org.apache.druid.segment.join

Class HashJoinSegmentStorageAdapter

java.lang.Object

org.apache.druid.segment.join.HashJoinSegmentStorageAdapter

All Implemented Interfaces:

Expr.InputBindingInspector, ColumnInspector, CursorFactory, StorageAdapter

public class HashJoinSegmentStorageAdapter
extends Object
implements StorageAdapter

Method Summary

Modifier and Type	Method and Description
boolean	canVectorize(Filter filter, VirtualColumns virtualColumns, boolean descending) Returns true if the provided combination of parameters can be handled by "makeVectorCursor".
Set<String>	determineBaseColumnsWithPreAndPostJoinVirtualColumns(VirtualColumns virtualColumns, List<VirtualColumn> preJoinVirtualColumns, List<VirtualColumn> postJoinVirtualColumns) Return a String set containing the name of columns that belong to the base table (including any pre-join virtual columns as well).
Indexed<String>	getAvailableDimensions()
Iterable<String>	getAvailableMetrics()
ColumnCapabilities	getColumnCapabilities(String column) Returns capabilities of a particular column, if known.
String	getColumnTypeName(String column) Like ColumnCapabilities.getType(), but may return a more descriptive string for complex columns.
int	getDimensionCardinality(String column) Returns the number of distinct values for the given column if known, or Integer.MAX_VALUE if unknown, e.
org.joda.time.Interval	getInterval()
org.joda.time.DateTime	getMaxIngestedEventTime()
org.joda.time.DateTime	getMaxTime()
Comparable	getMaxValue(String column)
Metadata	getMetadata()
org.joda.time.DateTime	getMinTime()
Comparable	getMinValue(String column)
int	getNumRows()
boolean	hasBuiltInFilters() Returns true if this storage adapter can filter some rows out.
boolean	isBaseColumn(String column) Returns whether "column" will be selected from "baseAdapter".
Sequence<Cursor>	makeCursors(Filter filter, org.joda.time.Interval interval, VirtualColumns virtualColumns, Granularity gran, boolean descending, QueryMetrics<?> queryMetrics) Creates a sequence of Cursors, one for each time-granular bucket (based on the provided Granularity).
VectorCursor	makeVectorCursor(Filter filter, org.joda.time.Interval interval, VirtualColumns virtualColumns, boolean descending, int vectorSize, QueryMetrics<?> queryMetrics) Creates a VectorCursor.

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.ColumnInspector

getType

Methods inherited from interface org.apache.druid.math.expr.Expr.InputBindingInspector

areNumeric, areNumeric, areScalar, areScalar, canVectorize, canVectorize

Method Detail

getInterval

public org.joda.time.Interval getInterval()

Specified by:

getInterval in interface StorageAdapter

getAvailableDimensions

public Indexed<String> getAvailableDimensions()

Specified by:

getAvailableDimensions in interface StorageAdapter

getAvailableMetrics

public Iterable<String> getAvailableMetrics()

Specified by:

getAvailableMetrics in interface StorageAdapter

getDimensionCardinality

public int getDimensionCardinality(String column)

Description copied from interface: StorageAdapter

Returns the number of distinct values for the given column if known, or Integer.MAX_VALUE if unknown, e. g. the column is numeric. If the column doesn't exist, returns 0.

Specified by:

getDimensionCardinality in interface StorageAdapter

getMinTime

public org.joda.time.DateTime getMinTime()

Specified by:

getMinTime in interface StorageAdapter

getMaxTime

public org.joda.time.DateTime getMaxTime()

Specified by:

getMaxTime in interface StorageAdapter

getMinValue

@Nullable
public Comparable getMinValue(String column)

Specified by:

getMinValue in interface StorageAdapter

getMaxValue

@Nullable
public Comparable getMaxValue(String column)

Specified by:

getMaxValue in interface StorageAdapter

getColumnCapabilities

@Nullable
public ColumnCapabilities getColumnCapabilities(String column)

Description copied from interface: StorageAdapter

Returns capabilities of a particular column, if known. May be null if the column doesn't exist, or if the column does exist but the capabilities are unknown. The latter is possible with dynamically discovered columns. Note that StorageAdapters are representations of "real" segments, so they are not aware of any virtual columns that may be involved in a query. In general, query engines should instead use the method ColumnSelectorFactory.getColumnCapabilities(String), which returns capabilities for virtual columns as well.

