Package org.apache.druid.frame.processor

Class TrackingDimensionSelector

java.lang.Object

org.apache.druid.frame.processor.TrackingDimensionSelector

All Implemented Interfaces:

HotLoopCallee, BaseDoubleColumnValueSelector, BaseFloatColumnValueSelector, BaseLongColumnValueSelector, BaseNullableColumnValueSelector, BaseObjectColumnValueSelector<Object>, ColumnValueSelector<Object>, DimensionDictionarySelector, DimensionSelector

public class TrackingDimensionSelector
extends Object
implements DimensionSelector

DimensionSelector that delegates to a selector obtained from a ColumnSelectorFactory supplier. The delegate is refreshed when the supplier returns a different factory (by identity). This allows the selector to remain valid across changes to the underlying data source, such as frame changes during merging or combining.

Nested Class Summary

Nested classes/interfaces inherited from interface org.apache.druid.segment.DimensionSelector

DimensionSelector.NullDimensionSelectorHolder

Field Summary

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

EMPTY_ARRAY

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

CARDINALITY_UNKNOWN

Constructor Summary

Constructors

Constructor	Description
TrackingDimensionSelector(DimensionSpec dimensionSpec, Supplier<ColumnSelectorFactory> factorySupplier)

Method Summary

Modifier and Type	Method	Description
Class<?>	classOfObject()	Most-specific class of object returned by BaseObjectColumnValueSelector.getObject(), if known in advance.
Object	getObject()	Returns the currently-selected object.
IndexedInts	getRow()	Returns the indexed values at the current position in this DimensionSelector.
int	getValueCardinality()	Value cardinality is the cardinality of the different occurring values.
IdLookup	idLookup()	Returns IdLookup if available for this DimensionSelector, or null.
void	inspectRuntimeShape(RuntimeShapeInspector inspector)	Implementations of this method should call inspector.visit() with all fields of this class, which meet two conditions: 1.
String	lookupName(int id)	Returns the value for a particular dictionary id as a Java String.
ByteBuffer	lookupNameUtf8(int id)	Returns the value for a particular dictionary id as UTF-8 bytes.
ValueMatcher	makeValueMatcher(String value)
ValueMatcher	makeValueMatcher(DruidPredicateFactory predicateFactory)
boolean	nameLookupPossibleInAdvance()	Returns true if it is possible to DimensionDictionarySelector.lookupName(int) by ids from 0 to DimensionDictionarySelector.getValueCardinality() before the rows with those ids are returned.
boolean	supportsLookupNameUtf8()	Returns whether this selector supports DimensionDictionarySelector.lookupNameUtf8(int).

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.DimensionSelector

defaultGetObject, getDouble, getFloat, getLong, isNull

Constructor Details

TrackingDimensionSelector

public TrackingDimensionSelector(DimensionSpec dimensionSpec,
 Supplier<ColumnSelectorFactory> factorySupplier)

Method Details

getRow

public IndexedInts getRow()

Description copied from interface: DimensionSelector

Returns the indexed values at the current position in this DimensionSelector. IMPORTANT. The returned IndexedInts object could generally be reused inside the implementation of DimensionSelector, i. e. this method could always return the same object for the same selector. Users of this API, such as Aggregator.aggregate(), BufferAggregator.aggregate(java.nio.ByteBuffer, int), AggregateCombiner.reset(org.apache.druid.segment.ColumnValueSelector), AggregateCombiner.fold(org.apache.druid.segment.ColumnValueSelector) should be prepared for that and not storing the object returned from this method in their state, assuming that the object will remain unchanged even when the position of the selector changes. This may not be the case.

Specified by:

getRow in interface DimensionSelector

makeValueMatcher

public ValueMatcher makeValueMatcher(@Nullable
 String value)

Specified by:

makeValueMatcher in interface DimensionSelector

Parameters:

value - nullable dimension value

makeValueMatcher

public ValueMatcher makeValueMatcher(DruidPredicateFactory predicateFactory)

Specified by:

makeValueMatcher in interface DimensionSelector

getObject

@Nullable
public Object getObject()

Description copied from interface: BaseObjectColumnValueSelector

Returns the currently-selected object. The behavior of this method depends on the type of selector, which can be determined by calling ColumnSelectorFactory.getColumnCapabilities(String) on the same ColumnSelectorFactory that you got this selector from. If the capabilties are nonnull, the selector type is given by TypeSignature.getType(). String selectors, where type is ColumnType.STRING, may return any type of object from this method, especially in cases where the selector is casting objects to string at selection time. Callers are encouraged to avoid the need to deal with various objects by using ColumnSelectorFactory.makeDimensionSelector(org.apache.druid.query.dimension.DimensionSpec) instead. Numeric selectors, where TypeSignature.isNumeric(), may return any type of Number. Callers that wish to deal with more specific types should treat the original ColumnValueSelector as a BaseLongColumnValueSelector, BaseDoubleColumnValueSelector, or BaseFloatColumnValueSelector instead. Array selectors, where TypeSignature.isArray(), must return Object[]. The array may contain null elements, and the array itself may also be null. Selectors of unknown type, where ColumnSelectorFactory.getColumnCapabilities(String) returns null, may return any type of object. Callers must be prepared for a wide variety of possible input objects. This case is common during ingestion, where selectors are built on top of external data.

Specified by:

getObject in interface BaseObjectColumnValueSelector<Object>

classOfObject

public Class<?> classOfObject()

Description copied from interface: BaseObjectColumnValueSelector

Most-specific class of object returned by BaseObjectColumnValueSelector.getObject(), if known in advance. This method returns Object when selectors do not know in advance what class of object they may return.

