Version bump to 0.4.4 [skip ci]

The compiler was complaining about a missing include due to the
addition of NaN/inf checks in 482a5f8b. Adding the include
silences the warnings.

.github/workflows/build.yml

@@ -8,6 +8,8 @@ jobs:
       fail-fast: false
       matrix:
         include:
+          - postgres: 16
+            os: ubuntu-22.04
           - postgres: 15
             os: ubuntu-22.04
           - postgres: 14
@@ -25,6 +27,8 @@ jobs:
           postgres-version: ${{ matrix.postgres }}
           dev-files: true
       - run: make
+        env:
+          PG_CFLAGS: -Wall -Wextra -Werror -Wno-unused-parameter -Wno-sign-compare
       - run: |
           export PG_CONFIG=`which pg_config`
           sudo --preserve-env=PG_CONFIG make install
@@ -44,6 +48,8 @@ jobs:
         with:
           postgres-version: 14
       - run: make
+        env:
+          PG_CFLAGS: -Wall -Wextra -Werror -Wno-unused-parameter
       - run: make install
       - run: make installcheck
       - if: ${{ failure() }}
@@ -54,6 +60,7 @@ jobs:
           wget -q https://github.com/postgres/postgres/archive/refs/tags/REL_14_5.tar.gz
           tar xf REL_14_5.tar.gz
       - run: make prove_installcheck PROVE_FLAGS="-I ./postgres-REL_14_5/src/test/perl" PERL5LIB="/Users/runner/perl5/lib/perl5"
+      - run: make clean && /usr/local/opt/llvm@15/bin/scan-build --status-bugs make
   windows:
     runs-on: windows-latest
     if: ${{ !startsWith(github.ref_name, 'mac') }}
@@ -70,3 +77,22 @@ jobs:
           nmake /NOLOGO /F Makefile.win clean && ^
           nmake /NOLOGO /F Makefile.win uninstall
         shell: cmd
+  i386:
+    runs-on: ubuntu-latest
+    container:
+      image: debian:11
+      options: --platform linux/386
+    steps:
+      - run: apt-get update && apt-get install -y build-essential git libipc-run-perl postgresql-13 postgresql-server-dev-13 sudo
+      - run: service postgresql start
+      - run: |
+          git clone https://github.com/${{ github.repository }}.git pgvector
+          cd pgvector
+          git checkout ${{ github.ref }}
+          make
+          make install
+          chown -R postgres .
+          sudo -u postgres make installcheck
+          sudo -u postgres make prove_installcheck
+        env:
+          PG_CFLAGS: -Wall -Wextra -Werror -Wno-unused-parameter -Wno-sign-compare

CHANGELOG.md

@@ -1,3 +1,20 @@
+## 0.4.4 (2023-06-12)
+
+- Improved error message for malformed vector literal
+- Fixed segmentation fault with text input
+- Fixed consecutive delimiters with text input
+
+## 0.4.3 (2023-06-10)
+
+- Improved cost estimation
+- Improved support for spaces with text input
+- Fixed infinite and NaN values with binary input
+- Fixed infinite values with vector addition and subtraction
+- Fixed infinite values with list centers
+- Fixed compilation error when `float8` is pass by reference
+- Fixed compilation error on PowerPC
+- Fixed segmentation fault with index creation on i386
+
 ## 0.4.2 (2023-05-13)
 
 - Added notice when index created with little data

LICENSE

@@ -1,4 +1,4 @@
-Portions Copyright (c) 1996-2022, PostgreSQL Global Development Group
+Portions Copyright (c) 1996-2023, PostgreSQL Global Development Group
 
 Portions Copyright (c) 1994, The Regents of the University of California
 

META.json

@@ -2,7 +2,7 @@
 	"name": "vector",
 	"abstract": "Open-source vector similarity search for Postgres",
 	"description": "Supports L2 distance, inner product, and cosine distance",
-	"version": "0.4.2",
+	"version": "0.4.4",
 	"maintainer": [
 		"Andrew Kane <andrew@ankane.org>"
 	],
@@ -20,7 +20,7 @@
 		"vector": {
 			"file": "sql/vector.sql",
 			"docfile": "README.md",
-			"version": "0.4.2",
+			"version": "0.4.4",
 			"abstract": "Open-source vector similarity search for Postgres"
 		}
 	},

Makefile

@@ -1,5 +1,5 @@
 EXTENSION = vector
-EXTVERSION = 0.4.2
+EXTVERSION = 0.4.4
 
 MODULE_big = vector
 DATA = $(wildcard sql/*--*.sql)
@@ -19,6 +19,11 @@ ifeq ($(shell uname -s), Darwin)
 	endif
 endif
 
+# PowerPC doesn't support -march=native
+ifneq ($(filter ppc64%, $(shell uname -m)), )
+	OPTFLAGS =
+endif
+
 # For auto-vectorization:
 # - GCC (needs -ftree-vectorize OR -O3) - https://gcc.gnu.org/projects/tree-ssa/vectorization.html
 # - Clang (could use pragma instead) - https://llvm.org/docs/Vectorizers.html

Makefile.win

@@ -1,5 +1,5 @@
 EXTENSION = vector
-EXTVERSION = 0.4.2
+EXTVERSION = 0.4.4
 
 OBJS = src\ivfbuild.obj src\ivfflat.obj src\ivfinsert.obj src\ivfkmeans.obj src\ivfscan.obj src\ivfutils.obj src\ivfvacuum.obj src\vector.obj
 

