private void addUncompressed(Iterable<Number> values, int count) throws IOException {
meta.writeVInt(count);
meta.writeByte(UNCOMPRESSED); // uncompressed byte[]
meta.writeLong(data.getFilePointer());
for (Number nv : values) {
data.writeByte(nv.byteValue());
}
}
@Override
public void write(
Directory directory,
SegmentInfo segmentInfo,
String segmentSuffix,
FieldInfos infos,
IOContext context)
throws IOException {
final String fileName =
IndexFileNames.segmentFileName(
segmentInfo.name, segmentSuffix, Lucene46FieldInfosFormat.EXTENSION);
try (IndexOutput output = directory.createOutput(fileName, context)) {
CodecUtil.writeHeader(
output, Lucene46FieldInfosFormat.CODEC_NAME, Lucene46FieldInfosFormat.FORMAT_CURRENT);
output.writeVInt(infos.size());
for (FieldInfo fi : infos) {
IndexOptions indexOptions = fi.getIndexOptions();
byte bits = 0x0;
if (fi.hasVectors()) bits |= Lucene46FieldInfosFormat.STORE_TERMVECTOR;
if (fi.omitsNorms()) bits |= Lucene46FieldInfosFormat.OMIT_NORMS;
if (fi.hasPayloads()) bits |= Lucene46FieldInfosFormat.STORE_PAYLOADS;
if (fi.getIndexOptions() != IndexOptions.NONE) {
bits |= Lucene46FieldInfosFormat.IS_INDEXED;
assert indexOptions.compareTo(IndexOptions.DOCS_AND_FREQS_AND_POSITIONS) >= 0
|| !fi.hasPayloads();
if (indexOptions == IndexOptions.DOCS) {
bits |= Lucene46FieldInfosFormat.OMIT_TERM_FREQ_AND_POSITIONS;
} else if (indexOptions == IndexOptions.DOCS_AND_FREQS_AND_POSITIONS_AND_OFFSETS) {
bits |= Lucene46FieldInfosFormat.STORE_OFFSETS_IN_POSTINGS;
} else if (indexOptions == IndexOptions.DOCS_AND_FREQS) {
bits |= Lucene46FieldInfosFormat.OMIT_POSITIONS;
}
}
output.writeString(fi.name);
output.writeVInt(fi.number);
output.writeByte(bits);
// pack the DV types in one byte
final byte dv = docValuesByte(fi.getDocValuesType());
final byte nrm = docValuesByte(fi.hasNorms() ? DocValuesType.NUMERIC : DocValuesType.NONE);
assert (dv & (~0xF)) == 0 && (nrm & (~0x0F)) == 0;
byte val = (byte) (0xff & ((nrm << 4) | dv));
output.writeByte(val);
output.writeLong(fi.getDocValuesGen());
output.writeStringStringMap(fi.attributes());
}
CodecUtil.writeFooter(output);
}
}
// Make sure we don't somehow use more than 1 descriptor
// when reading a CFS with many subs:
public void testManySubFiles() throws IOException {
final Directory d = newFSDirectory(_TestUtil.getTempDir("CFSManySubFiles"));
final int FILE_COUNT = 10000;
for (int fileIdx = 0; fileIdx < FILE_COUNT; fileIdx++) {
IndexOutput out = d.createOutput("file." + fileIdx);
out.writeByte((byte) fileIdx);
out.close();
}
final CompoundFileWriter cfw = new CompoundFileWriter(d, "c.cfs");
for (int fileIdx = 0; fileIdx < FILE_COUNT; fileIdx++) {
cfw.addFile("file." + fileIdx);
}
cfw.close();
final IndexInput[] ins = new IndexInput[FILE_COUNT];
final CompoundFileReader cfr = new CompoundFileReader(d, "c.cfs");
for (int fileIdx = 0; fileIdx < FILE_COUNT; fileIdx++) {
ins[fileIdx] = cfr.openInput("file." + fileIdx);
}
for (int fileIdx = 0; fileIdx < FILE_COUNT; fileIdx++) {
assertEquals((byte) fileIdx, ins[fileIdx].readByte());
}
