fix OOM (allocating too-large int[] in indexer for binary lengths)

git-svn-id: https://svn.apache.org/repos/asf/lucene/dev/branches/lucene4547@1440224 13f79535-47bb-0310-9956-ffa450edef68
2025-02-06 18:18:38 +00:00 · 2013-01-29 23:56:15 +00:00 · 2013-01-29 23:56:15 +00:00 · 34634ff4ec
commit 34634ff4ec
parent d08b259fba
16 changed files with 86 additions and 28 deletions
--- a/lucene/core/src/java/org/apache/lucene/index/BinaryDocValuesWriter.java
+++ b/lucene/core/src/java/org/apache/lucene/index/BinaryDocValuesWriter.java
@ -23,23 +23,32 @@ import java.io.IOException;
 import java.util.Iterator;

 import org.apache.lucene.codecs.DocValuesConsumer;
+import org.apache.lucene.store.RAMFile;
+import org.apache.lucene.store.RAMInputStream;
+import org.apache.lucene.store.RAMOutputStream;
 import org.apache.lucene.util.BytesRef;
-import org.apache.lucene.util.BytesRefArray;
 import org.apache.lucene.util.Counter;
+import org.apache.lucene.util.packed.AppendingLongBuffer;


 /** Buffers up pending byte[] per doc, then flushes when
 *  segment flushes. */
 class BinaryDocValuesWriter extends DocValuesWriter {

-  private final BytesRefArray bytesRefArray;
+  private final RAMFile bytes;
+  private final RAMOutputStream bytesWriter;
+  private final AppendingLongBuffer lengths;
  private final FieldInfo fieldInfo;
+  private final Counter iwBytesUsed;
+  private long bytesUsed;
  private int addedValues = 0;
-  private final BytesRef emptyBytesRef = new BytesRef();

  public BinaryDocValuesWriter(FieldInfo fieldInfo, Counter iwBytesUsed) {
    this.fieldInfo = fieldInfo;
-    this.bytesRefArray = new BytesRefArray(iwBytesUsed);
+    this.bytes = new RAMFile();
+    this.bytesWriter = new RAMOutputStream(bytes);
+    this.iwBytesUsed = iwBytesUsed;
+    this.lengths = new AppendingLongBuffer();
  }

  public void addValue(int docID, BytesRef value) {
@ -56,19 +65,41 @@ class BinaryDocValuesWriter extends DocValuesWriter {
    // Fill in any holes:
    while(addedValues < docID) {
      addedValues++;
-      bytesRefArray.append(emptyBytesRef);
+      lengths.add(0);
    }
    addedValues++;
-    bytesRefArray.append(value);
+    lengths.add(value.length);
+    try {
+      bytesWriter.writeBytes(value.bytes, value.offset, value.length);
+    } catch (IOException e) {
+      throw new RuntimeException(e);
+    }
+    
+    updateBytesUsed();
+  }
+  
+  private void updateBytesUsed() {
+    // nocommit not totally accurate, but just fix not to use RAMFile anyway
+    long numBuffers = (bytesWriter.getFilePointer() / 1024) + 1; // round up
+    long oversize = numBuffers * (1024 + 32); // fudge for arraylist/etc overhead
+    final long newBytesUsed = lengths.ramBytesUsed() + oversize;
+    iwBytesUsed.addAndGet(newBytesUsed - bytesUsed);
+    bytesUsed = newBytesUsed;
  }

  @Override
  public void finish(int maxDoc) {
+    try {
+      bytesWriter.close();
+    } catch (IOException e) {
+      throw new RuntimeException(e);
+    }
  }

  @Override
  public void flush(SegmentWriteState state, DocValuesConsumer dvConsumer) throws IOException {
    final int maxDoc = state.segmentInfo.getDocCount();
+    final int size = addedValues;

    dvConsumer.addBinaryField(fieldInfo,
                              new Iterable<BytesRef>() {
@ -76,8 +107,18 @@ class BinaryDocValuesWriter extends DocValuesWriter {
                                @Override
                                public Iterator<BytesRef> iterator() {
                                   return new Iterator<BytesRef>() {
+                                     RAMInputStream bytesReader;
+                                     AppendingLongBuffer.Iterator iter = lengths.iterator();
                                     BytesRef value = new BytesRef();
                                     int upto;
+                                     
+                                     {
+                                       try {
+                                         bytesReader = new RAMInputStream("bogus", bytes);
+                                       } catch (IOException e) {
+                                         throw new RuntimeException(e);
+                                       }
+                                     }

