private void updateDoc(IndexWriter modifier, int id, int value) throws IOException {
Document doc = new Document();
doc.add(new Field("content", "aaa", Field.Store.NO, Field.Index.ANALYZED));
doc.add(new Field("id", String.valueOf(id), Field.Store.YES, Field.Index.NOT_ANALYZED));
doc.add(new Field("value", String.valueOf(value), Field.Store.NO, Field.Index.NOT_ANALYZED));
modifier.updateDocument(new Term("id", String.valueOf(id)), doc);
}
/**
* We assume that the initial indexing has been done and a set of reference objects has been found
* and indexed in the separate directory. However further documents were added and they now need
* to get a ranked list of reference objects. So we (i) get all these new documents missing the
* field "ro-order" and (ii) add this field.
*
* @param indexPath the index to update
* @throws IOException
*/
public void updateIndex(String indexPath) throws IOException {
IndexReader reader = DirectoryReader.open(FSDirectory.open(new File(indexPath)));
int numDocs = reader.numDocs();
boolean hasDeletions = reader.hasDeletions();
int countUpdated = 0;
IndexReader readerRo = DirectoryReader.open(FSDirectory.open(new File(indexPath + "-ro")));
ImageSearcher searcher =
new GenericImageSearcher(numReferenceObjectsUsed, featureClass, featureFieldName);
Map<String, Analyzer> perField = new HashMap<String, Analyzer>(1);
perField.put("ro-order", new WhitespaceAnalyzer(LuceneUtils.LUCENE_VERSION));
PerFieldAnalyzerWrapper aWrapper =
new PerFieldAnalyzerWrapper(new SimpleAnalyzer(LuceneUtils.LUCENE_VERSION), perField);
IndexWriter iw =
new IndexWriter(
FSDirectory.open(new File(indexPath)),
new IndexWriterConfig(LuceneUtils.LUCENE_VERSION, aWrapper)
.setOpenMode(IndexWriterConfig.OpenMode.CREATE));
StringBuilder sb = new StringBuilder(256);
// Needed for check whether the document is deleted.
Bits liveDocs = MultiFields.getLiveDocs(reader);
for (int i = 0; i < numDocs; i++) {
if (reader.hasDeletions() && !liveDocs.get(i)) continue; // if it is deleted, just ignore it.
Document document = reader.document(i);
if (document.getField("ro-order") == null) { // if the field is not here we create it.
ImageSearchHits hits = searcher.search(document, readerRo);
sb.delete(0, sb.length());
for (int j = 0; j < numReferenceObjectsUsed; j++) {
sb.append(hits.doc(j).getValues("ro-id")[0]);
sb.append(' ');
}
// System.out.println(sb.toString());
document.add(new TextField("ro-order", sb.toString(), Field.Store.YES));
iw.updateDocument(
new Term(
DocumentBuilder.FIELD_NAME_IDENTIFIER,
document.getValues(DocumentBuilder.FIELD_NAME_IDENTIFIER)[0]),
document);
countUpdated++;
}
// progress report
progress.setNumDocsProcessed(progress.getNumDocsProcessed() + 1);
// debug:
System.out.println("countUpdated = " + countUpdated);
}
iw.commit();
iw.close();
}
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);
}
public void testWithPendingDeletes3() throws IOException {
// main directory
Directory dir = newDirectory();
// auxiliary directory
Directory aux = newDirectory();
setUpDirs(dir, aux);
IndexWriter writer =
newWriter(
dir,
newIndexWriterConfig(TEST_VERSION_CURRENT, new MockAnalyzer(random()))
.setOpenMode(OpenMode.APPEND));
// Adds 10 docs, then replaces them with another 10
// docs, so 10 pending deletes:
for (int i = 0; i < 20; i++) {
Document doc = new Document();
doc.add(newStringField("id", "" + (i % 10), Field.Store.NO));
doc.add(newTextField("content", "bbb " + i, Field.Store.NO));
writer.updateDocument(new Term("id", "" + (i % 10)), doc);
}
// Deletes one of the 10 added docs, leaving 9:
PhraseQuery q = new PhraseQuery();
q.add(new Term("content", "bbb"));
q.add(new Term("content", "14"));
writer.deleteDocuments(q);
writer.addIndexes(aux);
writer.forceMerge(1);
writer.commit();
verifyNumDocs(dir, 1039);
verifyTermDocs(dir, new Term("content", "aaa"), 1030);
verifyTermDocs(dir, new Term("content", "bbb"), 9);
writer.close();
dir.close();
aux.close();
}
@Override
public void run() {
try {
DirectoryReader open = null;
for (int i = 0; i < num; i++) {
Document doc = new Document(); // docs.nextDoc();
