Examples with WindowedStorageAdapter org.apache.druid.segment.realtime.firehose.WindowedStorageAdapter used on opensource projects

Example 1 with WindowedStorageAdapter

use of org.apache.druid.segment.realtime.firehose.WindowedStorageAdapter in project druid by druid-io.

the class CompactionTaskRunTest method getCSVFormatRowsFromSegments.

private List<String> getCSVFormatRowsFromSegments(List<DataSegment> segments) throws Exception {
    final File cacheDir = temporaryFolder.newFolder();
    final SegmentCacheManager segmentCacheManager = segmentCacheManagerFactory.manufacturate(cacheDir);
    List<Cursor> cursors = new ArrayList<>();
    for (DataSegment segment : segments) {
        final File segmentFile = segmentCacheManager.getSegmentFiles(segment);
        final WindowedStorageAdapter adapter = new WindowedStorageAdapter(new QueryableIndexStorageAdapter(testUtils.getTestIndexIO().loadIndex(segmentFile)), segment.getInterval());
        final Sequence<Cursor> cursorSequence = adapter.getAdapter().makeCursors(null, segment.getInterval(), VirtualColumns.EMPTY, Granularities.ALL, false, null);
        cursors.addAll(cursorSequence.toList());
    }
    List<String> rowsFromSegment = new ArrayList<>();
    for (Cursor cursor : cursors) {
        cursor.reset();
        while (!cursor.isDone()) {
            final DimensionSelector selector1 = cursor.getColumnSelectorFactory().makeDimensionSelector(new DefaultDimensionSpec("ts", "ts"));
            final DimensionSelector selector2 = cursor.getColumnSelectorFactory().makeDimensionSelector(new DefaultDimensionSpec("dim", "dim"));
            final DimensionSelector selector3 = cursor.getColumnSelectorFactory().makeDimensionSelector(new DefaultDimensionSpec("val", "val"));
            Object dimObject = selector2.getObject();
            String dimVal = null;
            if (dimObject instanceof String) {
                dimVal = (String) dimObject;
            } else if (dimObject instanceof List) {
                dimVal = String.join("|", (List<String>) dimObject);
            }
            rowsFromSegment.add(makeCSVFormatRow(selector1.getObject().toString(), dimVal, selector3.defaultGetObject().toString()));
            cursor.advance();
        }
    }
    return rowsFromSegment;
}

Example 2 with WindowedStorageAdapter

use of org.apache.druid.segment.realtime.firehose.WindowedStorageAdapter in project druid by druid-io.

the class IndexTaskTest method testTransformSpec.

