You are viewing a plain text version of this content. The canonical link for it is here.
Posted to commits@lucene.apache.org by mi...@apache.org on 2010/12/11 12:07:02 UTC
svn commit: r1044635 [2/2] - in /lucene/dev/trunk/lucene/src:
java/org/apache/lucene/index/ test/org/apache/lucene/
test/org/apache/lucene/index/ test/org/apache/lucene/search/
test/org/apache/lucene/store/ test/org/apache/lucene/util/
Modified: lucene/dev/trunk/lucene/src/java/org/apache/lucene/index/IndexWriter.java
URL: http://svn.apache.org/viewvc/lucene/dev/trunk/lucene/src/java/org/apache/lucene/index/IndexWriter.java?rev=1044635&r1=1044634&r2=1044635&view=diff
==============================================================================
--- lucene/dev/trunk/lucene/src/java/org/apache/lucene/index/IndexWriter.java (original)
+++ lucene/dev/trunk/lucene/src/java/org/apache/lucene/index/IndexWriter.java Sat Dec 11 11:07:01 2010
@@ -35,6 +35,7 @@ import org.apache.lucene.util.Bits;
import java.io.IOException;
import java.io.Closeable;
import java.io.PrintStream;
+import java.util.concurrent.atomic.AtomicInteger;
import java.util.List;
import java.util.Collection;
import java.util.ArrayList;
@@ -201,9 +202,8 @@ public class IndexWriter implements Clos
private final static int MERGE_READ_BUFFER_SIZE = 4096;
// Used for printing messages
- private static Object MESSAGE_ID_LOCK = new Object();
- private static int MESSAGE_ID = 0;
- private int messageID = -1;
+ private static final AtomicInteger MESSAGE_ID = new AtomicInteger();
+ private int messageID = MESSAGE_ID.getAndIncrement();
volatile private boolean hitOOM;
private final Directory directory; // where this index resides
@@ -218,7 +218,7 @@ public class IndexWriter implements Clos
volatile SegmentInfos pendingCommit; // set when a commit is pending (after prepareCommit() & before commit())
volatile long pendingCommitChangeCount;
- private final SegmentInfos segmentInfos; // the segments
+ final SegmentInfos segmentInfos; // the segments
private DocumentsWriter docWriter;
private IndexFileDeleter deleter;
@@ -245,10 +245,11 @@ public class IndexWriter implements Clos
private long mergeGen;
private boolean stopMerges;
- private int flushCount;
- private int flushDeletesCount;
+ private final AtomicInteger flushCount = new AtomicInteger();
+ private final AtomicInteger flushDeletesCount = new AtomicInteger();
final ReaderPool readerPool = new ReaderPool();
+ final BufferedDeletes bufferedDeletes;
// This is a "write once" variable (like the organic dye
// on a DVD-R that may or may not be heated by a laser and
@@ -402,20 +403,26 @@ public class IndexWriter implements Clos
/**
* Release the segment reader (i.e. decRef it and close if there
* are no more references.
+ * @return true if this release altered the index (eg
+ * the SegmentReader had pending changes to del docs and
+ * was closed). Caller must call checkpoint() if so.
* @param sr
* @throws IOException
*/
- public synchronized void release(SegmentReader sr) throws IOException {
- release(sr, false);
+ public synchronized boolean release(SegmentReader sr) throws IOException {
+ return release(sr, false);
}
/**
* Release the segment reader (i.e. decRef it and close if there
* are no more references.
+ * @return true if this release altered the index (eg
+ * the SegmentReader had pending changes to del docs and
+ * was closed). Caller must call checkpoint() if so.
* @param sr
* @throws IOException
*/
- public synchronized void release(SegmentReader sr, boolean drop) throws IOException {
+ public synchronized boolean release(SegmentReader sr, boolean drop) throws IOException {
final boolean pooled = readerMap.containsKey(sr.getSegmentInfo());
@@ -446,13 +453,10 @@ public class IndexWriter implements Clos
// not pooling readers, we release it:
readerMap.remove(sr.getSegmentInfo());
- if (hasChanges) {
- // Must checkpoint w/ deleter, because this
- // segment reader will have created new _X_N.del
- // file.
- deleter.checkpoint(segmentInfos, false);
- }
+ return hasChanges;
}
+
+ return false;
}
/** Remove all our references to readers, and commits
@@ -600,6 +604,8 @@ public class IndexWriter implements Clos
}
}
+
+
/**
* Obtain the number of deleted docs for a pooled reader.
* If the reader isn't being pooled, the segmentInfo's
@@ -646,15 +652,6 @@ public class IndexWriter implements Clos
infoStream.println("IW " + messageID + " [" + new Date() + "; " + Thread.currentThread().getName() + "]: " + message);
}
- private synchronized void setMessageID(PrintStream infoStream) {
- if (infoStream != null && messageID == -1) {
- synchronized(MESSAGE_ID_LOCK) {
- messageID = MESSAGE_ID++;
- }
- }
- this.infoStream = infoStream;
- }
-
CodecProvider codecs;
/**
@@ -690,7 +687,7 @@ public class IndexWriter implements Clos
config = (IndexWriterConfig) conf.clone();
directory = d;
analyzer = conf.getAnalyzer();
- setMessageID(defaultInfoStream);
+ infoStream = defaultInfoStream;
maxFieldLength = conf.getMaxFieldLength();
termIndexInterval = conf.getTermIndexInterval();
mergePolicy = conf.getMergePolicy();
@@ -699,6 +696,8 @@ public class IndexWriter implements Clos
mergedSegmentWarmer = conf.getMergedSegmentWarmer();
codecs = conf.getCodecProvider();
+ bufferedDeletes = new BufferedDeletes(messageID);
+ bufferedDeletes.setInfoStream(infoStream);
poolReaders = conf.getReaderPooling();
OpenMode mode = conf.getOpenMode();
@@ -766,7 +765,7 @@ public class IndexWriter implements Clos
setRollbackSegmentInfos(segmentInfos);
- docWriter = new DocumentsWriter(directory, this, conf.getIndexingChain(), conf.getMaxThreadStates(), getCurrentFieldInfos());
+ docWriter = new DocumentsWriter(directory, this, conf.getIndexingChain(), conf.getMaxThreadStates(), getCurrentFieldInfos(), bufferedDeletes);
docWriter.setInfoStream(infoStream);
docWriter.setMaxFieldLength(maxFieldLength);
@@ -785,7 +784,6 @@ public class IndexWriter implements Clos
segmentInfos.changed();
}
- docWriter.setMaxBufferedDeleteTerms(conf.getMaxBufferedDeleteTerms());
docWriter.setRAMBufferSizeMB(conf.getRAMBufferSizeMB());
docWriter.setMaxBufferedDocs(conf.getMaxBufferedDocs());
pushMaxBufferedDocs();
@@ -896,9 +894,10 @@ public class IndexWriter implements Clos
*/
public void setInfoStream(PrintStream infoStream) {
ensureOpen();
- setMessageID(infoStream);
+ this.infoStream = infoStream;
docWriter.setInfoStream(infoStream);
deleter.setInfoStream(infoStream);
+ bufferedDeletes.setInfoStream(infoStream);
if (infoStream != null)
messageState();
}
@@ -1029,8 +1028,6 @@ public class IndexWriter implements Clos
private void closeInternal(boolean waitForMerges) throws CorruptIndexException, IOException {
- docWriter.pauseAllThreads();
-
try {
if (infoStream != null)
message("now flush at close");
@@ -1085,8 +1082,6 @@ public class IndexWriter implements Clos
closing = false;
notifyAll();
if (!closed) {
- if (docWriter != null)
- docWriter.resumeAllThreads();
if (infoStream != null)
message("hit exception while closing");
}
@@ -1094,87 +1089,6 @@ public class IndexWriter implements Clos
}
}
- /** Tells the docWriter to close its currently open shared
- * doc stores (stored fields & vectors files).
- * Return value specifices whether new doc store files are compound or not.
