Moved bit functions to separate file

Redefining a typedef is a C11 feature:

    In file included from src/hnsw.c:10:
    src/hnsw.h:147:5: warning: redefinition of typedef 'HnswElementData' is a C11 feature [-Wtypedef-redefinition]
    }                       HnswElementData;
                            ^
    src/hnsw.h:118:32: note: previous definition is here
    typedef struct HnswElementData HnswElementData;
                                   ^
    src/hnsw.h:163:5: warning: redefinition of typedef 'HnswNeighborArray' is a C11 feature [-Wtypedef-redefinition]
    }                       HnswNeighborArray;
                            ^
    src/hnsw.h:119:34: note: previous definition is here
    typedef struct HnswNeighborArray HnswNeighborArray;
                                     ^
    2 warnings generated.

I got these warnings when I built PostgreSQL with "CC=clang
CFLAGS=-std=gnu99"; other similar options would surely produce the
warnings too.

.github/workflows/build.yml

@@ -73,6 +73,7 @@ jobs:
           postgres-version: 14
       - run: |
           call "C:\Program Files\Microsoft Visual Studio\2022\Enterprise\VC\Auxiliary\Build\vcvars64.bat" && ^
+          cd %TEMP% && ^
           nmake /NOLOGO /F Makefile.win && ^
           nmake /NOLOGO /F Makefile.win install && ^
           nmake /NOLOGO /F Makefile.win installcheck && ^

CHANGELOG.md

@@ -1,11 +1,18 @@
-## 0.6.2 (unreleased)
+## 0.7.0 (unreleased)
+
+- Added support for binary vectors to HNSW
+- Added `hamming_distance` function
+- Added `jaccard_distance` function
+- Added `quantize_binary` function
+
+## 0.6.2 (2024-03-18)
 
 - Reduced lock contention with parallel HNSW index builds
 
 ## 0.6.1 (2024-03-04)
 
 - Fixed error with `ANALYZE` and vectors with different dimensions
-- Fixed error with `shared_preload_libraries`
+- Fixed segmentation fault with `shared_preload_libraries`
 - Fixed vector subtraction being marked as commutative
 
 ## 0.6.0 (2024-01-29)

META.json

@@ -2,7 +2,7 @@
 	"name": "vector",
 	"abstract": "Open-source vector similarity search for Postgres",
 	"description": "Supports L2 distance, inner product, and cosine distance",
-	"version": "0.6.1",
+	"version": "0.6.2",
 	"maintainer": [
 		"Andrew Kane <andrew@ankane.org>"
 	],
@@ -20,7 +20,7 @@
 		"vector": {
 			"file": "sql/vector.sql",
 			"docfile": "README.md",
-			"version": "0.6.1",
+			"version": "0.6.2",
 			"abstract": "Open-source vector similarity search for Postgres"
 		}
 	},

Makefile

@@ -1,9 +1,9 @@
 EXTENSION = vector
-EXTVERSION = 0.6.1
+EXTVERSION = 0.6.2
 
 MODULE_big = vector
 DATA = $(wildcard sql/*--*.sql)
-OBJS = src/hnsw.o src/hnswbuild.o src/hnswinsert.o src/hnswscan.o src/hnswutils.o src/hnswvacuum.o src/ivfbuild.o src/ivfflat.o src/ivfinsert.o src/ivfkmeans.o src/ivfscan.o src/ivfutils.o src/ivfvacuum.o src/vector.o
+OBJS = src/bitvector.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/vector.o
 HEADERS = src/vector.h
 
 TESTS = $(wildcard test/sql/*.sql)

Makefile.win

@@ -1,7 +1,7 @@
 EXTENSION = vector
-EXTVERSION = 0.6.1
+EXTVERSION = 0.6.2
 
-OBJS = src\hnsw.obj src\hnswbuild.obj src\hnswinsert.obj src\hnswscan.obj src\hnswutils.obj src\hnswvacuum.obj src\ivfbuild.obj src\ivfflat.obj src\ivfinsert.obj src\ivfkmeans.obj src\ivfscan.obj src\ivfutils.obj src\ivfvacuum.obj src\vector.obj
+OBJS = src\bitvector.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\vector.obj
 HEADERS = src\vector.h
 
 REGRESS = btree cast copy functions input ivfflat_cosine ivfflat_ip ivfflat_l2 ivfflat_options ivfflat_unlogged

README.md

@@ -5,7 +5,7 @@ Open-source vector similarity search for Postgres
 Store your vectors with the rest of your data. Supports:
 
 - exact and approximate nearest neighbor search
-- L2 distance, inner product, and cosine distance
+- L2 distance, inner product, cosine distance, and more
 - any [language](#languages) with a Postgres client
 
 Plus [ACID](https://en.wikipedia.org/wiki/ACID) compliance, point-in-time recovery, JOINs, and all of the other [great features](https://www.postgresql.org/about/) of Postgres
@@ -20,13 +20,13 @@ Compile and install the extension (supports Postgres 12+)
 
 ```sh
 cd /tmp
-git clone --branch v0.6.1 https://github.com/pgvector/pgvector.git
+git clone --branch v0.6.2 https://github.com/pgvector/pgvector.git
 cd pgvector
 make
 make install # may need sudo
 ```
 
-See the [installation notes](#installation-notes) if you run into issues
+See the [installation notes](#installation-notes---linux-and-mac) if you run into issues
 
 You can also install it with [Docker](#docker), [Homebrew](#homebrew), [PGXN](#pgxn), [APT](#apt), [Yum](#yum), [pkg](#pkg), or [conda-forge](#conda-forge), and it comes preinstalled with [Postgres.app](#postgresapp) and many [hosted providers](#hosted-postgres). There are also instructions for [GitHub Actions](https://github.com/pgvector/setup-pgvector).
 
