Examples with IndexMerger io.druid.segment.IndexMerger used on opensource projects

Example 1 with IndexMerger

use of io.druid.segment.IndexMerger in project druid by druid-io.

the class AggregationTestHelper method createGroupByQueryAggregationTestHelper.

public static final AggregationTestHelper createGroupByQueryAggregationTestHelper(List<? extends Module> jsonModulesToRegister, GroupByQueryConfig config, TemporaryFolder tempFolder) {
    ObjectMapper mapper = new DefaultObjectMapper();
    GroupByQueryRunnerFactory factory = GroupByQueryRunnerTest.makeQueryRunnerFactory(mapper, config);
    IndexIO indexIO = new IndexIO(mapper, new ColumnConfig() {

        @Override
        public int columnCacheSizeBytes() {
            return 0;
        }
    });
    return new AggregationTestHelper(mapper, new IndexMerger(mapper, indexIO), indexIO, factory.getToolchest(), factory, tempFolder, jsonModulesToRegister);
}

Example 2 with IndexMerger

use of io.druid.segment.IndexMerger in project druid by druid-io.

the class AggregationTestHelper method createTopNQueryAggregationTestHelper.

public static final AggregationTestHelper createTopNQueryAggregationTestHelper(List<? extends Module> jsonModulesToRegister, TemporaryFolder tempFolder) {
    ObjectMapper mapper = new DefaultObjectMapper();
    TopNQueryQueryToolChest toolchest = new TopNQueryQueryToolChest(new TopNQueryConfig(), QueryRunnerTestHelper.NoopIntervalChunkingQueryRunnerDecorator());
    TopNQueryRunnerFactory factory = new TopNQueryRunnerFactory(new StupidPool<>("TopNQueryRunnerFactory-bufferPool", new Supplier<ByteBuffer>() {

        @Override
        public ByteBuffer get() {
            return ByteBuffer.allocate(10 * 1024 * 1024);
        }
    }), toolchest, QueryRunnerTestHelper.NOOP_QUERYWATCHER);
    IndexIO indexIO = new IndexIO(mapper, new ColumnConfig() {

        @Override
        public int columnCacheSizeBytes() {
            return 0;
        }
    });
    return new AggregationTestHelper(mapper, new IndexMerger(mapper, indexIO), indexIO, toolchest, factory, tempFolder, jsonModulesToRegister);
}

Example 3 with IndexMerger

use of io.druid.segment.IndexMerger in project druid by druid-io.

the class YeOldePlumberSchool method findPlumber.

@Override
public Plumber findPlumber(final DataSchema schema, final RealtimeTuningConfig config, final FireDepartmentMetrics metrics) {
    // There can be only one.
    final Sink theSink = new Sink(interval, schema, config.getShardSpec(), version, config.getMaxRowsInMemory(), config.isReportParseExceptions());
    // Temporary directory to hold spilled segments.
    final File persistDir = new File(tmpSegmentDir, theSink.getSegment().getIdentifier());
    // Set of spilled segments. Will be merged at the end.
    final Set<File> spilled = Sets.newHashSet();
    // IndexMerger implementation.
    final IndexMerger theIndexMerger = config.getBuildV9Directly() ? indexMergerV9 : indexMerger;
    return new Plumber() {

        @Override
        public Object startJob() {
            return null;
        }

        @Override
        public int add(InputRow row, Supplier<Committer> committerSupplier) throws IndexSizeExceededException {
            Sink sink = getSink(row.getTimestampFromEpoch());
            if (sink == null) {
                return -1;
            }
            final int numRows = sink.add(row);
            if (!sink.canAppendRow()) {
                persist(committerSupplier.get());
            }
            return numRows;
        }

        private Sink getSink(long timestamp) {
            if (theSink.getInterval().contains(timestamp)) {
                return theSink;
            } else {
                return null;
            }
        }

        @Override
        public <T> QueryRunner<T> getQueryRunner(Query<T> query) {
            throw new UnsupportedOperationException("Don't query me, bro.");
        }

        @Override
        public void persist(Committer committer) {
            spillIfSwappable();
            committer.run();
        }

