[GitHub] [lucene-solr] mikemccand commented on a change in pull request #2246: LUCENE-9694: New tool for creating a deterministic index

[GitBox <gi...@apache.org>]

mikemccand commented on a change in pull request #2246:
URL: https://github.com/apache/lucene-solr/pull/2246#discussion_r564484147



##########
File path: lucene/misc/src/java/org/apache/lucene/misc/index/IndexRearranger.java
##########
@@ -0,0 +1,132 @@
+/*
+ * 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.index;
+
+import java.io.IOException;
+import java.util.ArrayList;
+import java.util.List;
+import java.util.concurrent.Callable;
+import java.util.concurrent.ExecutorService;
+import java.util.concurrent.Executors;
+import java.util.concurrent.Future;
+import org.apache.lucene.index.CodecReader;
+import org.apache.lucene.index.DirectoryReader;
+import org.apache.lucene.index.FilterCodecReader;
+import org.apache.lucene.index.IndexReader;
+import org.apache.lucene.index.IndexWriter;
+import org.apache.lucene.index.IndexWriterConfig;
+import org.apache.lucene.index.LeafReaderContext;
+import org.apache.lucene.index.NoMergePolicy;
+import org.apache.lucene.store.Directory;
+import org.apache.lucene.util.BitSet;
+import org.apache.lucene.util.Bits;
+import org.apache.lucene.util.NamedThreadFactory;
+
+/**
+ * Copy and rearrange index according to document selectors, from input dir to output dir Length of
+ * documentSelectors determines how many segments there will be
+ */
+public class IndexRearranger {
+  protected final Directory input, output;
+  protected final IndexWriterConfig config;
+  protected final List<DocumentSelector> documentSelectors;
+
+  public IndexRearranger(
+      Directory input,
+      Directory output,
+      IndexWriterConfig config,
+      List<DocumentSelector> documentSelectors) {
+    this.input = input;
+    this.output = output;
+    this.config = config;
+    this.documentSelectors = documentSelectors;
+  }
+
+  public void execute() throws Exception {
+    config.setMergePolicy(
+        NoMergePolicy.INSTANCE); // do not merge since one addIndexes call create one segment
+    try (IndexWriter writer = new IndexWriter(output, config);
+        IndexReader reader = DirectoryReader.open(input)) {
+      ExecutorService executor =
+          Executors.newFixedThreadPool(
+              Math.min(Runtime.getRuntime().availableProcessors(), documentSelectors.size()),
+              new NamedThreadFactory("rearranger"));
+      ArrayList<Future<Void>> futures = new ArrayList<>();
+      for (DocumentSelector record : documentSelectors) {
+        Callable<Void> addSegment =
+            () -> {
+              addOneSegment(writer, reader, record);
+              return null;
+            };
+        futures.add(executor.submit(addSegment));
+      }
+      for (Future<Void> future : futures) {

Review comment:
       Yay, concurrent!  (thread-per-segment).  And ironically if you "optimize" (force-merge to one segment) your index, you lose all concurrency :) That's fine ... just don't force-merge!

##########
File path: lucene/CHANGES.txt
##########
@@ -11,6 +11,8 @@ New Features
 
 * LUCENE-9004: Approximate nearest vector search via NSW graphs
 
+* LUCENE-9694: New tool for creating a deterministic index (Haoyu Zhai)

Review comment:
       Probably we could also back-port this to Lucene 8.9, right?  We are not using any Lucene 9.x only APIs?
   
   If so, can you move this CHANGES entry down to that section?  (Hmm, if it does not yet exist, you can create a skeletal one).