@@ -44,12 +44,15 @@ Then use `nmake` to build:
 
 ```cmd
 set "PGROOT=C:\Program Files\PostgreSQL\16"
-git clone --branch v0.6.1 https://github.com/pgvector/pgvector.git
+cd %TEMP%
+git clone --branch v0.6.2 https://github.com/pgvector/pgvector.git
 cd pgvector
 nmake /F Makefile.win
 nmake /F Makefile.win install
 ```
 
+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).
 
 ## Getting Started
@@ -218,7 +221,19 @@ Cosine distance
 CREATE INDEX ON items USING hnsw (embedding vector_cosine_ops);
 ```
 
-Vectors with up to 2,000 dimensions can be indexed.
+Hamming distance - added in 0.7.0
+
+```sql
+CREATE INDEX ON items USING hnsw (embedding bit_hamming_ops);
+```
+
+Jaccard distance - added in 0.7.0
+
+```sql
+CREATE INDEX ON items USING hnsw (embedding bit_jaccard_ops);
+```
+
+Vectors with up to 2,000 dimensions can be indexed, or bit vectors with up to 64,000 dimensions.
 
 ### Index Options
 
@@ -410,13 +425,39 @@ You can use [Reciprocal Rank Fusion](https://github.com/pgvector/pgvector-python
 
 ## Performance
 
+### Tuning
+
+Use a tool like [PgTune](https://pgtune.leopard.in.ua/) to set initial values for Postgres server parameters.
+
+### Loading
+
+Use `COPY` for bulk loading data ([example](https://github.com/pgvector/pgvector-python/blob/master/examples/bulk_loading.py)).
+
+```sql
+COPY items (embedding) FROM STDIN WITH (FORMAT BINARY);
+```
+
+Add any indexes *after* loading the initial data for best performance.
+
+### Indexing
+
+See index build time for [HNSW](#index-build-time) and [IVFFlat](#index-build-time-1).
+
+In production environments, create indexes concurrently to avoid blocking writes.
+
+```sql
+CREATE INDEX CONCURRENTLY ...
+```
+
+### Querying
+
 Use `EXPLAIN ANALYZE` to debug performance.
 
 ```sql
 EXPLAIN ANALYZE SELECT * FROM items ORDER BY embedding <-> '[3,1,2]' LIMIT 5;
 ```
 
-### Exact Search
+#### Exact Search
 
 To speed up queries without an index, increase `max_parallel_workers_per_gather`.
 
@@ -430,7 +471,7 @@ If vectors are normalized to length 1 (like [OpenAI embeddings](https://platform
 SELECT * FROM items ORDER BY embedding <#> '[3,1,2]' LIMIT 5;
 ```
 
-### Approximate Search
+#### Approximate Search
 
 To speed up queries with an IVFFlat index, increase the number of inverted lists (at the expense of recall).
 
@@ -438,7 +479,7 @@ To speed up queries with an IVFFlat index, increase the number of inverted lists
 CREATE INDEX ON items USING ivfflat (embedding vector_l2_ops) WITH (lists = 1000);
 ```
 
-## Vacuuming
+### Vacuuming
 
 Vacuuming can take a while for HNSW indexes. Speed it up by reindexing first.
 
@@ -447,6 +488,41 @@ REINDEX INDEX CONCURRENTLY index_name;
 VACUUM table_name;
 ```
 
+## Monitoring
+
+Monitor performance with [pg_stat_statements](https://www.postgresql.org/docs/current/pgstatstatements.html) (be sure to add it to `shared_preload_libraries`).
+
+```sql
+CREATE EXTENSION pg_stat_statements;
+```
+
+Get the most time-consuming queries with:
+
+```sql
+SELECT query, calls, ROUND((total_plan_time + total_exec_time) / calls) AS avg_time_ms,
+    ROUND((total_plan_time + total_exec_time) / 60000) AS total_time_min
+    FROM pg_stat_statements ORDER BY total_plan_time + total_exec_time DESC LIMIT 20;
+```
+
+Note: Replace `total_plan_time + total_exec_time` with `total_time` for Postgres < 13
+
+Monitor recall by comparing results from approximate search with exact search.
+
+```sql
+BEGIN;
+SET LOCAL enable_indexscan = off; -- use exact search
+SELECT ...
+COMMIT;
+```
+
+## Scaling
+
+Scale pgvector the same way you scale Postgres.
+
+Scale vertically by increasing memory, CPU, and storage on a single instance. Use existing tools to [tune parameters](#tuning) and [monitor performance](#monitoring).
+
+Scale horizontally with [replicas](https://www.postgresql.org/docs/current/hot-standby.html), or use [Citus](https://github.com/citusdata/citus) or another approach for sharding ([example](https://github.com/pgvector/pgvector-python/blob/master/examples/citus.py)).
+
 ## Languages
 