for (int fileIdx = 0; fileIdx < FILE_COUNT; fileIdx++) {
ins[fileIdx].close();
}
cfr.close();
d.close();
}
private void demo_FSIndexInputBug(Directory fsdir, String file) throws IOException {
// Setup the test file - we need more than 1024 bytes
IndexOutput os = fsdir.createOutput(file);
for (int i = 0; i < 2000; i++) {
os.writeByte((byte) i);
}
os.close();
IndexInput in = fsdir.openInput(file);
// This read primes the buffer in IndexInput
in.readByte();
// Close the file
in.close();
// ERROR: this call should fail, but succeeds because the buffer
// is still filled
in.readByte();
// ERROR: this call should fail, but succeeds for some reason as well
in.seek(1099);
try {
// OK: this call correctly fails. We are now past the 1024 internal
// buffer, so an actual IO is attempted, which fails
in.readByte();
fail("expected readByte() to throw exception");
} catch (IOException e) {
// expected exception
}
}
private void writeTable(
Iterable<Number> values,
FormatAndBits compression,
int count,
NormMap uniqueValues,
int numOrds)
throws IOException {
data.writeVInt(PackedInts.VERSION_CURRENT);
data.writeVInt(compression.format.getId());
data.writeVInt(compression.bitsPerValue);
data.writeVInt(numOrds);
for (int i = 0; i < numOrds; i++) {
data.writeByte(uniqueValues.values[i]);
}
final PackedInts.Writer writer =
PackedInts.getWriterNoHeader(
data,
compression.format,
count,
compression.bitsPerValue,
PackedInts.DEFAULT_BUFFER_SIZE);
for (Number nv : values) {
int ord = uniqueValues.ord(nv.byteValue());
if (ord < numOrds) {
writer.add(ord);
} else {
writer.add(numOrds); // collapses all ords >= numOrds into a single value
}
}
writer.finish();
}
// encodes values as sparse array: keys[] and values[]
// access is log(N) where N = keys.length (slow!)
// so this is only appropriate as an exception table for patched, or when common value is 0 (wont
// be accessed by searching)
private void addIndirect(
FieldInfo field,
final Iterable<Number> values,
int count,
final NormMap uniqueValues,
final int minOrd)
throws IOException {
int commonCount = uniqueValues.freqs[minOrd];
meta.writeVInt(count - commonCount);
meta.writeByte(INDIRECT);
meta.writeLong(data.getFilePointer());
// write docs with value
writeDocsWithValue(values, uniqueValues, minOrd);
// write actual values
writeNormsField(
field,
new Iterable<Number>() {
@Override
public Iterator<Number> iterator() {
return new FilterIterator<Number, Number>(values.iterator()) {
@Override
protected boolean predicateFunction(Number value) {
return uniqueValues.ord(value.byteValue()) > minOrd;
}
};
}
},
1);
}
/** Creates a file of the specified size with random data. */
private void createRandomFile(Directory dir, String name, int size) throws IOException {
IndexOutput os = dir.createOutput(name);
for (int i = 0; i < size; i++) {
byte b = (byte) (Math.random() * 256);
os.writeByte(b);
}
os.close();
}
/**
* Creates a file of the specified size with sequential data. The first byte is written as the
* start byte provided. All subsequent bytes are computed as start + offset where offset is the
* number of the byte.