##########
File path: lucene/misc/src/test/org/apache/lucene/misc/index/TestIndexRearranger.java
##########
@@ -0,0 +1,144 @@
+/*
+ * 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.index;
+
+import java.io.IOException;
+import java.util.List;
+import org.apache.lucene.document.Document;
+import org.apache.lucene.document.NumericDocValuesField;
+import org.apache.lucene.index.CodecReader;
+import org.apache.lucene.index.DirectoryReader;
+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.NoMergePolicy;
+import org.apache.lucene.index.NumericDocValues;
+import org.apache.lucene.search.Sort;
+import org.apache.lucene.search.SortField;
+import org.apache.lucene.store.Directory;
+import org.apache.lucene.util.BitSet;
+import org.apache.lucene.util.Bits;
+import org.apache.lucene.util.FixedBitSet;
+import org.apache.lucene.util.LuceneTestCase;
+
+public class TestIndexRearranger extends LuceneTestCase {
+  public void testRearrange() throws Exception {
+    Directory inputDir = newDirectory();
+    IndexWriter w =
+        new IndexWriter(
+            inputDir,
+            new IndexWriterConfig(null)
+                .setOpenMode(IndexWriterConfig.OpenMode.CREATE)
+                .setMergePolicy(NoMergePolicy.INSTANCE)
+                .setIndexSort(new Sort(new SortField("ord", SortField.Type.INT))));
+    for (int i = 0; i < 100; i++) {
+      Document doc = new Document();
+      doc.add(new NumericDocValuesField("ord", i));
+      w.addDocument(doc);
+      if (i % 10 == 9) {
+        w.commit();
+      }
+    }
+    IndexReader reader = DirectoryReader.open(w);
+    assertEquals(10, reader.leaves().size());
+    reader.close();
+    w.close();
+
+    Directory outputDir = newDirectory();
+    IndexRearranger rearranger =
+        new IndexRearranger(
+            inputDir,
+            outputDir,
+            new IndexWriterConfig(null)
+                .setOpenMode(IndexWriterConfig.OpenMode.CREATE)
+                .setMergePolicy(NoMergePolicy.INSTANCE)
+                .setIndexSort(new Sort(new SortField("ord", SortField.Type.INT))),
+            List.of(new OddDocSelector(), new EvenDocSelector()));
+    rearranger.execute();
+    reader = DirectoryReader.open(outputDir);
+    assertEquals(2, reader.leaves().size());
+    LeafReader segment1 = reader.leaves().get(0).reader();
+    assertEquals(50, segment1.numDocs());
+    NumericDocValues numericDocValues = segment1.getNumericDocValues("ord");
+    assertTrue(numericDocValues.advanceExact(0));
+    boolean odd = numericDocValues.longValue() % 2 == 1;
+    for (int i = 1; i < 50; i++) {
+      assertTrue(numericDocValues.advanceExact(i));
+      if (odd) {
+        assertEquals(1, numericDocValues.longValue() % 2);
+      } else {
+        assertEquals(0, numericDocValues.longValue() % 2);
+      }
+    }
+    LeafReader segment2 = reader.leaves().get(0).reader();
+    assertEquals(50, segment2.numDocs());
+    numericDocValues = segment2.getNumericDocValues("ord");
+    assertTrue(numericDocValues.advanceExact(0));
+    odd = numericDocValues.longValue() % 2 == 1;
+    for (int i = 1; i < 50; i++) {
+      assertTrue(numericDocValues.advanceExact(i));
+      if (odd) {
+        assertEquals(1, numericDocValues.longValue() % 2);
+      } else {
+        assertEquals(0, numericDocValues.longValue() % 2);
+      }
+    }
+    reader.close();
+    inputDir.close();
+    outputDir.close();
+  }
+
+  private class OddDocSelector implements IndexRearranger.DocumentSelector {

Review comment:
       This looks very similar to "shard splitting" implementations!
   
   I think an interesting `DocumentSelector` might be one which assigns documents into a "perfect" logarithmic staircase according to some base and exponent, e.g. target 7 segments with 1M docs, 7 segments with 100K docs, 7 segments with 10K docs.  This would allow us to take any index and rearrange it into arbitrary multi-segment situations.  Fun!