- */
- private synchronized boolean flushDocStores() throws IOException {
-
- if (infoStream != null) {
- message("flushDocStores segment=" + docWriter.getDocStoreSegment());
- }
-
- boolean useCompoundDocStore = false;
- if (infoStream != null) {
- message("closeDocStores segment=" + docWriter.getDocStoreSegment());
- }
-
- String docStoreSegment;
-
- boolean success = false;
- try {
- docStoreSegment = docWriter.closeDocStore();
- success = true;
- } finally {
- if (!success && infoStream != null) {
- message("hit exception closing doc store segment");
- }
- }
-
- if (infoStream != null) {
- message("flushDocStores files=" + docWriter.closedFiles());
- }
-
- useCompoundDocStore = mergePolicy.useCompoundDocStore(segmentInfos);
-
- if (useCompoundDocStore && docStoreSegment != null && docWriter.closedFiles().size() != 0) {
- // Now build compound doc store file
-
- if (infoStream != null) {
- message("create compound file " + IndexFileNames.segmentFileName(docStoreSegment, "", IndexFileNames.COMPOUND_FILE_STORE_EXTENSION));
- }
-
- success = false;
-
- final int numSegments = segmentInfos.size();
- final String compoundFileName = IndexFileNames.segmentFileName(docStoreSegment, "", IndexFileNames.COMPOUND_FILE_STORE_EXTENSION);
-
- try {
- CompoundFileWriter cfsWriter = new CompoundFileWriter(directory, compoundFileName);
- for (final String file : docWriter.closedFiles() ) {
- cfsWriter.addFile(file);
- }
-
- // Perform the merge
- cfsWriter.close();
- success = true;
-
- } finally {
- if (!success) {
- if (infoStream != null)
- message("hit exception building compound file doc store for segment " + docStoreSegment);
- deleter.deleteFile(compoundFileName);
- docWriter.abort();
- }
- }
-
- for(int i=0;i<numSegments;i++) {
- SegmentInfo si = segmentInfos.info(i);
- if (si.getDocStoreOffset() != -1 &&
- si.getDocStoreSegment().equals(docStoreSegment))
- si.setDocStoreIsCompoundFile(true);
- }
-
- checkpoint();
-
- // In case the files we just merged into a CFS were
- // not previously checkpointed:
- deleter.deleteNewFiles(docWriter.closedFiles());
- }
-
- return useCompoundDocStore;
- }
-
/** Returns the Directory used by this index. */
public Directory getDirectory() {
// Pass false because the flush during closing calls getDirectory
@@ -1226,8 +1140,12 @@ public class IndexWriter implements Clos
public synchronized boolean hasDeletions() throws IOException {
ensureOpen();
- if (docWriter.hasDeletes())
+ if (bufferedDeletes.any()) {
+ return true;
+ }
+ if (docWriter.anyDeletions()) {
return true;
+ }
for (int i = 0; i < segmentInfos.size(); i++)
if (segmentInfos.info(i).hasDeletions())
return true;
@@ -1321,13 +1239,14 @@ public class IndexWriter implements Clos
boolean success = false;
try {
try {
- doFlush = docWriter.addDocument(doc, analyzer);
+ doFlush = docWriter.updateDocument(doc, analyzer, null);
success = true;
} finally {
if (!success) {
- if (infoStream != null)
+ if (infoStream != null) {
message("hit exception adding document");
+ }
synchronized (this) {
// If docWriter has some aborted files that were
@@ -1361,9 +1280,9 @@ public class IndexWriter implements Clos
public void deleteDocuments(Term term) throws CorruptIndexException, IOException {
ensureOpen();
try {
- boolean doFlush = docWriter.bufferDeleteTerm(term);
- if (doFlush)
+ if (docWriter.deleteTerm(term, false)) {
flush(true, false, false);
+ }
} catch (OutOfMemoryError oom) {
handleOOM(oom, "deleteDocuments(Term)");
}
@@ -1385,9 +1304,9 @@ public class IndexWriter implements Clos
public void deleteDocuments(Term... terms) throws CorruptIndexException, IOException {
ensureOpen();
try {
- boolean doFlush = docWriter.bufferDeleteTerms(terms);
- if (doFlush)
+ if (docWriter.deleteTerms(terms)) {
flush(true, false, false);
+ }
} catch (OutOfMemoryError oom) {
handleOOM(oom, "deleteDocuments(Term..)");
}
@@ -1406,9 +1325,13 @@ public class IndexWriter implements Clos
*/
public void deleteDocuments(Query query) throws CorruptIndexException, IOException {
ensureOpen();
- boolean doFlush = docWriter.bufferDeleteQuery(query);
- if (doFlush)
- flush(true, false, false);
+ try {
+ if (docWriter.deleteQuery(query)) {
+ flush(true, false, false);
+ }
+ } catch (OutOfMemoryError oom) {
+ handleOOM(oom, "deleteDocuments(Query)");
+ }
}
/**
@@ -1426,9 +1349,13 @@ public class IndexWriter implements Clos
*/
public void deleteDocuments(Query... queries) throws CorruptIndexException, IOException {
ensureOpen();
- boolean doFlush = docWriter.bufferDeleteQueries(queries);
- if (doFlush)
- flush(true, false, false);
+ try {
+ if (docWriter.deleteQueries(queries)) {
+ flush(true, false, false);
+ }
+ } catch (OutOfMemoryError oom) {
+ handleOOM(oom, "deleteDocuments(Query..)");
+ }
}
/**
@@ -1478,25 +1405,30 @@ public class IndexWriter implements Clos
boolean doFlush = false;
boolean success = false;
try {
- doFlush = docWriter.updateDocument(term, doc, analyzer);
+ doFlush = docWriter.updateDocument(doc, analyzer, term);
success = true;
} finally {
if (!success) {
- if (infoStream != null)
+ if (infoStream != null) {
message("hit exception updating document");
+ }
synchronized (this) {
// If docWriter has some aborted files that were
// never incref'd, then we clean them up here
- final Collection<String> files = docWriter.abortedFiles();
- if (files != null)
- deleter.deleteNewFiles(files);
+ if (docWriter != null) {
+ final Collection<String> files = docWriter.abortedFiles();
+ if (files != null) {
+ deleter.deleteNewFiles(files);
+ }
+ }
}
}
}
- if (doFlush)
+ if (doFlush) {
flush(true, false, false);
+ }
} catch (OutOfMemoryError oom) {
handleOOM(oom, "updateDocument");
}
@@ -1522,13 +1454,13 @@ public class IndexWriter implements Clos
}
// for test purpose
- final synchronized int getFlushCount() {
- return flushCount;
+ final int getFlushCount() {
+ return flushCount.get();
}
// for test purpose
- final synchronized int getFlushDeletesCount() {
- return flushDeletesCount;
+ final int getFlushDeletesCount() {
+ return flushDeletesCount.get();
}
final String newSegmentName() {
@@ -1660,8 +1592,10 @@ public class IndexWriter implements Clos
if (maxNumSegments < 1)
throw new IllegalArgumentException("maxNumSegments must be >= 1; got " + maxNumSegments);
- if (infoStream != null)
+ if (infoStream != null) {
message("optimize: index now " + segString());
+ message("now flush at optimize");
+ }
flush(true, false, true);
@@ -1944,8 +1878,6 @@ public class IndexWriter implements Clos
message("rollback");
}
- docWriter.pauseAllThreads();
-
try {
finishMerges(false);
@@ -1955,6 +1887,8 @@ public class IndexWriter implements Clos
mergePolicy.close();
mergeScheduler.close();
+ bufferedDeletes.clear();
+
synchronized(this) {
if (pendingCommit != null) {
@@ -1993,7 +1927,6 @@ public class IndexWriter implements Clos
} finally {
synchronized(this) {
if (!success) {
- docWriter.resumeAllThreads();
closing = false;
notifyAll();
if (infoStream != null)
@@ -2021,7 +1954,6 @@ public class IndexWriter implements Clos
* will receive {@link MergePolicy.MergeAbortedException}s.
*/
public synchronized void deleteAll() throws IOException {
- docWriter.pauseAllThreads();
try {
// Abort any running merges
@@ -2029,7 +1961,6 @@ public class IndexWriter implements Clos
// Remove any buffered docs
docWriter.abort();
- docWriter.setFlushedDocCount(0);
// Remove all segments
segmentInfos.clear();
@@ -2047,7 +1978,6 @@ public class IndexWriter implements Clos
} catch (OutOfMemoryError oom) {
handleOOM(oom, "deleteAll");
} finally {
- docWriter.resumeAllThreads();
if (infoStream != null) {
message("hit exception during deleteAll");
}
@@ -2123,7 +2053,7 @@ public class IndexWriter implements Clos
* the index files referenced exist (correctly) in the
* index directory.
*/
- private synchronized void checkpoint() throws IOException {
+ synchronized void checkpoint() throws IOException {
changeCount++;
segmentInfos.changed();
deleter.checkpoint(segmentInfos, false);
@@ -2259,9 +2189,6 @@ public class IndexWriter implements Clos
synchronized (this) {
ensureOpen();
segmentInfos.addAll(infos);
- // Notify DocumentsWriter that the flushed count just increased
- docWriter.updateFlushedDocCount(docCount);
-
checkpoint();
}
@@ -2324,10 +2251,6 @@ public class IndexWriter implements Clos
// Register the new segment
synchronized(this) {
segmentInfos.add(info);
-
- // Notify DocumentsWriter that the flushed count just increased
- docWriter.updateFlushedDocCount(docCount);
-
checkpoint();
}
} catch (OutOfMemoryError oom) {
@@ -2535,196 +2458,92 @@ public class IndexWriter implements Clos
// We can be called during close, when closing==true, so we must pass false to ensureOpen:
ensureOpen(false);
- if (doFlush(flushDocStores, flushDeletes) && triggerMerge)
+ if (doFlush(flushDocStores, flushDeletes) && triggerMerge) {
maybeMerge();
- }
-
- // TODO: this method should not have to be entirely
- // synchronized, ie, merges should be allowed to commit
- // even while a flush is happening
- private synchronized final boolean doFlush(boolean flushDocStores, boolean flushDeletes) throws CorruptIndexException, IOException {
- try {
- try {
- return doFlushInternal(flushDocStores, flushDeletes);
- } finally {
- docWriter.balanceRAM();
- }
- } finally {
- docWriter.clearFlushPending();
}
}
// TODO: this method should not have to be entirely
// synchronized, ie, merges should be allowed to commit
// even while a flush is happening
- private synchronized final boolean doFlushInternal(boolean flushDocStores, boolean flushDeletes) throws CorruptIndexException, IOException {
+ private synchronized final boolean doFlush(boolean closeDocStores, boolean applyAllDeletes) throws CorruptIndexException, IOException {
if (hitOOM) {
throw new IllegalStateException("this writer hit an OutOfMemoryError; cannot flush");
}
- ensureOpen(false);
+ doBeforeFlush();
assert testPoint("startDoFlush");
- doBeforeFlush();
-
- flushCount++;
+ // We may be flushing because it was triggered by doc
+ // count, del count, ram usage (in which case flush
+ // pending is already set), or we may be flushing
+ // due to external event eg getReader or commit is
+ // called (in which case we now set it, and this will
+ // pause all threads):
+ flushControl.setFlushPendingNoWait("explicit flush");
- // If we are flushing because too many deletes
- // accumulated, then we should apply the deletes to free
- // RAM:
- flushDeletes |= docWriter.doApplyDeletes();
-
- // Make sure no threads are actively adding a document.
- // Returns true if docWriter is currently aborting, in
- // which case we skip flushing this segment
- if (infoStream != null) {
- message("flush: now pause all indexing threads");
- }
- if (docWriter.pauseAllThreads()) {
- docWriter.resumeAllThreads();
- return false;
- }
+ boolean success = false;
try {
- SegmentInfo newSegment = null;
-
- final int numDocs = docWriter.getNumDocsInRAM();
-
- // Always flush docs if there are any
- boolean flushDocs = numDocs > 0;
-
- String docStoreSegment = docWriter.getDocStoreSegment();
-
- assert docStoreSegment != null || numDocs == 0: "dss=" + docStoreSegment + " numDocs=" + numDocs;
-
- if (docStoreSegment == null)
- flushDocStores = false;
-
- int docStoreOffset = docWriter.getDocStoreOffset();
-
- boolean docStoreIsCompoundFile = false;
-
if (infoStream != null) {
- message(" flush: segment=" + docWriter.getSegment() +
- " docStoreSegment=" + docWriter.getDocStoreSegment() +
- " docStoreOffset=" + docStoreOffset +
- " flushDocs=" + flushDocs +
- " flushDeletes=" + flushDeletes +
- " flushDocStores=" + flushDocStores +
- " numDocs=" + numDocs +
- " numBufDelTerms=" + docWriter.getNumBufferedDeleteTerms());
+ message(" start flush: applyAllDeletes=" + applyAllDeletes + " closeDocStores=" + closeDocStores);
message(" index before flush " + segString());
}
-
- // Check if the doc stores must be separately flushed
- // because other segments, besides the one we are about
- // to flush, reference it
- if (flushDocStores && (!flushDocs || !docWriter.getSegment().equals(docWriter.getDocStoreSegment()))) {
- // We must separately flush the doc store
- if (infoStream != null)
- message(" flush shared docStore segment " + docStoreSegment);
-
- docStoreIsCompoundFile = flushDocStores();
- flushDocStores = false;
- }
-
- String segment = docWriter.getSegment();
-
- // If we are flushing docs, segment must not be null:
- assert segment != null || !flushDocs;
-
- if (flushDocs) {
-
- boolean success = false;
- final int flushedDocCount;
-
- try {
- flushedDocCount = docWriter.flush(flushDocStores);
- if (infoStream != null) {
- message("flushedFiles=" + docWriter.getFlushedFiles());
- }
- success = true;
- } finally {
- if (!success) {
- if (infoStream != null)
- message("hit exception flushing segment " + segment);
- deleter.refresh(segment);
- }
- }
-
- if (0 == docStoreOffset && flushDocStores) {
- // This means we are flushing private doc stores
- // with this segment, so it will not be shared
- // with other segments
- assert docStoreSegment != null;
- assert docStoreSegment.equals(segment);
- docStoreOffset = -1;
- docStoreIsCompoundFile = false;
- docStoreSegment = null;
- }
-
- // Create new SegmentInfo, but do not add to our
- // segmentInfos until deletes are flushed
- // successfully.
