private void checkHits(ScoreDoc[] hits, int expectedCount, IndexSearcher searcher)
throws IOException {
assertEquals("total results", expectedCount, hits.length);
for (int i = 0; i < hits.length; i++) {
if (i < 10 || (i > 94 && i < 105)) {
Document d = searcher.doc(hits[i].doc);
assertEquals("check " + i, String.valueOf(i), d.get(ID_FIELD));
}
}
}
private void printHits(PrintWriter out, ScoreDoc[] hits, IndexSearcher searcher)
throws IOException {
out.println(hits.length + " total results\n");
for (int i = 0; i < hits.length; i++) {
if (i < 10 || (i > 94 && i < 105)) {
Document d = searcher.doc(hits[i].doc);
out.println(i + " " + d.get(ID_FIELD));
}
}
}
private void doTest(Random random, PrintWriter out, boolean useCompoundFiles, int MAX_DOCS)
throws Exception {
Directory directory = newDirectory();
Analyzer analyzer = new MockAnalyzer(random);
IndexWriterConfig conf = newIndexWriterConfig(analyzer);
final MergePolicy mp = conf.getMergePolicy();
mp.setNoCFSRatio(useCompoundFiles ? 1.0 : 0.0);
IndexWriter writer = new IndexWriter(directory, conf);
if (VERBOSE) {
System.out.println("TEST: now build index MAX_DOCS=" + MAX_DOCS);
}
for (int j = 0; j < MAX_DOCS; j++) {
Document d = new Document();
d.add(newTextField(PRIORITY_FIELD, HIGH_PRIORITY, Field.Store.YES));
d.add(newTextField(ID_FIELD, Integer.toString(j), Field.Store.YES));
writer.addDocument(d);
}
writer.close();
// try a search without OR
IndexReader reader = DirectoryReader.open(directory);
IndexSearcher searcher = newSearcher(reader);
Query query = new TermQuery(new Term(PRIORITY_FIELD, HIGH_PRIORITY));
out.println("Query: " + query.toString(PRIORITY_FIELD));
if (VERBOSE) {
System.out.println("TEST: search query=" + query);
}
final Sort sort = new Sort(SortField.FIELD_SCORE, new SortField(ID_FIELD, SortField.Type.INT));
ScoreDoc[] hits = searcher.search(query, null, MAX_DOCS, sort).scoreDocs;
printHits(out, hits, searcher);
checkHits(hits, MAX_DOCS, searcher);
// try a new search with OR
searcher = newSearcher(reader);
hits = null;
BooleanQuery booleanQuery = new BooleanQuery();
booleanQuery.add(
new TermQuery(new Term(PRIORITY_FIELD, HIGH_PRIORITY)), BooleanClause.Occur.SHOULD);
booleanQuery.add(
new TermQuery(new Term(PRIORITY_FIELD, MED_PRIORITY)), BooleanClause.Occur.SHOULD);
out.println("Query: " + booleanQuery.toString(PRIORITY_FIELD));
hits = searcher.search(booleanQuery, null, MAX_DOCS, sort).scoreDocs;
printHits(out, hits, searcher);
checkHits(hits, MAX_DOCS, searcher);
reader.close();
directory.close();
}
public void testTermDocsEnum() throws Exception {
Directory dir = newDirectory();
IndexWriter w =
new IndexWriter(
dir, newIndexWriterConfig(TEST_VERSION_CURRENT, new MockAnalyzer(random())));
Document d = new Document();
d.add(newStringField("f", "j", Field.Store.NO));
w.addDocument(d);
w.commit();
w.addDocument(d);
IndexReader r = w.getReader();
w.close();
DocsEnum de = MultiFields.getTermDocsEnum(r, null, "f", new BytesRef("j"));
assertEquals(0, de.nextDoc());
assertEquals(1, de.nextDoc());
assertEquals(DocIdSetIterator.NO_MORE_DOCS, de.nextDoc());
r.close();
dir.close();
}
public void testSeparateEnums() throws Exception {
Directory dir = newDirectory();
IndexWriter w =
new IndexWriter(
dir, newIndexWriterConfig(TEST_VERSION_CURRENT, new MockAnalyzer(random())));
Document d = new Document();
d.add(newStringField("f", "j", Field.Store.NO));
w.addDocument(d);
w.commit();
w.addDocument(d);
IndexReader r = w.getReader();
w.close();
DocsEnum d1 = _TestUtil.docs(random(), r, "f", new BytesRef("j"), null, null, 0);
DocsEnum d2 = _TestUtil.docs(random(), r, "f", new BytesRef("j"), null, null, 0);
assertEquals(0, d1.nextDoc());
assertEquals(0, d2.nextDoc());
r.close();
dir.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 void testRandom() throws Exception {
int num = atLeast(2);
for (int iter = 0; iter < num; iter++) {
if (VERBOSE) {
System.out.println("TEST: iter=" + iter);
}
Directory dir = newDirectory();
