You are viewing a plain text version of this content. The canonical link for it is here.
Posted to commits@lucene.apache.org by jp...@apache.org on 2024/01/26 14:47:07 UTC
(lucene) branch main updated: Early exit range queries on non-matching segments. (#13033)
This is an automated email from the ASF dual-hosted git repository.
jpountz pushed a commit to branch main
in repository https://gitbox.apache.org/repos/asf/lucene.git
The following commit(s) were added to refs/heads/main by this push:
new 082d318abd7 Early exit range queries on non-matching segments. (#13033)
082d318abd7 is described below
commit 082d318abd791e129283018a4a359e7f2081149a
Author: Adrien Grand <jp...@gmail.com>
AuthorDate: Fri Jan 26 15:47:01 2024 +0100
Early exit range queries on non-matching segments. (#13033)
This change makes `PointRangeQuery` exit early when it knows that it doesn't
match a segment. In the case when this query is part of a conjunction, this
helps make sure `ScorerSupplier#get` doesn't get called on other required
clauses, which is sometimes an expensive operation (e.g. multi-term queries).
This is especially relevant for time-based data combined with
`LogByteSizePolicy`, which gives non-adjacent ranges of timestamps to each
segment, which in-turn increases the likelihood that some segments won't match
a range query at all.
---
lucene/CHANGES.txt | 5 ++++
.../org/apache/lucene/search/PointRangeQuery.java | 18 ++++++++++++
.../org/apache/lucene/search/TestPointQueries.java | 33 ++++++++++++++++++++++
3 files changed, 56 insertions(+)
diff --git a/lucene/CHANGES.txt b/lucene/CHANGES.txt
index 4d042cb72c6..6a54fecb9d8 100644
--- a/lucene/CHANGES.txt
+++ b/lucene/CHANGES.txt
@@ -216,6 +216,11 @@ Optimizations
* GITHUB#12989: Split taxonomy facet arrays across reusable chunks of elements to reduce allocations. (Michael Froh, Stefan Vodita)
+* GITHUB#13033: PointRangeQuery now exits earlier on segments whose values
+ don't intersect with the query range. When a PointRangeQuery is a required
+ clause of a boolean query, this helps save work on other required clauses of
+ the same boolean query. (Adrien Grand)
+
Bug Fixes
---------------------
* GITHUB#12866: Prevent extra similarity computation for single-level HNSW graphs. (Kaival Parikh)
diff --git a/lucene/core/src/java/org/apache/lucene/search/PointRangeQuery.java b/lucene/core/src/java/org/apache/lucene/search/PointRangeQuery.java
index 60875bffe81..64fb00bdb34 100644
--- a/lucene/core/src/java/org/apache/lucene/search/PointRangeQuery.java
+++ b/lucene/core/src/java/org/apache/lucene/search/PointRangeQuery.java
@@ -290,6 +290,24 @@ public abstract class PointRangeQuery extends Query {
return null;
}
+ if (values.getDocCount() == 0) {
+ return null;
+ } else {
+ final byte[] fieldPackedLower = values.getMinPackedValue();
+ final byte[] fieldPackedUpper = values.getMaxPackedValue();
+ for (int i = 0; i < numDims; ++i) {
+ int offset = i * bytesPerDim;
+ if (comparator.compare(lowerPoint, offset, fieldPackedUpper, offset) > 0
+ || comparator.compare(upperPoint, offset, fieldPackedLower, offset) < 0) {
+ // If this query is a required clause of a boolean query, then returning null here
+ // will help make sure that we don't call ScorerSupplier#get on other required clauses
+ // of the same boolean query, which is an expensive operation for some queries (e.g.
+ // multi-term queries).
+ return null;
+ }
+ }
+ }
+
boolean allDocsMatch;
if (values.getDocCount() == reader.maxDoc()) {
final byte[] fieldPackedLower = values.getMinPackedValue();
diff --git a/lucene/core/src/test/org/apache/lucene/search/TestPointQueries.java b/lucene/core/src/test/org/apache/lucene/search/TestPointQueries.java
index 341f75e280e..93646bfafa7 100644
--- a/lucene/core/src/test/org/apache/lucene/search/TestPointQueries.java
+++ b/lucene/core/src/test/org/apache/lucene/search/TestPointQueries.java
@@ -2454,4 +2454,37 @@ public class TestPointQueries extends LuceneTestCase {
r.close();
dir.close();
}
+
+ public void testRangeQuerySkipsNonMatchingSegments() throws IOException {
+ Directory dir = newDirectory();
+ IndexWriter w = new IndexWriter(dir, newIndexWriterConfig());
+ Document doc = new Document();
+ doc.add(new IntPoint("field", 2));
+ doc.add(new IntPoint("field2d", 1, 3));
+ w.addDocument(doc);
+
+ DirectoryReader reader = DirectoryReader.open(w);
+ IndexSearcher searcher = newSearcher(reader);
+
+ Query query = IntPoint.newRangeQuery("field", 0, 1);
+ Weight weight =
+ searcher.createWeight(searcher.rewrite(query), ScoreMode.COMPLETE_NO_SCORES, 1f);
+ assertNull(weight.scorerSupplier(reader.leaves().get(0)));
+
+ query = IntPoint.newRangeQuery("field", 3, 4);
+ weight = searcher.createWeight(searcher.rewrite(query), ScoreMode.COMPLETE_NO_SCORES, 1f);
+ assertNull(weight.scorerSupplier(reader.leaves().get(0)));
+
+ query = IntPoint.newRangeQuery("field2d", new int[] {0, 0}, new int[] {2, 2});
+ weight = searcher.createWeight(searcher.rewrite(query), ScoreMode.COMPLETE_NO_SCORES, 1f);
+ assertNull(weight.scorerSupplier(reader.leaves().get(0)));
+
+ query = IntPoint.newRangeQuery("field2d", new int[] {2, 2}, new int[] {4, 4});
+ weight = searcher.createWeight(searcher.rewrite(query), ScoreMode.COMPLETE_NO_SCORES, 1f);
+ assertNull(weight.scorerSupplier(reader.leaves().get(0)));
+
+ reader.close();
+ w.close();
+ dir.close();
+ }
}