- newSegment = new SegmentInfo(segment,
- flushedDocCount,
- directory, false, docStoreOffset,
- docStoreSegment, docStoreIsCompoundFile,
- docWriter.hasProx(),
- docWriter.getSegmentCodecs());
-
- if (infoStream != null) {
- message("flush codec=" + docWriter.getSegmentCodecs());
- }
+
+ final SegmentInfo newSegment = docWriter.flush(this, closeDocStores, deleter, mergePolicy, segmentInfos);
+ if (newSegment != null) {
setDiagnostics(newSegment, "flush");
- }
-
- docWriter.pushDeletes();
-
- if (flushDocs) {
segmentInfos.add(newSegment);
checkpoint();
}
- if (flushDocs && mergePolicy.useCompoundFile(segmentInfos, newSegment)) {
- // Now build compound file
- boolean success = false;
- try {
- docWriter.createCompoundFile(segment);
- success = true;
- } finally {
- if (!success) {
- if (infoStream != null)
- message("hit exception creating compound file for newly flushed segment " + segment);
- deleter.deleteFile(IndexFileNames.segmentFileName(segment, "", IndexFileNames.COMPOUND_FILE_EXTENSION));
+ if (!applyAllDeletes) {
+ // If deletes alone are consuming > 1/2 our RAM
+ // buffer, force them all to apply now. This is to
+ // prevent too-frequent flushing of a long tail of
+ // tiny segments:
+ if (flushControl.getFlushDeletes() ||
+ (config.getRAMBufferSizeMB() != IndexWriterConfig.DISABLE_AUTO_FLUSH &&
+ bufferedDeletes.bytesUsed() > (1024*1024*config.getRAMBufferSizeMB()/2))) {
+ applyAllDeletes = true;
+ if (infoStream != null) {
+ message("force apply deletes bytesUsed=" + bufferedDeletes.bytesUsed() + " vs ramBuffer=" + (1024*1024*config.getRAMBufferSizeMB()));
}
}
-
- newSegment.setUseCompoundFile(true);
- checkpoint();
}
- if (flushDeletes) {
- applyDeletes();
+ if (applyAllDeletes) {
+ if (infoStream != null) {
+ message("apply all deletes during flush");
+ }
+ flushDeletesCount.incrementAndGet();
+ if (bufferedDeletes.applyDeletes(readerPool, segmentInfos, segmentInfos)) {
+ checkpoint();
+ }
+ flushControl.clearDeletes();
+ } else if (infoStream != null) {
+ message("don't apply deletes now delTermCount=" + bufferedDeletes.numTerms() + " bytesUsed=" + bufferedDeletes.bytesUsed());
}
-
- if (flushDocs)
- checkpoint();
doAfterFlush();
+ flushCount.incrementAndGet();
+
+ success = true;
- return flushDocs;
+ return newSegment != null;
} catch (OutOfMemoryError oom) {
handleOOM(oom, "doFlush");
// never hit
return false;
} finally {
- docWriter.clearFlushPending();
- docWriter.resumeAllThreads();
+ flushControl.clearFlushPending();
+ if (!success && infoStream != null) {
+ message("hit exception during flush");
+ }
}
}
@@ -2733,7 +2552,7 @@ public class IndexWriter implements Clos
*/
public final long ramSizeInBytes() {
ensureOpen();
- return docWriter.getRAMUsed();
+ return docWriter.bytesUsed() + bufferedDeletes.bytesUsed();
}
/** Expert: Return the number of documents currently
@@ -2776,7 +2595,7 @@ public class IndexWriter implements Clos
* saves the resulting deletes file (incrementing the
* delete generation for merge.info). If no deletes were
* flushed, no new deletes file is saved. */
- synchronized private void commitMergedDeletes(MergePolicy.OneMerge merge, SegmentReader mergeReader) throws IOException {
+ synchronized private void commitMergedDeletes(MergePolicy.OneMerge merge, SegmentReader mergedReader) throws IOException {
assert testPoint("startCommitMergeDeletes");
@@ -2815,7 +2634,7 @@ public class IndexWriter implements Clos
assert currentDelDocs.get(j);
else {
if (currentDelDocs.get(j)) {
- mergeReader.doDelete(docUpto);
+ mergedReader.doDelete(docUpto);
delCount++;
}
docUpto++;
@@ -2829,7 +2648,7 @@ public class IndexWriter implements Clos
// does:
for(int j=0; j<docCount; j++) {
if (currentDelDocs.get(j)) {
- mergeReader.doDelete(docUpto);
+ mergedReader.doDelete(docUpto);
delCount++;
}
docUpto++;
@@ -2839,13 +2658,13 @@ public class IndexWriter implements Clos
docUpto += info.docCount;
}
- assert mergeReader.numDeletedDocs() == delCount;
+ assert mergedReader.numDeletedDocs() == delCount;
- mergeReader.hasChanges = delCount > 0;
+ mergedReader.hasChanges = delCount > 0;
}
/* FIXME if we want to support non-contiguous segment merges */
- synchronized private boolean commitMerge(MergePolicy.OneMerge merge, SegmentMerger merger, int mergedDocCount, SegmentReader mergedReader) throws IOException {
+ synchronized private boolean commitMerge(MergePolicy.OneMerge merge, SegmentMerger merger, SegmentReader mergedReader) throws IOException {
assert testPoint("startCommitMerge");
@@ -2873,7 +2692,6 @@ public class IndexWriter implements Clos
final int start = ensureContiguousMerge(merge);
commitMergedDeletes(merge, mergedReader);
- docWriter.remapDeletes(segmentInfos, merger.getDocMaps(), merger.getDelCounts(), merge, mergedDocCount);
// If the doc store we are using has been closed and
// is in now compound format (but wasn't when we
@@ -2886,7 +2704,7 @@ public class IndexWriter implements Clos
segmentInfos.subList(start, start + merge.segments.size()).clear();
assert !segmentInfos.contains(merge.info);
segmentInfos.add(start, merge.info);
-
+
closeMergeReaders(merge, false);
// Must note the change to segmentInfos so any commits
@@ -2897,6 +2715,12 @@ public class IndexWriter implements Clos
// them so that they don't bother writing them to
// disk, updating SegmentInfo, etc.:
readerPool.clear(merge.segments);
+
+ // remove pending deletes of the segments
+ // that were merged, moving them onto the segment just
+ // before the merged segment
+ // Lock order: IW -> BD
+ bufferedDeletes.commitMerge(merge);
if (merge.optimize) {
// cascade the optimize:
@@ -3056,10 +2880,17 @@ public class IndexWriter implements Clos
final synchronized void mergeInit(MergePolicy.OneMerge merge) throws IOException {
boolean success = false;
try {
+ // Lock order: IW -> BD
+ if (bufferedDeletes.applyDeletes(readerPool, segmentInfos, merge.segments)) {
+ checkpoint();
+ }
_mergeInit(merge);
success = true;
} finally {
if (!success) {
+ if (infoStream != null) {
+ message("hit exception in mergeInit");
+ }
mergeFinish(merge);
}
}
@@ -3082,9 +2913,7 @@ public class IndexWriter implements Clos
if (merge.isAborted())
return;
-
- applyDeletes();
-
+
final SegmentInfos sourceSegments = merge.segments;
final int end = sourceSegments.size();
@@ -3274,10 +3103,11 @@ public class IndexWriter implements Clos
if (suppressExceptions) {
// Suppress any new exceptions so we throw the
// original cause
+ boolean anyChanges = false;
for (int i=0;i<numSegments;i++) {
if (merge.readers[i] != null) {
try {
- readerPool.release(merge.readers[i], false);
+ anyChanges |= readerPool.release(merge.readers[i], false);
} catch (Throwable t) {
}
merge.readers[i] = null;
@@ -3294,6 +3124,9 @@ public class IndexWriter implements Clos
merge.readersClone[i] = null;
}
}
+ if (anyChanges) {
+ checkpoint();
+ }
} else {
for (int i=0;i<numSegments;i++) {
if (merge.readers[i] != null) {
@@ -3521,15 +3354,21 @@ public class IndexWriter implements Clos
mergedSegmentWarmer.warm(mergedReader);
}
- if (!commitMerge(merge, merger, mergedDocCount, mergedReader)) {
+ if (!commitMerge(merge, merger, mergedReader)) {
// commitMerge will return false if this merge was aborted
return 0;
}
} finally {
synchronized(this) {
- readerPool.release(mergedReader);
+ if (readerPool.release(mergedReader)) {
+ // Must checkpoint after releasing the
+ // mergedReader since it may have written a new
+ // deletes file:
+ checkpoint();
+ }
}
}
+
success = true;
} finally {
@@ -3549,37 +3388,14 @@ public class IndexWriter implements Clos
mergeExceptions.add(merge);
}
- // Apply buffered deletes to all segments.
- private final synchronized boolean applyDeletes() throws CorruptIndexException, IOException {
- assert testPoint("startApplyDeletes");
- if (infoStream != null) {
- message("applyDeletes");
- }
- flushDeletesCount++;
- boolean success = false;
- boolean changed;
- try {
- changed = docWriter.applyDeletes(segmentInfos);
- success = true;
- } finally {
- if (!success && infoStream != null) {
- message("hit exception flushing deletes");
- }
- }
-
- if (changed)
- checkpoint();
- return changed;
- }
-
// For test purposes.
- final synchronized int getBufferedDeleteTermsSize() {
- return docWriter.getBufferedDeleteTerms().size();
+ final int getBufferedDeleteTermsSize() {
+ return docWriter.getPendingDeletes().terms.size();
}
// For test purposes.
- final synchronized int getNumBufferedDeleteTerms() {
- return docWriter.getNumBufferedDeleteTerms();
+ final int getNumBufferedDeleteTerms() {
+ return docWriter.getPendingDeletes().numTermDeletes.get();
}
// utility routines for tests
@@ -3819,14 +3635,13 @@ public class IndexWriter implements Clos
// finishStartCommit
// startCommitMergeDeletes
// startMergeInit
- // startApplyDeletes
// DocumentsWriter.ThreadState.init start
boolean testPoint(String name) {
return true;
}
synchronized boolean nrtIsCurrent(SegmentInfos infos) {
- return infos.version == segmentInfos.version && !docWriter.anyChanges();
+ return infos.version == segmentInfos.version && !docWriter.anyChanges() && !bufferedDeletes.any();
}
synchronized boolean isClosed() {
@@ -3863,7 +3678,6 @@ public class IndexWriter implements Clos
deleter.revisitPolicy();
}
-
/**
* Sets the {@link PayloadProcessorProvider} to use when merging payloads.
