Package io.deephaven.engine.table.impl.by

Class GroupByReaggregateOperator

java.lang.Object

io.deephaven.engine.table.impl.by.GroupByReaggregateOperator

All Implemented Interfaces:

GroupByOperator, IterativeChunkedAggregationOperator

public final class GroupByReaggregateOperator
extends Object
implements GroupByOperator

An IterativeChunkedAggregationOperator used to re-aggregate the results of an AggGroup as part of a rollup.

The operator is fundamentally different than the GroupByChunkedOperator. Rather than examining row keys, it listens to the rollup's base (or intermediate) level and reads the exposed RowSet column. The relevant RowSets are added to a random builder for each state while processing an update (or initialization). At the end of the update cycle, it builds the rowsets and updates an internal ObjectArraySource of RowSets.

The resulting column sources are once again AggregateColumnSource, which reuse the wrapped aggregated column source from the source table (thus each level of the rollup uses the original table's sources as the input to the AggregateColumnSources -- not the immediately prior level).

Nested Class Summary

Nested classes/interfaces inherited from interface io.deephaven.engine.table.impl.by.IterativeChunkedAggregationOperator

IterativeChunkedAggregationOperator.BucketedContext, IterativeChunkedAggregationOperator.SingletonContext

Field Summary

Fields inherited from interface io.deephaven.engine.table.impl.by.IterativeChunkedAggregationOperator

ZERO_LENGTH_ITERATIVE_CHUNKED_AGGREGATION_OPERATOR_ARRAY

Constructor Summary

Constructors

Constructor	Description
GroupByReaggregateOperator(@NotNull QueryTable inputTable, boolean registeredWithHelper, @Nullable String exposeRowSetsAs, @Nullable List<String> hiddenResults, @NotNull MatchPair... aggregatedColumnPairs)

