public void testSimpleSkip() throws IOException {
Directory dir = new CountingRAMDirectory(new RAMDirectory());
IndexWriter writer =
new IndexWriter(
dir,
newIndexWriterConfig(TEST_VERSION_CURRENT, new PayloadAnalyzer())
.setCodec(_TestUtil.alwaysPostingsFormat(new Lucene40PostingsFormat()))
.setMergePolicy(newLogMergePolicy()));
Term term = new Term("test", "a");
for (int i = 0; i < 5000; i++) {
Document d1 = new Document();
d1.add(newTextField(term.field(), term.text(), Field.Store.NO));
writer.addDocument(d1);
}
writer.commit();
writer.forceMerge(1);
writer.close();
AtomicReader reader = getOnlySegmentReader(DirectoryReader.open(dir));
for (int i = 0; i < 2; i++) {
counter = 0;
DocsAndPositionsEnum tp = reader.termPositionsEnum(term);
checkSkipTo(tp, 14, 185); // no skips
checkSkipTo(tp, 17, 190); // one skip on level 0
checkSkipTo(tp, 287, 200); // one skip on level 1, two on level 0
// this test would fail if we had only one skip level,
// because than more bytes would be read from the freqStream
checkSkipTo(tp, 4800, 250); // one skip on level 2
}
}
// LUCENE-1382
public void testCommitUserData() throws IOException {
Directory dir = newDirectory();
IndexWriter w =
new IndexWriter(
dir,
newIndexWriterConfig(TEST_VERSION_CURRENT, new MockAnalyzer(random()))
.setMaxBufferedDocs(2));
for (int j = 0; j < 17; j++) TestIndexWriter.addDoc(w);
w.close();
DirectoryReader r = DirectoryReader.open(dir);
// commit(Map) never called for this index
assertEquals(0, r.getIndexCommit().getUserData().size());
r.close();
w =
new IndexWriter(
dir,
newIndexWriterConfig(TEST_VERSION_CURRENT, new MockAnalyzer(random()))
.setMaxBufferedDocs(2));
for (int j = 0; j < 17; j++) TestIndexWriter.addDoc(w);
Map<String, String> data = new HashMap<String, String>();
data.put("label", "test1");
w.commit(data);
w.close();
r = DirectoryReader.open(dir);
assertEquals("test1", r.getIndexCommit().getUserData().get("label"));
r.close();
w =
new IndexWriter(
dir, newIndexWriterConfig(TEST_VERSION_CURRENT, new MockAnalyzer(random())));
w.forceMerge(1);
w.close();
dir.close();
}
/*
* Verify that calling forceMerge when writer is open for
* "commit on close" works correctly both for rollback()
* and close().
*/
public void testCommitOnCloseForceMerge() throws IOException {
MockDirectoryWrapper dir = newDirectory();
// Must disable throwing exc on double-write: this
// test uses IW.rollback which easily results in
// writing to same file more than once
dir.setPreventDoubleWrite(false);
IndexWriter writer =
new IndexWriter(
dir,
newIndexWriterConfig(TEST_VERSION_CURRENT, new MockAnalyzer(random()))
.setMaxBufferedDocs(10)
.setMergePolicy(newLogMergePolicy(10)));
for (int j = 0; j < 17; j++) {
TestIndexWriter.addDocWithIndex(writer, j);
}
writer.close();
writer =
new IndexWriter(
dir,
newIndexWriterConfig(TEST_VERSION_CURRENT, new MockAnalyzer(random()))
.setOpenMode(OpenMode.APPEND));
writer.forceMerge(1);
// Open a reader before closing (commiting) the writer:
DirectoryReader reader = DirectoryReader.open(dir);
// Reader should see index as multi-seg at this
// point:
assertTrue("Reader incorrectly sees one segment", reader.getSequentialSubReaders().size() > 1);
reader.close();
// Abort the writer:
writer.rollback();
TestIndexWriter.assertNoUnreferencedFiles(dir, "aborted writer after forceMerge");
// Open a reader after aborting writer:
reader = DirectoryReader.open(dir);
// Reader should still see index as multi-segment
assertTrue("Reader incorrectly sees one segment", reader.getSequentialSubReaders().size() > 1);
reader.close();
if (VERBOSE) {
System.out.println("TEST: do real full merge");
}
writer =
new IndexWriter(
dir,
newIndexWriterConfig(TEST_VERSION_CURRENT, new MockAnalyzer(random()))
.setOpenMode(OpenMode.APPEND));
writer.forceMerge(1);
writer.close();
if (VERBOSE) {
System.out.println("TEST: writer closed");
}
TestIndexWriter.assertNoUnreferencedFiles(dir, "aborted writer after forceMerge");
// Open a reader after aborting writer:
reader = DirectoryReader.open(dir);
// Reader should see index as one segment
assertEquals(
"Reader incorrectly sees more than one segment",
1,
reader.getSequentialSubReaders().size());
reader.close();
dir.close();
}
/*
* Verify that a writer with "commit on close" indeed
* cleans up the temp segments created after opening
* that are not referenced by the starting segments
* file. We check this by using MockDirectoryWrapper to
* measure max temp disk space used.