                                     @Override
                                     public boolean hasNext() {
@ -91,8 +132,15 @@ class BinaryDocValuesWriter extends DocValuesWriter {

                                     @Override
                                     public BytesRef next() {
-                                       if (upto < bytesRefArray.size()) {
-                                         bytesRefArray.get(value, upto);
+                                       if (upto < size) {
+                                         int length = (int) iter.next();
+                                         value.grow(length);
+                                         try {
+                                           bytesReader.readBytes(value.bytes, 0, length);
+                                         } catch (IOException e) {
+                                           throw new RuntimeException(e);
+                                         }
+                                         value.length = length;
                                       } else {
                                         value.length = 0;
                                       }
--- a/lucene/core/src/test/org/apache/lucene/index/Test2BBinaryDocValues.java
+++ b/lucene/core/src/test/org/apache/lucene/index/Test2BBinaryDocValues.java
@ -37,7 +37,6 @@ import com.carrotsearch.randomizedtesting.annotations.TimeoutSuite;
 public class Test2BBinaryDocValues extends LuceneTestCase {
  
  // indexes Integer.MAX_VALUE docs with a fixed binary field
-  // nocommit: broken ram accounting? ant test  -Dtestcase=Test2BBinaryDocValues -Dtests.method=testFixedBinary -Dtests.seed=5554AA830176B848 -Dtests.slow=true -Dtests.docvaluesformat=Disk -Dtests.locale=sr_RS_#Latn -Dtests.timezone=Africa/Luanda -Dtests.file.encoding=UTF-8
  public void testFixedBinary() throws Exception {
    BaseDirectoryWrapper dir = newFSDirectory(_TestUtil.getTempDir("2BFixedBinary"));
    if (dir instanceof MockDirectoryWrapper) {
@ -98,7 +97,6 @@ public class Test2BBinaryDocValues extends LuceneTestCase {
  }
  
  // indexes Integer.MAX_VALUE docs with a variable binary field
-  // nocommit: broken ram accounting? ant test  -Dtestcase=Test2BBinaryDocValues -Dtests.method=testVariableBinary -Dtests.seed=FD50D16920062578 -Dtests.slow=true -Dtests.docvaluesformat=Disk -Dtests.locale=sr_ME_#Latn -Dtests.timezone=America/Argentina/Tucuman -Dtests.file.encoding=UTF-8
  public void testVariableBinary() throws Exception {
    BaseDirectoryWrapper dir = newFSDirectory(_TestUtil.getTempDir("2BVariableBinary"));
    if (dir instanceof MockDirectoryWrapper) {
--- a/lucene/suggest/src/java/org/apache/lucene/search/suggest/BufferingTermFreqIteratorWrapper.java
+++ b/lucene/suggest/src/java/org/apache/lucene/search/suggest/BufferingTermFreqIteratorWrapper.java
@ -22,7 +22,6 @@ import java.util.Comparator;
 import org.apache.lucene.search.spell.TermFreqIterator;
 import org.apache.lucene.util.ArrayUtil;
 import org.apache.lucene.util.BytesRef;
-import org.apache.lucene.util.BytesRefArray;
 import org.apache.lucene.util.Counter;

 /**
--- a/lucene/suggest/src/java/org/apache/lucene/search/suggest/BytesRefArray.java
+++ b/lucene/suggest/src/java/org/apache/lucene/search/suggest/BytesRefArray.java
@ -1,4 +1,4 @@
-package org.apache.lucene.util;
+package org.apache.lucene.search.suggest;

 /*
 * Licensed to the Apache Software Foundation (ASF) under one or more
@ -20,6 +20,13 @@ package org.apache.lucene.util;
 import java.util.Arrays;
 import java.util.Comparator;

+import org.apache.lucene.util.ArrayUtil;
+import org.apache.lucene.util.BytesRef;
+import org.apache.lucene.util.ByteBlockPool;
+import org.apache.lucene.util.BytesRefIterator;
+import org.apache.lucene.util.Counter;
+import org.apache.lucene.util.RamUsageEstimator;
+import org.apache.lucene.util.SorterTemplate;

 /**
 * A simple append only random-access {@link BytesRef} array that stores full
@ -31,7 +38,7 @@ import java.util.Comparator;
 * @lucene.internal
 * @lucene.experimental
 */
-public final class BytesRefArray {
+final class BytesRefArray {
  private final ByteBlockPool pool;
  private int[] offsets = new int[1];
  private int lastElement = 0;
--- a/lucene/suggest/src/java/org/apache/lucene/search/suggest/fst/InMemorySorter.java
+++ b/lucene/suggest/src/java/org/apache/lucene/search/suggest/fst/InMemorySorter.java
@ -1,4 +1,4 @@
-package org.apache.lucene.search.suggest.fst;
+package org.apache.lucene.search.suggest;