README.md

@@ -8,6 +8,8 @@ Supports
 - L2 distance, inner product, and cosine distance
 - 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
+
 [![Build Status](https://github.com/pgvector/pgvector/workflows/build/badge.svg?branch=master)](https://github.com/pgvector/pgvector/actions)
 
 ## Installation
@@ -16,24 +18,24 @@ Compile and install the extension (supports Postgres 11+)
 
 ```sh
 cd /tmp
-git clone --branch v0.4.2 https://github.com/pgvector/pgvector.git
+git clone --branch v0.4.4 https://github.com/pgvector/pgvector.git
 cd pgvector
 make
 make install # may need sudo
 ```
 
-Then load it in databases where you want to use it
-
-```sql
-CREATE EXTENSION vector;
-```
-
 See the [installation notes](#installation-notes) if you run into issues
 
-You can also install it with [Docker](#docker), [Homebrew](#homebrew), [PGXN](#pgxn), [Yum](#yum), or [conda-forge](#conda-forge)
+You can also install it with [Docker](#docker), [Homebrew](#homebrew), [PGXN](#pgxn), [APT](#apt), [Yum](#yum), or [conda-forge](#conda-forge), and it comes preinstalled with [Postgres.app](#postgresapp) and many [hosted providers](#hosted-postgres)
 
 ## Getting Started
 
+Enable the extension (do this once in each database where you want to use it)
+
+```tsql
+CREATE EXTENSION vector;
+```
+
 Create a vector column with 3 dimensions
 
 ```sql
@@ -127,7 +129,7 @@ SELECT embedding <-> '[3,1,2]' AS distance FROM items;
 
 For inner product, multiply by -1 (since `<#>` returns the negative inner product)
 
-```sql
+```tsql
 SELECT (embedding <#> '[3,1,2]') * -1 AS inner_product FROM items;
 ```
 
@@ -161,7 +163,7 @@ Three keys to achieving good recall are:
 
 1. Create the index *after* the table has some data
 2. Choose an appropriate number of lists - a good place to start is `rows / 1000` for up to 1M rows and `sqrt(rows)` for over 1M rows
-3. When querying, specify an appropriate number of [probes](#query-options) (higher is better for recall, lower is better for speed) - a good place to start is `lists / 10` for up to 1M rows and `sqrt(lists)` for over 1M rows
+3. When querying, specify an appropriate number of [probes](#query-options) (higher is better for recall, lower is better for speed) - a good place to start is `sqrt(lists)`
 
 Add an index for each distance function you want to use.
 
@@ -221,26 +223,42 @@ The phases are:
 
 Note: `tuples_done` and `tuples_total` are only populated during the `loading tuples` phase
 
-### Partial Indexes
+### Filtering
 
-Consider [partial indexes](https://www.postgresql.org/docs/current/indexes-partial.html) for queries with a `WHERE` clause
+There are a few ways to index nearest neighbor queries with a `WHERE` clause
 
 ```sql
 SELECT * FROM items WHERE category_id = 123 ORDER BY embedding <-> '[3,1,2]' LIMIT 5;
 ```
 
-can be indexed with:
+Create an index on one [or more](https://www.postgresql.org/docs/current/indexes-multicolumn.html) of the `WHERE` columns for exact search
 
 ```sql
-CREATE INDEX ON items USING ivfflat (embedding vector_l2_ops) WITH (lists = 100) WHERE (category_id = 123);
+CREATE INDEX ON items (category_id);
 ```
 
-To index many different values of `category_id`, consider [partitioning](https://www.postgresql.org/docs/current/ddl-partitioning.html) on `category_id`.
+Or a [partial index](https://www.postgresql.org/docs/current/indexes-partial.html) on the vector column for approximate search
+
+```sql
+CREATE INDEX ON items USING ivfflat (embedding vector_l2_ops) WITH (lists = 100)
+    WHERE (category_id = 123);
+```
+
+Use [partitioning](https://www.postgresql.org/docs/current/ddl-partitioning.html) for approximate search on many different values of the `WHERE` columns
 
 ```sql
 CREATE TABLE items (embedding vector(3), category_id int) PARTITION BY LIST(category_id);
 ```
 
+## Hybrid Search
+
+Use together with Postgres [full-text search](https://www.postgresql.org/docs/current/textsearch-intro.html) for hybrid search ([Python example](https://github.com/pgvector/pgvector-python/blob/master/examples/hybrid_search.py)).
+
+```sql
+SELECT id, content FROM items, to_tsquery('hello & search') query
+    WHERE textsearch @@ query ORDER BY ts_rank_cd(textsearch, query) DESC LIMIT 5;
+```
+
 ## Performance
 
 Use `EXPLAIN ANALYZE` to debug performance.
@@ -259,7 +277,7 @@ SET max_parallel_workers_per_gather = 4;
 
 If vectors are normalized to length 1 (like [OpenAI embeddings](https://platform.openai.com/docs/guides/embeddings/which-distance-function-should-i-use)), use inner product for best performance.
 