 Use pgvector from any language with a Postgres client. You can even generate and store vectors in one language and query them in another.
@@ -540,6 +616,18 @@ and query with:
 SELECT * FROM items ORDER BY embedding::vector(3) <-> '[3,1,2]' LIMIT 5;
 ```
 
+#### Are binary vectors supported?
+
+You can store binary vectors and perform exact nearest neighbor search by Hamming distance in Postgres without an extension ([example](https://github.com/pgvector/pgvector-python/blob/master/examples/hash_image_search.py)).
+
+```tsql
+CREATE TABLE items (id bigserial PRIMARY KEY, embedding bit(3));
+INSERT INTO items (embedding) VALUES (B'000'), (B'111');
+SELECT * FROM items ORDER BY bit_count(embedding # B'101') LIMIT 5;
+```
+
+Indexing is not currently supported.
+
 #### Do indexes need to fit into memory?
 
 No, but like other index types, you’ll likely see better performance if they do. You can get the size of an index with:
@@ -552,7 +640,17 @@ SELECT pg_size_pretty(pg_relation_size('index_name'));
 
 #### Why isn’t a query using an index?
 
-The cost estimation in pgvector < 0.4.3 does not always work well with the planner. You can encourage the planner to use an index for a query with:
+The query needs to have an `ORDER BY` and `LIMIT`, and the `ORDER BY` must be the result of a distance operator, not an expression.
+
+```sql
+-- index
+ORDER BY embedding <=> '[3,1,2]' LIMIT 5;
+
+-- no index
+ORDER BY 1 - (embedding <=> '[3,1,2]') DESC LIMIT 5;
+```
+
+You can encourage the planner to use an index for a query with:
 
 ```sql
 BEGIN;
@@ -620,6 +718,7 @@ cosine_distance(vector, vector) → double precision | cosine distance |
 inner_product(vector, vector) → double precision | inner product |
 l2_distance(vector, vector) → double precision | Euclidean distance |
 l1_distance(vector, vector) → double precision | taxicab distance | 0.5.0
+quantize_binary(vector) → bit | quantize | 0.7.0
 vector_dims(vector) → integer | number of dimensions |
 vector_norm(vector) → double precision | Euclidean norm |
 
@@ -630,7 +729,21 @@ Function | Description | Added
 avg(vector) → vector | average |
 sum(vector) → vector | sum | 0.5.0
 
-## Installation Notes
+### Bit Operators
+
+Operator | Description | Added
+--- | --- | ---
+<~> | Hamming distance | 0.7.0
+<%> | Jaccard distance | 0.7.0
+
+### Bit Functions
+
+Function | Description | Added
+--- | --- | ---
+hamming_distance(bit, bit) → double precision | Hamming distance | 0.7.0
+jaccard_distance(bit, bit) → double precision | Jaccard distance | 0.7.0
+
+## Installation Notes - Linux and Mac
 
 ### Postgres Location
 
@@ -672,12 +785,24 @@ If compilation fails and the output includes `warning: no such sysroot directory
 
 ### Portability
 
+By default, pgvector compiles with `-march=native` on some platforms for best performance. However, this can lead to `Illegal instruction` errors if trying to run the compiled extension on a different machine.
+
 To compile for portability, use:
 
 ```sh
 make OPTFLAGS=""
 ```
 
+## Installation Notes - Windows
+
+### Missing Header
+
+If compilation fails with `Cannot open include file: 'postgres.h': No such file or directory`, make sure `PGROOT` is correct.
+
+### Permissions
+
+If installation fails with `Access is denied`, re-run the installation instructions as an administrator.
+
 ## Additional Installation Methods
 
 ### Docker
@@ -693,7 +818,7 @@ This adds pgvector to the [Postgres image](https://hub.docker.com/_/postgres) (r
 You can also build the image manually:
 
 ```sh
-git clone --branch v0.6.1 https://github.com/pgvector/pgvector.git
+git clone --branch v0.6.2 https://github.com/pgvector/pgvector.git
 cd pgvector
 docker build --build-arg PG_MAJOR=16 -t myuser/pgvector .
 ```
@@ -862,6 +987,12 @@ make installcheck REGRESS=functions                            # regression test
 make prove_installcheck PROVE_TESTS=test/t/001_ivfflat_wal.pl  # TAP test
 ```
 
+To enable assertions:
+
+```sh
+make clean && PG_CFLAGS="-DUSE_ASSERT_CHECKING" make && make install
+```
+
 To enable benchmarking:
 
 ```sh
@@ -874,12 +1005,6 @@ To show memory usage:
 make clean && PG_CFLAGS="-DHNSW_MEMORY -DIVFFLAT_MEMORY" make && make install
 ```
 
-To enable assertions:
-
-```sh
-make clean && PG_CFLAGS="-DUSE_ASSERT_CHECKING" make && make install
-```
-
 To get k-means metrics:
 