*/
private void createSequenceFile(Directory dir, String name, byte start, int size)
throws IOException {
IndexOutput os = dir.createOutput(name);
for (int i = 0; i < size; i++) {
os.writeByte(start);
start++;
}
os.close();
}
private void addTableCompressed(
Iterable<Number> values, FormatAndBits compression, int count, NormMap uniqueValues)
throws IOException {
meta.writeVInt(count);
meta.writeByte(TABLE_COMPRESSED); // table-compressed
meta.writeLong(data.getFilePointer());
writeTable(values, compression, count, uniqueValues, uniqueValues.size);
}
/**
* It creates a file with fixed size using a RepeatableLongByteSequence object to generate a
* repeatable content
*
* @param dir The Directory containing the file to create
* @param fileName The file name to create
* @param contentFileSize The size content file to create
* @throws IOException
*/
private void createFileWithRepeatableContent(
Directory dir, String fileName, final int contentFileSize) throws IOException {
IndexOutput indexOutput = dir.createOutput(fileName);
RepeatableLongByteSequence bytesGenerator = new RepeatableLongByteSequence();
for (int i = 0; i < contentFileSize; i++) {
indexOutput.writeByte(bytesGenerator.nextByte());
}
indexOutput.close();
}
private void writeField() throws IOException {
// remember where this field is written
currentField.tvfPointer = tvf.getFilePointer();
// System.out.println("Field Pointer: " + currentField.tvfPointer);
final int size = terms.size();
tvf.writeVInt(size);
boolean storePositions = currentField.storePositions;
boolean storeOffsets = currentField.storeOffsets;
byte bits = 0x0;
if (storePositions) bits |= STORE_POSITIONS_WITH_TERMVECTOR;
if (storeOffsets) bits |= STORE_OFFSET_WITH_TERMVECTOR;
tvf.writeByte(bits);
String lastTermText = "";
for (int i = 0; i < size; i++) {
TVTerm term = (TVTerm) terms.elementAt(i);
int start = StringHelper.stringDifference(lastTermText, term.termText);
int length = term.termText.length() - start;
tvf.writeVInt(start); // write shared prefix length
tvf.writeVInt(length); // write delta length
tvf.writeChars(term.termText, start, length); // write delta chars
tvf.writeVInt(term.freq);
lastTermText = term.termText;
if (storePositions) {
if (term.positions == null)
throw new IllegalStateException("Trying to write positions that are null!");
// use delta encoding for positions
int position = 0;
for (int j = 0; j < term.freq; j++) {
tvf.writeVInt(term.positions[j] - position);
position = term.positions[j];
}
}
if (storeOffsets) {
if (term.offsets == null)
throw new IllegalStateException("Trying to write offsets that are null!");
// use delta encoding for offsets
int position = 0;
for (int j = 0; j < term.freq; j++) {
tvf.writeVInt(term.offsets[j].getStartOffset() - position);
tvf.writeVInt(
term.offsets[j].getEndOffset()
- term.offsets[j].getStartOffset()); // Save the diff between the two.
position = term.offsets[j].getEndOffset();
}
}
}
}
@Test
public void testVerifyingIndexOutput() throws IOException {
Directory dir = newDirectory();
IndexOutput output = dir.createOutput("foo.bar", IOContext.DEFAULT);
int iters = scaledRandomIntBetween(10, 100);
for (int i = 0; i < iters; i++) {
BytesRef bytesRef = new BytesRef(TestUtil.randomRealisticUnicodeString(random(), 10, 1024));
output.writeBytes(bytesRef.bytes, bytesRef.offset, bytesRef.length);
}
CodecUtil.writeFooter(output);
output.close();
IndexInput indexInput = dir.openInput("foo.bar", IOContext.DEFAULT);
String checksum = Store.digestToString(CodecUtil.retrieveChecksum(indexInput));
indexInput.seek(0);
BytesRef ref = new BytesRef(scaledRandomIntBetween(1, 1024));
long length = indexInput.length();
IndexOutput verifyingOutput =
new Store.LuceneVerifyingIndexOutput(
new StoreFileMetaData("foo1.bar", length, checksum),