@Test
public void testTransformSpec() throws Exception {
    File tmpDir = temporaryFolder.newFolder();
    File tmpFile = File.createTempFile("druid", "index", tmpDir);
    try (BufferedWriter writer = Files.newWriter(tmpFile, StandardCharsets.UTF_8)) {
        writer.write("2014-01-01T00:00:10Z,a,an|array,1|2|3,1\n");
        writer.write("2014-01-01T01:00:20Z,b,another|array,3|4,1\n");
        writer.write("2014-01-01T02:00:30Z,c,and|another,0|1,1\n");
    }
    final DimensionsSpec dimensionsSpec = new DimensionsSpec(DimensionsSpec.getDefaultSchemas(Arrays.asList("ts", "dim", "dim_array", "dim_num_array", "dimt", "dimtarray1", "dimtarray2", "dimtnum_array")));
    final List<String> columns = Arrays.asList("ts", "dim", "dim_array", "dim_num_array", "val");
    final String listDelimiter = "|";
    final TransformSpec transformSpec = new TransformSpec(new SelectorDimFilter("dim", "b", null), ImmutableList.of(new ExpressionTransform("dimt", "concat(dim,dim)", ExprMacroTable.nil()), new ExpressionTransform("dimtarray1", "array(dim, dim)", ExprMacroTable.nil()), new ExpressionTransform("dimtarray2", "map(d -> concat(d, 'foo'), dim_array)", ExprMacroTable.nil()), new ExpressionTransform("dimtnum_array", "map(d -> d + 3, dim_num_array)", ExprMacroTable.nil())));
    final IndexTuningConfig tuningConfig = createTuningConfigWithMaxRowsPerSegment(2, false);
    final IndexIngestionSpec indexIngestionSpec;
    if (useInputFormatApi) {
        indexIngestionSpec = createIngestionSpec(jsonMapper, tmpDir, DEFAULT_TIMESTAMP_SPEC, dimensionsSpec, new CsvInputFormat(columns, listDelimiter, null, false, 0), transformSpec, null, tuningConfig, false, false);
    } else {
        indexIngestionSpec = createIngestionSpec(jsonMapper, tmpDir, new CSVParseSpec(DEFAULT_TIMESTAMP_SPEC, dimensionsSpec, listDelimiter, columns, false, 0), transformSpec, null, tuningConfig, false, false);
    }
    IndexTask indexTask = new IndexTask(null, null, indexIngestionSpec, null);
    Assert.assertEquals(indexTask.getId(), indexTask.getGroupId());
    final List<DataSegment> segments = runTask(indexTask).rhs;
    Assert.assertEquals(1, segments.size());
    DataSegment segment = segments.get(0);
    final File segmentFile = segmentCacheManager.getSegmentFiles(segment);
    final WindowedStorageAdapter adapter = new WindowedStorageAdapter(new QueryableIndexStorageAdapter(indexIO.loadIndex(segmentFile)), segment.getInterval());
    final Sequence<Cursor> cursorSequence = adapter.getAdapter().makeCursors(null, segment.getInterval(), VirtualColumns.EMPTY, Granularities.ALL, false, null);
    final List<Map<String, Object>> transforms = cursorSequence.map(cursor -> {
        final DimensionSelector selector1 = cursor.getColumnSelectorFactory().makeDimensionSelector(new DefaultDimensionSpec("dimt", "dimt"));
        final DimensionSelector selector2 = cursor.getColumnSelectorFactory().makeDimensionSelector(new DefaultDimensionSpec("dimtarray1", "dimtarray1"));
        final DimensionSelector selector3 = cursor.getColumnSelectorFactory().makeDimensionSelector(new DefaultDimensionSpec("dimtarray2", "dimtarray2"));
        final DimensionSelector selector4 = cursor.getColumnSelectorFactory().makeDimensionSelector(new DefaultDimensionSpec("dimtnum_array", "dimtnum_array"));
        Map<String, Object> row = new HashMap<>();
        row.put("dimt", selector1.defaultGetObject());
        row.put("dimtarray1", selector2.defaultGetObject());
        row.put("dimtarray2", selector3.defaultGetObject());
        row.put("dimtnum_array", selector4.defaultGetObject());
        cursor.advance();
        return row;
    }).toList();
    Assert.assertEquals(1, transforms.size());
    Assert.assertEquals("bb", transforms.get(0).get("dimt"));
    Assert.assertEquals(ImmutableList.of("b", "b"), transforms.get(0).get("dimtarray1"));
    Assert.assertEquals(ImmutableList.of("anotherfoo", "arrayfoo"), transforms.get(0).get("dimtarray2"));
    Assert.assertEquals(ImmutableList.of("6.0", "7.0"), transforms.get(0).get("dimtnum_array"));
    Assert.assertEquals(DATASOURCE, segments.get(0).getDataSource());
    Assert.assertEquals(Intervals.of("2014/P1D"), segments.get(0).getInterval());
    Assert.assertEquals(NumberedShardSpec.class, segments.get(0).getShardSpec().getClass());
    Assert.assertEquals(0, segments.get(0).getShardSpec().getPartitionNum());
}

Example 3 with WindowedStorageAdapter

use of org.apache.druid.segment.realtime.firehose.WindowedStorageAdapter in project hive by apache.

the class TestDruidRecordWriter method testWrite.