BytesRef br = new BytesRef("test");
doc.add(newStringField("id", br, Field.Store.NO));
writer.updateDocument(new Term("id", br), doc);
if (random().nextInt(3) == 0) {
if (open == null) {
open = DirectoryReader.open(writer, true);
}
DirectoryReader reader = DirectoryReader.openIfChanged(open);
if (reader != null) {
open.close();
open = reader;
}
assertEquals(
"iter: "
+ i
+ " numDocs: "
+ open.numDocs()
+ " del: "
+ open.numDeletedDocs()
+ " max: "
+ open.maxDoc(),
1,
open.numDocs());
}
}
if (open != null) {
open.close();
}
} catch (Exception e) {
throw new RuntimeException(e);
}
}
public static void updateDocument(String searchEngineId, long companyId, Document document)
throws SearchException {
if (isIndexReadOnly()) {
return;
}
if (_log.isDebugEnabled()) {
_log.debug("Document " + document.toString());
}
SearchEngine searchEngine = getSearchEngine(searchEngineId);
IndexWriter indexWriter = searchEngine.getIndexWriter();
_searchPermissionChecker.addPermissionFields(companyId, document);
SearchContext searchContext = new SearchContext();
searchContext.setCompanyId(companyId);
searchContext.setSearchEngineId(searchEngineId);
indexWriter.updateDocument(searchContext, document);
}
SingleIndex(String directoryName, String fileName, String indexPath) {
try {
Directory dir = FSDirectory.open(new File(indexPath)); // write to an directory for checking
Analyzer analyzer = new StandardAnalyzer(Version.LUCENE_48);
IndexWriterConfig iwc = new IndexWriterConfig(Version.LUCENE_48, analyzer);
IndexWriter writer = new IndexWriter(dir, iwc);
String[] files = directoryName.list();
FileInputStream fis;
try {
fis = new FileInputStream(fileName);
} catch (FileNotFoundException fnfe) {
// at least on windows, some temporary files raise this exception with an "access denied"
// message
// checking if the file can be read doesn't help
return;
}
try {
// make a new, empty document
Document doc = new Document();
// Add the path of the file as a field named "path". Use a
// field that is indexed (i.e. searchable), but don't tokenize
// the field into separate words and don't index term frequency
// or positional information:
Field pathField = new StringField("path", file.getPath(), Field.Store.YES);
doc.add(pathField);
// Add the last modified date of the file a field named "modified".
// Use a LongField that is indexed (i.e. efficiently filterable with
// NumericRangeFilter). This indexes to milli-second resolution, which
// is often too fine. You could instead create a number based on
// year/month/day/hour/minutes/seconds, down the resolution you require.
// For example the long value 2011021714 would mean
// February 17, 2011, 2-3 PM.
doc.add(new LongField("modified", file.lastModified(), Field.Store.NO));
// Add the contents of the file to a field named "contents". Specify a Reader,
// so that the text of the file is tokenized and indexed, but not stored.
// Note that FileReader expects the file to be in UTF-8 encoding.
// If that's not the case searching for special characters will fail.
doc.add(
new TextField(
"contents",
new BufferedReader(new InputStreamReader(fis, StandardCharsets.UTF_8))));
if (writer.getConfig().getOpenMode() == OpenMode.CREATE) {
// New index, so we just add the document (no old document can be there):
System.out.println("adding " + file);
writer.addDocument(doc);
} else {
// Existing index (an old copy of this document may have been indexed) so
// we use updateDocument instead to replace the old one matching the exact
// path, if present:
System.out.println("updating " + file);
writer.updateDocument(new Term("path", file.getPath()), doc);
}
} finally {
fis.close();
}
writer.close();
} catch (Exception e) {
e.printStackTrace();
}
}
/**
* Creates a set of reference objects and stores it in a new index (name "<indexPath>-ro"). Then
* creates ordered lists of reference object positions for each data item in the index with given
* feature. Finally a new index (name "<indexPath>-ms") is created where all the original
* documents as well as the new data are stored.