##########
File path: lucene/misc/src/java/org/apache/lucene/misc/index/IndexRearranger.java
##########
@@ -0,0 +1,132 @@
+/*
+ * 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.index;
+
+import java.io.IOException;
+import java.util.ArrayList;
+import java.util.List;
+import java.util.concurrent.Callable;
+import java.util.concurrent.ExecutorService;
+import java.util.concurrent.Executors;
+import java.util.concurrent.Future;
+import org.apache.lucene.index.CodecReader;
+import org.apache.lucene.index.DirectoryReader;
+import org.apache.lucene.index.FilterCodecReader;
+import org.apache.lucene.index.IndexReader;
+import org.apache.lucene.index.IndexWriter;
+import org.apache.lucene.index.IndexWriterConfig;
+import org.apache.lucene.index.LeafReaderContext;
+import org.apache.lucene.index.NoMergePolicy;
+import org.apache.lucene.store.Directory;
+import org.apache.lucene.util.BitSet;
+import org.apache.lucene.util.Bits;
+import org.apache.lucene.util.NamedThreadFactory;
+
+/**
+ * Copy and rearrange index according to document selectors, from input dir to output dir Length of
+ * documentSelectors determines how many segments there will be
+ */
+public class IndexRearranger {
+  protected final Directory input, output;
+  protected final IndexWriterConfig config;
+  protected final List<DocumentSelector> documentSelectors;
+
+  public IndexRearranger(
+      Directory input,
+      Directory output,
+      IndexWriterConfig config,
+      List<DocumentSelector> documentSelectors) {

Review comment:
       Maybe we could (followon PR maybe) have some sort of utility class implementing `DocumentSelector` by opening an existing index, crawling its docids to segments based on some specified primary key field or so?
   
   Then this command-line tool could take a pointer to source index (to extract `DocumentSelector` per segment), name of primary key field, a pointer to input index to rewrite, and a pointer to filesystem where the rearranged index should be written?

##########
File path: lucene/misc/src/java/org/apache/lucene/misc/index/IndexRearranger.java
##########
@@ -0,0 +1,132 @@
+/*
+ * 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.index;
+
+import java.io.IOException;
+import java.util.ArrayList;
+import java.util.List;
+import java.util.concurrent.Callable;
+import java.util.concurrent.ExecutorService;
+import java.util.concurrent.Executors;
+import java.util.concurrent.Future;
+import org.apache.lucene.index.CodecReader;
+import org.apache.lucene.index.DirectoryReader;
+import org.apache.lucene.index.FilterCodecReader;
+import org.apache.lucene.index.IndexReader;
+import org.apache.lucene.index.IndexWriter;
+import org.apache.lucene.index.IndexWriterConfig;
+import org.apache.lucene.index.LeafReaderContext;
+import org.apache.lucene.index.NoMergePolicy;
+import org.apache.lucene.store.Directory;
+import org.apache.lucene.util.BitSet;
+import org.apache.lucene.util.Bits;
+import org.apache.lucene.util.NamedThreadFactory;
+
+/**
+ * Copy and rearrange index according to document selectors, from input dir to output dir Length of
+ * documentSelectors determines how many segments there will be
+ */
+public class IndexRearranger {
+  protected final Directory input, output;
+  protected final IndexWriterConfig config;
+  protected final List<DocumentSelector> documentSelectors;
+
+  public IndexRearranger(
+      Directory input,
+      Directory output,
+      IndexWriterConfig config,
+      List<DocumentSelector> documentSelectors) {
+    this.input = input;
+    this.output = output;
+    this.config = config;
+    this.documentSelectors = documentSelectors;
+  }
+
+  public void execute() throws Exception {
+    config.setMergePolicy(
+        NoMergePolicy.INSTANCE); // do not merge since one addIndexes call create one segment
+    try (IndexWriter writer = new IndexWriter(output, config);
+        IndexReader reader = DirectoryReader.open(input)) {
+      ExecutorService executor =
+          Executors.newFixedThreadPool(
+              Math.min(Runtime.getRuntime().availableProcessors(), documentSelectors.size()),
+              new NamedThreadFactory("rearranger"));
+      ArrayList<Future<Void>> futures = new ArrayList<>();
+      for (DocumentSelector record : documentSelectors) {
+        Callable<Void> addSegment =
+            () -> {
+              addOneSegment(writer, reader, record);
+              return null;
+            };
+        futures.add(executor.submit(addSegment));
+      }
+      for (Future<Void> future : futures) {
+        future.get();
+      }
+      executor.shutdown();
+    }
+  }
+
+  private static void addOneSegment(IndexWriter writer, IndexReader reader, DocumentSelector record)