// Test is failing due to Guava dependency, Druid 0.13.0 should have less dependency on Guava
@Ignore
@Test
public void testWrite() throws IOException, SegmentLoadingException {
    final String dataSourceName = "testDataSource";
    final File segmentOutputDir = temporaryFolder.newFolder();
    final File workingDir = temporaryFolder.newFolder();
    Configuration config = new Configuration();
    final InputRowParser inputRowParser = new MapInputRowParser(new TimeAndDimsParseSpec(new TimestampSpec(DruidConstants.DEFAULT_TIMESTAMP_COLUMN, "auto", null), new DimensionsSpec(ImmutableList.of(new StringDimensionSchema("host")), null, null)));
    final Map<String, Object> parserMap = objectMapper.convertValue(inputRowParser, new TypeReference<Map<String, Object>>() {
    });
    DataSchema dataSchema = new DataSchema(dataSourceName, parserMap, new AggregatorFactory[] { new LongSumAggregatorFactory("visited_sum", "visited_sum"), new HyperUniquesAggregatorFactory("unique_hosts", "unique_hosts") }, new UniformGranularitySpec(Granularities.DAY, Granularities.NONE, ImmutableList.of(INTERVAL_FULL)), null, objectMapper);
    IndexSpec indexSpec = new IndexSpec(new RoaringBitmapSerdeFactory(true), null, null, null);
    RealtimeTuningConfig tuningConfig = new RealtimeTuningConfig(null, null, null, null, temporaryFolder.newFolder(), null, null, null, null, indexSpec, null, null, 0, 0, null, null, 0L, null, null);
    LocalFileSystem localFileSystem = FileSystem.getLocal(config);
    DataSegmentPusher dataSegmentPusher = new LocalDataSegmentPusher(new LocalDataSegmentPusherConfig() {

        @Override
        public File getStorageDirectory() {
            return segmentOutputDir;
        }
    });
    Path segmentDescriptorPath = new Path(workingDir.getAbsolutePath(), DruidStorageHandler.SEGMENTS_DESCRIPTOR_DIR_NAME);
    DruidRecordWriter druidRecordWriter = new DruidRecordWriter(dataSchema, tuningConfig, dataSegmentPusher, 20, segmentDescriptorPath, localFileSystem);
    List<DruidWritable> druidWritables = expectedRows.stream().map(input -> new DruidWritable(ImmutableMap.<String, Object>builder().putAll(input).put(Constants.DRUID_TIMESTAMP_GRANULARITY_COL_NAME, Granularities.DAY.bucketStart(new DateTime((long) input.get(DruidConstants.DEFAULT_TIMESTAMP_COLUMN))).getMillis()).build())).collect(Collectors.toList());
    for (DruidWritable druidWritable : druidWritables) {
        druidRecordWriter.write(druidWritable);
    }
    druidRecordWriter.close(false);
    List<DataSegment> dataSegmentList = DruidStorageHandlerUtils.getCreatedSegments(segmentDescriptorPath, config);
    Assert.assertEquals(1, dataSegmentList.size());
    File tmpUnzippedSegmentDir = temporaryFolder.newFolder();
    new LocalDataSegmentPuller().getSegmentFiles(dataSegmentList.get(0), tmpUnzippedSegmentDir);
    final QueryableIndex queryableIndex = DruidStorageHandlerUtils.INDEX_IO.loadIndex(tmpUnzippedSegmentDir);
    QueryableIndexStorageAdapter adapter = new QueryableIndexStorageAdapter(queryableIndex);
    Firehose firehose = new IngestSegmentFirehose(ImmutableList.of(new WindowedStorageAdapter(adapter, adapter.getInterval())), null, ImmutableList.of("host"), ImmutableList.of("visited_sum", "unique_hosts"), null);
    List<InputRow> rows = Lists.newArrayList();
    while (firehose.hasMore()) {
        rows.add(firehose.nextRow());
    }
    verifyRows(expectedRows, rows);
}

Example 4 with WindowedStorageAdapter

use of org.apache.druid.segment.realtime.firehose.WindowedStorageAdapter in project druid by druid-io.

the class IngestSegmentFirehoseFactory method connect.