 ```sh

sql/vector--0.6.1--0.6.2.sql

@@ -0,0 +1,2 @@
+-- complain if script is sourced in psql, rather than via CREATE EXTENSION
+\echo Use "ALTER EXTENSION vector UPDATE TO '0.6.2'" to load this file. \quit

sql/vector--0.6.2--0.7.0.sql

@@ -0,0 +1,31 @@
+-- complain if script is sourced in psql, rather than via CREATE EXTENSION
+\echo Use "ALTER EXTENSION vector UPDATE TO '0.7.0'" to load this file. \quit
+
+CREATE FUNCTION quantize_binary(vector) RETURNS bit
+	AS 'MODULE_PATHNAME' LANGUAGE C IMMUTABLE STRICT PARALLEL SAFE;
+
+CREATE FUNCTION hamming_distance(bit, bit) RETURNS float8
+	AS 'MODULE_PATHNAME' LANGUAGE C IMMUTABLE STRICT PARALLEL SAFE;
+
+CREATE FUNCTION jaccard_distance(bit, bit) RETURNS float8
+	AS 'MODULE_PATHNAME' LANGUAGE C IMMUTABLE STRICT PARALLEL SAFE;
+
+CREATE OPERATOR <~> (
+	LEFTARG = bit, RIGHTARG = bit, PROCEDURE = hamming_distance,
+	COMMUTATOR = '<~>'
+);
+
+CREATE OPERATOR <%> (
+	LEFTARG = bit, RIGHTARG = bit, PROCEDURE = jaccard_distance,
+	COMMUTATOR = '<%>'
+);
+
+CREATE OPERATOR CLASS bit_hamming_ops
+	FOR TYPE bit USING hnsw AS
+	OPERATOR 1 <~> (bit, bit) FOR ORDER BY float_ops,
+	FUNCTION 1 hamming_distance(bit, bit);
+
+CREATE OPERATOR CLASS bit_jaccard_ops
+	FOR TYPE bit USING hnsw AS
+	OPERATOR 1 <%> (bit, bit) FOR ORDER BY float_ops,
+	FUNCTION 1 jaccard_distance(bit, bit);

sql/vector.sql

@@ -58,6 +58,9 @@ CREATE FUNCTION vector_sub(vector, vector) RETURNS vector
 CREATE FUNCTION vector_mul(vector, vector) RETURNS vector
 	AS 'MODULE_PATHNAME' LANGUAGE C IMMUTABLE STRICT PARALLEL SAFE;
 
+CREATE FUNCTION quantize_binary(vector) RETURNS bit
+	AS 'MODULE_PATHNAME' LANGUAGE C IMMUTABLE STRICT PARALLEL SAFE;
+
 -- private functions
 
 CREATE FUNCTION vector_lt(vector, vector) RETURNS bool
@@ -287,3 +290,31 @@ CREATE OPERATOR CLASS vector_cosine_ops
 	OPERATOR 1 <=> (vector, vector) FOR ORDER BY float_ops,
 	FUNCTION 1 vector_negative_inner_product(vector, vector),
 	FUNCTION 2 vector_norm(vector);
+
+-- bit functions
+
+CREATE FUNCTION hamming_distance(bit, bit) RETURNS float8
+	AS 'MODULE_PATHNAME' LANGUAGE C IMMUTABLE STRICT PARALLEL SAFE;
+
+CREATE FUNCTION jaccard_distance(bit, bit) RETURNS float8
+	AS 'MODULE_PATHNAME' LANGUAGE C IMMUTABLE STRICT PARALLEL SAFE;
+
+CREATE OPERATOR <~> (
+	LEFTARG = bit, RIGHTARG = bit, PROCEDURE = hamming_distance,
+	COMMUTATOR = '<~>'
+);
+
+CREATE OPERATOR <%> (
+	LEFTARG = bit, RIGHTARG = bit, PROCEDURE = jaccard_distance,
+	COMMUTATOR = '<%>'
+);
+
+CREATE OPERATOR CLASS bit_hamming_ops
+	FOR TYPE bit USING hnsw AS
+	OPERATOR 1 <~> (bit, bit) FOR ORDER BY float_ops,
+	FUNCTION 1 hamming_distance(bit, bit);
+
+CREATE OPERATOR CLASS bit_jaccard_ops
+	FOR TYPE bit USING hnsw AS
+	OPERATOR 1 <%> (bit, bit) FOR ORDER BY float_ops,
+	FUNCTION 1 jaccard_distance(bit, bit);