* Note that the given <code>pcp</code> will be invoked for every segment that
@@ -3894,4 +3708,123 @@ public class IndexWriter implements Clos
return payloadProcessorProvider;
}
+ // decides when flushes happen
+ final class FlushControl {
+
+ private boolean flushPending;
+ private boolean flushDeletes;
+ private int delCount;
+ private int docCount;
+ private boolean flushing;
+
+ private synchronized boolean setFlushPending(String reason, boolean doWait) {
+ if (flushPending || flushing) {
+ if (doWait) {
+ while(flushPending || flushing) {
+ try {
+ wait();
+ } catch (InterruptedException ie) {
+ throw new ThreadInterruptedException(ie);
+ }
+ }
+ }
+ return false;
+ } else {
+ if (infoStream != null) {
+ message("now trigger flush reason=" + reason);
+ }
+ flushPending = true;
+ return flushPending;
+ }
+ }
+
+ public synchronized void setFlushPendingNoWait(String reason) {
+ setFlushPending(reason, false);
+ }
+
+ public synchronized boolean getFlushPending() {
+ return flushPending;
+ }
+
+ public synchronized boolean getFlushDeletes() {
+ return flushDeletes;
+ }
+
+ public synchronized void clearFlushPending() {
+ if (infoStream != null) {
+ message("clearFlushPending");
+ }
+ flushPending = false;
+ flushDeletes = false;
+ docCount = 0;
+ notifyAll();
+ }
+
+ public synchronized void clearDeletes() {
+ delCount = 0;
+ }
+
+ public synchronized boolean waitUpdate(int docInc, int delInc) {
+ return waitUpdate(docInc, delInc, false);
+ }
+
+ public synchronized boolean waitUpdate(int docInc, int delInc, boolean skipWait) {
+ while(flushPending) {
+ try {
+ wait();
+ } catch (InterruptedException ie) {
+ throw new ThreadInterruptedException(ie);
+ }
+ }
+
+ // skipWait is only used when a thread is BOTH adding
+ // a doc and buffering a del term, and, the adding of
+ // the doc already triggered a flush
+ if (skipWait) {
+ docCount += docInc;
+ delCount += delInc;
+ return false;
+ }
+
+ final int maxBufferedDocs = config.getMaxBufferedDocs();
+ if (maxBufferedDocs != IndexWriterConfig.DISABLE_AUTO_FLUSH &&
+ (docCount+docInc) >= maxBufferedDocs) {
+ return setFlushPending("maxBufferedDocs", true);
+ }
+ docCount += docInc;
+
+ final int maxBufferedDeleteTerms = config.getMaxBufferedDeleteTerms();
+ if (maxBufferedDeleteTerms != IndexWriterConfig.DISABLE_AUTO_FLUSH &&
+ (delCount+delInc) >= maxBufferedDeleteTerms) {
+ flushDeletes = true;
+ return setFlushPending("maxBufferedDeleteTerms", true);
+ }
+ delCount += delInc;
+
+ return flushByRAMUsage("add delete/doc");
+ }
+
+ public synchronized boolean flushByRAMUsage(String reason) {
+ final double ramBufferSizeMB = config.getRAMBufferSizeMB();
+ if (ramBufferSizeMB != IndexWriterConfig.DISABLE_AUTO_FLUSH) {
+ final long limit = (long) (ramBufferSizeMB*1024*1024);
+ long used = bufferedDeletes.bytesUsed() + docWriter.bytesUsed();
+ if (used >= limit) {
+
+ // DocumentsWriter may be able to free up some
+ // RAM:
+ // Lock order: FC -> DW
+ docWriter.balanceRAM();
+
+ used = bufferedDeletes.bytesUsed() + docWriter.bytesUsed();
+ if (used >= limit) {
+ return setFlushPending("ram full: " + reason, false);
+ }
+ }
+ }
+ return false;
+ }
+ }
+
+ final FlushControl flushControl = new FlushControl();
}
Modified: lucene/dev/trunk/lucene/src/java/org/apache/lucene/index/ParallelPostingsArray.java
URL: http://svn.apache.org/viewvc/lucene/dev/trunk/lucene/src/java/org/apache/lucene/index/ParallelPostingsArray.java?rev=1044635&r1=1044634&r2=1044635&view=diff
==============================================================================
--- lucene/dev/trunk/lucene/src/java/org/apache/lucene/index/ParallelPostingsArray.java (original)
+++ lucene/dev/trunk/lucene/src/java/org/apache/lucene/index/ParallelPostingsArray.java Sat Dec 11 11:07:01 2010
@@ -1,7 +1,5 @@
package org.apache.lucene.index;
-import org.apache.lucene.util.ArrayUtil;
-
/**
* Licensed to the Apache Software Foundation (ASF) under one or more
* contributor license agreements. See the NOTICE file distributed with
@@ -19,9 +17,11 @@ import org.apache.lucene.util.ArrayUtil;
* limitations under the License.
*/
+import org.apache.lucene.util.ArrayUtil;
+import org.apache.lucene.util.RamUsageEstimator;
class ParallelPostingsArray {
- final static int BYTES_PER_POSTING = 3 * DocumentsWriter.INT_NUM_BYTE;
+ final static int BYTES_PER_POSTING = 3 * RamUsageEstimator.NUM_BYTES_INT;
final int size;
final int[] textStarts;
Added: lucene/dev/trunk/lucene/src/java/org/apache/lucene/index/SegmentDeletes.java
URL: http://svn.apache.org/viewvc/lucene/dev/trunk/lucene/src/java/org/apache/lucene/index/SegmentDeletes.java?rev=1044635&view=auto
==============================================================================
--- lucene/dev/trunk/lucene/src/java/org/apache/lucene/index/SegmentDeletes.java (added)
+++ lucene/dev/trunk/lucene/src/java/org/apache/lucene/index/SegmentDeletes.java Sat Dec 11 11:07:01 2010
@@ -0,0 +1,188 @@
+package org.apache.lucene.index;
+
+/**
+ * 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.
+ */
+
+import java.util.ArrayList;
+import java.util.HashMap;
+import java.util.List;
+import java.util.Map;
+import java.util.SortedMap;
+import java.util.TreeMap;
+import java.util.concurrent.atomic.AtomicLong;
+import java.util.concurrent.atomic.AtomicInteger;
+
+import org.apache.lucene.search.Query;
+import org.apache.lucene.util.RamUsageEstimator;
+
+/** Holds buffered deletes, by docID, term or query for a
+ * single segment. This is used to hold buffered pending
+ * deletes against the to-be-flushed segment as well as
+ * per-segment deletes for each segment in the index. */
+
+// NOTE: we are sync'd by BufferedDeletes, ie, all access to
+// instances of this class is via sync'd methods on
+// BufferedDeletes
+class SegmentDeletes {
+
+ /* Rough logic: HashMap has an array[Entry] w/ varying
+ load factor (say 2 * POINTER). Entry is object w/ Term
+ key, Integer val, int hash, Entry next
+ (OBJ_HEADER + 3*POINTER + INT). Term is object w/
+ String field and String text (OBJ_HEADER + 2*POINTER).
+ We don't count Term's field since it's interned.
+ Term's text is String (OBJ_HEADER + 4*INT + POINTER +
+ OBJ_HEADER + string.length*CHAR). Integer is
+ OBJ_HEADER + INT. */
+ final static int BYTES_PER_DEL_TERM = 8*RamUsageEstimator.NUM_BYTES_OBJ_REF + 5*RamUsageEstimator.NUM_BYTES_OBJ_HEADER + 6*RamUsageEstimator.NUM_BYTES_INT;
+
+ /* Rough logic: del docIDs are List<Integer>. Say list
+ allocates ~2X size (2*POINTER). Integer is OBJ_HEADER
+ + int */
+ final static int BYTES_PER_DEL_DOCID = 2*RamUsageEstimator.NUM_BYTES_OBJ_REF + RamUsageEstimator.NUM_BYTES_OBJ_HEADER + RamUsageEstimator.NUM_BYTES_INT;
+
+ /* Rough logic: HashMap has an array[Entry] w/ varying
+ load factor (say 2 * POINTER). Entry is object w/
+ Query key, Integer val, int hash, Entry next
+ (OBJ_HEADER + 3*POINTER + INT). Query we often
+ undercount (say 24 bytes). Integer is OBJ_HEADER + INT. */
+ final static int BYTES_PER_DEL_QUERY = 5*RamUsageEstimator.NUM_BYTES_OBJ_REF + 2*RamUsageEstimator.NUM_BYTES_OBJ_HEADER + 2*RamUsageEstimator.NUM_BYTES_INT + 24;
+
+ // TODO: many of the deletes stored here will map to
+ // Integer.MAX_VALUE; we could be more efficient for this
+ // case ie use a SortedSet not a SortedMap. But: Java's
+ // SortedSet impls are simply backed by a Map so we won't
+ // save anything unless we do something custom...
+ final AtomicInteger numTermDeletes = new AtomicInteger();
+ final SortedMap<Term,Integer> terms = new TreeMap<Term,Integer>();
+ final Map<Query,Integer> queries = new HashMap<Query,Integer>();
+ final List<Integer> docIDs = new ArrayList<Integer>();
+
+ public static final Integer MAX_INT = Integer.valueOf(Integer.MAX_VALUE);
+
+ final AtomicLong bytesUsed = new AtomicLong();
+
+ private final static boolean VERBOSE_DELETES = false;
+
+ @Override
+ public String toString() {
+ if (VERBOSE_DELETES) {
+ return "SegmentDeletes [numTerms=" + numTermDeletes + ", terms=" + terms
+ + ", queries=" + queries + ", docIDs=" + docIDs + ", bytesUsed="
+ + bytesUsed + "]";
+ } else {
+ String s = "";
+ if (numTermDeletes.get() != 0) {
+ s += " " + numTermDeletes.get() + " deleted terms (unique count=" + terms.size() + ")";
+ }
+ if (queries.size() != 0) {
+ s += " " + queries.size() + " deleted queries";
+ }
+ if (docIDs.size() != 0) {
+ s += " " + docIDs.size() + " deleted docIDs";
+ }
+ if (bytesUsed.get() != 0) {
+ s += " bytesUsed=" + bytesUsed.get();
+ }
+
+ return s;
+ }
+ }
+
+ void update(SegmentDeletes in, boolean noLimit) {
+ numTermDeletes.addAndGet(in.numTermDeletes.get());
+ for (Map.Entry<Term,Integer> ent : in.terms.entrySet()) {
+ final Term term = ent.getKey();
+ if (!terms.containsKey(term)) {
+ // only incr bytesUsed if this term wasn't already buffered:
+ bytesUsed.addAndGet(BYTES_PER_DEL_TERM);
+ }
+ final Integer limit;
+ if (noLimit) {
+ limit = MAX_INT;
+ } else {
+ limit = ent.getValue();
+ }
+ terms.put(term, limit);
+ }
+
+ for (Map.Entry<Query,Integer> ent : in.queries.entrySet()) {
+ final Query query = ent.getKey();
+ if (!queries.containsKey(query)) {
+ // only incr bytesUsed if this query wasn't already buffered:
+ bytesUsed.addAndGet(BYTES_PER_DEL_QUERY);
+ }
+ final Integer limit;
+ if (noLimit) {
+ limit = MAX_INT;
+ } else {
+ limit = ent.getValue();
+ }
+ queries.put(query, limit);
+ }
+
+ // docIDs never move across segments and the docIDs
+ // should already be cleared
+ }
+
+ public void addQuery(Query query, int docIDUpto) {
+ queries.put(query, docIDUpto);
+ bytesUsed.addAndGet(BYTES_PER_DEL_QUERY);
+ }
+
+ public void addDocID(int docID) {
+ docIDs.add(Integer.valueOf(docID));
+ bytesUsed.addAndGet(BYTES_PER_DEL_DOCID);
+ }
+
+ public void addTerm(Term term, int docIDUpto) {
+ Integer current = terms.get(term);
+ if (current != null && docIDUpto < current) {
+ // Only record the new number if it's greater than the
+ // current one. This is important because if multiple
+ // threads are replacing the same doc at nearly the
+ // same time, it's possible that one thread that got a
+ // higher docID is scheduled before the other
+ // threads. If we blindly replace than we can get
+ // double-doc in the segment.