dir.createOutput("foo1.bar", IOContext.DEFAULT));
while (length > 0) {
if (random().nextInt(10) == 0) {
verifyingOutput.writeByte(indexInput.readByte());
length--;
} else {
int min = (int) Math.min(length, ref.bytes.length);
indexInput.readBytes(ref.bytes, ref.offset, min);
verifyingOutput.writeBytes(ref.bytes, ref.offset, min);
length -= min;
}
}
Store.verify(verifyingOutput);
verifyingOutput.writeByte((byte) 0x0);
try {
Store.verify(verifyingOutput);
fail("should be a corrupted index");
} catch (CorruptIndexException | IndexFormatTooOldException | IndexFormatTooNewException ex) {
// ok
}
IOUtils.close(indexInput, verifyingOutput, dir);
}
@Test
public void testWriteChunksDefaultChunks() throws Exception {
Cache cache = cacheManager.getCache();
Directory dir = DirectoryBuilder.newDirectoryInstance(cache, cache, cache, INDEXNAME).create();
final String testText = "This is some rubbish";
final byte[] testTextAsBytes = testText.getBytes();
IndexOutput io = dir.createOutput("MyNewFile.txt");
io.writeByte((byte) 1);
io.writeByte((byte) 2);
io.writeByte((byte) 3);
io.writeBytes(testTextAsBytes, testTextAsBytes.length);
io.close();
DirectoryIntegrityCheck.verifyDirectoryStructure(cache, INDEXNAME);
FileCacheKey fileCacheKey = new FileCacheKey(INDEXNAME, "MyNewFile.txt");
assert null != cache.get(fileCacheKey);
FileMetadata metadata = (FileMetadata) cache.get(fileCacheKey);
AssertJUnit.assertEquals(testTextAsBytes.length + 3, metadata.getSize());
assert null
!= cache.get(
new ChunkCacheKey(
INDEXNAME, "MyNewFile.txt", 0, DirectoryBuilderImpl.DEFAULT_BUFFER_SIZE));
// test contents by reading:
IndexInput ii = dir.openInput("MyNewFile.txt");
assert ii.readByte() == 1;
assert ii.readByte() == 2;
assert ii.readByte() == 3;
byte[] buf = new byte[testTextAsBytes.length];
ii.readBytes(buf, 0, testTextAsBytes.length);
ii.close();
assert testText.equals(new String(buf).trim());
dir.close();
DirectoryIntegrityCheck.verifyDirectoryStructure(cache, INDEXNAME);
}
/** finishes writing */
public void finish() throws IOException {
if (count != numValues) {
throw new IllegalStateException(
"Wrong number of values added, expected: " + numValues + ", got: " + count);
}
assert !finished;
flush();
// pad for fast io: we actually only need this for certain BPV, but its just 3 bytes...
for (int i = 0; i < 3; i++) {
output.writeByte((byte) 0);
}
finished = true;
}
private void addDeltaCompressed(Iterable<Number> values, int count) throws IOException {
meta.writeVInt(count);
meta.writeByte(DELTA_COMPRESSED); // delta-compressed
meta.writeLong(data.getFilePointer());
data.writeVInt(PackedInts.VERSION_CURRENT);
data.writeVInt(BLOCK_SIZE);
final BlockPackedWriter writer = new BlockPackedWriter(data, BLOCK_SIZE);
for (Number nv : values) {
writer.add(nv.longValue());
}
writer.finish();
}
public void testWriteChunks() throws Exception {
final int BUFFER_SIZE = 64;
Cache cache = cacheManager.getCache();
Directory dir =
DirectoryBuilder.newDirectoryInstance(cache, cache, cache, INDEXNAME)
.chunkSize(BUFFER_SIZE)
.create();
IndexOutput io = dir.createOutput("MyNewFile.txt");
io.writeByte((byte) 66);
io.writeByte((byte) 69);
io.flush();
io.close();
assert dir.fileExists("MyNewFile.txt");
assert null != cache.get(new ChunkCacheKey(INDEXNAME, "MyNewFile.txt", 0, BUFFER_SIZE));
// test contents by reading:
byte[] buf = new byte[9];
IndexInput ii = dir.openInput("MyNewFile.txt");
ii.readBytes(buf, 0, (int) ii.length());
ii.close();
assert new String(new byte[] {66, 69}).equals(new String(buf).trim());
String testText =
"This is some rubbish again that will span more than one chunk - one hopes. Who knows, maybe even three or four chunks.";
io = dir.createOutput("MyNewFile.txt");
io.seek(0);
io.writeBytes(testText.getBytes(), 0, testText.length());
io.close();
// now compare.