src/bitvector.c

@@ -0,0 +1,90 @@
+#include "postgres.h"
+
+#include "bitvector.h"
+#include "port/pg_bitutils.h"
+#include "utils/varbit.h"
+
+#if PG_VERSION_NUM >= 160000
+#include "varatt.h"
+#endif
+
+/*
+ * Allocate and initialize a new bit vector
+ */
+VarBit *
+InitBitVector(int dim)
+{
+	VarBit	   *result;
+	int			size;
+
+	size = VARBITTOTALLEN(dim);
+	result = (VarBit *) palloc0(size);
+	SET_VARSIZE(result, size);
+	VARBITLEN(result) = dim;
+
+	return result;
+}
+
+/*
+ * Ensure same number of bits
+ */
+static inline void
+CheckBitLengths(uint32 aLen, uint32 bLen)
+{
+	if (aLen != bLen)
+		ereport(ERROR,
+				(errcode(ERRCODE_DATA_EXCEPTION),
+				 errmsg("different bit lengths %u and %u", aLen, bLen)));
+}
+
+/*
+ * Get the Hamming distance between two bit strings
+ */
+PGDLLEXPORT PG_FUNCTION_INFO_V1(hamming_distance);
+Datum
+hamming_distance(PG_FUNCTION_ARGS)
+{
+	VarBit	   *a = PG_GETARG_VARBIT_P(0);
+	VarBit	   *b = PG_GETARG_VARBIT_P(1);
+	unsigned char *ax = VARBITS(a);
+	unsigned char *bx = VARBITS(b);
+	uint64		distance = 0;
+
+	CheckBitLengths(VARBITLEN(a), VARBITLEN(b));
+
+	/* TODO Improve performance */
+	for (uint32 i = 0; i < VARBITBYTES(a); i++)
+		distance += pg_number_of_ones[ax[i] ^ bx[i]];
+
+	PG_RETURN_FLOAT8((double) distance);
+}
+
+/*
+ * Get the Jaccard distance between two bit strings
+ */
+PGDLLEXPORT PG_FUNCTION_INFO_V1(jaccard_distance);
+Datum
+jaccard_distance(PG_FUNCTION_ARGS)
+{
+	VarBit	   *a = PG_GETARG_VARBIT_P(0);
+	VarBit	   *b = PG_GETARG_VARBIT_P(1);
+	unsigned char *ax = VARBITS(a);
+	unsigned char *bx = VARBITS(b);
+	uint64		ab = 0;
+	uint64		aa;
+	uint64		bb;
+
+	CheckBitLengths(VARBITLEN(a), VARBITLEN(b));
+
+	/* TODO Improve performance */
+	for (uint32 i = 0; i < VARBITBYTES(a); i++)
+		ab += pg_number_of_ones[ax[i] & bx[i]];
+
+	if (ab == 0)
+		PG_RETURN_FLOAT8(1);
+
+	aa = pg_popcount((char *) ax, VARBITBYTES(a));
+	bb = pg_popcount((char *) bx, VARBITBYTES(b));
+
+	PG_RETURN_FLOAT8(1 - (ab / ((double) (aa + bb - ab))));
+}

src/bitvector.h

@@ -0,0 +1,8 @@
+#ifndef BITVECTOR_H
+#define BITVECTOR_H
+
+#include "utils/varbit.h"
+
+VarBit	   *InitBitVector(int dim);
+
+#endif

src/hnsw.h

@@ -129,7 +129,7 @@ HnswPtrDeclare(HnswNeighborArray, HnswNeighborArrayRelptr, HnswNeighborArrayPtr)
 HnswPtrDeclare(HnswNeighborArrayPtr, HnswNeighborsRelptr, HnswNeighborsPtr);
 HnswPtrDeclare(char, DatumRelptr, DatumPtr);
 
-typedef struct HnswElementData
+struct HnswElementData
 {
 	HnswElementPtr next;
 	ItemPointerData heaptids[HNSW_HEAPTIDS];
@@ -144,7 +144,7 @@ typedef struct HnswElementData
 	BlockNumber neighborPage;
 	DatumPtr	value;
 	LWLock		lock;
-}			HnswElementData;
+};
 
 typedef HnswElementData * HnswElement;
 
@@ -155,12 +155,12 @@ typedef struct HnswCandidate
 	bool		closer;
 }			HnswCandidate;
 
-typedef struct HnswNeighborArray
+struct HnswNeighborArray
 {
 	int			length;
 	bool		closerSet;
 	HnswCandidate items[FLEXIBLE_ARRAY_MEMBER];
-}			HnswNeighborArray;
+};
 
 typedef struct HnswPairingHeapNode
 {

src/hnswbuild.c

@@ -44,6 +44,7 @@
 #include "access/xact.h"
 #include "access/xloginsert.h"
 #include "catalog/index.h"
+#include "catalog/pg_type_d.h"
 #include "commands/progress.h"
 #include "hnsw.h"
 #include "miscadmin.h"
@@ -436,9 +437,9 @@ InsertTupleInMemory(HnswBuildState * buildstate, HnswElement element)
 	int			m = buildstate->m;
 	char	   *base = buildstate->hnswarea;
 
-	/* Wait if another process needs exclusive lock */
-	if (LWLockAcquireOrWait(entryWaitLock, LW_EXCLUSIVE))
-		LWLockRelease(entryWaitLock);
+	/* Wait if another process needs exclusive lock on entry lock */
+	LWLockAcquire(entryWaitLock, LW_EXCLUSIVE);
+	LWLockRelease(entryWaitLock);
 
 	/* Get entry point */
 	LWLockAcquire(entryLock, LW_SHARED);
@@ -450,7 +451,7 @@ InsertTupleInMemory(HnswBuildState * buildstate, HnswElement element)
 		/* Release shared lock */
 		LWLockRelease(entryLock);
 
