IndexWriter getIndexWriter(Directory luceneDir)
throws CorruptIndexException, LockObtainFailedException, IOException, ProviderException {
IndexWriter writer = null;
IndexWriterConfig writerConfig = new IndexWriterConfig(Version.LUCENE_36, getLuceneAnalyzer());
writerConfig.setOpenMode(OpenMode.CREATE_OR_APPEND);
writer = new IndexWriter(luceneDir, writerConfig);
// writer.setInfoStream( System.out );
return writer;
}
protected DirectoryReader indexDocsWithLucene(TestDoc[] testDocs) throws IOException {
Map<String, Analyzer> mapping = new HashMap<String, Analyzer>();
for (TestFieldSetting field : testDocs[0].fieldSettings) {
if (field.storedPayloads) {
mapping.put(
field.name,
new Analyzer() {
@Override
protected TokenStreamComponents createComponents(String fieldName, Reader reader) {
Tokenizer tokenizer = new StandardTokenizer(Version.CURRENT.luceneVersion, reader);
TokenFilter filter = new LowerCaseFilter(Version.CURRENT.luceneVersion, tokenizer);
filter = new TypeAsPayloadTokenFilter(filter);
return new TokenStreamComponents(tokenizer, filter);
}
});
}
}
PerFieldAnalyzerWrapper wrapper =
new PerFieldAnalyzerWrapper(
new StandardAnalyzer(Version.CURRENT.luceneVersion, CharArraySet.EMPTY_SET), mapping);
Directory dir = new RAMDirectory();
IndexWriterConfig conf = new IndexWriterConfig(Version.CURRENT.luceneVersion, wrapper);
conf.setOpenMode(IndexWriterConfig.OpenMode.CREATE);
IndexWriter writer = new IndexWriter(dir, conf);
for (TestDoc doc : testDocs) {
Document d = new Document();
d.add(new Field("id", doc.id, StringField.TYPE_STORED));
for (int i = 0; i < doc.fieldContent.length; i++) {
FieldType type = new FieldType(TextField.TYPE_STORED);
TestFieldSetting fieldSetting = doc.fieldSettings[i];
type.setStoreTermVectorOffsets(fieldSetting.storedOffset);
type.setStoreTermVectorPayloads(fieldSetting.storedPayloads);
type.setStoreTermVectorPositions(
fieldSetting.storedPositions
|| fieldSetting.storedPayloads
|| fieldSetting.storedOffset);
type.setStoreTermVectors(true);
type.freeze();
d.add(new Field(fieldSetting.name, doc.fieldContent[i], type));
}
writer.updateDocument(new Term("id", doc.id), d);
writer.commit();
}
writer.close();
return DirectoryReader.open(dir);
}
/**
* Index a picture
*
* @param source
* @param picture_id
* @param conf
* @throws IOException
*/
public static void index(byte[] source, UUID picture_id, IndexWriterConfig conf)
throws IOException {
ByteArrayInputStream in = new ByteArrayInputStream(source);
BufferedImage image = ImageIO.read(in);
// Creating an Lucene IndexWriter
log.debug("Is Lucene configured? " + (conf == null));
if (conf == null) {
conf =
new IndexWriterConfig(
LuceneUtils.LUCENE_VERSION, new WhitespaceAnalyzer(LuceneUtils.LUCENE_VERSION));
conf.setOpenMode(IndexWriterConfig.OpenMode.CREATE_OR_APPEND);
}
luceneIndexer(
image,
picture_id,
FeatureEnumerate.AutoColorCorrelogram.getText(),
DocumentBuilderFactory.getAutoColorCorrelogramDocumentBuilder(),
conf);
luceneIndexer(
image,
picture_id,
FeatureEnumerate.CEDD.getText(),
DocumentBuilderFactory.getCEDDDocumentBuilder(),
conf);
luceneIndexer(
image,
picture_id,
FeatureEnumerate.ColorLayout.getText(),
DocumentBuilderFactory.getColorLayoutBuilder(),
conf);
luceneIndexer(
image,
picture_id,
FeatureEnumerate.EdgeHistogram.getText(),
DocumentBuilderFactory.getEdgeHistogramBuilder(),
conf);
luceneIndexer(
image,
picture_id,
FeatureEnumerate.ColorHistogram.getText(),
DocumentBuilderFactory.getColorHistogramDocumentBuilder(),
conf);