byte[] chunk1 =
(byte[]) cache.get(new ChunkCacheKey(INDEXNAME, "MyNewFile.txt", 0, BUFFER_SIZE));
byte[] chunk2 =
(byte[]) cache.get(new ChunkCacheKey(INDEXNAME, "MyNewFile.txt", 1, BUFFER_SIZE));
assert null != chunk1;
assert null != chunk2;
assert testText.equals(new String(chunk1) + new String(chunk2).trim());
dir.close();
DirectoryIntegrityCheck.verifyDirectoryStructure(cache, INDEXNAME);
}
@Test
public void testVerifyingIndexOutputWithBogusInput() throws IOException {
Directory dir = newDirectory();
int length = scaledRandomIntBetween(10, 1024);
IndexOutput verifyingOutput =
new Store.LuceneVerifyingIndexOutput(
new StoreFileMetaData("foo1.bar", length, ""),
dir.createOutput("foo1.bar", IOContext.DEFAULT));
try {
while (length > 0) {
verifyingOutput.writeByte((byte) random().nextInt());
length--;
}
fail("should be a corrupted index");
} catch (CorruptIndexException | IndexFormatTooOldException | IndexFormatTooNewException ex) {
// ok
}
IOUtils.close(verifyingOutput, dir);
}
private void insertData(CoherenceDirectory dir, String fileName) throws IOException {
byte[] test = new byte[] {1, 2, 3, 4, 5, 6, 7, 8};
IndexOutput indexOutput = dir.createOutput(fileName);
indexOutput.writeBytes(new byte[] {2, 4, 6, 7, 8}, 5);
indexOutput.writeInt(-1);
indexOutput.writeLong(10);
indexOutput.writeInt(0);
indexOutput.writeInt(0);
indexOutput.writeBytes(test, 8);
indexOutput.writeBytes(test, 5);
indexOutput.seek(0);
indexOutput.writeByte((byte) 8);
if (dir.getBucketSize() > 4) {
indexOutput.seek(2);
indexOutput.writeBytes(new byte[] {1, 2}, 2);
}
indexOutput.close();
}
// encodes common values in a table, and the rest of the values as exceptions using INDIRECT.
// the exceptions should not be accessed very often, since the values are uncommon
private void addPatchedTable(
FieldInfo field,
final Iterable<Number> values,
final int numCommonValues,
int commonValuesCount,
int count,
final NormMap uniqueValues)
throws IOException {
meta.writeVInt(count);
meta.writeByte(PATCHED_TABLE);
meta.writeLong(data.getFilePointer());
assert numCommonValues == 3 || numCommonValues == 15;
FormatAndBits compression = fastestFormatAndBits(numCommonValues);
writeTable(values, compression, count, uniqueValues, numCommonValues);
meta.writeVInt(field.number);
addIndirect(field, values, count - commonValuesCount, uniqueValues, numCommonValues);
}
private void insertData(ByteBufferDirectory dir, int bufferSizeInBytes) throws IOException {
byte[] test = new byte[] {1, 2, 3, 4, 5, 6, 7, 8};
IndexOutput indexOutput = dir.createOutput("value1", IOContext.DEFAULT);
indexOutput.writeBytes(new byte[] {2, 4, 6, 7, 8}, 5);
indexOutput.writeInt(-1);
indexOutput.writeLong(10);
indexOutput.writeInt(0);
indexOutput.writeInt(0);
indexOutput.writeBytes(test, 8);
indexOutput.writeBytes(test, 5);
indexOutput.seek(0);
indexOutput.writeByte((byte) 8);
if (bufferSizeInBytes > 4) {
indexOutput.seek(2);
indexOutput.writeBytes(new byte[] {1, 2}, 2);
}
indexOutput.close();
}
// encodes only uncommon values in a sparse bitset
// access is constant time, and the common case is predictable
// exceptions nest either to CONST (if there are only 2 values), or INDIRECT (if there are > 2
// values)
private void addPatchedBitset(
FieldInfo field, final Iterable<Number> values, int count, NormMap uniqueValues)
throws IOException {
int commonCount = uniqueValues.freqs[0];
meta.writeVInt(count - commonCount);
