use of org.apache.druid.java.util.common.io.smoosh.FileSmoosher in project druid by druid-io.
the class IndexMergerV9 method makeIndexFiles.
private File makeIndexFiles(final List<IndexableAdapter> adapters, @Nullable final AggregatorFactory[] metricAggs, final File outDir, final ProgressIndicator progress, final List<String> mergedDimensions, final List<String> mergedMetrics, final Function<List<TransformableRowIterator>, TimeAndDimsIterator> rowMergerFn, final boolean fillRowNumConversions, final IndexSpec indexSpec, @Nullable final SegmentWriteOutMediumFactory segmentWriteOutMediumFactory) throws IOException {
progress.start();
progress.progress();
List<Metadata> metadataList = Lists.transform(adapters, IndexableAdapter::getMetadata);
final Metadata segmentMetadata;
if (metricAggs != null) {
AggregatorFactory[] combiningMetricAggs = new AggregatorFactory[metricAggs.length];
for (int i = 0; i < metricAggs.length; i++) {
combiningMetricAggs[i] = metricAggs[i].getCombiningFactory();
}
segmentMetadata = Metadata.merge(metadataList, combiningMetricAggs);
} else {
segmentMetadata = Metadata.merge(metadataList, null);
}
Closer closer = Closer.create();
try {
final FileSmoosher v9Smoosher = new FileSmoosher(outDir);
FileUtils.mkdirp(outDir);
SegmentWriteOutMediumFactory omf = segmentWriteOutMediumFactory != null ? segmentWriteOutMediumFactory : defaultSegmentWriteOutMediumFactory;
log.debug("Using SegmentWriteOutMediumFactory[%s]", omf.getClass().getSimpleName());
SegmentWriteOutMedium segmentWriteOutMedium = omf.makeSegmentWriteOutMedium(outDir);
closer.register(segmentWriteOutMedium);
long startTime = System.currentTimeMillis();
Files.asByteSink(new File(outDir, "version.bin")).write(Ints.toByteArray(IndexIO.V9_VERSION));
log.debug("Completed version.bin in %,d millis.", System.currentTimeMillis() - startTime);
progress.progress();
startTime = System.currentTimeMillis();
try (FileOutputStream fos = new FileOutputStream(new File(outDir, "factory.json"))) {
SegmentizerFactory customSegmentLoader = indexSpec.getSegmentLoader();
if (customSegmentLoader != null) {
mapper.writeValue(fos, customSegmentLoader);
} else {
mapper.writeValue(fos, new MMappedQueryableSegmentizerFactory(indexIO));
}
}
log.debug("Completed factory.json in %,d millis", System.currentTimeMillis() - startTime);
progress.progress();
final Map<String, ValueType> metricsValueTypes = new TreeMap<>(Comparators.naturalNullsFirst());
final Map<String, String> metricTypeNames = new TreeMap<>(Comparators.naturalNullsFirst());
final List<ColumnCapabilities> dimCapabilities = Lists.newArrayListWithCapacity(mergedDimensions.size());
mergeCapabilities(adapters, mergedDimensions, metricsValueTypes, metricTypeNames, dimCapabilities);
final Map<String, DimensionHandler> handlers = makeDimensionHandlers(mergedDimensions, dimCapabilities);
final List<DimensionMergerV9> mergers = new ArrayList<>();
for (int i = 0; i < mergedDimensions.size(); i++) {
DimensionHandler handler = handlers.get(mergedDimensions.get(i));
mergers.add(handler.makeMerger(indexSpec, segmentWriteOutMedium, dimCapabilities.get(i), progress, closer));
}
/**
*********** Setup Dim Conversions *************
*/
progress.progress();
startTime = System.currentTimeMillis();
writeDimValuesAndSetupDimConversion(adapters, progress, mergedDimensions, mergers);
log.debug("Completed dim conversions in %,d millis.", System.currentTimeMillis() - startTime);
/**
*********** Walk through data sets, merge them, and write merged columns ************
*/
progress.progress();
final TimeAndDimsIterator timeAndDimsIterator = makeMergedTimeAndDimsIterator(adapters, mergedDimensions, mergedMetrics, rowMergerFn, handlers, mergers);
closer.register(timeAndDimsIterator);
final GenericColumnSerializer timeWriter = setupTimeWriter(segmentWriteOutMedium, indexSpec);
final ArrayList<GenericColumnSerializer> metricWriters = setupMetricsWriters(segmentWriteOutMedium, mergedMetrics, metricsValueTypes, metricTypeNames, indexSpec);
List<IntBuffer> rowNumConversions = mergeIndexesAndWriteColumns(adapters, progress, timeAndDimsIterator, timeWriter, metricWriters, mergers, fillRowNumConversions);
/**
********** Create Inverted Indexes and Finalize Build Columns ************
*/
final String section = "build inverted index and columns";
progress.startSection(section);
makeTimeColumn(v9Smoosher, progress, timeWriter, indexSpec);
makeMetricsColumns(v9Smoosher, progress, mergedMetrics, metricsValueTypes, metricTypeNames, metricWriters, indexSpec);
for (int i = 0; i < mergedDimensions.size(); i++) {
DimensionMergerV9 merger = mergers.get(i);
merger.writeIndexes(rowNumConversions);
if (merger.canSkip()) {
continue;
}
ColumnDescriptor columnDesc = merger.makeColumnDescriptor();
makeColumn(v9Smoosher, mergedDimensions.get(i), columnDesc);
}
progress.stopSection(section);
/**
*********** Make index.drd & metadata.drd files *************
*/
progress.progress();
makeIndexBinary(v9Smoosher, adapters, outDir, mergedDimensions, mergedMetrics, progress, indexSpec, mergers);
makeMetadataBinary(v9Smoosher, progress, segmentMetadata);
v9Smoosher.close();
progress.stop();
return outDir;
} catch (Throwable t) {
throw closer.rethrow(t);
} finally {
closer.close();
}
}
use of org.apache.druid.java.util.common.io.smoosh.FileSmoosher in project druid by druid-io.
