org.apache.druid.segment

Class IndexMergerV9

java.lang.Object

org.apache.druid.segment.IndexMergerV9

All Implemented Interfaces:

IndexMerger

public class IndexMergerV9
extends Object
implements IndexMerger

Field Summary

Fields

Modifier and Type	Field and Description
static ColumnCapabilities.CoercionLogic	DIMENSION_CAPABILITY_MERGE_LOGIC
static ColumnCapabilities.CoercionLogic	METRIC_CAPABILITY_MERGE_LOGIC

Fields inherited from interface org.apache.druid.segment.IndexMerger

INVALID_ROW, SERIALIZER_UTILS, UNLIMITED_MAX_COLUMNS_TO_MERGE

Constructor Summary

Constructors

Constructor and Description
IndexMergerV9(com.fasterxml.jackson.databind.ObjectMapper mapper, IndexIO indexIO, SegmentWriteOutMediumFactory defaultSegmentWriteOutMediumFactory) This constructor is used only for Hadoop ingestion and Tranquility as they do not support storing empty columns yet.
IndexMergerV9(com.fasterxml.jackson.databind.ObjectMapper mapper, IndexIO indexIO, SegmentWriteOutMediumFactory defaultSegmentWriteOutMediumFactory, boolean storeEmptyColumns)

Method Summary

Modifier and Type	Method and Description
File	merge(List<IndexableAdapter> indexes, boolean rollup, AggregatorFactory[] metricAggs, File outDir, DimensionsSpec dimensionsSpec, IndexSpec indexSpec, int maxColumnsToMerge) Only used as a convenience method in tests.
File	mergeQueryableIndex(List<QueryableIndex> indexes, boolean rollup, AggregatorFactory[] metricAggs, DimensionsSpec dimensionsSpec, File outDir, IndexSpec indexSpec, IndexSpec indexSpecForIntermediatePersists, ProgressIndicator progress, SegmentWriteOutMediumFactory segmentWriteOutMediumFactory, int maxColumnsToMerge) Merge a collection of QueryableIndex.
File	persist(IncrementalIndex index, org.joda.time.Interval dataInterval, File outDir, IndexSpec indexSpec, ProgressIndicator progress, SegmentWriteOutMediumFactory segmentWriteOutMediumFactory) Persist an IncrementalIndex to disk in such a way that it can be loaded back up as a QueryableIndex.

Methods inherited from class java.lang.Object

clone, equals, finalize, getClass, hashCode, notify, notifyAll, toString, wait, wait, wait

Methods inherited from interface org.apache.druid.segment.IndexMerger

getLexicographicMergedDimensions, getLongestSharedDimOrder, getMergedDimensions, getMergedDimensionsFromQueryableIndexes, isDimensionOrderingValid, makeRowIteratorWithConvertedDimensionColumns, mergeIndexed, mergeQueryableIndex, persist, persist, toIndexableAdapters, toMergedIndexRowIterator

Field Detail

DIMENSION_CAPABILITY_MERGE_LOGIC

public static final ColumnCapabilities.CoercionLogic DIMENSION_CAPABILITY_MERGE_LOGIC

METRIC_CAPABILITY_MERGE_LOGIC

public static final ColumnCapabilities.CoercionLogic METRIC_CAPABILITY_MERGE_LOGIC

Constructor Detail

IndexMergerV9

public IndexMergerV9(com.fasterxml.jackson.databind.ObjectMapper mapper,
                     IndexIO indexIO,
                     SegmentWriteOutMediumFactory defaultSegmentWriteOutMediumFactory,
                     boolean storeEmptyColumns)

IndexMergerV9

@Inject
public IndexMergerV9(com.fasterxml.jackson.databind.ObjectMapper mapper,
                             IndexIO indexIO,
                             SegmentWriteOutMediumFactory defaultSegmentWriteOutMediumFactory)

This constructor is used only for Hadoop ingestion and Tranquility as they do not support storing empty columns yet. See HadoopDruidIndexerConfig and PlumberSchool for hadoop ingestion and Tranquility, respectively.

Method Detail

persist

public File persist(IncrementalIndex index,
                    org.joda.time.Interval dataInterval,
                    File outDir,
                    IndexSpec indexSpec,
                    ProgressIndicator progress,
                    @Nullable
                    SegmentWriteOutMediumFactory segmentWriteOutMediumFactory)
             throws IOException

Description copied from interface: IndexMerger

Persist an IncrementalIndex to disk in such a way that it can be loaded back up as a QueryableIndex. This is *not* thread-safe and havoc will ensue if this is called and writes are still occurring on the IncrementalIndex object.

Specified by:

persist in interface IndexMerger

Parameters:

index - the IncrementalIndex to persist

dataInterval - the Interval that the data represents. Typically, this is the same as the interval from the corresponding SegmentId.

outDir - the directory to persist the data to

indexSpec - storage and compression options

progress - an object that will receive progress updates

segmentWriteOutMediumFactory - controls allocation of temporary data structures

Returns:

the index output directory

Throws:

IOException - if an IO error occurs persisting the index

mergeQueryableIndex

public File mergeQueryableIndex(List<QueryableIndex> indexes,
                                boolean rollup,
                                AggregatorFactory[] metricAggs,
                                @Nullable
                                DimensionsSpec dimensionsSpec,
                                File outDir,
                                IndexSpec indexSpec,
                                IndexSpec indexSpecForIntermediatePersists,
                                ProgressIndicator progress,
                                @Nullable
                                SegmentWriteOutMediumFactory segmentWriteOutMediumFactory,
                                int maxColumnsToMerge)
                         throws IOException

Description copied from interface: IndexMerger

Merge a collection of QueryableIndex.

Specified by:

mergeQueryableIndex in interface IndexMerger

Throws:

IOException

merge

public File merge(List<IndexableAdapter> indexes,
                  boolean rollup,
                  AggregatorFactory[] metricAggs,
                  File outDir,
                  DimensionsSpec dimensionsSpec,
                  IndexSpec indexSpec,
                  int maxColumnsToMerge)
           throws IOException

Description copied from interface: IndexMerger

Only used as a convenience method in tests. In production code, to merge multiple QueryableIndex, use IndexMerger.mergeQueryableIndex(List, boolean, AggregatorFactory[], DimensionsSpec, File, IndexSpec, IndexSpec, ProgressIndicator, SegmentWriteOutMediumFactory, int). To merge multiple IncrementalIndex, call one of the IndexMerger.persist(org.apache.druid.segment.incremental.IncrementalIndex, java.io.File, org.apache.druid.segment.IndexSpec, org.apache.druid.segment.writeout.SegmentWriteOutMediumFactory) methods and then merge the resulting QueryableIndex.

Specified by:

merge in interface IndexMerger

Throws:

IOException