meta.writeByte(PATCHED_BITSET);
meta.writeLong(data.getFilePointer());
// write docs with value
writeDocsWithValue(values, uniqueValues, 0);
// write exceptions: only two cases make sense
// bpv = 1 (folded into sparse bitset already)
// bpv > 1 (add indirect exception table)
meta.writeVInt(field.number);
if (uniqueValues.size == 2) {
// special case: implicit in bitset
addConstant(uniqueValues.values[1]);
} else {
// exception table
addIndirect(field, values, count, uniqueValues, 0);
}
}
@Override
public void write(int b) throws IOException {
output.writeByte((byte) b);
}
@Override
public void writeByte(byte b) throws IOException {
delegate.writeByte(b);
}
private void write(Directory directory) throws IOException {
long nextGeneration = getNextPendingGeneration();
String segmentFileName =
IndexFileNames.fileNameFromGeneration(IndexFileNames.PENDING_SEGMENTS, "", nextGeneration);
// Always advance the generation on write:
generation = nextGeneration;
IndexOutput segnOutput = null;
boolean success = false;
try {
segnOutput = directory.createOutput(segmentFileName, IOContext.DEFAULT);
CodecUtil.writeIndexHeader(
segnOutput,
"segments",
VERSION_CURRENT,
StringHelper.randomId(),
Long.toString(nextGeneration, Character.MAX_RADIX));
segnOutput.writeVInt(Version.LATEST.major);
segnOutput.writeVInt(Version.LATEST.minor);
segnOutput.writeVInt(Version.LATEST.bugfix);
segnOutput.writeLong(version);
segnOutput.writeInt(counter); // write counter
segnOutput.writeInt(size());
if (size() > 0) {
Version minSegmentVersion = null;
// We do a separate loop up front so we can write the minSegmentVersion before
// any SegmentInfo; this makes it cleaner to throw IndexFormatTooOldExc at read time:
for (SegmentCommitInfo siPerCommit : this) {
Version segmentVersion = siPerCommit.info.getVersion();
if (minSegmentVersion == null || segmentVersion.onOrAfter(minSegmentVersion) == false) {
minSegmentVersion = segmentVersion;
}
}
segnOutput.writeVInt(minSegmentVersion.major);
segnOutput.writeVInt(minSegmentVersion.minor);
segnOutput.writeVInt(minSegmentVersion.bugfix);
}
// write infos
for (SegmentCommitInfo siPerCommit : this) {
SegmentInfo si = siPerCommit.info;
segnOutput.writeString(si.name);
byte segmentID[] = si.getId();
// TODO: remove this in lucene 6, we don't need to include 4.x segments in commits anymore
if (segmentID == null) {
segnOutput.writeByte((byte) 0);
} else {
if (segmentID.length != StringHelper.ID_LENGTH) {
throw new IllegalStateException(
"cannot write segment: invalid id segment="
+ si.name
+ "id="
+ StringHelper.idToString(segmentID));
}
segnOutput.writeByte((byte) 1);
segnOutput.writeBytes(segmentID, segmentID.length);
}
segnOutput.writeString(si.getCodec().getName());
segnOutput.writeLong(siPerCommit.getDelGen());
int delCount = siPerCommit.getDelCount();
if (delCount < 0 || delCount > si.maxDoc()) {
throw new IllegalStateException(
"cannot write segment: invalid maxDoc segment="
+ si.name
+ " maxDoc="
+ si.maxDoc()
+ " delCount="
+ delCount);
}
segnOutput.writeInt(delCount);
segnOutput.writeLong(siPerCommit.getFieldInfosGen());
segnOutput.writeLong(siPerCommit.getDocValuesGen());
segnOutput.writeSetOfStrings(siPerCommit.getFieldInfosFiles());
final Map<Integer, Set<String>> dvUpdatesFiles = siPerCommit.getDocValuesUpdatesFiles();
segnOutput.writeInt(dvUpdatesFiles.size());
for (Entry<Integer, Set<String>> e : dvUpdatesFiles.entrySet()) {