the class GenericIndexedBenchmark method createGenericIndexed.
@Setup(Level.Trial)
public void createGenericIndexed() throws IOException {
GenericIndexedWriter<byte[]> genericIndexedWriter = new GenericIndexedWriter<>(new OffHeapMemorySegmentWriteOutMedium(), "genericIndexedBenchmark", BYTE_ARRAY_STRATEGY);
genericIndexedWriter.open();
// GenericIndexObject caches prevObject for comparison, so need two arrays for correct objectsSorted computation.
ByteBuffer[] elements = new ByteBuffer[2];
elements[0] = ByteBuffer.allocate(elementSize);
elements[1] = ByteBuffer.allocate(elementSize);
for (int i = 0; i < n; i++) {
ByteBuffer element = elements[i & 1];
element.putInt(0, i);
genericIndexedWriter.write(element.array());
}
smooshDir = FileUtils.createTempDir();
file = File.createTempFile("genericIndexedBenchmark", "meta");
try (FileChannel fileChannel = FileChannel.open(file.toPath(), StandardOpenOption.CREATE, StandardOpenOption.WRITE);
FileSmoosher fileSmoosher = new FileSmoosher(smooshDir)) {
genericIndexedWriter.writeTo(fileChannel, fileSmoosher);
}
FileChannel fileChannel = FileChannel.open(file.toPath());
MappedByteBuffer byteBuffer = fileChannel.map(FileChannel.MapMode.READ_ONLY, 0, file.length());
genericIndexed = GenericIndexed.read(byteBuffer, BYTE_ARRAY_STRATEGY, SmooshedFileMapper.load(smooshDir));
}
use of org.apache.druid.java.util.common.io.smoosh.FileSmoosher in project druid by druid-io.
the class CompressedVSizeColumnarIntsSerializerTest method checkV2SerializedSizeAndData.
private void checkV2SerializedSizeAndData(int chunkSize) throws Exception {
File tmpDirectory = temporaryFolder.newFolder();
FileSmoosher smoosher = new FileSmoosher(tmpDirectory);
final String columnName = "test";
GenericIndexedWriter genericIndexed = GenericIndexedWriter.ofCompressedByteBuffers(segmentWriteOutMedium, "test", compressionStrategy, Long.BYTES * 10000);
CompressedVSizeColumnarIntsSerializer writer = new CompressedVSizeColumnarIntsSerializer(columnName, segmentWriteOutMedium, vals.length > 0 ? Ints.max(vals) : 0, chunkSize, byteOrder, compressionStrategy, genericIndexed);
writer.open();
for (int val : vals) {
writer.addValue(val);
}
final SmooshedWriter channel = smoosher.addWithSmooshedWriter("test", writer.getSerializedSize());
writer.writeTo(channel, smoosher);
channel.close();
smoosher.close();
SmooshedFileMapper mapper = Smoosh.map(tmpDirectory);
CompressedVSizeColumnarIntsSupplier supplierFromByteBuffer = CompressedVSizeColumnarIntsSupplier.fromByteBuffer(mapper.mapFile("test"), byteOrder);
ColumnarInts columnarInts = supplierFromByteBuffer.get();
for (int i = 0; i < vals.length; ++i) {
Assert.assertEquals(vals[i], columnarInts.get(i));
}
CloseableUtils.closeAll(columnarInts, mapper);
}
use of org.apache.druid.java.util.common.io.smoosh.FileSmoosher in project druid by druid-io.
