Use fabs [skip ci]

CHANGELOG.md

@@ -1,6 +1,7 @@
 ## 0.4.2 (unreleased)
 
 - Added notice when index created with little data
+- Fixed installation error with Postgres 12.0-12.2
 
 ## 0.4.1 (2023-03-21)
 

README.md

@@ -2,7 +2,11 @@
 
 Open-source vector similarity search for Postgres
 
-Supports exact and approximate nearest neighbor search for L2 distance, inner product, and cosine distance
+Supports
+
+- exact and approximate nearest neighbor search
+- L2 distance, inner product, and cosine distance
+- any [language](#languages) with a Postgres client
 
 [![Build Status](https://github.com/pgvector/pgvector/workflows/build/badge.svg?branch=master)](https://github.com/pgvector/pgvector/actions)
 
@@ -36,7 +40,7 @@ Create a vector column with 3 dimensions
 CREATE TABLE items (id bigserial PRIMARY KEY, embedding vector(3));
 ```
 
-Insert values
+Insert vectors
 
 ```sql
 INSERT INTO items (embedding) VALUES ('[1,2,3]'), ('[4,5,6]');
@@ -124,7 +128,7 @@ SELECT embedding <-> '[3,1,2]' AS distance FROM items;
 For inner product, multiply by -1 (since `<#>` returns the negative inner product)
 
 ```sql
-SELECT -1 * (embedding <#> '[3,1,2]') AS inner_product FROM items;
+SELECT (embedding <#> '[3,1,2]') * -1 AS inner_product FROM items;
 ```
 
 For cosine similarity, use 1 - cosine distance
@@ -133,7 +137,7 @@ For cosine similarity, use 1 - cosine distance
 SELECT 1 - (embedding <=> '[3,1,2]') AS cosine_similarity FROM items;
 ```
 
-#### Averaging
+#### Aggregates
 
 Average vectors
 
@@ -156,8 +160,8 @@ You can add an index to use approximate nearest neighbor search, which trades so
 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)
+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
 
 Add an index for each distance function you want to use.
 
@@ -275,8 +279,10 @@ Language | Libraries / Examples
 --- | ---
 C++ | [pgvector-cpp](https://github.com/pgvector/pgvector-cpp)
 C# | [pgvector-dotnet](https://github.com/pgvector/pgvector-dotnet)
+Crystal | [pgvector-crystal](https://github.com/pgvector/pgvector-crystal)
 Elixir | [pgvector-elixir](https://github.com/pgvector/pgvector-elixir)
 Go | [pgvector-go](https://github.com/pgvector/pgvector-go)
+Haskell | [pgvector-haskell](https://github.com/pgvector/pgvector-haskell)
 Java, Scala | [pgvector-java](https://github.com/pgvector/pgvector-java)
 Julia | [pgvector-julia](https://github.com/pgvector/pgvector-julia)
 Lua | [pgvector-lua](https://github.com/pgvector/pgvector-lua)
@@ -287,6 +293,7 @@ Python | [pgvector-python](https://github.com/pgvector/pgvector-python)
 R | [pgvector-r](https://github.com/pgvector/pgvector-r)
 Ruby | [pgvector-ruby](https://github.com/pgvector/pgvector-ruby), [Neighbor](https://github.com/ankane/neighbor)
 Rust | [pgvector-rust](https://github.com/pgvector/pgvector-rust)
+Swift | [pgvector-swift](https://github.com/pgvector/pgvector-swift)
 
 ## Frequently Asked Questions
 
@@ -305,6 +312,10 @@ Two things you can try are:
 1. use dimensionality reduction
 2. compile Postgres with a larger block size (`./configure --with-blocksize=32`) and edit the limit in `src/ivfflat.h`
 
+#### Why am I seeing less results after adding an index?
+
+The index was likely created with too little data for the number of lists. Drop the index until the table has more data.
+
 ## Reference
 
 ### Vector Type
@@ -441,7 +452,7 @@ 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/app-framework-services/p/pgvector-extension-for-postgresql)
-- Render - vote or comment on [this page](https://feedback.render.com/features/p/add-pgvector-extension-to-postgresql)
+- Heroku Postgres - vote or comment on [this page](https://github.com/heroku/roadmap/issues/156)
 
 ## Upgrading
 

src/ivfflat.h

@@ -206,6 +206,7 @@ typedef struct IvfflatScanOpaqueData
 
 	/* Lists */
 	pairingheap *listQueue;
+	double		minDistance;
 	IvfflatScanList lists[FLEXIBLE_ARRAY_MEMBER];	/* must come last */
 }			IvfflatScanOpaqueData;
 

src/ivfscan.c

@@ -60,6 +60,9 @@ GetScanLists(IndexScanDesc scan, Datum value)
 			/* Use procinfo from the index instead of scan key for performance */
 			distance = DatumGetFloat8(FunctionCall2Coll(so->procinfo, so->collation, PointerGetDatum(&list->center), value));
 
+			if (distance < so->minDistance)
+				so->minDistance = distance;
+
 			if (listCount < so->probes)
 			{
 				scanlist = &so->lists[listCount];
@@ -124,12 +127,18 @@ GetScanItems(IndexScanDesc scan, Datum value)
 	 *
 	 * See postgres/src/backend/storage/buffer/README for description
 	 */
-	BufferAccessStrategy bas = GetAccessStrategy(BAS_NORMAL);
+	BufferAccessStrategy bas = GetAccessStrategy(BAS_BULKREAD);
 
 	/* Search closest probes lists */
 	while (!pairingheap_is_empty(so->listQueue))
 	{
-		searchPage = ((IvfflatScanList *) pairingheap_remove_first(so->listQueue))->startPage;
+		IvfflatScanList *scanlist = (IvfflatScanList *) pairingheap_remove_first(so->listQueue);
+
+		/* Query-aware dynamic pruning */
+		if (fabs(scanlist->distance) > 1.5 * fabs(so->minDistance))
+			continue;
+
+		searchPage = scanlist->startPage;
 
 		/* Search all entry pages for list */
 		while (BlockNumberIsValid(searchPage))
@@ -252,6 +261,7 @@ ivfflatrescan(IndexScanDesc scan, ScanKey keys, int nkeys, ScanKey orderbys, int
 
 	so->first = true;
 	pairingheap_reset(so->listQueue);
+	so->minDistance = DBL_MAX;
 
 	if (keys && scan->numberOfKeys > 0)
 		memmove(scan->keyData, keys, scan->numberOfKeys * sizeof(ScanKeyData));

src/vector.c

@@ -100,7 +100,7 @@ CheckStateArray(ArrayType *statearray, const char *caller)
 	return (float8 *) ARR_DATA_PTR(statearray);
 }
 
-#if PG_VERSION_NUM < 120000
+#if PG_VERSION_NUM < 120003
 static pg_noinline void
 float_overflow_error(void)
 {