segnOutput.writeInt(e.getKey());
segnOutput.writeSetOfStrings(e.getValue());
}
}
segnOutput.writeMapOfStrings(userData);
CodecUtil.writeFooter(segnOutput);
segnOutput.close();
directory.sync(Collections.singleton(segmentFileName));
success = true;
} finally {
if (success) {
pendingCommit = true;
} else {
// We hit an exception above; try to close the file
// but suppress any exception:
IOUtils.closeWhileHandlingException(segnOutput);
// Try not to leave a truncated segments_N file in
// the index:
IOUtils.deleteFilesIgnoringExceptions(directory, segmentFileName);
}
}
}
@Override
public void writeByte(byte b) throws IOException {
wrapped.writeByte(b);
}
private void addConstant(byte constant) throws IOException {
meta.writeVInt(0);
meta.writeByte(CONST_COMPRESSED);
meta.writeLong(constant);
}
@Test
public void testWriteUsingSeekMethod() throws IOException {
final int BUFFER_SIZE = 64;
Cache cache = cacheManager.getCache();
Directory dir =
DirectoryBuilder.newDirectoryInstance(cache, cache, cache, INDEXNAME)
.chunkSize(BUFFER_SIZE)
.create();
String fileName = "SomeText.txt";
IndexOutput io = dir.createOutput(fileName);
RepeatableLongByteSequence bytesGenerator = new RepeatableLongByteSequence();
// It writes repeatable text
final int REPEATABLE_BUFFER_SIZE = 1501;
for (int i = 0; i < REPEATABLE_BUFFER_SIZE; i++) {
io.writeByte(bytesGenerator.nextByte());
}
io.flush();
assert io.length() == REPEATABLE_BUFFER_SIZE;
// Text to write on file with repeatable text
final String someText = "This is some text";
final byte[] someTextAsBytes = someText.getBytes();
// 4 points in random order where writing someText: at begin of file, at end of file, within a
// single chunk,
// between 2 chunks
final int[] pointers = {0, 635, REPEATABLE_BUFFER_SIZE, 135};
for (int i = 0; i < pointers.length; i++) {
io.seek(pointers[i]);
io.writeBytes(someTextAsBytes, someTextAsBytes.length);
}
io.close();
bytesGenerator.reset();
final long finalSize = REPEATABLE_BUFFER_SIZE + someTextAsBytes.length;
assert io.length() == finalSize;
assert io.length()
== DirectoryIntegrityCheck.deepCountFileSize(new FileCacheKey(INDEXNAME, fileName), cache);
int indexPointer = 0;
Arrays.sort(pointers);
byte[] buffer = null;
int chunkIndex = -1;
// now testing the stream is equal to the produced repeatable but including the edits at pointed
// positions
for (int i = 0; i < REPEATABLE_BUFFER_SIZE + someTextAsBytes.length; i++) {
if (i % BUFFER_SIZE == 0) {
buffer =
(byte[]) cache.get(new ChunkCacheKey(INDEXNAME, fileName, ++chunkIndex, BUFFER_SIZE));
}
byte predictableByte = bytesGenerator.nextByte();
if (i < pointers[indexPointer]) {
// Assert predictable text
AssertJUnit.assertEquals(predictableByte, buffer[i % BUFFER_SIZE]);
} else if (pointers[indexPointer] <= i
&& i < pointers[indexPointer] + someTextAsBytes.length) {
// Assert someText
AssertJUnit.assertEquals(
someTextAsBytes[i - pointers[indexPointer]], buffer[i % BUFFER_SIZE]);
}
if (i == pointers[indexPointer] + someTextAsBytes.length) {
// Change pointer
indexPointer++;
}
}
dir.close();
DirectoryIntegrityCheck.verifyDirectoryStructure(cache, INDEXNAME);
}
@Override
public void writeByte(byte b) throws IOException {
tempOut.writeByte(b);
}
/**
* Called once per field per document if term vectors are enabled, to write the vectors to
* RAMOutputStream, which is then quickly flushed to the real term vectors files in the Directory.