-```sql
+```tsql
 SELECT * FROM items ORDER BY embedding <#> '[3,1,2]' LIMIT 5;
 ```
 
@@ -368,7 +386,7 @@ If compilation fails with `fatal error: postgres.h: No such file or directory`,
 For Ubuntu and Debian, use:
 
 ```sh
-sudo apt-get install postgresql-server-dev-15
+sudo apt install postgresql-server-dev-15
 ```
 
 Note: Replace `15` with your Postgres server version
@@ -379,7 +397,7 @@ Support for Windows is currently experimental. Use `nmake` to build:
 
 ```cmd
 set "PGROOT=C:\Program Files\PostgreSQL\15"
-git clone --branch v0.4.2 https://github.com/pgvector/pgvector.git
+git clone --branch v0.4.4 https://github.com/pgvector/pgvector.git
 cd pgvector
 nmake /F Makefile.win
 nmake /F Makefile.win install
@@ -400,9 +418,9 @@ This adds pgvector to the [Postgres image](https://hub.docker.com/_/postgres) (r
 You can also build the image manually:
 
 ```sh
-git clone --branch v0.4.2 https://github.com/pgvector/pgvector.git
+git clone --branch v0.4.4 https://github.com/pgvector/pgvector.git
 cd pgvector
-docker build -t pgvector .
+docker build --build-arg PG_MAJOR=15 -t myuser/pgvector .
 ```
 