Specified by:

classOfObject in interface BaseObjectColumnValueSelector<Object>

getValueCardinality

public int getValueCardinality()

Description copied from interface: DimensionDictionarySelector

Value cardinality is the cardinality of the different occurring values. If there were 4 rows: A,B A B A Value cardinality would be 2. Cardinality may be unknown (e.g. the selector used by IncrementalIndex while reading input rows), in which case this method will return -1. If cardinality is unknown, you should assume this dimension selector has no dictionary, and avoid storing ids, calling "lookupId", or calling "lookupName" outside of the context of operating on a single row. If cardinality is known then it is assumed that underlying dictionary is lexicographically sorted by the encoded value. For example if there are values "A" , "B" , "C" in a column with cardinality 3 then it is assumed that id("A") < id("B") < id("C")

Specified by:

getValueCardinality in interface DimensionDictionarySelector

Returns:

the value cardinality, or DimensionDictionarySelector.CARDINALITY_UNKNOWN if unknown.

lookupName

@Nullable
public String lookupName(int id)

Description copied from interface: DimensionDictionarySelector

Returns the value for a particular dictionary id as a Java String. For example, if a column has four rows: A,B A A,B B getRow() would return getRow(0) => [0 1] getRow(1) => [0] getRow(2) => [0 1] getRow(3) => [1] and then lookupName would return: lookupName(0) => A lookupName(1) => B Performance note: if you want a java.lang.String, always use this method. It will be at least as fast as calling DimensionDictionarySelector.lookupNameUtf8(int) and decoding the bytes. However, if you want UTF-8 bytes, then check if DimensionDictionarySelector.supportsLookupNameUtf8() returns true, and if it does, use DimensionDictionarySelector.lookupNameUtf8(int) instead.

Specified by:

lookupName in interface DimensionDictionarySelector

Parameters:

id - id to lookup the dictionary value for

Returns:

dictionary value for the given id, or null if the value is itself null

lookupNameUtf8

@Nullable
public ByteBuffer lookupNameUtf8(int id)

Description copied from interface: DimensionDictionarySelector

Returns the value for a particular dictionary id as UTF-8 bytes. The returned buffer is in big-endian order. It is not reused, so callers may modify the position, limit, byte order, etc of the buffer. The returned buffer may point to the original data, so callers must take care not to use it outside the valid lifetime of this selector. In particular, if the original data came from a reference-counted segment, callers must not use the returned ByteBuffer after releasing their reference to the relevant ReferenceCountedSegmentProvider. Performance note: if you want UTF-8 bytes, and DimensionDictionarySelector.supportsLookupNameUtf8() returns true, always use this method. It will be at least as fast as calling DimensionDictionarySelector.lookupName(int) and encoding the bytes. However, if you want a java.lang.String, then use DimensionDictionarySelector.lookupName(int) instead of this method.

Specified by:

lookupNameUtf8 in interface DimensionDictionarySelector

Parameters:

id - id to lookup the dictionary value for

Returns:

dictionary value for the given id, or null if the value is itself null

supportsLookupNameUtf8

public boolean supportsLookupNameUtf8()

Description copied from interface: DimensionDictionarySelector

Returns whether this selector supports DimensionDictionarySelector.lookupNameUtf8(int).

Specified by:

supportsLookupNameUtf8 in interface DimensionDictionarySelector

nameLookupPossibleInAdvance

public boolean nameLookupPossibleInAdvance()

Description copied from interface: DimensionDictionarySelector

Returns true if it is possible to DimensionDictionarySelector.lookupName(int) by ids from 0 to DimensionDictionarySelector.getValueCardinality() before the rows with those ids are returned.

Returns false if DimensionDictionarySelector.lookupName(int) could be called with ids, returned from the most recent row (or row vector) returned by this DimensionSelector, but not earlier. If DimensionDictionarySelector.getValueCardinality() of this selector additionally returns DimensionDictionarySelector.CARDINALITY_UNKNOWN, lookupName() couldn't be called with ids, returned by not the most recent row (or row vector), i. e. names for ids couldn't be looked up "later". If DimensionDictionarySelector.getValueCardinality() returns a non-negative number, lookupName() could be called with any ids, returned from rows (or row vectors) returned since the creation of this DimensionSelector.

If DimensionDictionarySelector.lookupName(int) is called with an ineligible id, result is undefined: exception could be thrown, or null returned, or some other random value.

Specified by:

nameLookupPossibleInAdvance in interface DimensionDictionarySelector

idLookup

@Nullable
public IdLookup idLookup()

Description copied from interface: DimensionDictionarySelector

Returns IdLookup if available for this DimensionSelector, or null.

Specified by:

idLookup in interface DimensionDictionarySelector

inspectRuntimeShape

public void inspectRuntimeShape(RuntimeShapeInspector inspector)

Description copied from interface: HotLoopCallee

Implementations of this method should call inspector.visit() with all fields of this class, which meet two conditions: 1. They are used in methods of this class, annotated with CalledFromHotLoop 2. They are either: a. Nullable objects b. Instances of HotLoopCallee c. Objects, which don't always have a specific class in runtime. For example, a field of type Set could be HashSet or TreeSet in runtime, depending on how this instance (the instance on which inspectRuntimeShape() is called) is configured. d. ByteBuffer or similar objects, where byte order matters e. boolean flags, affecting branch taking f. Arrays of objects, meeting any of conditions a-e.

Specified by:

inspectRuntimeShape in interface HotLoopCallee