*/
@Override
void finish() throws IOException {
assert docState.testPoint("TermVectorsTermsWriterPerField.finish start");
final int numPostings = termsHashPerField.bytesHash.size();
final BytesRef flushTerm = perThread.flushTerm;
assert numPostings >= 0;
if (!doVectors || numPostings == 0) return;
if (numPostings > maxNumPostings) maxNumPostings = numPostings;
final IndexOutput tvf = perThread.doc.perDocTvf;
// This is called once, after inverting all occurrences
// of a given field in the doc. At this point we flush
// our hash into the DocWriter.
assert fieldInfo.storeTermVector;
assert perThread.vectorFieldsInOrder(fieldInfo);
perThread.doc.addField(termsHashPerField.fieldInfo.number);
TermVectorsPostingsArray postings = (TermVectorsPostingsArray) termsHashPerField.postingsArray;
// TODO: we may want to make this sort in same order
// as Codec's terms dict?
final int[] termIDs =
termsHashPerField.sortPostings(BytesRef.getUTF8SortedAsUnicodeComparator());
tvf.writeVInt(numPostings);
byte bits = 0x0;
if (doVectorPositions) bits |= TermVectorsReader.STORE_POSITIONS_WITH_TERMVECTOR;
if (doVectorOffsets) bits |= TermVectorsReader.STORE_OFFSET_WITH_TERMVECTOR;
tvf.writeByte(bits);
int lastLen = 0;
byte[] lastBytes = null;
int lastStart = 0;
final ByteSliceReader reader = perThread.vectorSliceReader;
final ByteBlockPool termBytePool = perThread.termsHashPerThread.termBytePool;
for (int j = 0; j < numPostings; j++) {
final int termID = termIDs[j];
final int freq = postings.freqs[termID];
// Get BytesRef
termBytePool.setBytesRef(flushTerm, postings.textStarts[termID]);
// Compute common byte prefix between last term and
// this term
int prefix = 0;
if (j > 0) {
while (prefix < lastLen && prefix < flushTerm.length) {
if (lastBytes[lastStart + prefix] != flushTerm.bytes[flushTerm.offset + prefix]) {
break;
}
prefix++;
}
}
lastLen = flushTerm.length;
lastBytes = flushTerm.bytes;
lastStart = flushTerm.offset;
final int suffix = flushTerm.length - prefix;
tvf.writeVInt(prefix);
tvf.writeVInt(suffix);
tvf.writeBytes(flushTerm.bytes, lastStart + prefix, suffix);
tvf.writeVInt(freq);
if (doVectorPositions) {
termsHashPerField.initReader(reader, termID, 0);
reader.writeTo(tvf);
}
if (doVectorOffsets) {
termsHashPerField.initReader(reader, termID, 1);
reader.writeTo(tvf);
}
}
termsHashPerField.reset();
// NOTE: we clear, per-field, at the thread level,
// because term vectors fully write themselves on each
// field; this saves RAM (eg if large doc has two large
// fields w/ term vectors on) because we recycle/reuse
// all RAM after each field:
perThread.termsHashPerThread.reset(false);
}