+ return;
+ }
+
+ terms.put(term, Integer.valueOf(docIDUpto));
+ numTermDeletes.incrementAndGet();
+ if (current == null) {
+ bytesUsed.addAndGet(BYTES_PER_DEL_TERM + term.bytes.length);
+ }
+ }
+
+ void clear() {
+ terms.clear();
+ queries.clear();
+ docIDs.clear();
+ numTermDeletes.set(0);
+ bytesUsed.set(0);
+ }
+
+ void clearDocIDs() {
+ bytesUsed.addAndGet(-docIDs.size()*BYTES_PER_DEL_DOCID);
+ docIDs.clear();
+ }
+
+ boolean any() {
+ return terms.size() > 0 || docIDs.size() > 0 || queries.size() > 0;
+ }
+}
Modified: lucene/dev/trunk/lucene/src/java/org/apache/lucene/index/SegmentInfo.java
URL: http://svn.apache.org/viewvc/lucene/dev/trunk/lucene/src/java/org/apache/lucene/index/SegmentInfo.java?rev=1044635&r1=1044634&r2=1044635&view=diff
==============================================================================
--- lucene/dev/trunk/lucene/src/java/org/apache/lucene/index/SegmentInfo.java (original)
+++ lucene/dev/trunk/lucene/src/java/org/apache/lucene/index/SegmentInfo.java Sat Dec 11 11:07:01 2010
@@ -361,6 +361,10 @@ public final class SegmentInfo {
return docStoreSegment;
}
+ public void setDocStoreSegment(String segment) {
+ docStoreSegment = segment;
+ }
+
void setDocStoreOffset(int offset) {
docStoreOffset = offset;
clearFiles();
@@ -534,6 +538,12 @@ public final class SegmentInfo {
if (docStoreOffset != -1) {
s.append("->").append(docStoreSegment);
+ if (docStoreIsCompoundFile) {
+ s.append('c');
+ } else {
+ s.append('C');
+ }
+ s.append('+').append(docStoreOffset);
}
return s.toString();
Modified: lucene/dev/trunk/lucene/src/java/org/apache/lucene/index/TermVectorsTermsWriterPerField.java
URL: http://svn.apache.org/viewvc/lucene/dev/trunk/lucene/src/java/org/apache/lucene/index/TermVectorsTermsWriterPerField.java?rev=1044635&r1=1044634&r2=1044635&view=diff
==============================================================================
--- lucene/dev/trunk/lucene/src/java/org/apache/lucene/index/TermVectorsTermsWriterPerField.java (original)
+++ lucene/dev/trunk/lucene/src/java/org/apache/lucene/index/TermVectorsTermsWriterPerField.java Sat Dec 11 11:07:01 2010
@@ -24,6 +24,7 @@ import org.apache.lucene.document.Fielda
import org.apache.lucene.store.IndexOutput;
import org.apache.lucene.util.ByteBlockPool;
import org.apache.lucene.util.BytesRef;
+import org.apache.lucene.util.RamUsageEstimator;
final class TermVectorsTermsWriterPerField extends TermsHashConsumerPerField {
@@ -298,7 +299,7 @@ final class TermVectorsTermsWriterPerFie
@Override
int bytesPerPosting() {
- return super.bytesPerPosting() + 3 * DocumentsWriter.INT_NUM_BYTE;
+ return super.bytesPerPosting() + 3 * RamUsageEstimator.NUM_BYTES_INT;
}
}
}
Modified: lucene/dev/trunk/lucene/src/test/org/apache/lucene/TestDemo.java
URL: http://svn.apache.org/viewvc/lucene/dev/trunk/lucene/src/test/org/apache/lucene/TestDemo.java?rev=1044635&r1=1044634&r2=1044635&view=diff
==============================================================================
--- lucene/dev/trunk/lucene/src/test/org/apache/lucene/TestDemo.java (original)
+++ lucene/dev/trunk/lucene/src/test/org/apache/lucene/TestDemo.java Sat Dec 11 11:07:01 2010
@@ -50,6 +50,7 @@ public class TestDemo extends LuceneTest
// To store an index on disk, use this instead:
//Directory directory = FSDirectory.open("/tmp/testindex");
RandomIndexWriter iwriter = new RandomIndexWriter(random, directory);
+ iwriter.w.setInfoStream(VERBOSE ? System.out : null);
Document doc = new Document();
String longTerm = "longtermlongtermlongtermlongtermlongtermlongtermlongtermlongtermlongtermlongtermlongtermlongtermlongtermlongtermlongtermlongtermlongtermlongterm";
String text = "This is the text to be indexed. " + longTerm;
Modified: lucene/dev/trunk/lucene/src/test/org/apache/lucene/TestSearchForDuplicates.java
URL: http://svn.apache.org/viewvc/lucene/dev/trunk/lucene/src/test/org/apache/lucene/TestSearchForDuplicates.java?rev=1044635&r1=1044634&r2=1044635&view=diff
==============================================================================
--- lucene/dev/trunk/lucene/src/test/org/apache/lucene/TestSearchForDuplicates.java (original)
+++ lucene/dev/trunk/lucene/src/test/org/apache/lucene/TestSearchForDuplicates.java Sat Dec 11 11:07:01 2010
@@ -84,6 +84,10 @@ public class TestSearchForDuplicates ext
lmp.setUseCompoundFile(useCompoundFiles);
lmp.setUseCompoundDocStore(useCompoundFiles);
IndexWriter writer = new IndexWriter(directory, conf);
+ if (VERBOSE) {
+ System.out.println("TEST: now build index");
+ writer.setInfoStream(System.out);
+ }
final int MAX_DOCS = 225;
Modified: lucene/dev/trunk/lucene/src/test/org/apache/lucene/index/TestConcurrentMergeScheduler.java
URL: http://svn.apache.org/viewvc/lucene/dev/trunk/lucene/src/test/org/apache/lucene/index/TestConcurrentMergeScheduler.java?rev=1044635&r1=1044634&r2=1044635&view=diff
==============================================================================
--- lucene/dev/trunk/lucene/src/test/org/apache/lucene/index/TestConcurrentMergeScheduler.java (original)
+++ lucene/dev/trunk/lucene/src/test/org/apache/lucene/index/TestConcurrentMergeScheduler.java Sat Dec 11 11:07:01 2010
@@ -57,9 +57,9 @@ public class TestConcurrentMergeSchedule
isClose = true;
}
}
- if (isDoFlush && !isClose) {
+ if (isDoFlush && !isClose && random.nextBoolean()) {
hitExc = true;
- throw new IOException("now failing during flush");
+ throw new IOException(Thread.currentThread().getName() + ": now failing during flush");
}
}
}
@@ -73,12 +73,17 @@ public class TestConcurrentMergeSchedule
directory.failOn(failure);
IndexWriter writer = new IndexWriter(directory, newIndexWriterConfig(TEST_VERSION_CURRENT, new MockAnalyzer()).setMaxBufferedDocs(2));
+ writer.setInfoStream(VERBOSE ? System.out : null);
Document doc = new Document();
Field idField = newField("id", "", Field.Store.YES, Field.Index.NOT_ANALYZED);
doc.add(idField);
int extraCount = 0;
for(int i=0;i<10;i++) {
+ if (VERBOSE) {
+ System.out.println("TEST: iter=" + i);
+ }
+
for(int j=0;j<20;j++) {
idField.setValue(Integer.toString(i*20+j));
writer.addDocument(doc);
@@ -97,10 +102,14 @@ public class TestConcurrentMergeSchedule
}
extraCount++;
} catch (IOException ioe) {
+ if (VERBOSE) {
+ ioe.printStackTrace(System.out);
+ }
failure.clearDoFail();
break;
}
}
+ assertEquals(20*(i+1)+extraCount, writer.numDocs());
}
writer.close();
@@ -155,8 +164,12 @@ public class TestConcurrentMergeSchedule
IndexWriter writer = new IndexWriter(directory, newIndexWriterConfig(
TEST_VERSION_CURRENT, new MockAnalyzer())
.setMaxBufferedDocs(2));
+ writer.setInfoStream(VERBOSE ? System.out : null);
for(int iter=0;iter<7;iter++) {
+ if (VERBOSE) {
+ System.out.println("TEST: iter=" + iter);
+ }
for(int j=0;j<21;j++) {
Document doc = new Document();
Modified: lucene/dev/trunk/lucene/src/test/org/apache/lucene/index/TestIndexWriter.java
URL: http://svn.apache.org/viewvc/lucene/dev/trunk/lucene/src/test/org/apache/lucene/index/TestIndexWriter.java?rev=1044635&r1=1044634&r2=1044635&view=diff
==============================================================================
--- lucene/dev/trunk/lucene/src/test/org/apache/lucene/index/TestIndexWriter.java (original)
+++ lucene/dev/trunk/lucene/src/test/org/apache/lucene/index/TestIndexWriter.java Sat Dec 11 11:07:01 2010
@@ -1083,6 +1083,9 @@ public class TestIndexWriter extends Luc
doc.add(idField);
for(int pass=0;pass<2;pass++) {
+ if (VERBOSE) {
+ System.out.println("TEST: pass=" + pass);
+ }
IndexWriter writer = new IndexWriter(
directory,
@@ -1094,10 +1097,12 @@ public class TestIndexWriter extends Luc
// backed directory:
setMergePolicy(newLogMergePolicy(false, 10))
);
+ writer.setInfoStream(VERBOSE ? System.out : null);
- //System.out.println("TEST: pass=" + pass + " cms=" + (pass >= 2));
for(int iter=0;iter<10;iter++) {
- //System.out.println("TEST: iter=" + iter);
+ if (VERBOSE) {
+ System.out.println("TEST: iter=" + iter);
+ }
for(int j=0;j<199;j++) {
idField.setValue(Integer.toString(iter*201+j));
writer.addDocument(doc);
@@ -1142,8 +1147,9 @@ public class TestIndexWriter extends Luc
}
};
- if (failure.size() > 0)
+ if (failure.size() > 0) {
throw failure.get(0);
+ }
t1.start();
@@ -1156,6 +1162,7 @@ public class TestIndexWriter extends Luc
// Reopen
writer = new IndexWriter(directory, newIndexWriterConfig( TEST_VERSION_CURRENT, new MockAnalyzer()).setOpenMode(OpenMode.APPEND));
+ writer.setInfoStream(VERBOSE ? System.out : null);
}
writer.close();
}
@@ -2575,7 +2582,7 @@ public class TestIndexWriter extends Luc
Directory dir = newDirectory();
FlushCountingIndexWriter w = new FlushCountingIndexWriter(dir, newIndexWriterConfig( TEST_VERSION_CURRENT, new MockAnalyzer(MockTokenizer.WHITESPACE, true, false)).setRAMBufferSizeMB(0.5).setMaxBufferedDocs(-1).setMaxBufferedDeleteTerms(-1));
- //w.setInfoStream(System.out);
+ w.setInfoStream(VERBOSE ? System.out : null);
Document doc = new Document();
doc.add(newField("field", "go 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20", Field.Store.NO, Field.Index.ANALYZED));
int num = 6 * RANDOM_MULTIPLIER;
@@ -2583,6 +2590,9 @@ public class TestIndexWriter extends Luc
int count = 0;
final boolean doIndexing = r.nextBoolean();
+ if (VERBOSE) {
+ System.out.println("TEST: iter doIndexing=" + doIndexing);
+ }
if (doIndexing) {
// Add docs until a flush is triggered
final int startFlushCount = w.flushCount;
Modified: lucene/dev/trunk/lucene/src/test/org/apache/lucene/index/TestIndexWriterDelete.java
URL: http://svn.apache.org/viewvc/lucene/dev/trunk/lucene/src/test/org/apache/lucene/index/TestIndexWriterDelete.java?rev=1044635&r1=1044634&r2=1044635&view=diff