luceneIndexer(
image,
picture_id,
FeatureEnumerate.PHOG.getText(),
DocumentBuilderFactory.getPHOGDocumentBuilder(),
conf);
}
private static void deleteFromFeature(
UUID pictureId, Term term, String prefix, IndexWriterConfig conf) throws IOException {
File file = getPath(prefix);
// Creating an Lucene IndexWriter
log.debug("Is Lucene configured: " + (conf == null));
if (conf == null) {
conf =
new IndexWriterConfig(
LuceneUtils.LUCENE_VERSION, new WhitespaceAnalyzer(LuceneUtils.LUCENE_VERSION));
conf.setOpenMode(IndexWriterConfig.OpenMode.CREATE_OR_APPEND);
}
IndexWriter iw = new IndexWriter(FSDirectory.open(file), conf);
iw.deleteDocuments(term);
iw.close();
}
@Test
public void testRecoveryDiff() throws IOException, InterruptedException {
int numDocs = 2 + random().nextInt(100);
List<Document> docs = new ArrayList<>();
for (int i = 0; i < numDocs; i++) {
Document doc = new Document();
doc.add(
new StringField("id", "" + i, random().nextBoolean() ? Field.Store.YES : Field.Store.NO));
doc.add(
new TextField(
"body",
TestUtil.randomRealisticUnicodeString(random()),
random().nextBoolean() ? Field.Store.YES : Field.Store.NO));
doc.add(
new SortedDocValuesField(
"dv", new BytesRef(TestUtil.randomRealisticUnicodeString(random()))));
docs.add(doc);
}
long seed = random().nextLong();
Store.MetadataSnapshot first;
{
Random random = new Random(seed);
IndexWriterConfig iwc =
new IndexWriterConfig(new MockAnalyzer(random)).setCodec(actualDefaultCodec());
iwc.setMergePolicy(NoMergePolicy.INSTANCE);
iwc.setUseCompoundFile(random.nextBoolean());
iwc.setMaxThreadStates(1);
final ShardId shardId = new ShardId(new Index("index"), 1);
DirectoryService directoryService = new LuceneManagedDirectoryService(random);
Store store =
new Store(
shardId,
ImmutableSettings.EMPTY,
directoryService,
randomDistributor(random, directoryService),
new DummyShardLock(shardId));
IndexWriter writer = new IndexWriter(store.directory(), iwc);
final boolean lotsOfSegments = rarely(random);
for (Document d : docs) {
writer.addDocument(d);
if (lotsOfSegments && random.nextBoolean()) {
writer.commit();
} else if (rarely(random)) {
writer.commit();
}
}
writer.commit();
writer.close();
first = store.getMetadata();
assertDeleteContent(store, directoryService);
store.close();
}
long time = new Date().getTime();
while (time == new Date().getTime()) {
Thread.sleep(10); // bump the time
}
Store.MetadataSnapshot second;
Store store;
{
Random random = new Random(seed);
IndexWriterConfig iwc =
new IndexWriterConfig(new MockAnalyzer(random)).setCodec(actualDefaultCodec());
iwc.setMergePolicy(NoMergePolicy.INSTANCE);
iwc.setUseCompoundFile(random.nextBoolean());
iwc.setMaxThreadStates(1);
final ShardId shardId = new ShardId(new Index("index"), 1);
DirectoryService directoryService = new LuceneManagedDirectoryService(random);
store =
new Store(
shardId,
ImmutableSettings.EMPTY,
directoryService,
randomDistributor(random, directoryService),
new DummyShardLock(shardId));
IndexWriter writer = new IndexWriter(store.directory(), iwc);
final boolean lotsOfSegments = rarely(random);
for (Document d : docs) {
writer.addDocument(d);
if (lotsOfSegments && random.nextBoolean()) {
writer.commit();
} else if (rarely(random)) {
writer.commit();
}
}
writer.commit();
writer.close();
second = store.getMetadata();
}
Store.RecoveryDiff diff = first.recoveryDiff(second);
assertThat(first.size(), equalTo(second.size()));
for (StoreFileMetaData md : first) {
assertThat(second.get(md.name()), notNullValue());