IndexWriter w =
new IndexWriter(
dir,
newIndexWriterConfig(TEST_VERSION_CURRENT, new MockAnalyzer(random()))
.setMergePolicy(NoMergePolicy.COMPOUND_FILES));
_TestUtil.keepFullyDeletedSegments(w);
Map<BytesRef, List<Integer>> docs = new HashMap<BytesRef, List<Integer>>();
Set<Integer> deleted = new HashSet<Integer>();
List<BytesRef> terms = new ArrayList<BytesRef>();
int numDocs = _TestUtil.nextInt(random(), 1, 100 * RANDOM_MULTIPLIER);
Document doc = new Document();
Field f = newStringField("field", "", Field.Store.NO);
doc.add(f);
Field id = newStringField("id", "", Field.Store.NO);
doc.add(id);
boolean onlyUniqueTerms = random().nextBoolean();
if (VERBOSE) {
System.out.println("TEST: onlyUniqueTerms=" + onlyUniqueTerms + " numDocs=" + numDocs);
}
Set<BytesRef> uniqueTerms = new HashSet<BytesRef>();
for (int i = 0; i < numDocs; i++) {
if (!onlyUniqueTerms && random().nextBoolean() && terms.size() > 0) {
// re-use existing term
BytesRef term = terms.get(random().nextInt(terms.size()));
docs.get(term).add(i);
f.setStringValue(term.utf8ToString());
} else {
String s = _TestUtil.randomUnicodeString(random(), 10);
BytesRef term = new BytesRef(s);
if (!docs.containsKey(term)) {
docs.put(term, new ArrayList<Integer>());
}
docs.get(term).add(i);
terms.add(term);
uniqueTerms.add(term);
f.setStringValue(s);
}
id.setStringValue("" + i);
w.addDocument(doc);
if (random().nextInt(4) == 1) {
w.commit();
}
if (i > 0 && random().nextInt(20) == 1) {
int delID = random().nextInt(i);
deleted.add(delID);
w.deleteDocuments(new Term("id", "" + delID));
if (VERBOSE) {
System.out.println("TEST: delete " + delID);
}
}
}
if (VERBOSE) {
List<BytesRef> termsList = new ArrayList<BytesRef>(uniqueTerms);
Collections.sort(termsList, BytesRef.getUTF8SortedAsUTF16Comparator());
System.out.println("TEST: terms in UTF16 order:");
for (BytesRef b : termsList) {
System.out.println(" " + UnicodeUtil.toHexString(b.utf8ToString()) + " " + b);
for (int docID : docs.get(b)) {
if (deleted.contains(docID)) {
System.out.println(" " + docID + " (deleted)");
} else {
System.out.println(" " + docID);
}
}
}
}
IndexReader reader = w.getReader();
w.close();
if (VERBOSE) {
System.out.println("TEST: reader=" + reader);
}
Bits liveDocs = MultiFields.getLiveDocs(reader);
for (int delDoc : deleted) {
assertFalse(liveDocs.get(delDoc));
}
for (int i = 0; i < 100; i++) {
BytesRef term = terms.get(random().nextInt(terms.size()));
if (VERBOSE) {
System.out.println(
"TEST: seek term=" + UnicodeUtil.toHexString(term.utf8ToString()) + " " + term);
}
DocsEnum docsEnum = _TestUtil.docs(random(), reader, "field", term, liveDocs, null, 0);
assertNotNull(docsEnum);
for (int docID : docs.get(term)) {
if (!deleted.contains(docID)) {
assertEquals(docID, docsEnum.nextDoc());
}
}
assertEquals(DocIdSetIterator.NO_MORE_DOCS, docsEnum.nextDoc());
}
reader.close();
dir.close();
}
}
@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();
}
public static void main(String[] args) throws Exception {
// the IndexReader object is the main handle that will give you
// all the documents, terms and inverted index
IndexReader r = IndexReader.open(FSDirectory.open(new File("index")));
// You can figure out the number of documents using the maxDoc() function
System.out.println("The number of documents in this index is: " + r.maxDoc());
int i = 0;
// You can find out all the terms that have been indexed using the terms() function
TermEnum t = r.terms();
while (t.next()) {
// Since there are so many terms, let us try printing only term #100000-#100010
if (i > 100000) System.out.println("[" + i + "] " + t.term().text());
if (++i > 100010) break;
}
// You can create your own query terms by calling the Term constructor, with the field
// 'contents'
// In the following example, the query term is 'brute'
Term te = new Term("contents", "brute");
// You can also quickly find out the number of documents that have term t