==============================================================================
--- lucene/dev/trunk/lucene/src/test/org/apache/lucene/index/TestIndexWriterDelete.java (original)
+++ lucene/dev/trunk/lucene/src/test/org/apache/lucene/index/TestIndexWriterDelete.java Sat Dec 11 11:07:01 2010
@@ -114,6 +114,9 @@ public class TestIndexWriterDelete exten
Directory dir = newDirectory();
IndexWriter writer = new IndexWriter(dir, newIndexWriterConfig(
TEST_VERSION_CURRENT, new MockAnalyzer(MockTokenizer.WHITESPACE, false)).setMaxBufferedDeleteTerms(1));
+
+ writer.setInfoStream(VERBOSE ? System.out : null);
+ writer.addDocument(new Document());
writer.deleteDocuments(new Term("foobar", "1"));
writer.deleteDocuments(new Term("foobar", "1"));
writer.deleteDocuments(new Term("foobar", "1"));
@@ -125,11 +128,14 @@ public class TestIndexWriterDelete exten
// test when delete terms only apply to ram segments
public void testRAMDeletes() throws IOException {
for(int t=0;t<2;t++) {
+ if (VERBOSE) {
+ System.out.println("TEST: t=" + t);
+ }
Directory dir = newDirectory();
IndexWriter modifier = new IndexWriter(dir, newIndexWriterConfig(
TEST_VERSION_CURRENT, new MockAnalyzer(MockTokenizer.WHITESPACE, false)).setMaxBufferedDocs(4)
.setMaxBufferedDeleteTerms(4));
-
+ modifier.setInfoStream(VERBOSE ? System.out : null);
int id = 0;
int value = 100;
@@ -439,6 +445,9 @@ public class TestIndexWriterDelete exten
// Iterate w/ ever increasing free disk space:
while (!done) {
+ if (VERBOSE) {
+ System.out.println("TEST: cycle");
+ }
MockDirectoryWrapper dir = new MockDirectoryWrapper(random, new RAMDirectory(startDir));
dir.setPreventDoubleWrite(false);
IndexWriter modifier = new IndexWriter(dir,
@@ -448,6 +457,7 @@ public class TestIndexWriterDelete exten
.setMaxBufferedDeleteTerms(1000)
.setMergeScheduler(new ConcurrentMergeScheduler()));
((ConcurrentMergeScheduler) modifier.getConfig().getMergeScheduler()).setSuppressExceptions();
+ modifier.setInfoStream(VERBOSE ? System.out : null);
// For each disk size, first try to commit against
// dir that will hit random IOExceptions & disk
@@ -456,6 +466,9 @@ public class TestIndexWriterDelete exten
boolean success = false;
for (int x = 0; x < 2; x++) {
+ if (VERBOSE) {
+ System.out.println("TEST: x=" + x);
+ }
double rate = 0.1;
double diskRatio = ((double)diskFree) / diskUsage;
Modified: lucene/dev/trunk/lucene/src/test/org/apache/lucene/index/TestIndexWriterExceptions.java
URL: http://svn.apache.org/viewvc/lucene/dev/trunk/lucene/src/test/org/apache/lucene/index/TestIndexWriterExceptions.java?rev=1044635&r1=1044634&r2=1044635&view=diff
==============================================================================
--- lucene/dev/trunk/lucene/src/test/org/apache/lucene/index/TestIndexWriterExceptions.java (original)
+++ lucene/dev/trunk/lucene/src/test/org/apache/lucene/index/TestIndexWriterExceptions.java Sat Dec 11 11:07:01 2010
@@ -51,7 +51,7 @@ public class TestIndexWriterExceptions e
IndexWriter writer;
final Random r = new java.util.Random(47);
- Throwable failure;
+ volatile Throwable failure;
public IndexerThread(int i, IndexWriter writer) {
setName("Indexer " + i);
@@ -79,6 +79,9 @@ public class TestIndexWriterExceptions e
final long stopTime = System.currentTimeMillis() + 500;
do {
+ if (VERBOSE) {
+ System.out.println(Thread.currentThread().getName() + ": TEST: IndexerThread: cycle");
+ }
doFail.set(this);
final String id = ""+r.nextInt(50);
idField.setValue(id);
@@ -136,7 +139,7 @@ public class TestIndexWriterExceptions e
if (doFail.get() != null && !name.equals("startDoFlush") && r.nextInt(20) == 17) {
if (VERBOSE) {
System.out.println(Thread.currentThread().getName() + ": NOW FAIL: " + name);
- //new Throwable().printStackTrace(System.out);
+ new Throwable().printStackTrace(System.out);
}
throw new RuntimeException(Thread.currentThread().getName() + ": intentionally failing at " + name);
}
@@ -145,16 +148,23 @@ public class TestIndexWriterExceptions e
}
public void testRandomExceptions() throws Throwable {
+ if (VERBOSE) {
+ System.out.println("\nTEST: start testRandomExceptions");
+ }
MockDirectoryWrapper dir = newDirectory();
MockIndexWriter writer = new MockIndexWriter(dir, newIndexWriterConfig( TEST_VERSION_CURRENT, new MockAnalyzer())
.setRAMBufferSizeMB(0.1).setMergeScheduler(new ConcurrentMergeScheduler()));
((ConcurrentMergeScheduler) writer.getConfig().getMergeScheduler()).setSuppressExceptions();
//writer.setMaxBufferedDocs(10);
+ if (VERBOSE) {
+ System.out.println("TEST: initial commit");
+ }
writer.commit();
- if (VERBOSE)
+ if (VERBOSE) {
writer.setInfoStream(System.out);
+ }
IndexerThread thread = new IndexerThread(0, writer);
thread.run();
@@ -163,6 +173,9 @@ public class TestIndexWriterExceptions e
fail("thread " + thread.getName() + ": hit unexpected failure");
}
+ if (VERBOSE) {
+ System.out.println("TEST: commit after thread start");
+ }
writer.commit();
try {
@@ -192,8 +205,9 @@ public class TestIndexWriterExceptions e
//writer.setMaxBufferedDocs(10);
writer.commit();
- if (VERBOSE)
+ if (VERBOSE) {
writer.setInfoStream(System.out);
+ }
final int NUM_THREADS = 4;
@@ -294,6 +308,7 @@ public class TestIndexWriterExceptions e
public void testExceptionJustBeforeFlush() throws IOException {
Directory dir = newDirectory();
MockIndexWriter w = new MockIndexWriter(dir, newIndexWriterConfig( TEST_VERSION_CURRENT, new MockAnalyzer()).setMaxBufferedDocs(2));
+ w.setInfoStream(VERBOSE ? System.out : null);
Document doc = new Document();
doc.add(newField("field", "a field", Field.Store.YES,
Field.Index.ANALYZED));
Modified: lucene/dev/trunk/lucene/src/test/org/apache/lucene/index/TestIndexWriterOnDiskFull.java
URL: http://svn.apache.org/viewvc/lucene/dev/trunk/lucene/src/test/org/apache/lucene/index/TestIndexWriterOnDiskFull.java?rev=1044635&r1=1044634&r2=1044635&view=diff
==============================================================================
--- lucene/dev/trunk/lucene/src/test/org/apache/lucene/index/TestIndexWriterOnDiskFull.java (original)
+++ lucene/dev/trunk/lucene/src/test/org/apache/lucene/index/TestIndexWriterOnDiskFull.java Sat Dec 11 11:07:01 2010
@@ -47,29 +47,35 @@ public class TestIndexWriterOnDiskFull e
public void testAddDocumentOnDiskFull() throws IOException {
for(int pass=0;pass<2;pass++) {
- if (VERBOSE)
+ if (VERBOSE) {
System.out.println("TEST: pass=" + pass);
+ }
boolean doAbort = pass == 1;
long diskFree = 200;
while(true) {
- if (VERBOSE)
+ if (VERBOSE) {
System.out.println("TEST: cycle: diskFree=" + diskFree);
+ }
MockDirectoryWrapper dir = new MockDirectoryWrapper(random, new RAMDirectory());
dir.setMaxSizeInBytes(diskFree);
IndexWriter writer = new IndexWriter(dir, newIndexWriterConfig( TEST_VERSION_CURRENT, new MockAnalyzer()));
writer.setInfoStream(VERBOSE ? System.out : null);
MergeScheduler ms = writer.getConfig().getMergeScheduler();
- if (ms instanceof ConcurrentMergeScheduler)
+ if (ms instanceof ConcurrentMergeScheduler) {
// This test intentionally produces exceptions
// in the threads that CMS launches; we don't
// want to pollute test output with these.
((ConcurrentMergeScheduler) ms).setSuppressExceptions();
+ }
boolean hitError = false;
try {
for(int i=0;i<200;i++) {
addDoc(writer);
}
+ if (VERBOSE) {
+ System.out.println("TEST: done adding docs; now commit");
+ }
writer.commit();
} catch (IOException e) {
if (VERBOSE) {
@@ -81,13 +87,19 @@ public class TestIndexWriterOnDiskFull e
if (hitError) {
if (doAbort) {
+ if (VERBOSE) {
+ System.out.println("TEST: now rollback");
+ }
writer.rollback();
} else {
try {
+ if (VERBOSE) {
+ System.out.println("TEST: now close");
+ }
writer.close();
} catch (IOException e) {
if (VERBOSE) {
- System.out.println("TEST: exception on close");
+ System.out.println("TEST: exception on close; retry w/ no disk space limit");
e.printStackTrace(System.out);
}
dir.setMaxSizeInBytes(0);
Modified: lucene/dev/trunk/lucene/src/test/org/apache/lucene/index/TestIndexWriterWithThreads.java
URL: http://svn.apache.org/viewvc/lucene/dev/trunk/lucene/src/test/org/apache/lucene/index/TestIndexWriterWithThreads.java?rev=1044635&r1=1044634&r2=1044635&view=diff
==============================================================================
--- lucene/dev/trunk/lucene/src/test/org/apache/lucene/index/TestIndexWriterWithThreads.java (original)
+++ lucene/dev/trunk/lucene/src/test/org/apache/lucene/index/TestIndexWriterWithThreads.java Sat Dec 11 11:07:01 2010
@@ -106,6 +106,9 @@ public class TestIndexWriterWithThreads
int NUM_THREADS = 3;
for(int iter=0;iter<10;iter++) {
+ if (VERBOSE) {
+ System.out.println("\nTEST: iter=" + iter);
+ }
MockDirectoryWrapper dir = newDirectory();
IndexWriter writer = new IndexWriter(
dir,
@@ -116,6 +119,7 @@ public class TestIndexWriterWithThreads
);
((ConcurrentMergeScheduler) writer.getConfig().getMergeScheduler()).setSuppressExceptions();
dir.setMaxSizeInBytes(4*1024+20*iter);
+ writer.setInfoStream(VERBOSE ? System.out : null);
IndexerThread[] threads = new IndexerThread[NUM_THREADS];
Added: lucene/dev/trunk/lucene/src/test/org/apache/lucene/index/TestPerSegmentDeletes.java
URL: http://svn.apache.org/viewvc/lucene/dev/trunk/lucene/src/test/org/apache/lucene/index/TestPerSegmentDeletes.java?rev=1044635&view=auto
==============================================================================
--- lucene/dev/trunk/lucene/src/test/org/apache/lucene/index/TestPerSegmentDeletes.java (added)
+++ lucene/dev/trunk/lucene/src/test/org/apache/lucene/index/TestPerSegmentDeletes.java Sat Dec 11 11:07:01 2010
@@ -0,0 +1,299 @@
+package org.apache.lucene.index;
+
+/**
+ * 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.