// si files are different - containing timestamps etc
assertThat(second.get(md.name()).isSame(md), equalTo(false));
}
assertThat(diff.different.size(), equalTo(first.size()));
assertThat(
diff.identical.size(),
equalTo(0)); // in lucene 5 nothing is identical - we use random ids in file headers
assertThat(diff.missing, empty());
// check the self diff
Store.RecoveryDiff selfDiff = first.recoveryDiff(first);
assertThat(selfDiff.identical.size(), equalTo(first.size()));
assertThat(selfDiff.different, empty());
assertThat(selfDiff.missing, empty());
// lets add some deletes
Random random = new Random(seed);
IndexWriterConfig iwc =
new IndexWriterConfig(new MockAnalyzer(random)).setCodec(actualDefaultCodec());
iwc.setMergePolicy(NoMergePolicy.INSTANCE);
iwc.setUseCompoundFile(random.nextBoolean());
iwc.setMaxThreadStates(1);
iwc.setOpenMode(IndexWriterConfig.OpenMode.APPEND);
IndexWriter writer = new IndexWriter(store.directory(), iwc);
writer.deleteDocuments(new Term("id", Integer.toString(random().nextInt(numDocs))));
writer.commit();
writer.close();
Store.MetadataSnapshot metadata = store.getMetadata();
StoreFileMetaData delFile = null;
for (StoreFileMetaData md : metadata) {
if (md.name().endsWith(".liv")) {
delFile = md;
break;
}
}
Store.RecoveryDiff afterDeleteDiff = metadata.recoveryDiff(second);
if (delFile != null) {
assertThat(
afterDeleteDiff.identical.size(), equalTo(metadata.size() - 2)); // segments_N + del file
assertThat(afterDeleteDiff.different.size(), equalTo(0));
assertThat(afterDeleteDiff.missing.size(), equalTo(2));
} else {
// an entire segment must be missing (single doc segment got dropped)
assertThat(afterDeleteDiff.identical.size(), greaterThan(0));
assertThat(afterDeleteDiff.different.size(), equalTo(0));
assertThat(afterDeleteDiff.missing.size(), equalTo(1)); // the commit file is different
}
// check the self diff
selfDiff = metadata.recoveryDiff(metadata);
assertThat(selfDiff.identical.size(), equalTo(metadata.size()));
assertThat(selfDiff.different, empty());
assertThat(selfDiff.missing, empty());
// add a new commit
iwc = new IndexWriterConfig(new MockAnalyzer(random)).setCodec(actualDefaultCodec());
iwc.setMergePolicy(NoMergePolicy.INSTANCE);
iwc.setUseCompoundFile(
true); // force CFS - easier to test here since we know it will add 3 files
iwc.setMaxThreadStates(1);
iwc.setOpenMode(IndexWriterConfig.OpenMode.APPEND);
writer = new IndexWriter(store.directory(), iwc);
writer.addDocument(docs.get(0));
writer.close();
Store.MetadataSnapshot newCommitMetaData = store.getMetadata();
Store.RecoveryDiff newCommitDiff = newCommitMetaData.recoveryDiff(metadata);
if (delFile != null) {
assertThat(
newCommitDiff.identical.size(),
equalTo(
newCommitMetaData.size()
- 5)); // segments_N, del file, cfs, cfe, si for the new segment
assertThat(newCommitDiff.different.size(), equalTo(1)); // the del file must be different
assertThat(newCommitDiff.different.get(0).name(), endsWith(".liv"));
assertThat(
newCommitDiff.missing.size(), equalTo(4)); // segments_N,cfs, cfe, si for the new segment
} else {
assertThat(
newCommitDiff.identical.size(),
equalTo(newCommitMetaData.size() - 4)); // segments_N, cfs, cfe, si for the new segment
assertThat(newCommitDiff.different.size(), equalTo(0));
assertThat(
newCommitDiff.missing.size(),
equalTo(
4)); // an entire segment must be missing (single doc segment got dropped) plus the
// commit is different
}
store.deleteContent();
IOUtils.close(store);
}