 /*
 * Licensed to the Apache Software Foundation (ASF) under one or more
@ -19,9 +19,9 @@ package org.apache.lucene.search.suggest.fst;

 import java.util.Comparator;

+import org.apache.lucene.search.suggest.fst.BytesRefSorter;
 import org.apache.lucene.util.BytesRef;
 import org.apache.lucene.util.BytesRefIterator;
-import org.apache.lucene.util.BytesRefArray;
 import org.apache.lucene.util.Counter;

 /**
--- a/lucene/suggest/src/java/org/apache/lucene/search/suggest/fst/Sort.java
+++ b/lucene/suggest/src/java/org/apache/lucene/search/suggest/fst/Sort.java
@ -1,4 +1,4 @@
-package org.apache.lucene.search.suggest.fst;
+package org.apache.lucene.search.suggest;

 /*
 * Licensed to the Apache Software Foundation (ASF) under one or more
--- a/lucene/suggest/src/java/org/apache/lucene/search/suggest/SortedTermFreqIteratorWrapper.java
+++ b/lucene/suggest/src/java/org/apache/lucene/search/suggest/SortedTermFreqIteratorWrapper.java
@ -22,9 +22,8 @@ import java.io.IOException;
 import java.util.Comparator;

 import org.apache.lucene.search.spell.TermFreqIterator;
-import org.apache.lucene.search.suggest.fst.Sort;
-import org.apache.lucene.search.suggest.fst.Sort.ByteSequencesReader;
-import org.apache.lucene.search.suggest.fst.Sort.ByteSequencesWriter;
+import org.apache.lucene.search.suggest.Sort.ByteSequencesReader;
+import org.apache.lucene.search.suggest.Sort.ByteSequencesWriter;
 import org.apache.lucene.store.ByteArrayDataInput;
 import org.apache.lucene.store.ByteArrayDataOutput;
 import org.apache.lucene.util.ArrayUtil;
--- a/lucene/suggest/src/java/org/apache/lucene/search/suggest/analyzing/AnalyzingSuggester.java
+++ b/lucene/suggest/src/java/org/apache/lucene/search/suggest/analyzing/AnalyzingSuggester.java
@ -34,7 +34,7 @@ import org.apache.lucene.analysis.TokenStream;
 import org.apache.lucene.analysis.TokenStreamToAutomaton;
 import org.apache.lucene.search.spell.TermFreqIterator;
 import org.apache.lucene.search.suggest.Lookup;
-import org.apache.lucene.search.suggest.fst.Sort;
+import org.apache.lucene.search.suggest.Sort;
 import org.apache.lucene.store.ByteArrayDataInput;
 import org.apache.lucene.store.ByteArrayDataOutput;
 import org.apache.lucene.store.DataInput;
--- a/lucene/suggest/src/java/org/apache/lucene/search/suggest/fst/ExternalRefSorter.java
+++ b/lucene/suggest/src/java/org/apache/lucene/search/suggest/fst/ExternalRefSorter.java
@ -20,7 +20,8 @@ package org.apache.lucene.search.suggest.fst;
 import java.io.*;
 import java.util.Comparator;

-import org.apache.lucene.search.suggest.fst.Sort.ByteSequencesReader;
+import org.apache.lucene.search.suggest.Sort;
+import org.apache.lucene.search.suggest.Sort.ByteSequencesReader;
 import org.apache.lucene.util.BytesRef;
 import org.apache.lucene.util.BytesRefIterator;
 import org.apache.lucene.util.IOUtils;
--- a/lucene/suggest/src/java/org/apache/lucene/search/suggest/fst/FSTCompletionBuilder.java
+++ b/lucene/suggest/src/java/org/apache/lucene/search/suggest/fst/FSTCompletionBuilder.java
@ -20,6 +20,7 @@ package org.apache.lucene.search.suggest.fst;
 import java.io.Closeable;
 import java.io.IOException;