-		/* Get exclusive lock */
+		/* Tell other processes to wait and get exclusive lock */
 		LWLockAcquire(entryWaitLock, LW_EXCLUSIVE);
 		LWLockAcquire(entryLock, LW_EXCLUSIVE);
 		LWLockRelease(entryWaitLock);
@@ -671,6 +672,12 @@ HnswSharedMemoryAlloc(Size size, void *state)
 static void
 InitBuildState(HnswBuildState * buildstate, Relation heap, Relation index, IndexInfo *indexInfo, ForkNumber forkNum)
 {
+	int			maxDimensions = HNSW_MAX_DIM;
+	Oid			typid = TupleDescAttr(index->rd_att, 0)->atttypid;
+
+	if (typid == BITOID || typid == VARBITOID)
+		maxDimensions *= 32;
+
 	buildstate->heap = heap;
 	buildstate->index = index;
 	buildstate->indexInfo = indexInfo;
@@ -684,8 +691,8 @@ InitBuildState(HnswBuildState * buildstate, Relation heap, Relation index, Index
 	if (buildstate->dimensions < 0)
 		elog(ERROR, "column does not have dimensions");
 
-	if (buildstate->dimensions > HNSW_MAX_DIM)
-		elog(ERROR, "column cannot have more than %d dimensions for hnsw index", HNSW_MAX_DIM);
+	if (buildstate->dimensions > maxDimensions)
+		elog(ERROR, "column cannot have more than %d dimensions for hnsw index", maxDimensions);
 
 	if (buildstate->efConstruction < 2 * buildstate->m)
 		elog(ERROR, "ef_construction must be greater than or equal to 2 * m");

src/hnswscan.c

@@ -1,6 +1,8 @@
 #include "postgres.h"
 
 #include "access/relscan.h"
+#include "bitvector.h"
+#include "catalog/pg_type_d.h"
 #include "hnsw.h"
 #include "pgstat.h"
 #include "storage/bufmgr.h"
@@ -73,7 +75,15 @@ GetScanValue(IndexScanDesc scan)
 	Datum		value;
 
 	if (scan->orderByData->sk_flags & SK_ISNULL)
-		value = PointerGetDatum(InitVector(GetDimensions(scan->indexRelation)));
+	{
+		Oid			typid = TupleDescAttr(scan->indexRelation->rd_att, 0)->atttypid;
+		int			dimensions = GetDimensions(scan->indexRelation);
+
+		if (typid == BITOID || typid == VARBITOID)
+			value = PointerGetDatum(InitBitVector(dimensions));
+		else
+			value = PointerGetDatum(InitVector(dimensions));
+	}
 	else
 	{
 		value = scan->orderByData->sk_argument;

src/vector.c

@@ -2,6 +2,7 @@
 
 #include <math.h>
 
+#include "bitvector.h"
 #include "catalog/pg_type.h"
 #include "common/shortest_dec.h"
 #include "fmgr.h"
@@ -860,6 +861,26 @@ vector_mul(PG_FUNCTION_ARGS)
 	PG_RETURN_POINTER(result);
 }
 
+/*
+ * Quantize a vector
+ */
+PGDLLEXPORT PG_FUNCTION_INFO_V1(quantize_binary);
+Datum
+quantize_binary(PG_FUNCTION_ARGS)
+{
+	Vector	   *a = PG_GETARG_VECTOR_P(0);
+	float	   *ax = a->x;
+	VarBit	   *result = InitBitVector(a->dim);
+	unsigned char *rx = VARBITS(result);
+
+	/* TODO Improve */
+	for (int i = 0; i < a->dim; i++)
+		rx[i / 8] |= (ax[i] > 0) << (7 - (i % 8));
+
+	PG_RETURN_VARBIT_P(result);
+}
+
+
 /*
  * Internal helper to compare vectors
  */

test/expected/bit_functions.out

@@ -0,0 +1,64 @@
+SELECT hamming_distance(B'111', B'111');
+ hamming_distance 
+------------------
+                0
+(1 row)
+
+SELECT hamming_distance(B'111', B'110');
+ hamming_distance 
+------------------
+                1
+(1 row)
+
+SELECT hamming_distance(B'111', B'100');
+ hamming_distance 
+------------------
+                2
+(1 row)
+
+SELECT hamming_distance(B'111', B'000');
+ hamming_distance 
+------------------
+                3
+(1 row)
+
+SELECT hamming_distance(B'111', B'00');
+ERROR:  different bit lengths 3 and 2
+SELECT jaccard_distance(B'1111', B'1111');
+ jaccard_distance 
+------------------
+                0
+(1 row)
+
+SELECT jaccard_distance(B'1111', B'1110');
+ jaccard_distance 
+------------------
+             0.25
+(1 row)
+
+SELECT jaccard_distance(B'1111', B'1100');
+ jaccard_distance 
+------------------
+              0.5
+(1 row)
+
+SELECT jaccard_distance(B'1111', B'1000');
+ jaccard_distance 
+------------------
+             0.75
+(1 row)
+
+SELECT jaccard_distance(B'1111', B'0000');
+ jaccard_distance 
+------------------
+                1
+(1 row)
+
+SELECT jaccard_distance(B'1100', B'1000');
+ jaccard_distance 
+------------------
+              0.5
+(1 row)
+
+SELECT jaccard_distance(B'1111', B'000');
+ERROR:  different bit lengths 4 and 3

test/expected/functions.out

@@ -208,6 +208,18 @@ SELECT l1_distance('[3e38]', '[-3e38]');
     Infinity
 (1 row)
 