Method Summary

Modifier and Type	Method	Description
void	addChunk(IterativeChunkedAggregationOperator.BucketedContext bucketedContext, Chunk<? extends Values> values, LongChunk<? extends RowKeys> inputRowKeys, @NotNull IntChunk<RowKeys> destinations, @NotNull IntChunk<ChunkPositions> startPositions, @NotNull IntChunk<ChunkLengths> length, @NotNull WritableBooleanChunk<Values> stateModified)	Aggregate a chunk of data into the result columns.
boolean	addChunk(IterativeChunkedAggregationOperator.SingletonContext singletonContext, int chunkSize, Chunk<? extends Values> values, @NotNull LongChunk<? extends RowKeys> inputRowKeys, long destination)	Aggregate a chunk of data into the result columns.
void	ensureCapacity(long tableSize)	Ensure that this operator can handle destinations up to tableSize - 1.
MatchPair[]	getAggregatedColumnPairs()
String	getExposedRowSetsAs()
List<String>	getHiddenResults()
Map<String,? extends ColumnSource<?>>	getInputResultColumns()	Get a map from input column names to the corresponding output ColumnSource.
Map<String,? extends ColumnSource<?>>	getResultColumns()	Return a map of result columns produced by this operator.
boolean	hasModifications(boolean columnsModified)	Determine whether to propagate changes when input columns have been modified.
UnaryOperator<ModifiedColumnSet>	initializeRefreshing(@NotNull QueryTable resultTable, @NotNull LivenessReferent aggregationUpdateListener)	Initialize refreshing result support for this operator.
void	modifyChunk(IterativeChunkedAggregationOperator.BucketedContext bucketedContext, Chunk<? extends Values> previousValues, Chunk<? extends Values> newValues, LongChunk<? extends RowKeys> postShiftRowKeys, @NotNull IntChunk<RowKeys> destinations, @NotNull IntChunk<ChunkPositions> startPositions, @NotNull IntChunk<ChunkLengths> length, @NotNull WritableBooleanChunk<Values> stateModified)	Modify a chunk of data previously aggregated into the result columns using a parallel chunk of new values.
boolean	modifyChunk(IterativeChunkedAggregationOperator.SingletonContext singletonContext, int chunkSize, Chunk<? extends Values> previousValues, Chunk<? extends Values> newValues, LongChunk<? extends RowKeys> postShiftRowKeys, long destination)	Modify a chunk of data previously aggregated into the result columns using a parallel chunk of new values.
void	propagateInitialState(@NotNull QueryTable resultTable, int startingDestinationsCount)	Perform any internal state keeping needed for destinations that were added during initialization.
void	propagateUpdates(@NotNull TableUpdate downstream, @NotNull RowSet newDestinations)	Perform any internal state keeping needed for destinations that were added (went from 0 keys to > 0), removed (went from > 0 keys to 0), or modified (keys added or removed, or keys modified) by this iteration.
void	removeChunk(IterativeChunkedAggregationOperator.BucketedContext bucketedContext, Chunk<? extends Values> values, LongChunk<? extends RowKeys> inputRowKeys, @NotNull IntChunk<RowKeys> destinations, @NotNull IntChunk<ChunkPositions> startPositions, @NotNull IntChunk<ChunkLengths> length, @NotNull WritableBooleanChunk<Values> stateModified)	Remove a chunk of data previously aggregated into the result columns.
boolean	removeChunk(IterativeChunkedAggregationOperator.SingletonContext singletonContext, int chunkSize, Chunk<? extends Values> values, @NotNull LongChunk<? extends RowKeys> inputRowKeys, long destination)	Remove a chunk of data previously aggregated into the result columns.
boolean	resetForStep(@NotNull TableUpdate upstream, int startingDestinationsCount)	Reset any per-step internal state.
@NotNull IterativeChunkedAggregationOperator	resultExtractor(List<Pair> resultPairs)
void	shiftChunk(IterativeChunkedAggregationOperator.BucketedContext bucketedContext, Chunk<? extends Values> previousValues, Chunk<? extends Values> newValues, @NotNull LongChunk<? extends RowKeys> preShiftRowKeys, @NotNull LongChunk<? extends RowKeys> postShiftRowKeys, @NotNull IntChunk<RowKeys> destinations, @NotNull IntChunk<ChunkPositions> startPositions, @NotNull IntChunk<ChunkLengths> length, @NotNull WritableBooleanChunk<Values> stateModified)	Called with shifted row keys when IterativeChunkedAggregationOperator.requiresRowKeys() returns true, including shifted same-slot modifies.
boolean	shiftChunk(IterativeChunkedAggregationOperator.SingletonContext singletonContext, Chunk<? extends Values> previousValues, Chunk<? extends Values> newValues, @NotNull LongChunk<? extends RowKeys> preShiftRowKeys, @NotNull LongChunk<? extends RowKeys> postShiftRowKeys, long destination)	Shift a chunk of data previously aggregated into the result columns, including shifted same-slot modifies.
void	startTrackingPrevValues()	Called after initialization; when the operator's result columns must have previous tracking enabled.

Methods inherited from class java.lang.Object

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

Methods inherited from interface io.deephaven.engine.table.impl.by.IterativeChunkedAggregationOperator

addRowSet, makeBucketedContext, makeSingletonContext, modifyRowKeys, modifyRowKeys, propagateFailure, requiresRowKeys, requiresRunFinds, unchunkedRowSet

Constructor Details

GroupByReaggregateOperator

public GroupByReaggregateOperator(@NotNull
 @NotNull QueryTable inputTable,
 boolean registeredWithHelper,
 @Nullable
 @Nullable String exposeRowSetsAs,
 @Nullable
 @Nullable List<String> hiddenResults,
 @NotNull
 @NotNull MatchPair... aggregatedColumnPairs)

Method Details

addChunk

public void addChunk(IterativeChunkedAggregationOperator.BucketedContext bucketedContext,
 Chunk<? extends Values> values,
 LongChunk<? extends RowKeys> inputRowKeys,
 @NotNull
 @NotNull IntChunk<RowKeys> destinations,
 @NotNull
 @NotNull IntChunk<ChunkPositions> startPositions,
 @NotNull
 @NotNull IntChunk<ChunkLengths> length,
 @NotNull
 @NotNull WritableBooleanChunk<Values> stateModified)

Description copied from interface: IterativeChunkedAggregationOperator

Aggregate a chunk of data into the result columns.