*
* @param indexPath the path to the original index
* @throws IOException
*/
public void createIndex(String indexPath) throws IOException {
IndexReader reader = DirectoryReader.open(FSDirectory.open(new File(indexPath)));
int numDocs = reader.numDocs();
if (numDocs < numReferenceObjects) {
throw new UnsupportedOperationException("Too few documents in index.");
}
// progress report
progress.setNumDocsAll(numDocs);
progress.setCurrentState(State.RoSelection);
boolean hasDeletions = reader.hasDeletions();
// init reference objects:
IndexWriter iw = LuceneUtils.createIndexWriter(indexPath + "-ro", true);
HashSet<Integer> referenceObjsIds = new HashSet<Integer>(numReferenceObjects);
double numDocsDouble = (double) numDocs;
while (referenceObjsIds.size() < numReferenceObjects) {
referenceObjsIds.add((int) (numDocsDouble * Math.random()));
}
int count = 0;
if (hasDeletions) {
System.err.println(
"WARNING: There are deleted docs in your index. You should "
+ "optimize your index before using this method.");
}
// progress report
progress.setCurrentState(State.RoIndexing);
// find them in the index and put them into a separate index:
for (int i : referenceObjsIds) {
count++;
Document document = reader.document(i);
document.add(new Field("ro-id", count + "", StringField.TYPE_STORED));
iw.addDocument(document);
}
iw.commit();
iw.close();
// progress report
progress.setCurrentState(State.Indexing);
// now find the reference objects for each entry ;)
IndexReader readerRo = DirectoryReader.open(FSDirectory.open(new File(indexPath + "-ro")));
ImageSearcher searcher =
new GenericImageSearcher(numReferenceObjectsUsed, featureClass, featureFieldName);
Map<String, Analyzer> analyzerPerField = new HashMap<String, Analyzer>();
analyzerPerField.put("ro-order", new WhitespaceAnalyzer(LuceneUtils.LUCENE_VERSION));
PerFieldAnalyzerWrapper aWrapper =
new PerFieldAnalyzerWrapper(
new SimpleAnalyzer(LuceneUtils.LUCENE_VERSION), analyzerPerField);
iw =
new IndexWriter(
FSDirectory.open(new File(indexPath)),
new IndexWriterConfig(LuceneUtils.LUCENE_VERSION, aWrapper)
.setOpenMode(IndexWriterConfig.OpenMode.CREATE));
StringBuilder sb = new StringBuilder(256);
// Needed for check whether the document is deleted.
Bits liveDocs = MultiFields.getLiveDocs(reader);
for (int i = 0; i < numDocs; i++) {
if (reader.hasDeletions() && !liveDocs.get(i)) continue; // if it is deleted, just ignore it.
Document document = reader.document(i);
ImageSearchHits hits = searcher.search(document, readerRo);
sb.delete(0, sb.length());
for (int j = 0; j < numReferenceObjectsUsed; j++) {
sb.append(hits.doc(j).getValues("ro-id")[0]);
sb.append(' ');
}
// System.out.println(sb.toString());
document.add(new TextField("ro-order", sb.toString(), Field.Store.YES));
iw.updateDocument(
new Term(
DocumentBuilder.FIELD_NAME_IDENTIFIER,
document.getValues(DocumentBuilder.FIELD_NAME_IDENTIFIER)[0]),
document);
// progress report
progress.setNumDocsProcessed(progress.getNumDocsProcessed() + 1);
}
iw.commit();
iw.close();
// progress report
progress.setCurrentState(State.Idle);
}
@Test
public void testMixedChecksums() throws IOException {
final ShardId shardId = new ShardId(new Index("index"), 1);
DirectoryService directoryService = new LuceneManagedDirectoryService(random());
Store store =
new Store(
shardId,
ImmutableSettings.EMPTY,
directoryService,
randomDistributor(directoryService),
new DummyShardLock(shardId));
// this time random codec....
IndexWriter writer =
new IndexWriter(
store.directory(),
newIndexWriterConfig(random(), new MockAnalyzer(random()))
.setCodec(actualDefaultCodec()));
int docs = 1 + random().nextInt(100);
for (int i = 0; i < docs; i++) {
Document doc = new Document();
doc.add(
new TextField("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()))));
writer.addDocument(doc);
}
if (random().nextBoolean()) {
for (int i = 0; i < docs; i++) {
if (random().nextBoolean()) {
Document doc = new Document();
doc.add(
new TextField(
"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));
writer.updateDocument(new Term("id", "" + i), doc);
}
}
}
if (random().nextBoolean()) {
DirectoryReader.open(writer, random().nextBoolean()).close(); // flush
}
Store.MetadataSnapshot metadata;
// check before we committed
try {
store.getMetadata();
fail("no index present - expected exception");
} catch (IndexNotFoundException ex) {
// expected
}
assertThat(store.getMetadataOrEmpty(), is(Store.MetadataSnapshot.EMPTY)); // nothing committed
writer.commit();
writer.close();
Store.LegacyChecksums checksums = new Store.LegacyChecksums();
metadata = store.getMetadata();
assertThat(metadata.asMap().isEmpty(), is(false));
for (StoreFileMetaData meta : metadata) {
try (IndexInput input = store.directory().openInput(meta.name(), IOContext.DEFAULT)) {
if (meta.checksum() == null) {
String checksum = null;
try {
CodecUtil.retrieveChecksum(input);
fail("expected a corrupt index - posting format has not checksums");
} catch (CorruptIndexException
