostgreSQL::Test::Cluster->new('node');
+$node->init;
+$node->start;
+
+# Create table
+$node->safe_psql("postgres", "CREATE EXTENSION vector;");
+$node->safe_psql("postgres", "CREATE TABLE tst (v minivec(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();
+$node->safe_psql("postgres", "CREATE INDEX idx ON tst USING hnsw (v minivec_l2_ops);");
+test_duplicates();
+
+# Reset
+$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",
+	{
+		"042_hnsw_minivec_duplicates" => "INSERT INTO tst VALUES ('[1,1,1]');"
+	}
+);
+
+done_testing();

test/t/042_ivfflat_iterative_search_recall.pl

@@ -1,125 +0,0 @@
-use strict;
-use warnings FATAL => 'all';
-use PostgreSQL::Test::Cluster;
-use PostgreSQL::Test::Utils;
-use Test::More;
-
-my $node;
-my @queries = ();
-my @expected;
-my $limit = 20;
-my @cs = (100, 1000);
-
-sub test_recall
-{
-	my ($c, $probes, $min, $operator) = @_;
-	my $correct = 0;
-	my $total = 0;
-
-	my $explain = $node->safe_psql("postgres", qq(
-		SET enable_seqscan = off;
-		SET ivfflat.probes = $probes;
-		SET ivfflat.iterative_search = on;
-		EXPLAIN ANALYZE SELECT i FROM tst WHERE i % $c = 0 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;
-			SET ivfflat.iterative_search = on;
-			SELECT i FROM tst WHERE i % $c = 0 ORDER BY v $operator '$queries[$i]' LIMIT $limit;
-		));
-		my @actual_ids = split("\n", $actual);
-
-		my @expected_ids = split("\n", $expected[$i]);
-		my %expected_set = map { $_ => 1 } @expected_ids;
-
-		foreach (@actual_ids)
-		{
-			if (exists($expected_set{$_}))
-			{
-				$correct++;
-			}
-		}
-
-		$total += $limit;
-	}
-
-	cmp_ok($correct / $total, ">=", $min, $operator);
-}
-
-# Initialize node
-$node = PostgreSQL::Test::Cluster->new('node');
-$node->init;
-$node->start;
-
-# Create table
-$node->safe_psql("postgres", "CREATE EXTENSION vector;");
-$node->safe_psql("postgres", "CREATE TABLE tst (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_cosine_ops");
-
-for my $i (0 .. $#operators)
-{
-	my $operator = $operators[$i];
-	my $opclass = $opclasses[$i];
-
-	$node->safe_psql("postgres", "CREATE INDEX idx ON tst USING ivfflat (v $opclass);");
-
-	foreach (@cs)
-	{
-		my $c = $_;
-
-		# Get exact results
-		@expected = ();
-		foreach (@queries)
-		{
-			my $res = $node->safe_psql("postgres", qq(
-				SET enable_indexscan = off;
-				WITH top AS (
-					SELECT v $operator '$_' AS distance FROM tst WHERE i % $c = 0 ORDER BY distance LIMIT $limit
-				)
-				SELECT i FROM tst WHERE (v $operator '$_') <= (SELECT MAX(distance) FROM top)
-			));
-			push(@expected, $res);
-		}
-
-		if ($c == 100)
-		{
-			test_recall($c, 1, 0.58, $operator);
-			test_recall($c, 10, 0.98, $operator);
-		}
-		else
-		{
-			if ($operator eq "<->")
-			{
-				test_recall($c, 1, 0.80, $operator);
-			}
-			else
-			{
-				test_recall($c, 1, 0.88, $operator);
-			}
-		}
-	}
-
-	$node->safe_psql("postgres", "DROP INDEX idx;");
-}
-
-done_testing();