 ### Homebrew
@@ -423,6 +441,16 @@ Install from the [PostgreSQL Extension Network](https://pgxn.org/dist/vector) wi
 pgxn install vector
 ```
 
+### APT
+
+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-15-pgvector
+```
+
+Note: Replace `15` 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:
@@ -445,6 +473,10 @@ conda install -c conda-forge pgvector
 
 This method is [community-maintained](https://github.com/conda-forge/pgvector-feedstock) by [@mmcauliffe](https://github.com/mmcauliffe)
 
+### Postgres.app
+
+Download the [latest release](https://postgresapp.com/downloads.html) with Postgres 15+.
+
 ## Hosted Postgres
 
 pgvector is available on [these providers](https://github.com/pgvector/pgvector/issues/54).
@@ -452,7 +484,6 @@ pgvector is available on [these providers](https://github.com/pgvector/pgvector/
 To request a new extension on other providers:
 
 - Google Cloud SQL - vote or comment on [this page](https://issuetracker.google.com/issues/265172065)
-- Azure Database - vote or comment on [this page](https://feedback.azure.com/d365community/idea/7b423322-6189-ed11-a81b-000d3ae49307)
 - DigitalOcean Managed Databases - vote or comment on [this page](https://ideas.digitalocean.com/managed-database/p/pgvector-extension-for-postgresql)
 - Heroku Postgres - vote or comment on [this page](https://github.com/heroku/roadmap/issues/156)
 

sql/vector--0.4.2--0.4.3.sql

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

sql/vector--0.4.3--0.4.4.sql

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

src/ivfbuild.c

@@ -147,7 +147,7 @@ AddTupleToSort(Relation index, ItemPointer tid, Datum *values, IvfflatBuildState
 {
 	double		distance;
 	double		minDistance = DBL_MAX;
-	int			closestCenter = -1;
+	int			closestCenter = 0;
 	VectorArray centers = buildstate->centers;
 	TupleTableSlot *slot = buildstate->slot;
 	int			i;
@@ -568,6 +568,21 @@ PrintKmeansMetrics(IvfflatBuildState * buildstate)
 }
 #endif
 
+/*
+ * Scan table for tuples to index
+ */
+static void
+ScanTable(IvfflatBuildState * buildstate)
+{
+#if PG_VERSION_NUM >= 120000
+	buildstate->reltuples = table_index_build_scan(buildstate->heap, buildstate->index, buildstate->indexInfo,
+												   true, true, BuildCallback, (void *) buildstate, NULL);
+#else
+	buildstate->reltuples = IndexBuildHeapScan(buildstate->heap, buildstate->index, buildstate->indexInfo,
+											   true, BuildCallback, (void *) buildstate, NULL);
+#endif
+}
+
 /*
  * Create entry pages
  */
@@ -575,7 +590,7 @@ static void
 CreateEntryPages(IvfflatBuildState * buildstate, ForkNumber forkNum)
 {
 	AttrNumber	attNums[] = {1};
-	Oid			sortOperators[] = {Float8LessOperator};
+	Oid			sortOperators[] = {Int4LessOperator};
 	Oid			sortCollations[] = {InvalidOid};
 	bool		nullsFirstFlags[] = {false};
 
@@ -585,25 +600,17 @@ CreateEntryPages(IvfflatBuildState * buildstate, ForkNumber forkNum)
 
 	/* Add tuples to sort */
 	if (buildstate->heap != NULL)
-	{
-#if PG_VERSION_NUM >= 120000
-		buildstate->reltuples = table_index_build_scan(buildstate->heap, buildstate->index, buildstate->indexInfo,
-													   true, true, BuildCallback, (void *) buildstate, NULL);
-#else
-		buildstate->reltuples = IndexBuildHeapScan(buildstate->heap, buildstate->index, buildstate->indexInfo,
-												   true, BuildCallback, (void *) buildstate, NULL);
-#endif
-	}
+		IvfflatBench("assign tuples", ScanTable(buildstate));
 
 	/* Sort */
-	tuplesort_performsort(buildstate->sortstate);
+	IvfflatBench("sort tuples", tuplesort_performsort(buildstate->sortstate));
 
 #ifdef IVFFLAT_KMEANS_DEBUG
 	PrintKmeansMetrics(buildstate);
 #endif
 
 	/* Insert */
-	InsertTuples(buildstate->index, buildstate, forkNum);
+	IvfflatBench("load tuples", InsertTuples(buildstate->index, buildstate, forkNum));
 	tuplesort_end(buildstate->sortstate);
 }
 
@@ -621,7 +628,7 @@ BuildIndex(Relation heap, Relation index, IndexInfo *indexInfo,
 	/* Create pages */
 	CreateMetaPage(index, buildstate->dimensions, buildstate->lists, forkNum);
 	CreateListPages(index, buildstate->centers, buildstate->dimensions, buildstate->lists, forkNum, &buildstate->listInfo);
-	IvfflatBench("CreateEntryPages", CreateEntryPages(buildstate, forkNum));
+	CreateEntryPages(buildstate, forkNum);
 
 	FreeBuildState(buildstate);
 }

src/ivfflat.c

@@ -7,6 +7,7 @@
 #include "ivfflat.h"
 #include "utils/guc.h"
 #include "utils/selfuncs.h"
+#include "utils/spccache.h"
 
 #if PG_VERSION_NUM >= 120000
 #include "commands/progress.h"
@@ -63,13 +64,13 @@ ivfflatbuildphasename(int64 phasenum)
 static void
 ivfflatcostestimate(PlannerInfo *root, IndexPath *path, double loop_count,
 					Cost *indexStartupCost, Cost *indexTotalCost,
