Repository: lucene-solr
Updated Branches:
refs/heads/master 46d05afda -> 44324d3df
LUCENE-7050: Cache TermsQuery and point queries more aggressively.
Project: http://git-wip-us.apache.org/repos/asf/lucene-solr/repo
Commit: http://git-wip-us.apache.org/repos/asf/lucene-solr/commit/44324d3d
Tree: http://git-wip-us.apache.org/repos/asf/lucene-solr/tree/44324d3d
Diff: http://git-wip-us.apache.org/repos/asf/lucene-solr/diff/44324d3d
Branch: refs/heads/master
Commit: 44324d3dfe34fb436595f8c15bfc97eb39564b1f
Parents: 46d05af
Author: Adrien Grand <jp...@gmail.com>
Authored: Mon Feb 29 11:02:57 2016 +0100
Committer: Adrien Grand <jp...@gmail.com>
Committed: Mon Feb 29 11:02:57 2016 +0100
----------------------------------------------------------------------
lucene/CHANGES.txt | 3 +++
.../search/UsageTrackingQueryCachingPolicy.java | 16 ++++++++++++-
.../apache/lucene/queries/TermsQueryTest.java | 25 ++++++++++++++++++++
3 files changed, 43 insertions(+), 1 deletion(-)
----------------------------------------------------------------------
http://git-wip-us.apache.org/repos/asf/lucene-solr/blob/44324d3d/lucene/CHANGES.txt
----------------------------------------------------------------------
diff --git a/lucene/CHANGES.txt b/lucene/CHANGES.txt
index 7060170..c0925e8 100644
--- a/lucene/CHANGES.txt
+++ b/lucene/CHANGES.txt
@@ -124,6 +124,9 @@ Optimizations
* LUCENE-6793: LegacyNumericRangeQuery.hashCode() is now less subject to hash
collisions. (J.B. Langston via Adrien Grand)
+* LUCENE-7050: TermsQuery is now cached more aggressively by the default
+ query caching policy. (Adrien Grand)
+
Changes in Runtime Behavior
* LUCENE-6789: IndexSearcher's default Similarity is changed to BM25Similarity.
http://git-wip-us.apache.org/repos/asf/lucene-solr/blob/44324d3d/lucene/core/src/java/org/apache/lucene/search/UsageTrackingQueryCachingPolicy.java
----------------------------------------------------------------------
diff --git a/lucene/core/src/java/org/apache/lucene/search/UsageTrackingQueryCachingPolicy.java b/lucene/core/src/java/org/apache/lucene/search/UsageTrackingQueryCachingPolicy.java
index 1f48645..417de58 100644
--- a/lucene/core/src/java/org/apache/lucene/search/UsageTrackingQueryCachingPolicy.java
+++ b/lucene/core/src/java/org/apache/lucene/search/UsageTrackingQueryCachingPolicy.java
@@ -37,13 +37,27 @@ public final class UsageTrackingQueryCachingPolicy implements QueryCachingPolicy
// the hash code that we use as a sentinel in the ring buffer.
private static final int SENTINEL = Integer.MIN_VALUE;
+ private static boolean isPointQuery(Query query) {
+ // we need to check for super classes because we occasionally use anonymous
+ // sub classes of eg. PointRangeQuery
+ for (Class<?> clazz = query.getClass(); clazz != Query.class; clazz = clazz.getSuperclass()) {
+ final String simpleName = clazz.getSimpleName();
+ if (simpleName.startsWith("Point") && simpleName.endsWith("Query")) {
+ return true;
+ }
+ }
+ return false;
+ }
+
static boolean isCostly(Query query) {
// This does not measure the cost of iterating over the filter (for this we
// already have the DocIdSetIterator#cost API) but the cost to build the
// DocIdSet in the first place
return query instanceof MultiTermQuery ||
query instanceof MultiTermQueryConstantScoreWrapper ||
- query instanceof PointRangeQuery;
+ isPointQuery(query) ||
+ // can't refer to TermsQuery directly as it is in another module
+ "TermsQuery".equals(query.getClass().getSimpleName());
}
static boolean isCheap(Query query) {
http://git-wip-us.apache.org/repos/asf/lucene-solr/blob/44324d3d/lucene/queries/src/test/org/apache/lucene/queries/TermsQueryTest.java
----------------------------------------------------------------------
diff --git a/lucene/queries/src/test/org/apache/lucene/queries/TermsQueryTest.java b/lucene/queries/src/test/org/apache/lucene/queries/TermsQueryTest.java
index 7b1d4c9..c14d543 100644
--- a/lucene/queries/src/test/org/apache/lucene/queries/TermsQueryTest.java
+++ b/lucene/queries/src/test/org/apache/lucene/queries/TermsQueryTest.java
@@ -34,6 +34,8 @@ import org.apache.lucene.index.Fields;
import org.apache.lucene.index.FilterDirectoryReader;
import org.apache.lucene.index.FilterLeafReader;
import org.apache.lucene.index.IndexReader;
+import org.apache.lucene.index.IndexWriter;
+import org.apache.lucene.index.IndexWriterConfig;
import org.apache.lucene.index.LeafReader;
import org.apache.lucene.index.RandomIndexWriter;
import org.apache.lucene.index.Term;
@@ -49,6 +51,7 @@ import org.apache.lucene.search.QueryUtils;
import org.apache.lucene.search.Sort;
import org.apache.lucene.search.TermQuery;
import org.apache.lucene.search.TopDocs;
+import org.apache.lucene.search.UsageTrackingQueryCachingPolicy;
import org.apache.lucene.store.Directory;
import org.apache.lucene.util.BytesRef;
import org.apache.lucene.util.IOUtils;
@@ -325,4 +328,26 @@ public class TermsQueryTest extends LuceneTestCase {
TermsQuery query = new TermsQuery(new Term("field", new BytesRef(new byte[] { (byte) 0xff, (byte) 0xfe })));
assertEquals("field:[ff fe]", query.toString());
}
+
+ public void testIsConsideredCostlyByQueryCache() throws IOException {
+ Directory dir = newDirectory();
+ IndexWriterConfig iwc = newIndexWriterConfig();
+ IndexWriter w = new IndexWriter(dir, iwc);
+ Document doc = new Document();
+ for (int i = 0; i < 10000; ++i) {
+ w.addDocument(doc);
+ }
+ w.forceMerge(1);
+ DirectoryReader reader = DirectoryReader.open(w);
+ w.close();
+ TermsQuery query = new TermsQuery(new Term("foo", "bar"), new Term("foo", "baz"));
+ UsageTrackingQueryCachingPolicy policy = new UsageTrackingQueryCachingPolicy();
+ assertFalse(policy.shouldCache(query, getOnlySegmentReader(reader).getContext()));
+ policy.onUse(query);
+ policy.onUse(query);
+ // cached after two uses
+ assertTrue(policy.shouldCache(query, getOnlySegmentReader(reader).getContext()));
+ reader.close();
+ dir.close();
+ }
}