Examples with SegmentWriteOutMedium org.apache.druid.segment.writeout.SegmentWriteOutMedium used on opensource projects

Example 1 with SegmentWriteOutMedium

use of org.apache.druid.segment.writeout.SegmentWriteOutMedium in project druid by druid-io.

the class V3CompressedVSizeColumnarMultiIntsSerializerTest method generateV2SerializedSizeAndData.

private void generateV2SerializedSizeAndData(long numRows, int maxValue, int maxValuesPerRow, int offsetChunkFactor, int valueChunkFactor) throws Exception {
    File tmpDirectory = FileUtils.createTempDir(StringUtils.format("CompressedVSizeIndexedV3WriterTest_%d_%d", offsetChunkFactor, offsetChunkFactor));
    FileSmoosher smoosher = new FileSmoosher(tmpDirectory);
    try (SegmentWriteOutMedium segmentWriteOutMedium = TmpFileSegmentWriteOutMediumFactory.instance().makeSegmentWriteOutMedium(temporaryFolder.newFolder())) {
        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 (long l = 0L; l < numRows; l++) {
            writer.addValues(new ArrayBasedIndexedInts(generateRow(rand, maxValue, maxValuesPerRow)));
        }
        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, mapper);
        ColumnarMultiInts columnarMultiInts = supplierFromByteBuffer.get();
        Assert.assertEquals(columnarMultiInts.size(), numRows);
        Random verifier = new Random(0);
        for (int i = 0; i < numRows; ++i) {
            IndexedInts subVals = columnarMultiInts.get(i);
            int[] expected = generateRow(verifier, maxValue, maxValuesPerRow);
            Assert.assertEquals(subVals.size(), expected.length);
            for (int j = 0, size = subVals.size(); j < size; ++j) {
                Assert.assertEquals(subVals.get(j), expected[j]);
            }
        }
        CloseableUtils.closeAll(columnarMultiInts, mapper);
    }
}

Example 2 with SegmentWriteOutMedium

use of org.apache.druid.segment.writeout.SegmentWriteOutMedium in project druid by druid-io.

the class LargeColumnSupportedComplexColumnSerializerTest method testSanity.

@Test
public void testSanity() throws IOException {
    HyperUniquesSerdeForTest serde = new HyperUniquesSerdeForTest(Hashing.murmur3_128());
    int[] cases = { 1000, 5000, 10000, 20000 };
    int[] columnSizes = { Integer.MAX_VALUE, Integer.MAX_VALUE / 2, Integer.MAX_VALUE / 4, 5000 * Long.BYTES, 2500 * Long.BYTES };
    for (int columnSize : columnSizes) {
        for (int aCase : cases) {
            File tmpFile = temporaryFolder.newFolder();
            HyperLogLogCollector baseCollector = HyperLogLogCollector.makeLatestCollector();
            try (SegmentWriteOutMedium segmentWriteOutMedium = new OffHeapMemorySegmentWriteOutMedium();
                FileSmoosher v9Smoosher = new FileSmoosher(tmpFile)) {
                LargeColumnSupportedComplexColumnSerializer serializer = LargeColumnSupportedComplexColumnSerializer.createWithColumnSize(segmentWriteOutMedium, "test", serde.getObjectStrategy(), columnSize);
                serializer.open();
                for (int i = 0; i < aCase; i++) {
                    HyperLogLogCollector collector = HyperLogLogCollector.makeLatestCollector();
                    byte[] hashBytes = fn.hashLong(i).asBytes();
                    collector.add(hashBytes);
                    baseCollector.fold(collector);
                    serializer.serialize(new ObjectColumnSelector() {

                        @Nullable
                        @Override
                        public Object getObject() {
                            return collector;
                        }

                        @Override
                        public Class classOfObject() {
                            return HyperLogLogCollector.class;
                        }

                        @Override
                        public void inspectRuntimeShape(RuntimeShapeInspector inspector) {
                        // doesn't matter in tests
                        }
                    });
                }
                try (final SmooshedWriter channel = v9Smoosher.addWithSmooshedWriter("test", serializer.getSerializedSize())) {
                    serializer.writeTo(channel, v9Smoosher);
                }
            }
            SmooshedFileMapper mapper = Smoosh.map(tmpFile);
            final ColumnBuilder builder = new ColumnBuilder().setType(ValueType.COMPLEX).setHasMultipleValues(false).setFileMapper(mapper);
            serde.deserializeColumn(mapper.mapFile("test"), builder, null);
            ColumnHolder columnHolder = builder.build();
            ComplexColumn complexColumn = (ComplexColumn) columnHolder.getColumn();
            HyperLogLogCollector collector = HyperLogLogCollector.makeLatestCollector();
            for (int i = 0; i < aCase; i++) {
                collector.fold((HyperLogLogCollector) complexColumn.getRowValue(i));
            }
            Assert.assertEquals(baseCollector.estimateCardinality(), collector.estimateCardinality(), 0.0);
        }
    }
}

Example 3 with SegmentWriteOutMedium

use of org.apache.druid.segment.writeout.SegmentWriteOutMedium in project druid by druid-io.

the class CompressedColumnarIntsSerializerTest method testTooManyValues.

// this test takes ~30 minutes to run
@Ignore
@Test
public void testTooManyValues() throws IOException {
    expectedException.expect(ColumnCapacityExceededException.class);
    expectedException.expectMessage(ColumnCapacityExceededException.formatMessage("test"));
    try (SegmentWriteOutMedium segmentWriteOutMedium = TmpFileSegmentWriteOutMediumFactory.instance().makeSegmentWriteOutMedium(temporaryFolder.newFolder())) {
        CompressedColumnarIntsSerializer serializer = new CompressedColumnarIntsSerializer("test", segmentWriteOutMedium, "test", CompressedColumnarIntsSupplier.MAX_INTS_IN_BUFFER, byteOrder, compressionStrategy);
        serializer.open();
        final long numRows = Integer.MAX_VALUE + 100L;
        for (long i = 0L; i < numRows; i++) {
            serializer.addValue(ThreadLocalRandom.current().nextInt(0, Integer.MAX_VALUE));
        }
    }
}

Example 4 with SegmentWriteOutMedium

use of org.apache.druid.segment.writeout.SegmentWriteOutMedium in project druid by druid-io.

the class CompressedDoublesSerdeTest method testTooManyValues.

// this test takes ~45 minutes to run
@Ignore
@Test
public void testTooManyValues() throws IOException {
    expectedException.expect(ColumnCapacityExceededException.class);
    expectedException.expectMessage(ColumnCapacityExceededException.formatMessage("test"));
    try (SegmentWriteOutMedium segmentWriteOutMedium = TmpFileSegmentWriteOutMediumFactory.instance().makeSegmentWriteOutMedium(temporaryFolder.newFolder())) {
        ColumnarDoublesSerializer serializer = CompressionFactory.getDoubleSerializer("test", segmentWriteOutMedium, "test", order, compressionStrategy);
        serializer.open();
        final long numRows = Integer.MAX_VALUE + 100L;
        for (long i = 0L; i < numRows; i++) {
            serializer.add(ThreadLocalRandom.current().nextDouble());
        }
    }
}

Example 5 with SegmentWriteOutMedium

use of org.apache.druid.segment.writeout.SegmentWriteOutMedium 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();
    }
}