System.out.println("Number of documents with the word 'brute' is: " + r.docFreq(te));
// You can use the inverted index to find out all the documents that contain the term 'brute'
// by using the termDocs function
TermDocs td = r.termDocs(te);
while (td.next()) {
System.out.println(
"Document number ["
+ td.doc()
+ "] contains the term 'brute' "
+ td.freq()
+ " time(s).");
}
// You can find the URL of the a specific document number using the document() function
// For example, the URL for document number 14191 is:
Document d = r.document(14191);
String url =
d.getFieldable("path")
.stringValue(); // the 'path' field of the Document object holds the URL
System.out.println(url.replace("%%", "/"));
// -------- Now let us use all of the functions above to make something useful --------
// The following bit of code is a worked out example of how to get a bunch of documents
// in response to a query and show them (without ranking them according to TF/IDF)
Scanner sc = new Scanner(System.in);
String str = "";
System.out.print("query> ");
while (!(str = sc.nextLine()).equals("quit")) {
String[] terms = str.split("\\s+");
for (String word : terms) {
Term term = new Term("contents", word);
TermDocs tdocs = r.termDocs(term);
while (tdocs.next()) {
String d_url =
r.document(tdocs.doc()).getFieldable("path").stringValue().replace("%%", "/");
System.out.println("[" + tdocs.doc() + "] " + d_url);
}
}
System.out.print("query> ");
}
}
@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);
}
@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);
}
@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);
}
// 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);
}
// private static int[] oldToNew(IndexReader reader, Searcher searcher) throws IOException {
private static DocScore[] newToOld(IndexReader reader, Searcher searcher) throws IOException {
int readerMax = reader.maxDoc();
DocScore[] newToOld = new DocScore[readerMax];
// use site, an indexed, un-tokenized field to get boost
// byte[] boosts = reader.norms("site"); TODO MC
/* TODO MC */
Document docMeta;
Pattern includes = Pattern.compile("\\|");
String value = NutchConfiguration.create().get(INCLUDE_EXTENSIONS_KEY, "");
String includeExtensions[] = includes.split(value);
Hashtable<String, Boolean> validExtensions = new Hashtable<String, Boolean>();
for (int i = 0; i < includeExtensions.length; i++) {
validExtensions.put(includeExtensions[i], true);
System.out.println("extension boosted " + includeExtensions[i]);
}
/* TODO MC */
for (int oldDoc = 0; oldDoc < readerMax; oldDoc++) {
float score;
if (reader.isDeleted(oldDoc)) {
// score = 0.0f;
score = -1f; // TODO MC
} else {
// score = Similarity.decodeNorm(boosts[oldDoc]); TODO MC
/* TODO MC */
docMeta = searcher.doc(oldDoc);
if (validExtensions.get(docMeta.get("subType"))
== null) { // searched extensions will have higher scores
score = -0.5f;
} else {
score = Integer.parseInt(docMeta.get("inlinks"));
/*
if (score==0) {
score=0.001f; // TODO MC - to not erase
}
*/
}
/* TODO MC */
// System.out.println("Score for old document "+oldDoc+" is "+score+" and type
// "+docMeta.get("subType")); // TODO MC debug remove
}
DocScore docScore = new DocScore();
docScore.doc = oldDoc;
docScore.score = score;
newToOld[oldDoc] = docScore;
}
System.out.println("Sorting " + newToOld.length + " documents.");
Arrays.sort(newToOld);
// HeapSorter.sort(newToOld); // TODO MC - due to the lack of space
/* TODO MC
int[] oldToNew = new int[readerMax];
for (int newDoc = 0; newDoc < readerMax; newDoc++) {
DocScore docScore = newToOld[newDoc];
//oldToNew[docScore.oldDoc] = docScore.score > 0.0f ? newDoc : -1; // TODO MC
oldToNew[docScore.oldDoc] = newDoc; // TODO MC
}
*/
/* TODO MC *
for (int newDoc = 0; newDoc < readerMax; newDoc++) {
DocScore docScore = newToOld[newDoc];
System.out.println("Score for new document "+newDoc+" is "+docScore.score); // TODO MC debug remove
}
* TODO MC */
// return oldToNew; TODO MC
return newToOld; // TODO MC
}