| IndexFormatTooOldException
| IndexFormatTooNewException ex) {
try (ChecksumIndexInput checksumIndexInput =
store.directory().openChecksumInput(meta.name(), IOContext.DEFAULT)) {
checksumIndexInput.seek(meta.length());
checksum = Store.digestToString(checksumIndexInput.getChecksum());
}
// fine - it's a postings format without checksums
checksums.add(new StoreFileMetaData(meta.name(), meta.length(), checksum, null));
}
} else {
String checksum = Store.digestToString(CodecUtil.retrieveChecksum(input));
assertThat(
"File: " + meta.name() + " has a different checksum",
meta.checksum(),
equalTo(checksum));
assertThat(meta.hasLegacyChecksum(), equalTo(false));
assertThat(meta.writtenBy(), equalTo(Version.LATEST));
}
}
}
assertConsistent(store, metadata);
checksums.write(store);
metadata = store.getMetadata();
assertThat(metadata.asMap().isEmpty(), is(false));
for (StoreFileMetaData meta : metadata) {
assertThat(
"file: " + meta.name() + " has a null checksum", meta.checksum(), not(nullValue()));
if (meta.hasLegacyChecksum()) {
try (ChecksumIndexInput checksumIndexInput =
store.directory().openChecksumInput(meta.name(), IOContext.DEFAULT)) {
checksumIndexInput.seek(meta.length());
assertThat(
meta.checksum(), equalTo(Store.digestToString(checksumIndexInput.getChecksum())));
}
} else {
try (IndexInput input = store.directory().openInput(meta.name(), IOContext.DEFAULT)) {
String checksum = Store.digestToString(CodecUtil.retrieveChecksum(input));
assertThat(
"File: " + meta.name() + " has a different checksum",
meta.checksum(),
equalTo(checksum));
assertThat(meta.hasLegacyChecksum(), equalTo(false));
assertThat(meta.writtenBy(), equalTo(Version.LATEST));
}
}
}
assertConsistent(store, metadata);
TestUtil.checkIndex(store.directory());
assertDeleteContent(store, directoryService);
IOUtils.close(store);
}
/**
* Make sure if modifier tries to commit but hits disk full that modifier remains consistent and
* usable. Similar to TestIndexReader.testDiskFull().
*/
private void testOperationsOnDiskFull(boolean updates) throws IOException {
boolean debug = false;
Term searchTerm = new Term("content", "aaa");
int START_COUNT = 157;
int END_COUNT = 144;
// First build up a starting index:
MockRAMDirectory startDir = new MockRAMDirectory();
IndexWriter writer =
new IndexWriter(
startDir, new WhitespaceAnalyzer(), true, IndexWriter.MaxFieldLength.UNLIMITED);
for (int i = 0; i < 157; i++) {
Document d = new Document();
d.add(new Field("id", Integer.toString(i), Field.Store.YES, Field.Index.NOT_ANALYZED));
d.add(new Field("content", "aaa " + i, Field.Store.NO, Field.Index.ANALYZED));
writer.addDocument(d);
}
writer.close();
long diskUsage = startDir.sizeInBytes();
long diskFree = diskUsage + 10;
IOException err = null;
boolean done = false;
// Iterate w/ ever increasing free disk space:
while (!done) {
MockRAMDirectory dir = new MockRAMDirectory(startDir);
dir.setPreventDoubleWrite(false);
IndexWriter modifier =
new IndexWriter(dir, new WhitespaceAnalyzer(), IndexWriter.MaxFieldLength.UNLIMITED);
modifier.setMaxBufferedDocs(1000); // use flush or close
modifier.setMaxBufferedDeleteTerms(1000); // use flush or close
// For each disk size, first try to commit against
// dir that will hit random IOExceptions & disk
// full; after, give it infinite disk space & turn
// off random IOExceptions & retry w/ same reader:
boolean success = false;
for (int x = 0; x < 2; x++) {
double rate = 0.1;
double diskRatio = ((double) diskFree) / diskUsage;
long thisDiskFree;
String testName;
if (0 == x) {
thisDiskFree = diskFree;
if (diskRatio >= 2.0) {
rate /= 2;
}
if (diskRatio >= 4.0) {
rate /= 2;
}
if (diskRatio >= 6.0) {
rate = 0.0;
}
if (debug) {
System.out.println("\ncycle: " + diskFree + " bytes");
}
testName = "disk full during reader.close() @ " + thisDiskFree + " bytes";
} else {
thisDiskFree = 0;
rate = 0.0;
if (debug) {
System.out.println("\ncycle: same writer: unlimited disk space");
}
testName = "reader re-use after disk full";
}
dir.setMaxSizeInBytes(thisDiskFree);
dir.setRandomIOExceptionRate(rate, diskFree);
try {
if (0 == x) {
int docId = 12;
for (int i = 0; i < 13; i++) {
if (updates) {
Document d = new Document();
d.add(
new Field(
"id", Integer.toString(i), Field.Store.YES, Field.Index.NOT_ANALYZED));
d.add(new Field("content", "bbb " + i, Field.Store.NO, Field.Index.ANALYZED));
modifier.updateDocument(new Term("id", Integer.toString(docId)), d);
} else { // deletes
modifier.deleteDocuments(new Term("id", Integer.toString(docId)));
// modifier.setNorm(docId, "contents", (float)2.0);
}
docId += 12;
}
}
modifier.close();
success = true;
if (0 == x) {
done = true;
}
} catch (IOException e) {
if (debug) {
System.out.println(" hit IOException: " + e);
e.printStackTrace(System.out);
}
err = e;
if (1 == x) {
e.printStackTrace();
fail(testName + " hit IOException after disk space was freed up");
}
}
// If the close() succeeded, make sure there are
// no unreferenced files.