Review comment:
       Hmm maybe rename `record` to `selector`?

##########
File path: lucene/misc/src/java/org/apache/lucene/misc/index/IndexRearranger.java
##########
@@ -0,0 +1,132 @@
+/*
+ * 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.index;
+
+import java.io.IOException;
+import java.util.ArrayList;
+import java.util.List;
+import java.util.concurrent.Callable;
+import java.util.concurrent.ExecutorService;
+import java.util.concurrent.Executors;
+import java.util.concurrent.Future;
+import org.apache.lucene.index.CodecReader;
+import org.apache.lucene.index.DirectoryReader;
+import org.apache.lucene.index.FilterCodecReader;
+import org.apache.lucene.index.IndexReader;
+import org.apache.lucene.index.IndexWriter;
+import org.apache.lucene.index.IndexWriterConfig;
+import org.apache.lucene.index.LeafReaderContext;
+import org.apache.lucene.index.NoMergePolicy;
+import org.apache.lucene.store.Directory;
+import org.apache.lucene.util.BitSet;
+import org.apache.lucene.util.Bits;
+import org.apache.lucene.util.NamedThreadFactory;
+
+/**
+ * Copy and rearrange index according to document selectors, from input dir to output dir Length of
+ * documentSelectors determines how many segments there will be
+ */
+public class IndexRearranger {

Review comment:
       Maybe add a `// TODO` here and roughly describe the alternative solution @s1monw had proposed on the dev list thread?  That solution sounded compelling but a bit tricky to implement by wrangling `IndexWriter` to flush/merge in exactly the right way, but, it would probably be faster than this two-phased approach.

##########
File path: lucene/misc/src/java/org/apache/lucene/misc/index/IndexRearranger.java
##########
@@ -0,0 +1,132 @@
+/*
+ * 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.index;
+
+import java.io.IOException;
+import java.util.ArrayList;
+import java.util.List;
+import java.util.concurrent.Callable;
+import java.util.concurrent.ExecutorService;
+import java.util.concurrent.Executors;
+import java.util.concurrent.Future;
+import org.apache.lucene.index.CodecReader;
+import org.apache.lucene.index.DirectoryReader;
+import org.apache.lucene.index.FilterCodecReader;
+import org.apache.lucene.index.IndexReader;
+import org.apache.lucene.index.IndexWriter;
+import org.apache.lucene.index.IndexWriterConfig;
+import org.apache.lucene.index.LeafReaderContext;
+import org.apache.lucene.index.NoMergePolicy;
+import org.apache.lucene.store.Directory;
+import org.apache.lucene.util.BitSet;
+import org.apache.lucene.util.Bits;
+import org.apache.lucene.util.NamedThreadFactory;
+
+/**
+ * Copy and rearrange index according to document selectors, from input dir to output dir Length of

Review comment:
       Add period after `output dir` before `Length`?




----------------------------------------------------------------
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.

For queries about this service, please contact Infrastructure at:
users@infra.apache.org



---------------------------------------------------------------------
To unsubscribe, e-mail: issues-unsubscribe@lucene.apache.org
For additional commands, e-mail: issues-help@lucene.apache.org