test/t/043_hnsw_iterative_search.pl

@@ -1,67 +0,0 @@
-use strict;
-use warnings FATAL => 'all';
-use PostgreSQL::Test::Cluster;
-use PostgreSQL::Test::Utils;
-use Test::More;
-
-my $dim = 3;
-my $array_sql = join(",", ('random()') x $dim);
-
-# Initialize node
-my $node = PostgreSQL::Test::Cluster->new('node');
-$node->init;
-$node->start;
-
-# Create table
-$node->safe_psql("postgres", "CREATE EXTENSION vector;");
-$node->safe_psql("postgres", "CREATE TABLE tst (i int4 PRIMARY KEY, v vector($dim));");
-$node->safe_psql("postgres",
-	"INSERT INTO tst SELECT i, ARRAY[$array_sql] FROM generate_series(1, 100000) i;"
-);
-$node->safe_psql("postgres", qq(
-	SET maintenance_work_mem = '128MB';
-	SET max_parallel_maintenance_workers = 2;
-	CREATE INDEX ON tst USING hnsw (v vector_l2_ops)
-));
-
-my $count = $node->safe_psql("postgres", qq(
-	SET enable_seqscan = off;
-	SET hnsw.iterative_search = on;
-	SET work_mem = '8MB';
-	SELECT COUNT(*) FROM (SELECT v FROM tst WHERE i % 10000 = 0 ORDER BY v <-> (SELECT v FROM tst LIMIT 1) LIMIT 11) t;
-));
-is($count, 10);
-
-foreach ((30000, 50000, 70000))
-{
-	my $max_tuples = $_;
-	my $expected = $max_tuples / 10000;
-	my $sum = 0;
-
-	for my $i (1 .. 20)
-	{
-		$count = $node->safe_psql("postgres", qq(
-			SET enable_seqscan = off;
-			SET hnsw.iterative_search = on;
-			SET hnsw.iterative_search_max_tuples = $max_tuples;
-			SET work_mem = '8MB';
-			SELECT COUNT(*) FROM (SELECT v FROM tst WHERE i % 10000 = 0 ORDER BY v <-> (SELECT v FROM tst WHERE i = $i) LIMIT 11) t;
-		));
-		$sum += $count;
-	}
-
-	my $avg = $sum / 20;
-	cmp_ok($avg, '>', $expected - 2);
-	cmp_ok($avg, '<', $expected + 2);
-}
-
-my ($ret, $stdout, $stderr) = $node->psql("postgres", qq(
-	SET enable_seqscan = off;
-	SET hnsw.iterative_search = on;
-	SET client_min_messages = debug1;
-	SET work_mem = '2MB';
-	SELECT COUNT(*) FROM (SELECT v FROM tst WHERE i % 10000 = 0 ORDER BY v <-> (SELECT v FROM tst LIMIT 1) LIMIT 11) t;
-));
-like($stderr, qr/hnsw index scan exceeded work_mem after \d+ tuples/);
-
-done_testing();

test/t/043_ivfflat_minivec_build_recall.pl

@@ -0,0 +1,154 @@
+use strict;
+use warnings FATAL => 'all';
+use PostgreSQL::Test::Cluster;
+use PostgreSQL::Test::Utils;
+use Test::More;
+
+my $node;
+my @queries = ();
+my @expected;
+my $limit = 20;
+my $dim = 10;
+my $array_sql = join(",", ('random()') x $dim);
+
+sub test_recall
+{
+	my ($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 @expected_ids = split("\n", $expected[$i]);
+		my %expected_set = map { $_ => 1 } @expected_ids;
+
+		foreach (@actual_ids)
+		{
+			if (exists($expected_set{$_}))
+			{
+				$correct++;
+			}
+		}
+
+		$total += $limit;
+	}
+
+	cmp_ok($correct / $total, ">=", $min, $operator);
+}
+
+# Initialize node
+$node = PostgreSQL::Test::Cluster->new('node');
+$node->init;
+$node->start;
+
+# Create table
+$node->safe_psql("postgres", "CREATE EXTENSION vector;");
+$node->safe_psql("postgres", "CREATE TABLE tst (i int4, v minivec($dim));");
+$node->safe_psql("postgres",
+	"INSERT INTO tst SELECT i, ARRAY[$array_sql] FROM generate_series(1, 100000) i;"
+);
+
+# Generate queries
+for (1 .. 20)
+{
+	my @r = ();
+	for (1 .. $dim)
+	{
+		push(@r, rand());
+	}
+	push(@queries, "[" . join(",", @r) . "]");
+}
+
+# Check each index type
+my @operators = ("<->", "<#>", "<=>");
+my @opclasses = ("minivec_l2_ops", "minivec_ip_ops", "minivec_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", qq(
+			WITH top AS (
+				SELECT v $operator '$_' AS distance FROM tst ORDER BY distance LIMIT $limit
+			)
+			SELECT i FROM tst WHERE (v $operator '$_') <= (SELECT MAX(distance) FROM top)
+		));
+		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, 0.33, $operator);
+		test_recall(10, 0.93, $operator);
+	}
+
+	# Test probes equals lists
+	if ($operator eq "<=>")
+	{
+		test_recall(100, 0.98, $operator);
+	}
+	else
+	{
+		test_recall(100, 1.00, $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, 0.33, $operator);
+		test_recall(10, 0.93, $operator);
+	}
+
+	# Test probes equals lists
+	if ($operator eq "<=>")
+	{
+		test_recall(100, 0.98, $operator);
+	}
+	else
+	{
+		test_recall(100, 1.00, $operator);
+	}
+
+	$node->safe_psql("postgres", "DROP INDEX idx;");
+}
+
+done_testing();