+ */
+
+import java.io.IOException;
+import java.util.ArrayList;
+import java.util.List;
+import java.util.Random;
+import java.util.Set;
+
+import org.apache.lucene.analysis.MockAnalyzer;
+import org.apache.lucene.document.Document;
+import org.apache.lucene.index.TermsEnum.SeekStatus;
+import org.apache.lucene.store.Directory;
+import org.apache.lucene.store.MockDirectoryWrapper;
+import org.apache.lucene.store.RAMDirectory;
+import org.apache.lucene.util.ArrayUtil;
+import org.apache.lucene.util.Bits;
+import org.apache.lucene.util.BytesRef;
+import org.apache.lucene.util.LuceneTestCase;
+import org.apache.lucene.util.Version;
+
+public class TestPerSegmentDeletes extends LuceneTestCase {
+ public void testDeletes1() throws Exception {
+ //IndexWriter.debug2 = System.out;
+ Directory dir = new MockDirectoryWrapper(new Random(), new RAMDirectory());
+ IndexWriterConfig iwc = new IndexWriterConfig(Version.LUCENE_CURRENT,
+ new MockAnalyzer());
+ iwc.setMergeScheduler(new SerialMergeScheduler());
+ iwc.setMaxBufferedDocs(5000);
+ iwc.setRAMBufferSizeMB(100);
+ RangeMergePolicy fsmp = new RangeMergePolicy(false);
+ iwc.setMergePolicy(fsmp);
+ IndexWriter writer = new IndexWriter(dir, iwc);
+ for (int x = 0; x < 5; x++) {
+ writer.addDocument(TestIndexWriterReader.createDocument(x, "1", 2));
+ //System.out.println("numRamDocs(" + x + ")" + writer.numRamDocs());
+ }
+ //System.out.println("commit1");
+ writer.commit();
+ assertEquals(1, writer.segmentInfos.size());
+ for (int x = 5; x < 10; x++) {
+ writer.addDocument(TestIndexWriterReader.createDocument(x, "2", 2));
+ //System.out.println("numRamDocs(" + x + ")" + writer.numRamDocs());
+ }
+ //System.out.println("commit2");
+ writer.commit();
+ assertEquals(2, writer.segmentInfos.size());
+
+ for (int x = 10; x < 15; x++) {
+ writer.addDocument(TestIndexWriterReader.createDocument(x, "3", 2));
+ //System.out.println("numRamDocs(" + x + ")" + writer.numRamDocs());
+ }
+
+ writer.deleteDocuments(new Term("id", "1"));
+
+ writer.deleteDocuments(new Term("id", "11"));
+
+ // flushing without applying deletes means
+ // there will still be deletes in the segment infos
+ writer.flush(false, false, false);
+ assertTrue(writer.bufferedDeletes.any());
+
+ // get reader flushes pending deletes
+ // so there should not be anymore
+ IndexReader r1 = writer.getReader();
+ assertFalse(writer.bufferedDeletes.any());
+ r1.close();
+
+ // delete id:2 from the first segment
+ // merge segments 0 and 1
+ // which should apply the delete id:2
+ writer.deleteDocuments(new Term("id", "2"));
+ writer.flush(false, false, false);
+ fsmp.doMerge = true;
+ fsmp.start = 0;
+ fsmp.length = 2;
+ writer.maybeMerge();
+
+ assertEquals(2, writer.segmentInfos.size());
+
+ // id:2 shouldn't exist anymore because
+ // it's been applied in the merge and now it's gone
+ IndexReader r2 = writer.getReader();
+ int[] id2docs = toDocsArray(new Term("id", "2"), null, r2);
+ assertTrue(id2docs == null);
+ r2.close();
+
+ /**
+ // added docs are in the ram buffer
+ for (int x = 15; x < 20; x++) {
+ writer.addDocument(TestIndexWriterReader.createDocument(x, "4", 2));
+ System.out.println("numRamDocs(" + x + ")" + writer.numRamDocs());
+ }
+ assertTrue(writer.numRamDocs() > 0);
+ // delete from the ram buffer
+ writer.deleteDocuments(new Term("id", Integer.toString(13)));
+
+ Term id3 = new Term("id", Integer.toString(3));
+
+ // delete from the 1st segment
+ writer.deleteDocuments(id3);
+
+ assertTrue(writer.numRamDocs() > 0);
+
+ //System.out
+ // .println("segdels1:" + writer.docWriter.deletesToString());
+
+ //assertTrue(writer.docWriter.segmentDeletes.size() > 0);
+
+ // we cause a merge to happen
+ fsmp.doMerge = true;
+ fsmp.start = 0;
+ fsmp.length = 2;
+ System.out.println("maybeMerge "+writer.segmentInfos);
+
+ SegmentInfo info0 = writer.segmentInfos.get(0);
+ SegmentInfo info1 = writer.segmentInfos.get(1);
+
+ writer.maybeMerge();
+ System.out.println("maybeMerge after "+writer.segmentInfos);
+ // there should be docs in RAM
+ assertTrue(writer.numRamDocs() > 0);
+
+ // assert we've merged the 1 and 2 segments
+ // and still have a segment leftover == 2
+ assertEquals(2, writer.segmentInfos.size());
+ assertFalse(segThere(info0, writer.segmentInfos));
+ assertFalse(segThere(info1, writer.segmentInfos));
+
+ //System.out.println("segdels2:" + writer.docWriter.deletesToString());
+
+ //assertTrue(writer.docWriter.segmentDeletes.size() > 0);
+
+ IndexReader r = writer.getReader();
+ IndexReader r1 = r.getSequentialSubReaders()[0];
+ printDelDocs(r1.getDeletedDocs());
+ int[] docs = toDocsArray(id3, null, r);
+ System.out.println("id3 docs:"+Arrays.toString(docs));
+ // there shouldn't be any docs for id:3
+ assertTrue(docs == null);
+ r.close();
+
+ part2(writer, fsmp);
+ **/
+ // System.out.println("segdels2:"+writer.docWriter.segmentDeletes.toString());
+ //System.out.println("close");
+ writer.close();
+ dir.close();
+ }
+
+ /**
+ static boolean hasPendingDeletes(SegmentInfos infos) {
+ for (SegmentInfo info : infos) {
+ if (info.deletes.any()) {
+ return true;
+ }
+ }
+ return false;
+ }
+ **/
+ void part2(IndexWriter writer, RangeMergePolicy fsmp) throws Exception {
+ for (int x = 20; x < 25; x++) {
+ writer.addDocument(TestIndexWriterReader.createDocument(x, "5", 2));
+ //System.out.println("numRamDocs(" + x + ")" + writer.numRamDocs());
+ }
+ writer.flush(false, true, false);
+ for (int x = 25; x < 30; x++) {
+ writer.addDocument(TestIndexWriterReader.createDocument(x, "5", 2));
+ //System.out.println("numRamDocs(" + x + ")" + writer.numRamDocs());
+ }
+ writer.flush(false, true, false);
+
+ //System.out.println("infos3:"+writer.segmentInfos);
+
+ Term delterm = new Term("id", "8");
+ writer.deleteDocuments(delterm);
+ //System.out.println("segdels3:" + writer.docWriter.deletesToString());
+
+ fsmp.doMerge = true;
+ fsmp.start = 1;
+ fsmp.length = 2;
+ writer.maybeMerge();
+
+ // deletes for info1, the newly created segment from the
+ // merge should have no deletes because they were applied in
+ // the merge
+ //SegmentInfo info1 = writer.segmentInfos.get(1);
+ //assertFalse(exists(info1, writer.docWriter.segmentDeletes));
+
+ //System.out.println("infos4:"+writer.segmentInfos);
+ //System.out.println("segdels4:" + writer.docWriter.deletesToString());
+ }
+
+ boolean segThere(SegmentInfo info, SegmentInfos infos) {
+ for (SegmentInfo si : infos) {
+ if (si.name.equals(info.name)) return true;
+ }
+ return false;
+ }
+
+ public static void printDelDocs(Bits bits) {
+ if (bits == null) return;
+ for (int x = 0; x < bits.length(); x++) {
+ System.out.println(x + ":" + bits.get(x));
+ }
+ }
+
+ public static int[] toDocsArray(Term term, Bits bits, IndexReader reader)
+ throws IOException {
+ Fields fields = MultiFields.getFields(reader);
+ Terms cterms = fields.terms(term.field);
+ TermsEnum ctermsEnum = cterms.iterator();
+ SeekStatus ss = ctermsEnum.seek(new BytesRef(term.text()), false);
+ if (ss.equals(SeekStatus.FOUND)) {
+ DocsEnum docsEnum = ctermsEnum.docs(bits, null);
+ return toArray(docsEnum);
+ }
+ return null;
+ }
+
+ public static int[] toArray(DocsEnum docsEnum) throws IOException {
+ List<Integer> docs = new ArrayList<Integer>();
+ while (docsEnum.nextDoc() != DocsEnum.NO_MORE_DOCS) {
+ int docID = docsEnum.docID();
+ docs.add(docID);
+ }
+ return ArrayUtil.toIntArray(docs);
+ }
+
+ public class RangeMergePolicy extends MergePolicy {
+ boolean doMerge = false;
+ int start;
+ int length;
+
+ private final boolean useCompoundFile;
+
+ private RangeMergePolicy(boolean useCompoundFile) {
+ this.useCompoundFile = useCompoundFile;
+ }
+
+ @Override
+ public void close() {}
+
+ public MergeSpecification findMerges(SegmentInfos segmentInfos)
+ throws CorruptIndexException, IOException {
+ MergeSpecification ms = new MergeSpecification();
+ if (doMerge) {
+ SegmentInfos mergeInfos = new SegmentInfos();
+ for (int x=start; x < (start+length); x++) {
+ mergeInfos.add(segmentInfos.get(x));
+ }
+ OneMerge om = new OneMerge(mergeInfos);
+ ms.add(om);
+ doMerge = false;
+ return ms;
+ }
+ return null;
+ }
+
+ @Override
+ public MergeSpecification findMergesForOptimize(SegmentInfos segmentInfos,
+ int maxSegmentCount, Set<SegmentInfo> segmentsToOptimize)
+ throws CorruptIndexException, IOException {
+ return null;
+ }
+
+ @Override
+ public MergeSpecification findMergesToExpungeDeletes(
+ SegmentInfos segmentInfos) throws CorruptIndexException, IOException {
+ return null;
+ }
+
+ @Override
+ public boolean useCompoundDocStore(SegmentInfos segments) {
+ return useCompoundFile;
+ }
+
+ @Override
+ public boolean useCompoundFile(SegmentInfos segments, SegmentInfo newSegment) {
+ return useCompoundFile;
+ }
+ }
+}
Modified: lucene/dev/trunk/lucene/src/test/org/apache/lucene/index/TestStressIndexing2.java