-					Selectivity *indexSelectivity, double *indexCorrelation
-					,double *indexPages
-)
+					Selectivity *indexSelectivity, double *indexCorrelation,
+					double *indexPages)
 {
 	GenericCosts costs;
 	int			lists;
 	double		ratio;
+	double		spc_seq_page_cost;
 	Relation	indexRel;
 #if PG_VERSION_NUM < 120000
 	List	   *qinfos;
@@ -88,6 +89,22 @@ ivfflatcostestimate(PlannerInfo *root, IndexPath *path, double loop_count,
 
 	MemSet(&costs, 0, sizeof(costs));
 
+	indexRel = index_open(path->indexinfo->indexoid, NoLock);
+	lists = IvfflatGetLists(indexRel);
+	index_close(indexRel, NoLock);
+
+	/* Get the ratio of lists that we need to visit */
+	ratio = ((double) ivfflat_probes) / lists;
+	if (ratio > 1.0)
+		ratio = 1.0;
+
+	/*
+	 * This gives us the subset of tuples to visit. This value is passed into
+	 * the generic cost estimator to determine the number of pages to visit
+	 * during the index scan.
+	 */
+	costs.numIndexTuples = path->indexinfo->tuples * ratio;
+
 #if PG_VERSION_NUM >= 120000
 	genericcostestimate(root, path, loop_count, &costs);
 #else
@@ -95,17 +112,31 @@ ivfflatcostestimate(PlannerInfo *root, IndexPath *path, double loop_count,
 	genericcostestimate(root, path, loop_count, qinfos, &costs);
 #endif
 
-	indexRel = index_open(path->indexinfo->indexoid, NoLock);
-	lists = IvfflatGetLists(indexRel);
-	index_close(indexRel, NoLock);
+	get_tablespace_page_costs(path->indexinfo->reltablespace, NULL, &spc_seq_page_cost);
 
-	ratio = ((double) ivfflat_probes) / lists;
-	if (ratio > 1)
-		ratio = 1;
+	/* Adjust cost if needed since TOAST not included in seq scan cost */
+	if (costs.numIndexPages > path->indexinfo->rel->pages && ratio < 0.5)
+	{
+		/* Change all page cost from random to sequential */
+		costs.indexTotalCost -= costs.numIndexPages * (costs.spc_random_page_cost - spc_seq_page_cost);
 
-	costs.indexTotalCost *= ratio;
+		/* Remove cost of extra pages */
+		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);
+	}
 
-	/* Startup cost and total cost are same */
+	/*
+	 * 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;

src/ivfinsert.c

@@ -23,6 +23,10 @@ FindInsertPage(Relation rel, Datum *values, BlockNumber *insertPage, ListInfo *
 	OffsetNumber offno;
 	OffsetNumber maxoffno;
 
+	/* Avoid compiler warning */
+	listInfo->blkno = nextblkno;
+	listInfo->offno = FirstOffsetNumber;
+
 	procinfo = index_getprocinfo(rel, 1, IVFFLAT_DISTANCE_PROC);
 	collation = rel->rd_indcollation[0];
 
@@ -39,7 +43,7 @@ FindInsertPage(Relation rel, Datum *values, BlockNumber *insertPage, ListInfo *
 			list = (IvfflatList) PageGetItem(cpage, PageGetItemId(cpage, offno));
 			distance = DatumGetFloat8(FunctionCall2Coll(procinfo, collation, values[0], PointerGetDatum(&list->center)));
 
-			if (distance < minDistance)
+			if (distance < minDistance || !BlockNumberIsValid(*insertPage))
 			{
 				*insertPage = list->insertPage;
 				listInfo->blkno = nextblkno;

src/ivfkmeans.c

@@ -1,6 +1,7 @@
 #include "postgres.h"
 
 #include <float.h>
+#include <math.h>
 
 #include "ivfflat.h"
 #include "miscadmin.h"
@@ -211,7 +212,7 @@ ElkanKmeans(Relation index, VectorArray samples, VectorArray centers)
 
 	/* Check memory requirements */
 	/* Add one to error message to ceil */
-	if (totalSize / 1024 > maintenance_work_mem)
+	if (totalSize > (Size) maintenance_work_mem * 1024L)
 		ereport(ERROR,
 				(errcode(ERRCODE_PROGRAM_LIMIT_EXCEEDED),
 				 errmsg("memory required is %zu MB, maintenance_work_mem is %d MB",
@@ -251,7 +252,7 @@ ElkanKmeans(Relation index, VectorArray samples, VectorArray centers)
 	for (j = 0; j < numSamples; j++)
 	{
 		minDistance = DBL_MAX;
-		closestCenter = -1;
+		closestCenter = 0;
 
 		/* Find closest center */
 		for (k = 0; k < numCenters; k++)
@@ -398,6 +399,14 @@ ElkanKmeans(Relation index, VectorArray samples, VectorArray centers)
 
 			if (centerCounts[j] > 0)
 			{
+				/* Double avoids overflow, but requires more memory */
+				/* TODO Update bounds */
+				for (k = 0; k < dimensions; k++)
+				{
+					if (isinf(vec->x[k]))
+						vec->x[k] = vec->x[k] > 0 ? FLT_MAX : -FLT_MAX;
+				}
+
 				for (k = 0; k < dimensions; k++)
 					vec->x[k] /= centerCounts[j];
 			}
@@ -461,12 +470,29 @@ CheckCenters(Relation index, VectorArray centers)
 {
 	FmgrInfo   *normprocinfo;
 	Oid			collation;
+	Vector	   *vec;
 	int			i;
+	int			j;
 	double		norm;
 
 	if (centers->length != centers->maxlen)
 		elog(ERROR, "Not enough centers. Please report a bug.");
 
+	/* Ensure no NaN or infinite values */
+	for (i = 0; i < centers->length; i++)
+	{
+		vec = VectorArrayGet(centers, i);
+
+		for (j = 0; j < vec->dim; j++)
+		{
+			if (isnan(vec->x[j]))
