javanna commented on code in PR #15795: URL: https://github.com/apache/lucene/pull/15795#discussion_r3274125232
########## lucene/misc/src/java/org/apache/lucene/misc/search/MemoryAccountingBitsetCollectorManager.java: ########## @@ -0,0 +1,92 @@ +/* + * 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, FixedBitSet> { + + private final CollectorMemoryTracker tracker; + private long totalBytesUsed; Review Comment: I am not sure it's a good idea to have such a mutable member in the collector manager. Collectors managers may be reused across multiple executions, although in many cases they are not. It feels like either they should be reset at the end of reduce, or even better, whatever is needed externally after reduction should be returned as part of the reduce method itself. ########## lucene/misc/src/java/org/apache/lucene/misc/search/MemoryAccountingBitsetCollectorManager.java: ########## @@ -0,0 +1,92 @@ +/* + * 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, FixedBitSet> { + + private final CollectorMemoryTracker tracker; + private long totalBytesUsed; + + public MemoryAccountingBitsetCollectorManager(CollectorMemoryTracker tracker) { + this.tracker = tracker; + } + + @Override + public MemoryAccountingBitsetCollector newCollector() { + return new MemoryAccountingBitsetCollector( + new CollectorMemoryTracker(tracker.getName() + "-collector", tracker.getMemoryLimit()), Review Comment: Could you expand on why a new collector memory tracker is created per slice, and there's also a global instance tracker member ? Seems like the tracker is thread-safe, hence it could be shared across the collectors? Also, if it is not shared, the memory limit becomes the number of collectors * the number of slices which is very different from the single threaded search scenario? ########## lucene/misc/src/java/org/apache/lucene/misc/search/MemoryAccountingBitsetCollectorManager.java: ########## @@ -0,0 +1,92 @@ +/* + * 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, FixedBitSet> { + + private final CollectorMemoryTracker tracker; + private long totalBytesUsed; + + public MemoryAccountingBitsetCollectorManager(CollectorMemoryTracker tracker) { + this.tracker = tracker; + } + + @Override + public MemoryAccountingBitsetCollector newCollector() { + return new MemoryAccountingBitsetCollector( + new CollectorMemoryTracker(tracker.getName() + "-collector", tracker.getMemoryLimit()), + true); + } + + @Override + public FixedBitSet reduce(Collection<MemoryAccountingBitsetCollector> collectors) { + if (collectors.isEmpty()) { + return new FixedBitSet(0); + } + + if (collectors.size() == 1) { Review Comment: are the empty and size 1 special cases necessary ? Can we simplify reduce? ########## 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: I wonder if we should make the concurrent case the only case, given that we are going to remove the single threaded search method. ########## lucene/misc/src/java/org/apache/lucene/misc/search/MemoryAccountingBitsetCollectorManager.java: ########## @@ -0,0 +1,92 @@ +/* + * 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, FixedBitSet> { + + private final CollectorMemoryTracker tracker; + private long totalBytesUsed; + + public MemoryAccountingBitsetCollectorManager(CollectorMemoryTracker tracker) { + this.tracker = tracker; + } + + @Override + public MemoryAccountingBitsetCollector newCollector() { + return new MemoryAccountingBitsetCollector( + new CollectorMemoryTracker(tracker.getName() + "-collector", tracker.getMemoryLimit()), + true); + } + + @Override + public FixedBitSet reduce(Collection<MemoryAccountingBitsetCollector> collectors) { + if (collectors.isEmpty()) { + return new FixedBitSet(0); + } + + if (collectors.size() == 1) { + MemoryAccountingBitsetCollector collector = collectors.iterator().next(); + this.totalBytesUsed = collector.tracker.getBytes(); + + if (collector.getMinDocBase() == 0) { + return collector.bitSet; + } + + FixedBitSet result = new FixedBitSet(collector.getMaxDocEnd()); + int length = collector.getMaxDocEnd() - collector.getMinDocBase(); + FixedBitSet.orRange(collector.bitSet, 0, result, collector.getMinDocBase(), length); + return result; + } + + // Find global doc range across all collectors + int globalMinDocBase = Integer.MAX_VALUE; + int globalMaxDocEnd = 0; + for (MemoryAccountingBitsetCollector collector : collectors) { + globalMinDocBase = Math.min(globalMinDocBase, collector.getMinDocBase()); + globalMaxDocEnd = Math.max(globalMaxDocEnd, collector.getMaxDocEnd()); + long collectorBytes = collector.tracker.getBytes(); + this.totalBytesUsed += collectorBytes; + } + + 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: Do note that with intra-segment concurrency enabled, using maxdoc we would allocate a bitset for the whole segment, even when searching a portion of it. This is worth at least a comment, or a TODO. ########## lucene/misc/src/java/org/apache/lucene/misc/search/MemoryAccountingBitsetCollectorManager.java: ########## @@ -0,0 +1,92 @@ +/* + * 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, FixedBitSet> { + + private final CollectorMemoryTracker tracker; + private long totalBytesUsed; + + public MemoryAccountingBitsetCollectorManager(CollectorMemoryTracker tracker) { + this.tracker = tracker; + } + + @Override + public MemoryAccountingBitsetCollector newCollector() { + return new MemoryAccountingBitsetCollector( + new CollectorMemoryTracker(tracker.getName() + "-collector", tracker.getMemoryLimit()), + true); + } + + @Override + public FixedBitSet reduce(Collection<MemoryAccountingBitsetCollector> collectors) { + if (collectors.isEmpty()) { + return new FixedBitSet(0); + } + + if (collectors.size() == 1) { + MemoryAccountingBitsetCollector collector = collectors.iterator().next(); + this.totalBytesUsed = collector.tracker.getBytes(); + + if (collector.getMinDocBase() == 0) { + return collector.bitSet; + } + + FixedBitSet result = new FixedBitSet(collector.getMaxDocEnd()); + int length = collector.getMaxDocEnd() - collector.getMinDocBase(); + FixedBitSet.orRange(collector.bitSet, 0, result, collector.getMinDocBase(), length); + return result; + } + + // Find global doc range across all collectors + int globalMinDocBase = Integer.MAX_VALUE; + int globalMaxDocEnd = 0; + for (MemoryAccountingBitsetCollector collector : collectors) { + globalMinDocBase = Math.min(globalMinDocBase, collector.getMinDocBase()); + globalMaxDocEnd = Math.max(globalMaxDocEnd, collector.getMaxDocEnd()); + long collectorBytes = collector.tracker.getBytes(); + this.totalBytesUsed += collectorBytes; + } + + 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: If I understand correctly, the idea is to allocate a bitset that holds the range of docs that each collector effectively sees, as opposed to the entire range of doc ids which is wasteful. That's a good idea! ########## lucene/misc/src/test/org/apache/lucene/misc/search/TestMemoryAccountingBitsetCollector.java: ########## @@ -57,31 +58,43 @@ public void tearDown() throws Exception { dir.close(); } - public void testMemoryAccountingBitsetCollectorMemoryLimit() { + public void testMemoryAccountingBitsetCollectorMemoryLimit() throws Exception { Review Comment: Do all existing tests already use newSearcher and a random index writer ? -- 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]