URL: http://svn.apache.org/viewvc/lucene/dev/trunk/lucene/src/test/org/apache/lucene/index/TestStressIndexing2.java?rev=1044635&r1=1044634&r2=1044635&view=diff
==============================================================================
--- lucene/dev/trunk/lucene/src/test/org/apache/lucene/index/TestStressIndexing2.java (original)
+++ lucene/dev/trunk/lucene/src/test/org/apache/lucene/index/TestStressIndexing2.java Sat Dec 11 11:07:01 2010
@@ -95,6 +95,9 @@ public class TestStressIndexing2 extends
int num = 3 * RANDOM_MULTIPLIER;
for (int i = 0; i < num; i++) { // increase iterations for better testing
+ if (VERBOSE) {
+ System.out.println("\n\nTEST: top iter=" + i);
+ }
sameFieldOrder=random.nextBoolean();
mergeFactor=random.nextInt(3)+2;
maxBufferedDocs=random.nextInt(3)+2;
@@ -107,10 +110,17 @@ public class TestStressIndexing2 extends
int range=random.nextInt(20)+1;
Directory dir1 = newDirectory();
Directory dir2 = newDirectory();
+ if (VERBOSE) {
+ System.out.println(" nThreads=" + nThreads + " iter=" + iter + " range=" + range + " doPooling=" + doReaderPooling + " maxThreadStates=" + maxThreadStates + " sameFieldOrder=" + sameFieldOrder + " mergeFactor=" + mergeFactor);
+ }
Map<String,Document> docs = indexRandom(nThreads, iter, range, dir1, maxThreadStates, doReaderPooling);
- //System.out.println("TEST: index serial");
+ if (VERBOSE) {
+ System.out.println("TEST: index serial");
+ }
indexSerial(random, docs, dir2);
- //System.out.println("TEST: verify");
+ if (VERBOSE) {
+ System.out.println("TEST: verify");
+ }
verifyEquals(dir1, dir2, "id");
dir1.close();
dir2.close();
@@ -140,6 +150,7 @@ public class TestStressIndexing2 extends
IndexWriter w = new MockIndexWriter(dir, newIndexWriterConfig(
TEST_VERSION_CURRENT, new MockAnalyzer()).setOpenMode(OpenMode.CREATE).setRAMBufferSizeMB(
0.1).setMaxBufferedDocs(maxBufferedDocs));
+ w.setInfoStream(VERBOSE ? System.out : null);
w.commit();
LogMergePolicy lmp = (LogMergePolicy) w.getConfig().getMergePolicy();
lmp.setUseCompoundFile(false);
@@ -190,10 +201,14 @@ public class TestStressIndexing2 extends
boolean doReaderPooling) throws IOException, InterruptedException {
Map<String,Document> docs = new HashMap<String,Document>();
for(int iter=0;iter<3;iter++) {
+ if (VERBOSE) {
+ System.out.println("TEST: iter=" + iter);
+ }
IndexWriter w = new MockIndexWriter(dir, newIndexWriterConfig(
TEST_VERSION_CURRENT, new MockAnalyzer()).setOpenMode(OpenMode.CREATE)
.setRAMBufferSizeMB(0.1).setMaxBufferedDocs(maxBufferedDocs).setMaxThreadStates(maxThreadStates)
.setReaderPooling(doReaderPooling));
+ w.setInfoStream(VERBOSE ? System.out : null);
LogMergePolicy lmp = (LogMergePolicy) w.getConfig().getMergePolicy();
lmp.setUseCompoundFile(false);
lmp.setUseCompoundDocStore(false);
@@ -272,9 +287,33 @@ public class TestStressIndexing2 extends
r2.close();
}
+ private static void printDocs(IndexReader r) throws Throwable {
+ IndexReader[] subs = r.getSequentialSubReaders();
+ for(IndexReader sub : subs) {
+ Bits delDocs = sub.getDeletedDocs();
+ System.out.println(" " + ((SegmentReader) sub).getSegmentInfo());
+ for(int docID=0;docID<sub.maxDoc();docID++) {
+ Document doc = sub.document(docID);
+ if (delDocs == null || !delDocs.get(docID)) {
+ System.out.println(" docID=" + docID + " id:" + doc.get("id"));
+ } else {
+ System.out.println(" DEL docID=" + docID + " id:" + doc.get("id"));
+ }
+ }
+ }
+ }
+
public static void verifyEquals(IndexReader r1, IndexReader r2, String idField) throws Throwable {
- assertEquals(r1.numDocs(), r2.numDocs());
+ if (VERBOSE) {
+ System.out.println("\nr1 docs:");
+ printDocs(r1);
+ System.out.println("\nr2 docs:");
+ printDocs(r2);
+ }
+ if (r1.numDocs() != r2.numDocs()) {
+ assert false: "r1.numDocs()=" + r1.numDocs() + " vs r2.numDocs()=" + r2.numDocs();
+ }
boolean hasDeletes = !(r1.maxDoc()==r2.maxDoc() && r1.numDocs()==r1.maxDoc());
int[] r2r1 = new int[r2.maxDoc()]; // r2 id to r1 id mapping
@@ -692,19 +731,28 @@ public class TestStressIndexing2 extends
for (int i=0; i<fields.size(); i++) {
d.add(fields.get(i));
}
+ if (VERBOSE) {
+ System.out.println(Thread.currentThread().getName() + ": indexing id:" + idString);
+ }
w.updateDocument(idTerm.createTerm(idString), d);
- // System.out.println("indexing "+d);
+ //System.out.println(Thread.currentThread().getName() + ": indexing "+d);
docs.put(idString, d);
}
public void deleteDoc() throws IOException {
String idString = getIdString();
+ if (VERBOSE) {
+ System.out.println(Thread.currentThread().getName() + ": del id:" + idString);
+ }
w.deleteDocuments(idTerm.createTerm(idString));
docs.remove(idString);
}
public void deleteByQuery() throws IOException {
String idString = getIdString();
+ if (VERBOSE) {
+ System.out.println(Thread.currentThread().getName() + ": del query id:" + idString);
+ }
w.deleteDocuments(new TermQuery(idTerm.createTerm(idString)));
docs.remove(idString);
}
Modified: lucene/dev/trunk/lucene/src/test/org/apache/lucene/search/TestTermVectors.java
URL: http://svn.apache.org/viewvc/lucene/dev/trunk/lucene/src/test/org/apache/lucene/search/TestTermVectors.java?rev=1044635&r1=1044634&r2=1044635&view=diff
==============================================================================
--- lucene/dev/trunk/lucene/src/test/org/apache/lucene/search/TestTermVectors.java (original)
+++ lucene/dev/trunk/lucene/src/test/org/apache/lucene/search/TestTermVectors.java Sat Dec 11 11:07:01 2010
@@ -353,12 +353,19 @@ public class TestTermVectors extends Luc
RandomIndexWriter writer = new RandomIndexWriter(random, directory,
newIndexWriterConfig(TEST_VERSION_CURRENT, new MockAnalyzer(MockTokenizer.SIMPLE, true))
.setOpenMode(OpenMode.CREATE));
+ writer.w.setInfoStream(VERBOSE ? System.out : null);
+ if (VERBOSE) {
+ System.out.println("TEST: now add non-vectors");
+ }
for (int i = 0; i < 100; i++) {
Document doc = new Document();
doc.add(new Field("field", English.intToEnglish(i),
Field.Store.YES, Field.Index.ANALYZED, Field.TermVector.NO));
writer.addDocument(doc);
}
+ if (VERBOSE) {
+ System.out.println("TEST: now add vectors");
+ }
for(int i=0;i<10;i++) {
Document doc = new Document();
doc.add(new Field("field", English.intToEnglish(100+i),
@@ -366,6 +373,9 @@ public class TestTermVectors extends Luc
writer.addDocument(doc);
}
+ if (VERBOSE) {
+ System.out.println("TEST: now getReader");
+ }
IndexReader reader = writer.getReader();
writer.close();
searcher = new IndexSearcher(reader);
@@ -374,6 +384,7 @@ public class TestTermVectors extends Luc
ScoreDoc[] hits = searcher.search(query, null, 1000).scoreDocs;
assertEquals(10, hits.length);
for (int i = 0; i < hits.length; i++) {
+
TermFreqVector [] vector = searcher.reader.getTermFreqVectors(hits[i].doc);
assertTrue(vector != null);
assertTrue(vector.length == 1);
Modified: lucene/dev/trunk/lucene/src/test/org/apache/lucene/store/MockIndexOutputWrapper.java
URL: http://svn.apache.org/viewvc/lucene/dev/trunk/lucene/src/test/org/apache/lucene/store/MockIndexOutputWrapper.java?rev=1044635&r1=1044634&r2=1044635&view=diff
==============================================================================
--- lucene/dev/trunk/lucene/src/test/org/apache/lucene/store/MockIndexOutputWrapper.java (original)
+++ lucene/dev/trunk/lucene/src/test/org/apache/lucene/store/MockIndexOutputWrapper.java Sat Dec 11 11:07:01 2010
@@ -19,6 +19,8 @@ package org.apache.lucene.store;
import java.io.IOException;
+import org.apache.lucene.util.LuceneTestCase;
+
/**
* Used by MockRAMDirectory to create an output stream that
* will throw an IOException on fake disk full, track max
@@ -102,6 +104,9 @@ public class MockIndexOutputWrapper exte
message += "; wrote " + freeSpace + " of " + len + " bytes";
}
message += ")";
+ if (LuceneTestCase.VERBOSE) {
+ System.out.println(Thread.currentThread().getName() + ": MDW: now throw fake disk full");
+ }
throw new IOException(message);
} else {
if (dir.randomState.nextBoolean()) {
Modified: lucene/dev/trunk/lucene/src/test/org/apache/lucene/util/LuceneTestCase.java
URL: http://svn.apache.org/viewvc/lucene/dev/trunk/lucene/src/test/org/apache/lucene/util/LuceneTestCase.java?rev=1044635&r1=1044634&r2=1044635&view=diff
==============================================================================
--- lucene/dev/trunk/lucene/src/test/org/apache/lucene/util/LuceneTestCase.java (original)
+++ lucene/dev/trunk/lucene/src/test/org/apache/lucene/util/LuceneTestCase.java Sat Dec 11 11:07:01 2010
@@ -143,7 +143,7 @@ public abstract class LuceneTestCase ext
public static final boolean TEST_NIGHTLY = Boolean.parseBoolean(System.getProperty("tests.nightly", "false"));
/** the line file used by LineFileDocs */
public static final String TEST_LINE_DOCS_FILE = System.getProperty("tests.linedocsfile", "europarl.lines.txt.gz");
-
+
private static final Pattern codecWithParam = Pattern.compile("(.*)\\(\\s*(\\d+)\\s*\\)");
/**