if (success) {
_TestUtil.checkIndex(dir);
TestIndexWriter.assertNoUnreferencedFiles(dir, "after writer.close");
}
// Finally, verify index is not corrupt, and, if
// we succeeded, we see all docs changed, and if
// we failed, we see either all docs or no docs
// changed (transactional semantics):
IndexReader newReader = null;
try {
newReader = IndexReader.open(dir, true);
} catch (IOException e) {
e.printStackTrace();
fail(
testName + ":exception when creating IndexReader after disk full during close: " + e);
}
IndexSearcher searcher = new IndexSearcher(newReader);
ScoreDoc[] hits = null;
try {
hits = searcher.search(new TermQuery(searchTerm), null, 1000).scoreDocs;
} catch (IOException e) {
e.printStackTrace();
fail(testName + ": exception when searching: " + e);
}
int result2 = hits.length;
if (success) {
if (x == 0 && result2 != END_COUNT) {
fail(
testName
+ ": method did not throw exception but hits.length for search on term 'aaa' is "
+ result2
+ " instead of expected "
+ END_COUNT);
} else if (x == 1 && result2 != START_COUNT && result2 != END_COUNT) {
// It's possible that the first exception was
// "recoverable" wrt pending deletes, in which
// case the pending deletes are retained and
// then re-flushing (with plenty of disk
// space) will succeed in flushing the
// deletes:
fail(
testName
+ ": method did not throw exception but hits.length for search on term 'aaa' is "
+ result2
+ " instead of expected "
+ START_COUNT
+ " or "
+ END_COUNT);
}
} else {
// On hitting exception we still may have added
// all docs:
if (result2 != START_COUNT && result2 != END_COUNT) {
err.printStackTrace();
fail(
testName
+ ": method did throw exception but hits.length for search on term 'aaa' is "
+ result2
+ " instead of expected "
+ START_COUNT
+ " or "
+ END_COUNT);
}
}
searcher.close();
newReader.close();
if (result2 == END_COUNT) {
break;
}
}
dir.close();
// Try again with 10 more bytes of free space:
diskFree += 10;
}
}
// IF THIS TEST FAILS ON UPGRADE GO LOOK AT THE
// OldSIMockingCodec!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
@Test
public void testWriteLegacyChecksums() throws IOException {
final ShardId shardId = new ShardId(new Index("index"), 1);
DirectoryService directoryService = new LuceneManagedDirectoryService(random());
Store store =
new Store(
shardId,
ImmutableSettings.EMPTY,
directoryService,
randomDistributor(directoryService),
new DummyShardLock(shardId));
// set default codec - all segments need checksums
final boolean usesOldCodec = randomBoolean();
IndexWriter writer =
new IndexWriter(
store.directory(),
newIndexWriterConfig(random(), new MockAnalyzer(random()))
.setCodec(usesOldCodec ? new OldSIMockingCodec() : actualDefaultCodec()));
int docs = 1 + random().nextInt(100);
for (int i = 0; i < docs; i++) {
Document doc = new Document();
doc.add(
new TextField("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()))));
writer.addDocument(doc);
}
if (random().nextBoolean()) {
for (int i = 0; i < docs; i++) {
if (random().nextBoolean()) {
Document doc = new Document();
doc.add(
new TextField(
"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));
writer.updateDocument(new Term("id", "" + i), doc);
}
}
}
if (random().nextBoolean()) {
DirectoryReader.open(writer, random().nextBoolean()).close(); // flush
}
Store.MetadataSnapshot metadata;
// check before we committed
try {
store.getMetadata();
fail("no index present - expected exception");
} catch (IndexNotFoundException ex) {
// expected
}
assertThat(store.getMetadataOrEmpty(), is(Store.MetadataSnapshot.EMPTY)); // nothing committed
writer.close();
Store.LegacyChecksums checksums = new Store.LegacyChecksums();
Map<String, StoreFileMetaData> legacyMeta = new HashMap<>();
for (String file : store.directory().listAll()) {
if (file.equals("write.lock") || file.equals(IndexFileNames.OLD_SEGMENTS_GEN)) {
continue;
}
BytesRef hash = new BytesRef();
if (file.startsWith("segments")) {
hash = Store.MetadataSnapshot.hashFile(store.directory(), file);
}
StoreFileMetaData storeFileMetaData =
new StoreFileMetaData(
file, store.directory().fileLength(file), file + "checksum", null, hash);
legacyMeta.put(file, storeFileMetaData);