+				elog(ERROR, "NaN detected. Please report a bug.");
+
+			if (isinf(vec->x[j]))
+				elog(ERROR, "Infinite value detected. Please report a bug.");
+		}
+	}
+
 	/* Ensure no duplicate centers */
 	/* Fine to sort in-place */
 	qsort(centers->items, centers->length, VECTOR_SIZE(centers->dim), CompareVectors);

src/ivfscan.c

@@ -170,6 +170,8 @@ GetScanItems(IndexScanDesc scan, Datum value)
 		}
 	}
 
+	FreeAccessStrategy(bas);
+
 	/* TODO Scan more lists */
 	if (tuples < 100)
 		ereport(DEBUG1,

src/ivfvacuum.c

@@ -132,6 +132,8 @@ ivfflatbulkdelete(IndexVacuumInfo *info, IndexBulkDeleteResult *stats,
 		}
 	}
 
+	FreeAccessStrategy(bas);
+
 	return stats;
 }
 

src/vector.c

@@ -42,7 +42,7 @@ CheckDims(Vector * a, Vector * b)
 }
 
 /*
- * Ensure expected dimension
+ * Ensure expected dimensions
  */
 static inline void
 CheckExpectedDim(int32 typmod, int dim)
@@ -53,7 +53,9 @@ CheckExpectedDim(int32 typmod, int dim)
 				 errmsg("expected %d dimensions, not %d", typmod, dim)));
 }
 
-
+/*
+ * Ensure valid dimensions
+ */
 static inline void
 CheckDim(int dim)
 {
@@ -79,13 +81,28 @@ CheckElement(float value)
 				(errcode(ERRCODE_DATA_EXCEPTION),
 				 errmsg("NaN not allowed in vector")));
 
-
 	if (isinf(value))
 		ereport(ERROR,
 				(errcode(ERRCODE_DATA_EXCEPTION),
 				 errmsg("infinite value not allowed in vector")));
 }
 
+/*
+ * Check for whitespace, since array_isspace() is static
+ */
+static inline bool
+vector_isspace(char ch)
+{
+	if (ch == ' ' ||
+		ch == '\t' ||
+		ch == '\n' ||
+		ch == '\r' ||
+		ch == '\v' ||
+		ch == '\f')
+		return true;
+	return false;
+}
+
 /*
  * Check state array
  */
@@ -110,30 +127,6 @@ float_overflow_error(void)
 }
 #endif
 
-/*
- * Print vector - useful for debugging
- */
-void
-PrintVector(char *msg, Vector * vector)
-{
-	StringInfoData buf;
-	int			dim = vector->dim;
-	int			i;
-
-	initStringInfo(&buf);
-
-	appendStringInfoChar(&buf, '[');
-	for (i = 0; i < dim; i++)
-	{
-		if (i > 0)
-			appendStringInfoString(&buf, ",");
-		appendStringInfoString(&buf, float8out_internal(vector->x[i]));
-	}
-	appendStringInfoChar(&buf, ']');
-
-	elog(INFO, "%s = %s", msg, buf.data);
-}
-
 /*
  * Convert textual representation to internal representation
  */
@@ -149,11 +142,15 @@ vector_in(PG_FUNCTION_ARGS)
 	char	   *pt;
 	char	   *stringEnd;
 	Vector	   *result;
+	char	   *lit = pstrdup(str);
+
+	while (vector_isspace(*str))
+		str++;
 
 	if (*str != '[')
 		ereport(ERROR,
 				(errcode(ERRCODE_INVALID_TEXT_REPRESENTATION),
-				 errmsg("malformed vector literal: \"%s\"", str),
+				 errmsg("malformed vector literal: \"%s\"", lit),
 				 errdetail("Vector contents must start with \"[\".")));
 
 	str++;
@@ -167,6 +164,15 @@ vector_in(PG_FUNCTION_ARGS)
 					(errcode(ERRCODE_PROGRAM_LIMIT_EXCEEDED),
 					 errmsg("vector cannot have more than %d dimensions", VECTOR_MAX_DIM)));
 
+		while (vector_isspace(*pt))
+			pt++;
+
+		/* Check for empty string like float4in */
+		if (*pt == '\0')
+			ereport(ERROR,
+					(errcode(ERRCODE_INVALID_TEXT_REPRESENTATION),
+					 errmsg("invalid input syntax for type vector: \"%s\"", lit)));
+
 		/* Use strtof like float4in to avoid a double-rounding problem */
 		x[dim] = strtof(pt, &stringEnd);
 		CheckElement(x[dim]);
@@ -175,33 +181,53 @@ vector_in(PG_FUNCTION_ARGS)
 		if (stringEnd == pt)
 			ereport(ERROR,
 					(errcode(ERRCODE_INVALID_TEXT_REPRESENTATION),
-					 errmsg("invalid input syntax for type vector: \"%s\"", pt)));
+					 errmsg("invalid input syntax for type vector: \"%s\"", lit)));
+
+		while (vector_isspace(*stringEnd))
+			stringEnd++;
 
 		if (*stringEnd != '\0' && *stringEnd != ']')
 			ereport(ERROR,
 					(errcode(ERRCODE_INVALID_TEXT_REPRESENTATION),
-					 errmsg("invalid input syntax for type vector: \"%s\"", pt)));
+					 errmsg("invalid input syntax for type vector: \"%s\"", lit)));
 
 		pt = strtok(NULL, ",");
 	}
 
-	if (*stringEnd != ']')
+	if (stringEnd == NULL || *stringEnd != ']')
 		ereport(ERROR,
 				(errcode(ERRCODE_INVALID_TEXT_REPRESENTATION),
-				 errmsg("malformed vector literal"),
+				 errmsg("malformed vector literal: \"%s\"", lit),
 				 errdetail("Unexpected end of input.")));
 
-	if (stringEnd[1] != '\0')
+	stringEnd++;
+
+	/* Only whitespace is allowed after the closing brace */
+	while (vector_isspace(*stringEnd))
+		stringEnd++;
+
+	if (*stringEnd != '\0')
 		ereport(ERROR,
 				(errcode(ERRCODE_INVALID_TEXT_REPRESENTATION),
-				 errmsg("malformed vector literal"),
+				 errmsg("malformed vector literal: \"%s\"", lit),
 				 errdetail("Junk after closing right brace.")));
 
+	/* Ensure no consecutive delimiters since strtok skips */
+	for (pt = lit + 1; *pt != '\0'; pt++)
+	{