+SELECT quantize_binary('[1,0,-1]');
+ quantize_binary 
+-----------------
+ 100
+(1 row)
+
+SELECT quantize_binary('[0,0.1,-0.2,-0.3,0.4,0.5,0.6,-0.7,0.8,-0.9,1]');
+ quantize_binary 
+-----------------
+ 01001110101
+(1 row)
+
 SELECT avg(v) FROM unnest(ARRAY['[1,2,3]'::vector, '[3,5,7]']) v;
     avg    
 -----------

test/expected/hnsw_hamming.out

@@ -0,0 +1,21 @@
+SET enable_seqscan = off;
+CREATE TABLE t (val bit(3));
+INSERT INTO t (val) VALUES (B'000'), (B'100'), (B'111'), (NULL);
+CREATE INDEX ON t USING hnsw (val bit_hamming_ops);
+INSERT INTO t (val) VALUES (B'110');
+SELECT * FROM t ORDER BY val <~> B'111';
+ val 
+-----
+ 111
+ 110
+ 100
+ 000
+(4 rows)
+
+SELECT COUNT(*) FROM (SELECT * FROM t ORDER BY val <~> (SELECT NULL::bit)) t2;
+ count 
+-------
+     4
+(1 row)
+
+DROP TABLE t;

test/expected/hnsw_jaccard.out

@@ -0,0 +1,21 @@
+SET enable_seqscan = off;
+CREATE TABLE t (val bit(4));
+INSERT INTO t (val) VALUES (B'0000'), (B'1100'), (B'1111'), (NULL);
+CREATE INDEX ON t USING hnsw (val bit_jaccard_ops);
+INSERT INTO t (val) VALUES (B'1110');
+SELECT * FROM t ORDER BY val <%> B'1111';
+ val  
+------
+ 1111
+ 1110
+ 1100
+ 0000
+(4 rows)
+
+SELECT COUNT(*) FROM (SELECT * FROM t ORDER BY val <%> (SELECT NULL::bit)) t2;
+ count 
+-------
+     4
+(1 row)
+
+DROP TABLE t;

test/expected/input.out

@@ -116,8 +116,30 @@ SELECT '[1, ,3]'::vector;
 ERROR:  invalid input syntax for type vector: "[1, ,3]"
 LINE 1: SELECT '[1, ,3]'::vector;
                ^
+SELECT '[1,2,3]'::vector(3);
+ vector  
+---------
+ [1,2,3]
+(1 row)
+
 SELECT '[1,2,3]'::vector(2);
 ERROR:  expected 2 dimensions, not 3
+SELECT '[1,2,3]'::vector(3, 2);
+ERROR:  invalid type modifier
+LINE 1: SELECT '[1,2,3]'::vector(3, 2);
+                          ^
+SELECT '[1,2,3]'::vector('a');
+ERROR:  invalid input syntax for type integer: "a"
+LINE 1: SELECT '[1,2,3]'::vector('a');
+                          ^
+SELECT '[1,2,3]'::vector(0);
+ERROR:  dimensions for type vector must be at least 1
+LINE 1: SELECT '[1,2,3]'::vector(0);
+                          ^
+SELECT '[1,2,3]'::vector(16001);
+ERROR:  dimensions for type vector cannot exceed 16000
+LINE 1: SELECT '[1,2,3]'::vector(16001);
+                          ^
 SELECT unnest('{"[1,2,3]", "[4,5,6]"}'::vector[]);
  unnest  
 ---------

test/sql/bit_functions.sql

@@ -0,0 +1,13 @@
+SELECT hamming_distance(B'111', B'111');
+SELECT hamming_distance(B'111', B'110');
+SELECT hamming_distance(B'111', B'100');
+SELECT hamming_distance(B'111', B'000');
+SELECT hamming_distance(B'111', B'00');
+
+SELECT jaccard_distance(B'1111', B'1111');
+SELECT jaccard_distance(B'1111', B'1110');
+SELECT jaccard_distance(B'1111', B'1100');
+SELECT jaccard_distance(B'1111', B'1000');
+SELECT jaccard_distance(B'1111', B'0000');
+SELECT jaccard_distance(B'1100', B'1000');
+SELECT jaccard_distance(B'1111', B'000');

test/sql/functions.sql

@@ -48,6 +48,9 @@ SELECT l1_distance('[0,0]', '[0,1]');
 SELECT l1_distance('[1,2]', '[3]');
 SELECT l1_distance('[3e38]', '[-3e38]');
 
+SELECT quantize_binary('[1,0,-1]');
+SELECT quantize_binary('[0,0.1,-0.2,-0.3,0.4,0.5,0.6,-0.7,0.8,-0.9,1]');
+
 SELECT avg(v) FROM unnest(ARRAY['[1,2,3]'::vector, '[3,5,7]']) v;
 SELECT avg(v) FROM unnest(ARRAY['[1,2,3]'::vector, '[3,5,7]', NULL]) v;
 SELECT avg(v) FROM unnest(ARRAY[]::vector[]) v;

test/sql/hnsw_hamming.sql

@@ -0,0 +1,12 @@
+SET enable_seqscan = off;
+
+CREATE TABLE t (val bit(3));
+INSERT INTO t (val) VALUES (B'000'), (B'100'), (B'111'), (NULL);
+CREATE INDEX ON t USING hnsw (val bit_hamming_ops);
+
+INSERT INTO t (val) VALUES (B'110');
+
+SELECT * FROM t ORDER BY val <~> B'111';
+SELECT COUNT(*) FROM (SELECT * FROM t ORDER BY val <~> (SELECT NULL::bit)) t2;
+
+DROP TABLE t;