@Override
public Firehose connect(InputRowParser inputRowParser, File temporaryDirectory) throws ParseException {
    log.debug("Connecting firehose: dataSource[%s], interval[%s], segmentIds[%s]", dataSource, interval, segmentIds);
    final List<TimelineObjectHolder<String, DataSegment>> timeLineSegments = getTimeline();
    // Download all segments locally.
    // Note: this requires enough local storage space to fit all of the segments, even though
    // IngestSegmentFirehose iterates over the segments in series. We may want to change this
    // to download files lazily, perhaps sharing code with PrefetchableTextFilesFirehoseFactory.
    final SegmentCacheManager segmentCacheManager = segmentCacheManagerFactory.manufacturate(temporaryDirectory);
    Map<DataSegment, File> segmentFileMap = Maps.newLinkedHashMap();
    for (TimelineObjectHolder<String, DataSegment> holder : timeLineSegments) {
        for (PartitionChunk<DataSegment> chunk : holder.getObject()) {
            final DataSegment segment = chunk.getObject();
            segmentFileMap.computeIfAbsent(segment, k -> {
                try {
                    return segmentCacheManager.getSegmentFiles(segment);
                } catch (SegmentLoadingException e) {
                    throw new RuntimeException(e);
                }
            });
        }
    }
    final List<String> dims = ReingestionTimelineUtils.getDimensionsToReingest(dimensions, inputRowParser.getParseSpec().getDimensionsSpec(), timeLineSegments);
    final List<String> metricsList = metrics == null ? ReingestionTimelineUtils.getUniqueMetrics(timeLineSegments) : metrics;
    final List<WindowedStorageAdapter> adapters = Lists.newArrayList(Iterables.concat(Iterables.transform(timeLineSegments, new Function<TimelineObjectHolder<String, DataSegment>, Iterable<WindowedStorageAdapter>>() {

        @Override
        public Iterable<WindowedStorageAdapter> apply(final TimelineObjectHolder<String, DataSegment> holder) {
            return Iterables.transform(holder.getObject(), new Function<PartitionChunk<DataSegment>, WindowedStorageAdapter>() {

                @Override
                public WindowedStorageAdapter apply(final PartitionChunk<DataSegment> input) {
                    final DataSegment segment = input.getObject();
                    try {
                        return new WindowedStorageAdapter(new QueryableIndexStorageAdapter(indexIO.loadIndex(Preconditions.checkNotNull(segmentFileMap.get(segment), "File for segment %s", segment.getId()))), holder.getInterval());
                    } catch (IOException e) {
                        throw new RuntimeException(e);
                    }
                }
            });
        }
    })));
    final TransformSpec transformSpec = TransformSpec.fromInputRowParser(inputRowParser);
    return new IngestSegmentFirehose(adapters, transformSpec, dims, metricsList, dimFilter);
}

Example 5 with WindowedStorageAdapter

use of org.apache.druid.segment.realtime.firehose.WindowedStorageAdapter in project druid by druid-io.

the class DruidSegmentReader method intermediateRowIterator.

@Override
protected CloseableIterator<Map<String, Object>> intermediateRowIterator() throws IOException {
    final CleanableFile segmentFile = source.fetch(temporaryDirectory, null);
    final WindowedStorageAdapter storageAdapter = new WindowedStorageAdapter(new QueryableIndexStorageAdapter(indexIO.loadIndex(segmentFile.file())), source.getIntervalFilter());
    final Sequence<Cursor> cursors = storageAdapter.getAdapter().makeCursors(Filters.toFilter(dimFilter), storageAdapter.getInterval(), VirtualColumns.EMPTY, Granularities.ALL, false, null);
    // Retain order of columns from the original segments. Useful for preserving dimension order if we're in
    // schemaless mode.
    final Set<String> columnsToRead = Sets.newLinkedHashSet(Iterables.filter(storageAdapter.getAdapter().getRowSignature().getColumnNames(), columnsFilter::apply));
    final Sequence<Map<String, Object>> sequence = Sequences.concat(Sequences.map(cursors, cursor -> cursorToSequence(cursor, columnsToRead)));
    return makeCloseableIteratorFromSequenceAndSegmentFile(sequence, segmentFile);
}