*/
public void testCommitOnCloseDiskUsage() throws IOException {
// MemoryCodec, since it uses FST, is not necessarily
// "additive", ie if you add up N small FSTs, then merge
// them, the merged result can easily be larger than the
// sum because the merged FST may use array encoding for
// some arcs (which uses more space):
final String idFormat = _TestUtil.getPostingsFormat("id");
final String contentFormat = _TestUtil.getPostingsFormat("content");
assumeFalse(
"This test cannot run with Memory codec",
idFormat.equals("Memory") || contentFormat.equals("Memory"));
MockDirectoryWrapper dir = newDirectory();
Analyzer analyzer;
if (random().nextBoolean()) {
// no payloads
analyzer =
new Analyzer() {
@Override
public TokenStreamComponents createComponents(String fieldName, Reader reader) {
return new TokenStreamComponents(
new MockTokenizer(reader, MockTokenizer.WHITESPACE, true));
}
};
} else {
// fixed length payloads
final int length = random().nextInt(200);
analyzer =
new Analyzer() {
@Override
public TokenStreamComponents createComponents(String fieldName, Reader reader) {
Tokenizer tokenizer = new MockTokenizer(reader, MockTokenizer.WHITESPACE, true);
return new TokenStreamComponents(
tokenizer, new MockFixedLengthPayloadFilter(random(), tokenizer, length));
}
};
}
IndexWriter writer =
new IndexWriter(
dir,
newIndexWriterConfig(TEST_VERSION_CURRENT, analyzer)
.setMaxBufferedDocs(10)
.setReaderPooling(false)
.setMergePolicy(newLogMergePolicy(10)));
for (int j = 0; j < 30; j++) {
TestIndexWriter.addDocWithIndex(writer, j);
}
writer.close();
dir.resetMaxUsedSizeInBytes();
dir.setTrackDiskUsage(true);
long startDiskUsage = dir.getMaxUsedSizeInBytes();
writer =
new IndexWriter(
dir,
newIndexWriterConfig(TEST_VERSION_CURRENT, analyzer)
.setOpenMode(OpenMode.APPEND)
.setMaxBufferedDocs(10)
.setMergeScheduler(new SerialMergeScheduler())
.setReaderPooling(false)
.setMergePolicy(newLogMergePolicy(10)));
for (int j = 0; j < 1470; j++) {
TestIndexWriter.addDocWithIndex(writer, j);
}
long midDiskUsage = dir.getMaxUsedSizeInBytes();
dir.resetMaxUsedSizeInBytes();
writer.forceMerge(1);
writer.close();
DirectoryReader.open(dir).close();
long endDiskUsage = dir.getMaxUsedSizeInBytes();
// Ending index is 50X as large as starting index; due
// to 3X disk usage normally we allow 150X max
// transient usage. If something is wrong w/ deleter
// and it doesn't delete intermediate segments then it
// will exceed this 150X:
// System.out.println("start " + startDiskUsage + "; mid " + midDiskUsage + ";end " +
// endDiskUsage);
assertTrue(
"writer used too much space while adding documents: mid="
+ midDiskUsage
+ " start="
+ startDiskUsage
+ " end="
+ endDiskUsage
+ " max="
+ (startDiskUsage * 150),
midDiskUsage < 150 * startDiskUsage);
assertTrue(
"writer used too much space after close: endDiskUsage="
+ endDiskUsage
+ " startDiskUsage="
+ startDiskUsage
+ " max="
+ (startDiskUsage * 150),
endDiskUsage < 150 * startDiskUsage);
dir.close();
}