test/sql/hnsw_jaccard.sql

@@ -0,0 +1,12 @@
+SET enable_seqscan = off;
+
+CREATE TABLE t (val bit(4));
+INSERT INTO t (val) VALUES (B'0000'), (B'1100'), (B'1111'), (NULL);
+CREATE INDEX ON t USING hnsw (val bit_jaccard_ops);
+
+INSERT INTO t (val) VALUES (B'1110');
+
+SELECT * FROM t ORDER BY val <%> B'1111';
+SELECT COUNT(*) FROM (SELECT * FROM t ORDER BY val <%> (SELECT NULL::bit)) t2;
+
+DROP TABLE t;

test/sql/input.sql

@@ -22,7 +22,13 @@ SELECT '[1,]'::vector;
 SELECT '[1a]'::vector;
 SELECT '[1,,3]'::vector;
 SELECT '[1, ,3]'::vector;
+
+SELECT '[1,2,3]'::vector(3);
 SELECT '[1,2,3]'::vector(2);
+SELECT '[1,2,3]'::vector(3, 2);
+SELECT '[1,2,3]'::vector('a');
+SELECT '[1,2,3]'::vector(0);
+SELECT '[1,2,3]'::vector(16001);
 
 SELECT unnest('{"[1,2,3]", "[4,5,6]"}'::vector[]);
 SELECT '{"[1,2,3]"}'::vector(2)[];

test/t/020_hnsw_bit_build_recall.pl

@@ -0,0 +1,137 @@
+use strict;
+use warnings;
+use PostgresNode;
+use TestLib;
+use Test::More;
+
+my $node;
+my @queries = ();
+my @expected;
+my $limit = 20;
+my $dim = 52;
+my $max = 2**$dim;
+
+sub test_recall
+{
+	my ($min, $operator) = @_;
+	my $correct = 0;
+	my $total = 0;
+
+	my $explain = $node->safe_psql("postgres", qq(
+		SET enable_seqscan = off;
+		SET hnsw.ef_search = 100;
+		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;
+			SET hnsw.ef_search = 100;
+			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 = get_new_node('node');
+$node->init;
+$node->start;
+
+# Create table
+$node->safe_psql("postgres", "CREATE EXTENSION vector;");
+$node->safe_psql("postgres", "CREATE TABLE tst (i int4, v bit($dim));");
+$node->safe_psql("postgres",
+	"INSERT INTO tst SELECT i, (random() * $max)::bigint::bit($dim) FROM generate_series(1, 10000) i;"
+);
+
+# Generate queries
+for (1 .. 20)
+{
+	my $r = int(rand() * $max);
+	push(@queries, "${r}::bigint::bit($dim)");
+}
+
+# Check each index type
+my @operators = ("<~>", "<\%>");
+my @opclasses = ("bit_hamming_ops", "bit_jaccard_ops");
+
+for my $i (0 .. $#operators)
+{
+	my $operator = $operators[$i];
+	my $opclass = $opclasses[$i];
+
+	# Get exact results
+	@expected = ();
+	foreach (@queries)
+	{
+		# Handle ties
+		my $res = $node->safe_psql("postgres", qq(
+			WITH top AS (
+				SELECT v $operator $_ AS distance FROM tst ORDER BY v $operator $_ 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 hnsw (v $opclass);
+	));
+
+	# Test approximate results
+	my $min = $operator eq "<\%>" ? 0.96 : 0.99;
+	test_recall($min, $operator);
+
+	$node->safe_psql("postgres", "DROP INDEX idx;");
+
+	# Build index in parallel in memory
+	my ($ret, $stdout, $stderr) = $node->psql("postgres", qq(
+		SET client_min_messages = DEBUG;
+		SET min_parallel_table_scan_size = 1;
+		CREATE INDEX idx ON tst USING hnsw (v $opclass);
+	));
+	is($ret, 0, $stderr);
+	like($stderr, qr/using \d+ parallel workers/);
+
+	# Test approximate results
+	test_recall($min, $operator);
+
+	$node->safe_psql("postgres", "DROP INDEX idx;");
+
+	# Build index in parallel on disk
+	# Set parallel_workers on table to use workers with low maintenance_work_mem
+	($ret, $stdout, $stderr) = $node->psql("postgres", qq(
+		ALTER TABLE tst SET (parallel_workers = 2);
+		SET client_min_messages = DEBUG;
+		SET maintenance_work_mem = '4MB';
+		CREATE INDEX idx ON tst USING hnsw (v $opclass);
+		ALTER TABLE tst RESET (parallel_workers);
+	));
+	is($ret, 0, $stderr);
+	like($stderr, qr/using \d+ parallel workers/);
+	like($stderr, qr/hnsw graph no longer fits into maintenance_work_mem/);
+
+	$node->safe_psql("postgres", "DROP INDEX idx;");
+}
+
+done_testing();

vector.control

@@ -1,4 +1,4 @@
 comment = 'vector data type and ivfflat and hnsw access methods'
-default_version = '0.6.1'
+default_version = '0.6.2'
 module_pathname = '$libdir/vector'
 relocatable = true