        @Override
        public void finishJob() {
            // The segment we will upload
            File fileToUpload = null;
            try {
                // User should have persisted everything by now.
                Preconditions.checkState(!theSink.swappable(), "All data must be persisted before fininshing the job!");
                if (spilled.size() == 0) {
                    throw new IllegalStateException("Nothing indexed?");
                } else if (spilled.size() == 1) {
                    fileToUpload = Iterables.getOnlyElement(spilled);
                } else {
                    List<QueryableIndex> indexes = Lists.newArrayList();
                    for (final File oneSpill : spilled) {
                        indexes.add(indexIO.loadIndex(oneSpill));
                    }
                    fileToUpload = new File(tmpSegmentDir, "merged");
                    theIndexMerger.mergeQueryableIndex(indexes, schema.getGranularitySpec().isRollup(), schema.getAggregators(), fileToUpload, config.getIndexSpec());
                }
                // Map merged segment so we can extract dimensions
                final QueryableIndex mappedSegment = indexIO.loadIndex(fileToUpload);
                final DataSegment segmentToUpload = theSink.getSegment().withDimensions(ImmutableList.copyOf(mappedSegment.getAvailableDimensions())).withBinaryVersion(SegmentUtils.getVersionFromDir(fileToUpload));
                dataSegmentPusher.push(fileToUpload, segmentToUpload);
                log.info("Uploaded segment[%s]", segmentToUpload.getIdentifier());
            } catch (Exception e) {
                log.warn(e, "Failed to merge and upload");
                throw Throwables.propagate(e);
            } finally {
                try {
                    if (fileToUpload != null) {
                        log.info("Deleting Index File[%s]", fileToUpload);
                        FileUtils.deleteDirectory(fileToUpload);
                    }
                } catch (IOException e) {
                    log.warn(e, "Error deleting directory[%s]", fileToUpload);
                }
            }
        }

        private void spillIfSwappable() {
            if (theSink.swappable()) {
                final FireHydrant indexToPersist = theSink.swap();
                final int rowsToPersist = indexToPersist.getIndex().size();
                final File dirToPersist = getSpillDir(indexToPersist.getCount());
                log.info("Spilling index[%d] with rows[%d] to: %s", indexToPersist.getCount(), rowsToPersist, dirToPersist);
                try {
                    theIndexMerger.persist(indexToPersist.getIndex(), dirToPersist, config.getIndexSpec());
                    indexToPersist.swapSegment(null);
                    metrics.incrementRowOutputCount(rowsToPersist);
                    spilled.add(dirToPersist);
                } catch (Exception e) {
                    log.warn(e, "Failed to spill index[%d]", indexToPersist.getCount());
                    throw Throwables.propagate(e);
                }
            }
        }

        private File getSpillDir(final int n) {
            return new File(persistDir, String.format("spill%d", n));
        }
    };
}

Example 4 with IndexMerger

use of io.druid.segment.IndexMerger in project druid by druid-io.

the class AggregationTestHelper method createSelectQueryAggregationTestHelper.

public static final AggregationTestHelper createSelectQueryAggregationTestHelper(List<? extends Module> jsonModulesToRegister, TemporaryFolder tempFolder) {
    ObjectMapper mapper = new DefaultObjectMapper();
    mapper.setInjectableValues(new InjectableValues.Std().addValue(SelectQueryConfig.class, new SelectQueryConfig(true)));
    Supplier<SelectQueryConfig> configSupplier = Suppliers.ofInstance(new SelectQueryConfig(true));
    SelectQueryQueryToolChest toolchest = new SelectQueryQueryToolChest(new DefaultObjectMapper(), QueryRunnerTestHelper.NoopIntervalChunkingQueryRunnerDecorator(), configSupplier);
    SelectQueryRunnerFactory factory = new SelectQueryRunnerFactory(new SelectQueryQueryToolChest(new DefaultObjectMapper(), QueryRunnerTestHelper.NoopIntervalChunkingQueryRunnerDecorator(), configSupplier), new SelectQueryEngine(configSupplier), QueryRunnerTestHelper.NOOP_QUERYWATCHER);
    IndexIO indexIO = new IndexIO(mapper, new ColumnConfig() {

        @Override
        public int columnCacheSizeBytes() {
            return 0;
        }
    });
    return new AggregationTestHelper(mapper, new IndexMerger(mapper, indexIO), indexIO, toolchest, factory, tempFolder, jsonModulesToRegister);
}

Example 5 with IndexMerger

use of io.druid.segment.IndexMerger in project druid by druid-io.

the class AggregationTestHelper method createTimeseriesQueryAggregationTestHelper.

public static final AggregationTestHelper createTimeseriesQueryAggregationTestHelper(List<? extends Module> jsonModulesToRegister, TemporaryFolder tempFolder) {
    ObjectMapper mapper = new DefaultObjectMapper();
    TimeseriesQueryQueryToolChest toolchest = new TimeseriesQueryQueryToolChest(QueryRunnerTestHelper.NoopIntervalChunkingQueryRunnerDecorator());
    TimeseriesQueryRunnerFactory factory = new TimeseriesQueryRunnerFactory(toolchest, new TimeseriesQueryEngine(), QueryRunnerTestHelper.NOOP_QUERYWATCHER);
    IndexIO indexIO = new IndexIO(mapper, new ColumnConfig() {

        @Override
        public int columnCacheSizeBytes() {
            return 0;
        }
    });
    return new AggregationTestHelper(mapper, new IndexMerger(mapper, indexIO), indexIO, toolchest, factory, tempFolder, jsonModulesToRegister);
}