test/t/044_hnsw_iterative_search_recall.pl

@@ -1,131 +0,0 @@
-use strict;
-use warnings FATAL => 'all';
-use PostgreSQL::Test::Cluster;
-use PostgreSQL::Test::Utils;
-use Test::More;
-
-my $node;
-my @queries = ();
-my @expected;
-my $limit = 20;
-my $dim = 3;
-my $array_sql = join(",", ('random()') x $dim);
-my @cs = (100, 1000);
-
-sub test_recall
-{
-	my ($c, $ef_search, $min, $operator) = @_;
-	my $correct = 0;
-	my $total = 0;
-
-	my $explain = $node->safe_psql("postgres", qq(
-		SET enable_seqscan = off;
-		SET hnsw.ef_search = $ef_search;
-		SET hnsw.iterative_search = on;
-		EXPLAIN ANALYZE SELECT i FROM tst WHERE i % $c = 0 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 hnsw.ef_search = $ef_search;
-			SET hnsw.iterative_search = on;
-			SELECT i FROM tst WHERE i % $c = 0 ORDER BY v $operator '$queries[$i]' LIMIT $limit;
-		));
-		my @actual_ids = split("\n", $actual);
-
-		my @expected_ids = split("\n", $expected[$i]);
-		my %expected_set = map { $_ => 1 } @expected_ids;
-
-		foreach (@actual_ids)
-		{
-			if (exists($expected_set{$_}))
-			{
-				$correct++;
-			}
-		}
-
-		$total += $limit;
-	}
-
-	cmp_ok($correct / $total, ">=", $min, $operator);
-}
-
-# Initialize node
-$node = PostgreSQL::Test::Cluster->new('node');
-$node->init;
-$node->start;
-
-# Create table
-$node->safe_psql("postgres", "CREATE EXTENSION vector;");
-$node->safe_psql("postgres", "CREATE TABLE tst (i int4, v vector($dim));");
-$node->safe_psql("postgres",
-	"INSERT INTO tst SELECT i, ARRAY[$array_sql] FROM generate_series(1, 100000) i;"
-);
-
-# Generate queries
-for (1 .. 20)
-{
-	my @r = ();
-	for (1 .. $dim)
-	{
-		push(@r, rand());
-	}
-	push(@queries, "[" . join(",", @r) . "]");
-}
-
-# Check each index type
-my @operators = ("<->", "<=>");
-my @opclasses = ("vector_l2_ops", "vector_cosine_ops");
-
-for my $i (0 .. $#operators)
-{
-	my $operator = $operators[$i];
-	my $opclass = $opclasses[$i];
-
-	$node->safe_psql("postgres", qq(
-		SET maintenance_work_mem = '128MB';
-		CREATE INDEX idx ON tst USING hnsw (v $opclass);
-	));
-
-	foreach (@cs)
-	{
-		my $c = $_;
-
-		# Get exact results
-		@expected = ();
-		foreach (@queries)
-		{
-			my $res = $node->safe_psql("postgres", qq(
-				SET enable_indexscan = off;
-				WITH top AS (
-					SELECT v $operator '$_' AS distance FROM tst WHERE i % $c = 0 ORDER BY distance LIMIT $limit
-				)
-				SELECT i FROM tst WHERE (v $operator '$_') <= (SELECT MAX(distance) FROM top)
-			));
-			push(@expected, $res);
-		}
-
-		if ($c == 100)
-		{
-			test_recall($c, 40, 0.99, $operator);
-		}
-		else
-		{
-			if ($operator eq "<->")
-			{
-				test_recall($c, 40, 0.99, $operator);
-			}
-			else
-			{
-				test_recall($c, 40, 0.99, $operator);
-			}
-		}
-	}
-
-	$node->safe_psql("postgres", "DROP INDEX idx;");
-}
-
-done_testing();