checksums.add(storeFileMetaData);
}
checksums.write(store);
metadata = store.getMetadata();
Map<String, StoreFileMetaData> stringStoreFileMetaDataMap = metadata.asMap();
assertThat(legacyMeta.size(), equalTo(stringStoreFileMetaDataMap.size()));
if (usesOldCodec) {
for (StoreFileMetaData meta : legacyMeta.values()) {
assertTrue(meta.toString(), stringStoreFileMetaDataMap.containsKey(meta.name()));
assertEquals(meta.name() + "checksum", meta.checksum());
assertTrue(
meta + " vs. " + stringStoreFileMetaDataMap.get(meta.name()),
stringStoreFileMetaDataMap.get(meta.name()).isSame(meta));
}
} else {
// even if we have a legacy checksum - if we use a new codec we should reuse
for (StoreFileMetaData meta : legacyMeta.values()) {
assertTrue(meta.toString(), stringStoreFileMetaDataMap.containsKey(meta.name()));
assertFalse(
meta + " vs. " + stringStoreFileMetaDataMap.get(meta.name()),
stringStoreFileMetaDataMap.get(meta.name()).isSame(meta));
StoreFileMetaData storeFileMetaData = metadata.get(meta.name());
try (IndexInput input =
store.openVerifyingInput(meta.name(), IOContext.DEFAULT, storeFileMetaData)) {
assertTrue(storeFileMetaData.toString(), input instanceof Store.VerifyingIndexInput);
input.seek(meta.length());
Store.verify(input);
}
}
}
assertDeleteContent(store, directoryService);
IOUtils.close(store);
}
@Test
public void testCleanupFromSnapshot() throws IOException {
final ShardId shardId = new ShardId(new Index("index"), 1);
DirectoryService directoryService = new LuceneManagedDirectoryService(random());
Store store =
new Store(
shardId,
ImmutableSettings.EMPTY,
directoryService,
randomDistributor(directoryService),
new DummyShardLock(shardId));
// this time random codec....
IndexWriterConfig indexWriterConfig =
newIndexWriterConfig(random(), new MockAnalyzer(random())).setCodec(actualDefaultCodec());
// we keep all commits and that allows us clean based on multiple snapshots
indexWriterConfig.setIndexDeletionPolicy(NoDeletionPolicy.INSTANCE);
IndexWriter writer = new IndexWriter(store.directory(), indexWriterConfig);
int docs = 1 + random().nextInt(100);
int numCommits = 0;
for (int i = 0; i < docs; i++) {
if (i > 0 && randomIntBetween(0, 10) == 0) {
writer.commit();
numCommits++;
}
Document doc = new Document();
doc.add(
new TextField("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()))));
writer.addDocument(doc);
}
if (numCommits < 1) {
writer.commit();
Document doc = new Document();
doc.add(
new TextField(
"id", "" + docs++, 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()))));
writer.addDocument(doc);
}
Store.MetadataSnapshot firstMeta = store.getMetadata();
if (random().nextBoolean()) {
for (int i = 0; i < docs; i++) {
if (random().nextBoolean()) {
Document doc = new Document();
doc.add(
new TextField(
"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));
writer.updateDocument(new Term("id", "" + i), doc);
}
}
}
writer.commit();
writer.close();
Store.MetadataSnapshot secondMeta = store.getMetadata();
Store.LegacyChecksums checksums = new Store.LegacyChecksums();
Map<String, StoreFileMetaData> legacyMeta = new HashMap<>();
for (String file : store.directory().listAll()) {
if (file.equals("write.lock") || file.equals(IndexFileNames.OLD_SEGMENTS_GEN)) {
continue;
}
BytesRef hash = new BytesRef();
if (file.startsWith("segments")) {
hash = Store.MetadataSnapshot.hashFile(store.directory(), file);
}
StoreFileMetaData storeFileMetaData =
new StoreFileMetaData(
file, store.directory().fileLength(file), file + "checksum", null, hash);
legacyMeta.put(file, storeFileMetaData);
checksums.add(storeFileMetaData);
}
checksums.write(
store); // write one checksum file here - we expect it to survive all the cleanups
if (randomBoolean()) {
store.cleanupAndVerify("test", firstMeta);
String[] strings = store.directory().listAll();
int numChecksums = 0;
int numNotFound = 0;
for (String file : strings) {
assertTrue(firstMeta.contains(file) || Store.isChecksum(file));
if (Store.isChecksum(file)) {
numChecksums++;
} else if (secondMeta.contains(file) == false) {
numNotFound++;
}
}
assertTrue(
"at least one file must not be in here since we have two commits?", numNotFound > 0);
assertEquals(