the class V3CompressedVSizeColumnarMultiIntsSerializerTest method checkV2SerializedSizeAndData.
private void checkV2SerializedSizeAndData(int offsetChunkFactor, int valueChunkFactor) throws Exception {
File tmpDirectory = FileUtils.createTempDir(StringUtils.format("CompressedVSizeIndexedV3WriterTest_%d_%d", offsetChunkFactor, offsetChunkFactor));
FileSmoosher smoosher = new FileSmoosher(tmpDirectory);
int maxValue = vals.size() > 0 ? getMaxValue(vals) : 0;
try (SegmentWriteOutMedium segmentWriteOutMedium = new OffHeapMemorySegmentWriteOutMedium()) {
CompressedColumnarIntsSerializer offsetWriter = new CompressedColumnarIntsSerializer(TEST_COLUMN_NAME, segmentWriteOutMedium, offsetChunkFactor, byteOrder, compressionStrategy, GenericIndexedWriter.ofCompressedByteBuffers(segmentWriteOutMedium, "offset", compressionStrategy, Long.BYTES * 250000));
GenericIndexedWriter genericIndexed = GenericIndexedWriter.ofCompressedByteBuffers(segmentWriteOutMedium, "value", compressionStrategy, Long.BYTES * 250000);
CompressedVSizeColumnarIntsSerializer valueWriter = new CompressedVSizeColumnarIntsSerializer(TEST_COLUMN_NAME, segmentWriteOutMedium, maxValue, valueChunkFactor, byteOrder, compressionStrategy, genericIndexed);
V3CompressedVSizeColumnarMultiIntsSerializer writer = new V3CompressedVSizeColumnarMultiIntsSerializer(TEST_COLUMN_NAME, offsetWriter, valueWriter);
writer.open();
for (int[] val : vals) {
writer.addValues(new ArrayBasedIndexedInts(val));
}
final SmooshedWriter channel = smoosher.addWithSmooshedWriter("test", writer.getSerializedSize());
writer.writeTo(channel, smoosher);
channel.close();
smoosher.close();
SmooshedFileMapper mapper = Smoosh.map(tmpDirectory);
V3CompressedVSizeColumnarMultiIntsSupplier supplierFromByteBuffer = V3CompressedVSizeColumnarMultiIntsSupplier.fromByteBuffer(mapper.mapFile("test"), byteOrder);
ColumnarMultiInts columnarMultiInts = supplierFromByteBuffer.get();
Assert.assertEquals(columnarMultiInts.size(), vals.size());
for (int i = 0; i < vals.size(); ++i) {
IndexedInts subVals = columnarMultiInts.get(i);
Assert.assertEquals(subVals.size(), vals.get(i).length);
for (int j = 0, size = subVals.size(); j < size; ++j) {
Assert.assertEquals(subVals.get(j), vals.get(i)[j]);
}
}
CloseableUtils.closeAll(columnarMultiInts, mapper);
}
}
use of org.apache.druid.java.util.common.io.smoosh.FileSmoosher in project druid by druid-io.
the class CompressedColumnarIntsSerializerTest method checkV2SerializedSizeAndData.
private void checkV2SerializedSizeAndData(int chunkFactor) throws Exception {
File tmpDirectory = FileUtils.createTempDir(StringUtils.format("CompressedIntsIndexedWriterTest_%d", chunkFactor));
FileSmoosher smoosher = new FileSmoosher(tmpDirectory);
CompressedColumnarIntsSerializer writer = new CompressedColumnarIntsSerializer("test", segmentWriteOutMedium, chunkFactor, byteOrder, compressionStrategy, GenericIndexedWriter.ofCompressedByteBuffers(segmentWriteOutMedium, "test", compressionStrategy, Long.BYTES * 10000));
writer.open();
for (int val : vals) {
writer.addValue(val);
}
final SmooshedWriter channel = smoosher.addWithSmooshedWriter("test", writer.getSerializedSize());
writer.writeTo(channel, smoosher);
channel.close();
smoosher.close();
SmooshedFileMapper mapper = Smoosh.map(tmpDirectory);
// read from ByteBuffer and check values
CompressedColumnarIntsSupplier supplierFromByteBuffer = CompressedColumnarIntsSupplier.fromByteBuffer(mapper.mapFile("test"), byteOrder);
ColumnarInts columnarInts = supplierFromByteBuffer.get();
Assert.assertEquals(vals.length, columnarInts.size());
for (int i = 0; i < vals.length; ++i) {
Assert.assertEquals(vals[i], columnarInts.get(i));
}
CloseableUtils.closeAndWrapExceptions(columnarInts);
mapper.close();
}
Aggregations