+		if (pt[-1] == ',' && *pt == ',')
+			ereport(ERROR,
+					(errcode(ERRCODE_INVALID_TEXT_REPRESENTATION),
+					 errmsg("malformed vector literal: \"%s\"", lit)));
+	}
+
 	if (dim < 1)
 		ereport(ERROR,
 				(errcode(ERRCODE_DATA_EXCEPTION),
 				 errmsg("vector must have at least 1 dimension")));
 
+	pfree(lit);
+
 	CheckExpectedDim(typmod, dim);
 
 	result = InitVector(dim);
@@ -272,6 +298,18 @@ vector_out(PG_FUNCTION_ARGS)
 	PG_RETURN_CSTRING(buf);
 }
 
+/*
+ * Print vector - useful for debugging
+ */
+void
+PrintVector(char *msg, Vector * vector)
+{
+	char	   *out = DatumGetPointer(DirectFunctionCall1(vector_out, PointerGetDatum(vector)));
+
+	elog(INFO, "%s = %s", msg, out);
+	pfree(out);
+}
+
 /*
  * Convert type modifier
  */
@@ -330,7 +368,10 @@ vector_recv(PG_FUNCTION_ARGS)
 
 	result = InitVector(dim);
 	for (i = 0; i < dim; i++)
+	{
 		result->x[i] = pq_getmsgfloat4(buf);
+		CheckElement(result->x[i]);
+	}
 
 	PG_RETURN_POINTER(result);
 }
@@ -407,6 +448,7 @@ array_to_vector(PG_FUNCTION_ARGS)
 					(errcode(ERRCODE_NULL_VALUE_NOT_ALLOWED),
 					 errmsg("array must not containing NULLs")));
 
+		/* TODO Move outside loop in 0.5.0 */
 		if (ARR_ELEMTYPE(array) == INT4OID)
 			result->x[i] = DatumGetInt32(elemsp[i]);
 		else if (ARR_ELEMTYPE(array) == FLOAT8OID)
@@ -434,17 +476,19 @@ Datum
 vector_to_float4(PG_FUNCTION_ARGS)
 {
 	Vector	   *vec = PG_GETARG_VECTOR_P(0);
-	Datum	   *d;
+	Datum	   *datums;
 	ArrayType  *result;
 	int			i;
 
-	d = (Datum *) palloc(sizeof(Datum) * vec->dim);
+	datums = (Datum *) palloc(sizeof(Datum) * vec->dim);
 
 	for (i = 0; i < vec->dim; i++)
-		d[i] = Float4GetDatum(vec->x[i]);
+		datums[i] = Float4GetDatum(vec->x[i]);
 
 	/* Use TYPALIGN_INT for float4 */
-	result = construct_array(d, vec->dim, FLOAT4OID, sizeof(float4), true, TYPALIGN_INT);
+	result = construct_array(datums, vec->dim, FLOAT4OID, sizeof(float4), true, TYPALIGN_INT);
+
+	pfree(datums);
 
 	PG_RETURN_POINTER(result);
 }
@@ -656,6 +700,13 @@ vector_add(PG_FUNCTION_ARGS)
 	for (int i = 0, imax = a->dim; i < imax; i++)
 		rx[i] = ax[i] + bx[i];
 
+	/* Check for overflow */
+	for (int i = 0, imax = a->dim; i < imax; i++)
+	{
+		if (isinf(rx[i]))
+			float_overflow_error();
+	}
+
 	PG_RETURN_POINTER(result);
 }
 
@@ -682,6 +733,13 @@ vector_sub(PG_FUNCTION_ARGS)
 	for (int i = 0, imax = a->dim; i < imax; i++)
 		rx[i] = ax[i] - bx[i];
 
+	/* Check for overflow */
+	for (int i = 0, imax = a->dim; i < imax; i++)
+	{
+		if (isinf(rx[i]))
+			float_overflow_error();
+	}
+
 	PG_RETURN_POINTER(result);
 }
 
@@ -827,12 +885,12 @@ vector_accum(PG_FUNCTION_ARGS)
 	n = statevalues[0] + 1.0;
 
 	statedatums = CreateStateDatums(dim);
-	statedatums[0] = Float8GetDatumFast(n);
+	statedatums[0] = Float8GetDatum(n);
 
 	if (newarr)
 	{
 		for (int i = 0; i < dim; i++)
-			statedatums[i + 1] = Float8GetDatumFast((double) x[i]);
+			statedatums[i + 1] = Float8GetDatum((double) x[i]);
 	}
 	else
 	{
@@ -840,10 +898,11 @@ vector_accum(PG_FUNCTION_ARGS)
 		{
 			double		v = statevalues[i + 1] + x[i];
 
+			/* Check for overflow */
 			if (isinf(v))
 				float_overflow_error();
 
-			statedatums[i + 1] = Float8GetDatumFast(v);
+			statedatums[i + 1] = Float8GetDatum(v);
 		}
 	}
 
@@ -888,7 +947,7 @@ vector_combine(PG_FUNCTION_ARGS)
 		dim = STATE_DIMS(statearray2);
 		statedatums = CreateStateDatums(dim);
 		for (int i = 1; i <= dim; i++)
-			statedatums[i] = Float8GetDatumFast(statevalues2[i]);
+			statedatums[i] = Float8GetDatum(statevalues2[i]);
 	}
 	else if (n2 == 0.0)
 	{
@@ -896,7 +955,7 @@ vector_combine(PG_FUNCTION_ARGS)
 		dim = STATE_DIMS(statearray1);
 		statedatums = CreateStateDatums(dim);
 		for (int i = 1; i <= dim; i++)
-			statedatums[i] = Float8GetDatumFast(statevalues1[i]);
+			statedatums[i] = Float8GetDatum(statevalues1[i]);
 	}
 	else
 	{
@@ -908,14 +967,15 @@ vector_combine(PG_FUNCTION_ARGS)
 		{
 			double		v = statevalues1[i] + statevalues2[i];
 
+			/* Check for overflow */
 			if (isinf(v))
 				float_overflow_error();
 
-			statedatums[i] = Float8GetDatumFast(v);
+			statedatums[i] = Float8GetDatum(v);
 		}
 	}
 
-	statedatums[0] = Float8GetDatumFast(n);
+	statedatums[0] = Float8GetDatum(n);
 
 	result = construct_array(statedatums, dim + 1,
 							 FLOAT8OID,
@@ -938,7 +998,6 @@ vector_avg(PG_FUNCTION_ARGS)
 	float8		n;
 	uint16		dim;
 	Vector	   *result;
-	float		v;
 
 	/* Check array before using */
 	statevalues = CheckStateArray(statearray, "vector_avg");
@@ -954,9 +1013,8 @@ vector_avg(PG_FUNCTION_ARGS)
 	result = InitVector(dim);
 	for (int i = 0; i < dim; i++)
 	{
-		v = statevalues[i + 1] / n;
-		CheckElement(v);
-		result->x[i] = v;
+		result->x[i] = statevalues[i + 1] / n;