"we wrote one checksum but it's gone now? - checksums are supposed to be kept",
numChecksums,
1);
} else {
store.cleanupAndVerify("test", secondMeta);
String[] strings = store.directory().listAll();
int numChecksums = 0;
int numNotFound = 0;
for (String file : strings) {
assertTrue(secondMeta.contains(file) || Store.isChecksum(file));
if (Store.isChecksum(file)) {
numChecksums++;
} else if (firstMeta.contains(file) == false) {
numNotFound++;
}
}
assertTrue(
"at least one file must not be in here since we have two commits?", numNotFound > 0);
assertEquals(
"we wrote one checksum but it's gone now? - checksums are supposed to be kept",
numChecksums,
1);
}
store.deleteContent();
IOUtils.close(store);
}
public void indexDoc() throws IOException {
Document d = new Document();
FieldType customType1 = new FieldType(TextField.TYPE_STORED);
customType1.setTokenized(false);
customType1.setOmitNorms(true);
ArrayList<Field> fields = new ArrayList<>();
String idString = getIdString();
Field idField = newField("id", idString, customType1);
fields.add(idField);
int nFields = nextInt(maxFields);
for (int i = 0; i < nFields; i++) {
FieldType customType = new FieldType();
switch (nextInt(4)) {
case 0:
break;
case 1:
customType.setStoreTermVectors(true);
break;
case 2:
customType.setStoreTermVectors(true);
customType.setStoreTermVectorPositions(true);
break;
case 3:
customType.setStoreTermVectors(true);
customType.setStoreTermVectorOffsets(true);
break;
}
switch (nextInt(4)) {
case 0:
customType.setStored(true);
customType.setOmitNorms(true);
customType.setIndexed(true);
fields.add(newField("f" + nextInt(100), getString(1), customType));
break;
case 1:
customType.setIndexed(true);
customType.setTokenized(true);
fields.add(newField("f" + nextInt(100), getString(0), customType));
break;
case 2:
customType.setStored(true);
customType.setStoreTermVectors(false);
customType.setStoreTermVectorOffsets(false);
customType.setStoreTermVectorPositions(false);
fields.add(newField("f" + nextInt(100), getString(0), customType));
break;
case 3:
customType.setStored(true);
customType.setIndexed(true);
customType.setTokenized(true);
fields.add(newField("f" + nextInt(100), getString(bigFieldSize), customType));
break;
}
}
if (sameFieldOrder) {
Collections.sort(fields, fieldNameComparator);
} else {
// random placement of id field also
Collections.swap(fields, nextInt(fields.size()), 0);
}
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(new Term("id", idString), d);
// System.out.println(Thread.currentThread().getName() + ": indexing "+d);
docs.put(idString, d);
}
protected void updateDocument(Term term, IndexDocument doc) throws Exception {
writer.updateDocument(term, doc);
}
/**
* Calls {@link IndexWriter#updateDocument(Term,Iterable)} and returns the generation that
* reflects this change.
*/
public long updateDocument(Term t, Iterable<? extends IndexableField> d) throws IOException {
writer.updateDocument(t, d);
// Return gen as of when indexing finished:
return indexingGen.get();
}
@Test
public void testRollingUpdates() throws Exception {
Random random = new Random(random().nextLong());
final BaseDirectoryWrapper dir = newDirectory();
// test checks for no unref'ed files with the IW helper method, which isn't aware of "tried to
// delete files"
if (dir instanceof MockDirectoryWrapper) {
((MockDirectoryWrapper) dir).setEnableVirusScanner(false);
}
final LineFileDocs docs = new LineFileDocs(random, true);
// provider.register(new MemoryCodec());
if (random().nextBoolean()) {
Codec.setDefault(
TestUtil.alwaysPostingsFormat(
new MemoryPostingsFormat(random().nextBoolean(), random.nextFloat())));
}
MockAnalyzer analyzer = new MockAnalyzer(random());
analyzer.setMaxTokenLength(TestUtil.nextInt(random(), 1, IndexWriter.MAX_TERM_LENGTH));
final IndexWriter w = new IndexWriter(dir, newIndexWriterConfig(analyzer));
final int SIZE = atLeast(20);
int id = 0;
IndexReader r = null;
IndexSearcher s = null;
final int numUpdates =
(int)
(SIZE * (2 + (TEST_NIGHTLY ? 200 * random().nextDouble() : 5 * random().nextDouble())));
if (VERBOSE) {
System.out.println("TEST: numUpdates=" + numUpdates);
}
int updateCount = 0;
// TODO: sometimes update ids not in order...
