javanna commented on code in PR #15795:
URL: https://github.com/apache/lucene/pull/15795#discussion_r3347373928
##########
lucene/misc/src/java/org/apache/lucene/misc/search/MemoryAccountingBitsetCollector.java:
##########
@@ -24,23 +24,46 @@
import org.apache.lucene.search.SimpleCollector;
import org.apache.lucene.util.FixedBitSet;
-/** Bitset collector which supports memory tracking */
+/**
+ * Bitset collector which supports memory tracking. Can operate in two modes:
Full index mode:
+ * allocates bitset for entire index (for single-threaded search),
Segment-local mode: allocates
+ * bitset only for segments processed (memory-efficient for concurrent search)
+ */
public class MemoryAccountingBitsetCollector extends SimpleCollector {
final CollectorMemoryTracker tracker;
+ final boolean segmentLocal;
Review Comment:
The segmentLocal flag could be dropped entirely by always using the
relative-offset logic. When a single-threaded collector processes all leaves,
minDocBase is always 0 (the first leaf always has docBase=0), so docBase -
minDocBase + doc reduces to docBase + doc, and maxDocEnd - minDocBase equals
the cumulative length += maxDoc. The two paths are numerically identical in the
single-threaded case — single-threaded search is just the degenerate case of
concurrent search where one collector sees all segments.
I do not see how this would be a breaking change. Does this make sense? I
may have misunderstood your concerns around bw comp.
##########
lucene/misc/src/java/org/apache/lucene/misc/search/MemoryAccountingBitsetCollectorManager.java:
##########
@@ -0,0 +1,69 @@
+/*
+ * Licensed to the Apache Software Foundation (ASF) under one or more
+ * contributor license agreements. See the NOTICE file distributed with
+ * this work for additional information regarding copyright ownership.
+ * The ASF licenses this file to You under the Apache License, Version 2.0
+ * (the "License"); you may not use this file except in compliance with
+ * the License. You may obtain a copy of the License at
+ *
+ * http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+package org.apache.lucene.misc.search;
+
+import java.util.Collection;
+import org.apache.lucene.misc.CollectorMemoryTracker;
+import org.apache.lucene.search.CollectorManager;
+import org.apache.lucene.util.FixedBitSet;
+
+/**
+ * CollectorManager for MemoryAccountingBitsetCollector that supports
concurrent search.
+ *
+ * <p>Creates multiple collectors for concurrent execution, each collector
only allocates bitset for
+ * segments it processes, then merges with proper offset in reduce().
+ */
+public class MemoryAccountingBitsetCollectorManager
+ implements CollectorManager<
+ MemoryAccountingBitsetCollector,
MemoryAccountingBitsetCollectorManager.Result> {
+
+ /** The result of a search, containing the matched document IDs and total
memory used. */
+ public record Result(FixedBitSet bitSet, long totalBytesUsed) {}
+
+ private final CollectorMemoryTracker tracker;
+
+ public MemoryAccountingBitsetCollectorManager(CollectorMemoryTracker
tracker) {
+ this.tracker = tracker;
+ }
+
+ @Override
+ public MemoryAccountingBitsetCollector newCollector() {
+ return new MemoryAccountingBitsetCollector(tracker, true);
+ }
+
+ @Override
+ public Result reduce(Collection<MemoryAccountingBitsetCollector> collectors)
{
+ int globalMaxDocEnd = 0;
+ for (MemoryAccountingBitsetCollector collector : collectors) {
+ globalMaxDocEnd = Math.max(globalMaxDocEnd, collector.getMaxDocEnd());
+ }
+
+ // TODO: with intra-segment concurrency enabled, globalMaxDocEnd equals
the full index maxDoc
+ // even when only a portion of the index was searched, causing
over-allocation of the result
+ // bitset.
+ FixedBitSet result = new FixedBitSet(globalMaxDocEnd);
+
+ for (MemoryAccountingBitsetCollector collector : collectors) {
+ if (collector.bitSet != null && collector.bitSet.length() > 0) {
+ int length = collector.getMaxDocEnd() - collector.getMinDocBase();
+ FixedBitSet.orRange(collector.bitSet, 0, result,
collector.getMinDocBase(), length);
Review Comment:
Don't we need to account for the memory used by this additional bitset too?
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: [email protected]
For queries about this service, please contact Infrastructure at:
[email protected]
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