+		CheckElement(result->x[i]);
 	}
 
 	PG_RETURN_POINTER(result);

test/expected/functions.out

@@ -4,12 +4,16 @@ SELECT '[1,2,3]'::vector + '[4,5,6]';
  [5,7,9]
 (1 row)
 
+SELECT '[3e38]'::vector + '[3e38]';
+ERROR:  value out of range: overflow
 SELECT '[1,2,3]'::vector - '[4,5,6]';
   ?column?  
 ------------
  [-3,-3,-3]
 (1 row)
 
+SELECT '[-3e38]'::vector - '[3e38]';
+ERROR:  value out of range: overflow
 SELECT vector_dims('[1,2,3]');
  vector_dims 
 -------------

test/expected/input.out

@@ -4,10 +4,22 @@ SELECT '[1,2,3]'::vector;
  [1,2,3]
 (1 row)
 
-SELECT '[-1,2,3]'::vector;
-  vector  
-----------
- [-1,2,3]
+SELECT '[-1,-2,-3]'::vector;
+   vector   
+------------
+ [-1,-2,-3]
+(1 row)
+
+SELECT '[1.,2.,3.]'::vector;
+ vector  
+---------
+ [1,2,3]
+(1 row)
+
+SELECT ' [ 1,  2 ,    3  ] '::vector;
+ vector  
+---------
+ [1,2,3]
 (1 row)
 
 SELECT '[1.23456]'::vector;
@@ -17,7 +29,7 @@ SELECT '[1.23456]'::vector;
 (1 row)
 
 SELECT '[hello,1]'::vector;
-ERROR:  invalid input syntax for type vector: "hello"
+ERROR:  invalid input syntax for type vector: "[hello,1]"
 LINE 1: SELECT '[hello,1]'::vector;
                ^
 SELECT '[NaN,1]'::vector;
@@ -32,13 +44,35 @@ SELECT '[-Infinity,1]'::vector;
 ERROR:  infinite value not allowed in vector
 LINE 1: SELECT '[-Infinity,1]'::vector;
                ^
+SELECT '[1.5e38,-1.5e38]'::vector;
+       vector       
+--------------------
+ [1.5e+38,-1.5e+38]
+(1 row)
+
+SELECT '[1.5e+38,-1.5e+38]'::vector;
+       vector       
+--------------------
+ [1.5e+38,-1.5e+38]
+(1 row)
+
+SELECT '[1.5e-38,-1.5e-38]'::vector;
+       vector       
+--------------------
+ [1.5e-38,-1.5e-38]
+(1 row)
+
+SELECT '[4e38,1]'::vector;
+ERROR:  infinite value not allowed in vector
+LINE 1: SELECT '[4e38,1]'::vector;
+               ^
 SELECT '[1,2,3'::vector;
-ERROR:  malformed vector literal
+ERROR:  malformed vector literal: "[1,2,3"
 LINE 1: SELECT '[1,2,3'::vector;
                ^
 DETAIL:  Unexpected end of input.
 SELECT '[1,2,3]9'::vector;
-ERROR:  malformed vector literal
+ERROR:  malformed vector literal: "[1,2,3]9"
 LINE 1: SELECT '[1,2,3]9'::vector;
                ^
 DETAIL:  Junk after closing right brace.
@@ -47,14 +81,36 @@ ERROR:  malformed vector literal: "1,2,3"
 LINE 1: SELECT '1,2,3'::vector;
                ^
 DETAIL:  Vector contents must start with "[".
+SELECT '['::vector;
+ERROR:  malformed vector literal: "["
+LINE 1: SELECT '['::vector;
+               ^
+DETAIL:  Unexpected end of input.
+SELECT '[,'::vector;
+ERROR:  malformed vector literal: "[,"
+LINE 1: SELECT '[,'::vector;
+               ^
+DETAIL:  Unexpected end of input.
 SELECT '[]'::vector;
 ERROR:  vector must have at least 1 dimension
 LINE 1: SELECT '[]'::vector;
                ^
 SELECT '[1,]'::vector;
-ERROR:  invalid input syntax for type vector: "]"
+ERROR:  invalid input syntax for type vector: "[1,]"
 LINE 1: SELECT '[1,]'::vector;
                ^
+SELECT '[1a]'::vector;
+ERROR:  invalid input syntax for type vector: "[1a]"
+LINE 1: SELECT '[1a]'::vector;
+               ^
+SELECT '[1,,3]'::vector;
+ERROR:  malformed vector literal: "[1,,3]"
+LINE 1: SELECT '[1,,3]'::vector;
+               ^
+SELECT '[1, ,3]'::vector;
+ERROR:  invalid input syntax for type vector: "[1, ,3]"
+LINE 1: SELECT '[1, ,3]'::vector;
+               ^
 SELECT '[1,2,3]'::vector(2);
 ERROR:  expected 2 dimensions, not 3
 SELECT unnest('{"[1,2,3]", "[4,5,6]"}'::vector[]);

test/sql/functions.sql

@@ -1,5 +1,7 @@
 SELECT '[1,2,3]'::vector + '[4,5,6]';
+SELECT '[3e38]'::vector + '[3e38]';
 SELECT '[1,2,3]'::vector - '[4,5,6]';
+SELECT '[-3e38]'::vector - '[3e38]';
 
 SELECT vector_dims('[1,2,3]');
 

test/sql/input.sql

@@ -1,15 +1,26 @@
 SELECT '[1,2,3]'::vector;
-SELECT '[-1,2,3]'::vector;
+SELECT '[-1,-2,-3]'::vector;
+SELECT '[1.,2.,3.]'::vector;
+SELECT ' [ 1,  2 ,    3  ] '::vector;
 SELECT '[1.23456]'::vector;
 SELECT '[hello,1]'::vector;
 SELECT '[NaN,1]'::vector;
 SELECT '[Infinity,1]'::vector;
 SELECT '[-Infinity,1]'::vector;
+SELECT '[1.5e38,-1.5e38]'::vector;
+SELECT '[1.5e+38,-1.5e+38]'::vector;
+SELECT '[1.5e-38,-1.5e-38]'::vector;
+SELECT '[4e38,1]'::vector;
 SELECT '[1,2,3'::vector;
 SELECT '[1,2,3]9'::vector;
 SELECT '1,2,3'::vector;
+SELECT '['::vector;
+SELECT '[,'::vector;
 SELECT '[]'::vector;
 SELECT '[1,]'::vector;
+SELECT '[1a]'::vector;
+SELECT '[1,,3]'::vector;
+SELECT '[1, ,3]'::vector;
 SELECT '[1,2,3]'::vector(2);
 
 SELECT unnest('{"[1,2,3]", "[4,5,6]"}'::vector[]);

vector.control

@@ -1,4 +1,4 @@
 comment = 'vector data type and ivfflat access method'
-default_version = '0.4.2'
+default_version = '0.4.4'
 module_pathname = '$libdir/vector'
 relocatable = true