for (int docIter = 0; docIter < numUpdates; docIter++) {
final Document doc = docs.nextDoc();
final String myID = Integer.toString(id);
if (id == SIZE - 1) {
id = 0;
} else {
id++;
}
if (VERBOSE) {
System.out.println(" docIter=" + docIter + " id=" + id);
}
((Field) doc.getField("docid")).setStringValue(myID);
Term idTerm = new Term("docid", myID);
final boolean doUpdate;
if (s != null && updateCount < SIZE) {
TopDocs hits = s.search(new TermQuery(idTerm), 1);
assertEquals(1, hits.totalHits);
doUpdate = !w.tryDeleteDocument(r, hits.scoreDocs[0].doc);
if (VERBOSE) {
if (doUpdate) {
System.out.println(" tryDeleteDocument failed");
} else {
System.out.println(" tryDeleteDocument succeeded");
}
}
} else {
doUpdate = true;
if (VERBOSE) {
System.out.println(" no searcher: doUpdate=true");
}
}
updateCount++;
if (doUpdate) {
if (random().nextBoolean()) {
w.updateDocument(idTerm, doc);
} else {
// It's OK to not be atomic for this test (no separate thread reopening readers):
w.deleteDocuments(new TermQuery(idTerm));
w.addDocument(doc);
}
} else {
w.addDocument(doc);
}
if (docIter >= SIZE && random().nextInt(50) == 17) {
if (r != null) {
r.close();
}
final boolean applyDeletions = random().nextBoolean();
if (VERBOSE) {
System.out.println("TEST: reopen applyDeletions=" + applyDeletions);
}
r = w.getReader(applyDeletions);
if (applyDeletions) {
s = newSearcher(r);
} else {
s = null;
}
assertTrue(
"applyDeletions=" + applyDeletions + " r.numDocs()=" + r.numDocs() + " vs SIZE=" + SIZE,
!applyDeletions || r.numDocs() == SIZE);
updateCount = 0;
}
}
if (r != null) {
r.close();
}
w.commit();
assertEquals(SIZE, w.numDocs());
w.close();
TestIndexWriter.assertNoUnreferencedFiles(dir, "leftover files after rolling updates");
docs.close();
// LUCENE-4455:
SegmentInfos infos = SegmentInfos.readLatestCommit(dir);
long totalBytes = 0;
for (SegmentCommitInfo sipc : infos) {
totalBytes += sipc.sizeInBytes();
}
long totalBytes2 = 0;
for (String fileName : dir.listAll()) {
if (IndexFileNames.CODEC_FILE_PATTERN.matcher(fileName).matches()) {
totalBytes2 += dir.fileLength(fileName);
}
}
assertEquals(totalBytes2, totalBytes);
dir.close();
}
@Test
public void testNewChecksums() throws IOException {
final ShardId shardId = new ShardId(new Index("index"), 1);
DirectoryService directoryService = new LuceneManagedDirectoryService(random());
Store store =
new Store(
shardId,
ImmutableSettings.EMPTY,
directoryService,
randomDistributor(directoryService),
new DummyShardLock(shardId));
// set default codec - all segments need checksums
IndexWriter writer =
new IndexWriter(
store.directory(),
newIndexWriterConfig(random(), new MockAnalyzer(random()))
.setCodec(actualDefaultCodec()));
int docs = 1 + random().nextInt(100);
for (int i = 0; i < docs; i++) {
Document doc = new Document();
doc.add(
new TextField("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()))));
writer.addDocument(doc);
}
if (random().nextBoolean()) {
for (int i = 0; i < docs; i++) {
if (random().nextBoolean()) {
Document doc = new Document();
doc.add(
new TextField(
"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));
writer.updateDocument(new Term("id", "" + i), doc);
}
}
}
if (random().nextBoolean()) {
DirectoryReader.open(writer, random().nextBoolean()).close(); // flush
}
Store.MetadataSnapshot metadata;
// check before we committed
try {
store.getMetadata();
fail("no index present - expected exception");
} catch (IndexNotFoundException ex) {
// expected
}
assertThat(store.getMetadataOrEmpty(), is(Store.MetadataSnapshot.EMPTY)); // nothing committed
writer.commit();
writer.close();
metadata = store.getMetadata();
assertThat(metadata.asMap().isEmpty(), is(false));
for (StoreFileMetaData meta : metadata) {
try (IndexInput input = store.directory().openInput(meta.name(), IOContext.DEFAULT)) {
String checksum = Store.digestToString(CodecUtil.retrieveChecksum(input));
assertThat(
"File: " + meta.name() + " has a different checksum",
meta.checksum(),
equalTo(checksum));
assertThat(meta.hasLegacyChecksum(), equalTo(false));
assertThat(meta.writtenBy(), equalTo(Version.LATEST));
if (meta.name().endsWith(".si") || meta.name().startsWith("segments_")) {
assertThat(meta.hash().length, greaterThan(0));
}
}
}
assertConsistent(store, metadata);
TestUtil.checkIndex(store.directory());
assertDeleteContent(store, directoryService);
IOUtils.close(store);
}