test/t/045_hnsw_hqann.pl

@@ -1,113 +0,0 @@
-use strict;
-use warnings FATAL => 'all';
-use PostgreSQL::Test::Cluster;
-use PostgreSQL::Test::Utils;
-use Test::More;
-
-my $node;
-my @queries = ();
-my @cs = ();
-my @expected;
-my $limit = 20;
-my $dim = 3;
-my $array_sql = join(",", ('random()') x $dim);
-my $nc = 1000;
-
-sub test_recall
-{
-	my ($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 WHERE c = $cs[0] ORDER BY v $operator '$queries[0]' LIMIT $limit;
-	));
-	like($explain, qr/Index Cond/);
-
-	for my $i (0 .. $#queries)
-	{
-		my $actual = $node->safe_psql("postgres", qq(
-			SET enable_seqscan = off;
-			SELECT i FROM tst WHERE c = $cs[$i] ORDER BY v $operator '$queries[$i]' LIMIT $limit;
-		));
-		my @actual_ids = split("\n", $actual);
-		my %actual_set = map { $_ => 1 } @actual_ids;
-
-		is(scalar(@actual_ids), $limit);
-
-		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 = PostgreSQL::Test::Cluster->new('node');
-$node->init;
-$node->start;
-
-# Create table
-$node->safe_psql("postgres", "CREATE EXTENSION vector;");
-$node->safe_psql("postgres", "CREATE TABLE tst (i int4, v vector($dim), c int4);");
-$node->safe_psql("postgres",
-	"INSERT INTO tst SELECT i, ARRAY[$array_sql], i % $nc FROM generate_series(1, 20000) i;"
-);
-
-# Generate queries
-for (1 .. 20)
-{
-	my @r = ();
-	for (1 .. $dim)
-	{
-		push(@r, rand());
-	}
-	push(@queries, "[" . join(",", @r) . "]");
-	push(@cs, int(rand() * $nc));
-}
-
-# Get exact results
-@expected = ();
-for my $i (0 .. $#queries)
-{
-	my $res = $node->safe_psql("postgres", "SELECT i FROM tst WHERE c = $cs[$i] ORDER BY v <=> '$queries[$i]' LIMIT $limit;");
-	push(@expected, $res);
-}
-
-# Add index
-$node->safe_psql("postgres", qq(
-	SET maintenance_work_mem = '256MB';
-	SET max_parallel_maintenance_workers = 2;
-	CREATE INDEX ON tst USING hnsw (v vector_cosine_ops, c);
-));
-
-# Test recall
-test_recall(0.99, '<=>');
-
-# Test vacuum
-$node->safe_psql("postgres", "DELETE FROM tst WHERE c > 5;");
-$node->safe_psql("postgres", "VACUUM tst;");
-
-# Test columns
-my ($ret, $stdout, $stderr) = $node->psql("postgres", "CREATE INDEX ON tst USING hnsw (c);");
-like($stderr, qr/first column must be a vector/);
-
-($ret, $stdout, $stderr) = $node->psql("postgres", "CREATE INDEX ON tst USING hnsw (c, v vector_cosine_ops);");
-like($stderr, qr/first column must be a vector/);
-
-($ret, $stdout, $stderr) = $node->psql("postgres", "CREATE INDEX ON tst USING hnsw (v vector_cosine_ops, c, c);");
-like($stderr, qr/index cannot have more than two columns/);
-
-($ret, $stdout, $stderr) = $node->psql("postgres", "CREATE INDEX ON tst USING hnsw (v vector_cosine_ops, v vector_cosine_ops);");
-like($stderr, qr/column 2 cannot be a vector/);
-
-done_testing();