+import org.apache.lucene.search.suggest.InMemorySorter;
 import org.apache.lucene.util.BytesRef;
 import org.apache.lucene.util.BytesRefIterator;
 import org.apache.lucene.util.IntsRef;
--- a/lucene/suggest/src/java/org/apache/lucene/search/suggest/fst/FSTCompletionLookup.java
+++ b/lucene/suggest/src/java/org/apache/lucene/search/suggest/fst/FSTCompletionLookup.java
@ -26,8 +26,9 @@ import java.util.List;

 import org.apache.lucene.search.spell.TermFreqIterator;
 import org.apache.lucene.search.suggest.Lookup;
+import org.apache.lucene.search.suggest.Sort;
+import org.apache.lucene.search.suggest.Sort.SortInfo;
 import org.apache.lucene.search.suggest.fst.FSTCompletion.Completion;
-import org.apache.lucene.search.suggest.fst.Sort.SortInfo;
 import org.apache.lucene.search.suggest.tst.TSTLookup;
 import org.apache.lucene.store.ByteArrayDataInput;
 import org.apache.lucene.store.ByteArrayDataOutput;
--- a/lucene/suggest/src/java/org/apache/lucene/search/suggest/fst/WFSTCompletionLookup.java
+++ b/lucene/suggest/src/java/org/apache/lucene/search/suggest/fst/WFSTCompletionLookup.java
@ -28,7 +28,7 @@ import java.util.List;
 import org.apache.lucene.search.spell.TermFreqIterator;
 import org.apache.lucene.search.suggest.Lookup;
 import org.apache.lucene.search.suggest.SortedTermFreqIteratorWrapper;
-import org.apache.lucene.search.suggest.fst.Sort.ByteSequencesWriter;
+import org.apache.lucene.search.suggest.Sort.ByteSequencesWriter;
 import org.apache.lucene.store.ByteArrayDataInput;
 import org.apache.lucene.store.ByteArrayDataOutput;
 import org.apache.lucene.store.InputStreamDataInput;
--- a/lucene/suggest/src/test/org/apache/lucene/search/suggest/TestBytesRefArray.java
+++ b/lucene/suggest/src/test/org/apache/lucene/search/suggest/TestBytesRefArray.java
@ -1,4 +1,4 @@
-package org.apache.lucene.util;
+package org.apache.lucene.search.suggest;

 /*
 * Licensed to the Apache Software Foundation (ASF) under one or more
@ -22,7 +22,7 @@ import java.util.*;

 import org.apache.lucene.util.BytesRef;
 import org.apache.lucene.util.BytesRefIterator;
-import org.apache.lucene.util.BytesRefArray;
+import org.apache.lucene.util.Counter;
 import org.apache.lucene.util.LuceneTestCase;
 import org.apache.lucene.util._TestUtil;

--- a/lucene/suggest/src/test/org/apache/lucene/search/suggest/fst/BytesRefSortersTest.java
+++ b/lucene/suggest/src/test/org/apache/lucene/search/suggest/fst/BytesRefSortersTest.java
@ -17,6 +17,8 @@ package org.apache.lucene.search.suggest.fst;
 * limitations under the License.
 */

+import org.apache.lucene.search.suggest.InMemorySorter;
+import org.apache.lucene.search.suggest.Sort;
 import org.apache.lucene.util.BytesRef;
 import org.apache.lucene.util.BytesRefIterator;
 import org.apache.lucene.util.LuceneTestCase;
--- a/lucene/suggest/src/test/org/apache/lucene/search/suggest/fst/LargeInputFST.java
+++ b/lucene/suggest/src/test/org/apache/lucene/search/suggest/fst/LargeInputFST.java
@ -19,6 +19,7 @@ package org.apache.lucene.search.suggest.fst;

 import java.io.*;

+import org.apache.lucene.search.suggest.Sort;
 import org.apache.lucene.util.BytesRef;

 /**
--- a/lucene/suggest/src/test/org/apache/lucene/search/suggest/fst/TestSort.java
+++ b/lucene/suggest/src/test/org/apache/lucene/search/suggest/fst/TestSort.java
@ -22,9 +22,10 @@ import java.util.ArrayList;
 import java.util.Arrays;
 import java.util.Comparator;

-import org.apache.lucene.search.suggest.fst.Sort.BufferSize;
-import org.apache.lucene.search.suggest.fst.Sort.ByteSequencesWriter;
-import org.apache.lucene.search.suggest.fst.Sort.SortInfo;
+import org.apache.lucene.search.suggest.Sort;
+import org.apache.lucene.search.suggest.Sort.BufferSize;
+import org.apache.lucene.search.suggest.Sort.ByteSequencesWriter;
+import org.apache.lucene.search.suggest.Sort.SortInfo;
 import org.apache.lucene.util.*;
 import org.junit.*;