Specified by:

addChunk in interface IterativeChunkedAggregationOperator

Parameters:

bucketedContext - the operator-specific context

values - a chunk of values to aggregate

inputRowKeys - the input row keys, in post-shift space

destinations - the destinations in resultColumn to aggregate into, parallel with startPositions and length

startPositions - the starting positions in the chunk for each destination

length - the number of values in the chunk for each destination

stateModified - a boolean output array, parallel to destinations, which is set to true if the corresponding destination has been modified

removeChunk

public void removeChunk(IterativeChunkedAggregationOperator.BucketedContext bucketedContext,
 Chunk<? extends Values> values,
 LongChunk<? extends RowKeys> inputRowKeys,
 @NotNull
 @NotNull IntChunk<RowKeys> destinations,
 @NotNull
 @NotNull IntChunk<ChunkPositions> startPositions,
 @NotNull
 @NotNull IntChunk<ChunkLengths> length,
 @NotNull
 @NotNull WritableBooleanChunk<Values> stateModified)

Description copied from interface: IterativeChunkedAggregationOperator

Remove a chunk of data previously aggregated into the result columns.

Specified by:

removeChunk in interface IterativeChunkedAggregationOperator

Parameters:

bucketedContext - the operator-specific context

values - a chunk of values that have been previously aggregated.

inputRowKeys - the input row keys, in pre-shift space

destinations - the destinations in resultColumn to remove the values from, parallel with startPositions and length

startPositions - the starting positions in the chunk for each destination

length - the number of values in the chunk for each destination

stateModified - a boolean output array, parallel to destinations, which is set to true if the corresponding destination has been modified

modifyChunk

public void modifyChunk(IterativeChunkedAggregationOperator.BucketedContext bucketedContext,
 Chunk<? extends Values> previousValues,
 Chunk<? extends Values> newValues,
 LongChunk<? extends RowKeys> postShiftRowKeys,
 @NotNull
 @NotNull IntChunk<RowKeys> destinations,
 @NotNull
 @NotNull IntChunk<ChunkPositions> startPositions,
 @NotNull
 @NotNull IntChunk<ChunkLengths> length,
 @NotNull
 @NotNull WritableBooleanChunk<Values> stateModified)

Description copied from interface: IterativeChunkedAggregationOperator

Modify a chunk of data previously aggregated into the result columns using a parallel chunk of new values. Never includes modifies that have been shifted if IterativeChunkedAggregationOperator.requiresRowKeys() returns true - those are handled in IterativeChunkedAggregationOperator.shiftChunk(BucketedContext, Chunk, Chunk, LongChunk, LongChunk, IntChunk, IntChunk, IntChunk, WritableBooleanChunk).

Specified by:

modifyChunk in interface IterativeChunkedAggregationOperator

Parameters:

bucketedContext - the operator-specific context

previousValues - a chunk of values that have been previously aggregated

newValues - a chunk of values to aggregate

postShiftRowKeys - the input row keys, in post-shift space

destinations - the destinations in resultColumn to remove the values from, parallel with startPositions and length

startPositions - the starting positions in the chunk for each destination

length - the number of values in the chunk for each destination

stateModified - a boolean output array, parallel to destinations, which is set to true if the corresponding destination has been modified

shiftChunk

public void shiftChunk(IterativeChunkedAggregationOperator.BucketedContext bucketedContext,
 Chunk<? extends Values> previousValues,
 Chunk<? extends Values> newValues,
 @NotNull
 @NotNull LongChunk<? extends RowKeys> preShiftRowKeys,
 @NotNull
 @NotNull LongChunk<? extends RowKeys> postShiftRowKeys,
 @NotNull
 @NotNull IntChunk<RowKeys> destinations,
 @NotNull
 @NotNull IntChunk<ChunkPositions> startPositions,
 @NotNull
 @NotNull IntChunk<ChunkLengths> length,
 @NotNull
 @NotNull WritableBooleanChunk<Values> stateModified)

Description copied from interface: IterativeChunkedAggregationOperator

Called with shifted row keys when IterativeChunkedAggregationOperator.requiresRowKeys() returns true, including shifted same-slot modifies.