Specified by:

getColumnCapabilities in interface ColumnInspector

Specified by:

getColumnCapabilities in interface StorageAdapter

Parameters:

column - column name

Returns:

capabilities, or null

getColumnTypeName

@Nullable
public String getColumnTypeName(String column)

Description copied from interface: StorageAdapter

Like ColumnCapabilities.getType(), but may return a more descriptive string for complex columns.

Specified by:

getColumnTypeName in interface StorageAdapter

Parameters:

column - column name

Returns:

type name

getNumRows

public int getNumRows()

Specified by:

getNumRows in interface StorageAdapter

getMaxIngestedEventTime

public org.joda.time.DateTime getMaxIngestedEventTime()

Specified by:

getMaxIngestedEventTime in interface StorageAdapter

getMetadata

public Metadata getMetadata()

Specified by:

getMetadata in interface StorageAdapter

hasBuiltInFilters

public boolean hasBuiltInFilters()

Description copied from interface: StorageAdapter

Returns true if this storage adapter can filter some rows out. The actual column cardinality can be lower than what StorageAdapter.getDimensionCardinality(java.lang.String) returns if this returns true. Dimension selectors for such storage adapter can return non-contiguous dictionary IDs because the dictionary IDs in filtered rows will not be returned. Note that the number of rows accessible via this storage adapter will not necessarily decrease because of the built-in filters. For inner joins, for example, the number of joined rows can be larger than the number of rows in the base adapter even though this method returns true.

Specified by:

hasBuiltInFilters in interface StorageAdapter

canVectorize

public boolean canVectorize(@Nullable
                            Filter filter,
                            VirtualColumns virtualColumns,
                            boolean descending)

Description copied from interface: CursorFactory

Returns true if the provided combination of parameters can be handled by "makeVectorCursor". Query engines should use this before running in vectorized mode, and be prepared to fall back to non-vectorized mode if this method returns false.

Specified by:

canVectorize in interface CursorFactory

makeVectorCursor

@Nullable
public VectorCursor makeVectorCursor(@Nullable
                                               Filter filter,
                                               org.joda.time.Interval interval,
                                               VirtualColumns virtualColumns,
                                               boolean descending,
                                               int vectorSize,
                                               @Nullable
                                               QueryMetrics<?> queryMetrics)

Description copied from interface: CursorFactory

Creates a VectorCursor. Unlike the Cursor returned by "makeCursor", there is just one of these. Hence, this method does not take a "granularity" parameter. Before calling this method, check "canVectorize" to see if the call you are about to make will throw an error or not. Returns null if there is no data to walk over (for example, if the "interval" does not overlap the data interval of this segment).

Specified by:

makeVectorCursor in interface CursorFactory

makeCursors

public Sequence<Cursor> makeCursors(@Nullable
                                    Filter filter,
                                    @Nonnull
                                    org.joda.time.Interval interval,
                                    @Nonnull
                                    VirtualColumns virtualColumns,
                                    @Nonnull
                                    Granularity gran,
                                    boolean descending,
                                    @Nullable
                                    QueryMetrics<?> queryMetrics)

Description copied from interface: CursorFactory

Creates a sequence of Cursors, one for each time-granular bucket (based on the provided Granularity).

Specified by:

makeCursors in interface CursorFactory

isBaseColumn

public boolean isBaseColumn(String column)

Returns whether "column" will be selected from "baseAdapter". This is true if it is not shadowed by any joinables (i.e. if it does not start with any of their prefixes).

determineBaseColumnsWithPreAndPostJoinVirtualColumns

public Set<String> determineBaseColumnsWithPreAndPostJoinVirtualColumns(VirtualColumns virtualColumns,
                                                                        @Nullable
                                                                        List<VirtualColumn> preJoinVirtualColumns,
                                                                        @Nullable
                                                                        List<VirtualColumn> postJoinVirtualColumns)

Return a String set containing the name of columns that belong to the base table (including any pre-join virtual columns as well). Additionally, if the preJoinVirtualColumns and/or postJoinVirtualColumns arguments are provided, this method will add each VirtualColumn in the provided virtualColumns to either preJoinVirtualColumns or postJoinVirtualColumns based on whether the virtual column is pre-join or post-join.

Parameters:

virtualColumns - List of virtual columns from the query

preJoinVirtualColumns - If provided, virtual columns determined to be pre-join will be added to this list

postJoinVirtualColumns - If provided, virtual columns determined to be post-join will be added to this list

Returns:

The set of base column names, including any pre-join virtual columns.