Specified by:

shiftChunk in interface IterativeChunkedAggregationOperator

Parameters:

bucketedContext - the operator-specific context

previousValues - a chunk of values that have been previously aggregated.

newValues - a chunk of values to aggregate

preShiftRowKeys - the input row keys, in pre-shift space

postShiftRowKeys - the input row keys, in post-shift space

destinations - the destinations in resultColumn to aggregate into, parallel with startPositions and length

startPositions - the starting positions in the chunk for each destination

length - the number of values in the chunk for each destination

stateModified - a boolean output array, parallel to destinations, which is set to true if the corresponding destination has been modified

addChunk

public boolean addChunk(IterativeChunkedAggregationOperator.SingletonContext singletonContext,
 int chunkSize,
 Chunk<? extends Values> values,
 @NotNull
 @NotNull LongChunk<? extends RowKeys> inputRowKeys,
 long destination)

Description copied from interface: IterativeChunkedAggregationOperator

Aggregate a chunk of data into the result columns.

Specified by:

addChunk in interface IterativeChunkedAggregationOperator

Parameters:

singletonContext - the operator-specific context

chunkSize - the size of the addition

values - the values to aggregate

inputRowKeys - the input row keys, in post-shift space

destination - the destination in the result columns

Returns:

true if the state was modified, false otherwise

removeChunk

public boolean removeChunk(IterativeChunkedAggregationOperator.SingletonContext singletonContext,
 int chunkSize,
 Chunk<? extends Values> values,
 @NotNull
 @NotNull LongChunk<? extends RowKeys> inputRowKeys,
 long destination)

Description copied from interface: IterativeChunkedAggregationOperator

Remove a chunk of data previously aggregated into the result columns.

Specified by:

removeChunk in interface IterativeChunkedAggregationOperator

Parameters:

singletonContext - the operator-specific context

chunkSize - the size of the removal

values - the values to remove from the aggregation

inputRowKeys - the input row keys, in pre-shift space

destination - the destination in the result columns

Returns:

true if the state was modified, false otherwise

modifyChunk

public boolean modifyChunk(IterativeChunkedAggregationOperator.SingletonContext singletonContext,
 int chunkSize,
 Chunk<? extends Values> previousValues,
 Chunk<? extends Values> newValues,
 LongChunk<? extends RowKeys> postShiftRowKeys,
 long destination)

Description copied from interface: IterativeChunkedAggregationOperator

Modify a chunk of data previously aggregated into the result columns using a parallel chunk of new values. Never includes modifies that have been shifted if IterativeChunkedAggregationOperator.requiresRowKeys() returns true - those are handled in IterativeChunkedAggregationOperator.shiftChunk(SingletonContext, Chunk, Chunk, LongChunk, LongChunk, long).

Specified by:

modifyChunk in interface IterativeChunkedAggregationOperator

Parameters:

singletonContext - the operator-specific context

chunkSize - the size of the modification

previousValues - a chunk of values that have been previously aggregated.

newValues - a chunk of values to aggregate

postShiftRowKeys - the input row keys, in post-shift space

Returns:

true if the state was modified, false otherwise

shiftChunk

public boolean shiftChunk(IterativeChunkedAggregationOperator.SingletonContext singletonContext,
 Chunk<? extends Values> previousValues,
 Chunk<? extends Values> newValues,
 @NotNull
 @NotNull LongChunk<? extends RowKeys> preShiftRowKeys,
 @NotNull
 @NotNull LongChunk<? extends RowKeys> postShiftRowKeys,
 long destination)

Description copied from interface: IterativeChunkedAggregationOperator

Shift a chunk of data previously aggregated into the result columns, including shifted same-slot modifies.

Specified by:

shiftChunk in interface IterativeChunkedAggregationOperator

Parameters:

singletonContext - the operator-specific context

previousValues - a chunk of values that have been previously aggregated.

newValues - a chunk of values to aggregate

preShiftRowKeys - the input row keys, in pre-shift space

postShiftRowKeys - the input row keys, in post-shift space

destination - the destination in the result columns

Returns:

true if the result should be considered modified

ensureCapacity

public void ensureCapacity(long tableSize)

Description copied from interface: IterativeChunkedAggregationOperator

Ensure that this operator can handle destinations up to tableSize - 1.

Specified by:

ensureCapacity in interface IterativeChunkedAggregationOperator

Parameters:

tableSize - the new size of the table

getResultColumns

public Map<String,? extends ColumnSource<?>> getResultColumns()

Description copied from interface: IterativeChunkedAggregationOperator

Return a map of result columns produced by this operator.

Specified by:

getResultColumns in interface IterativeChunkedAggregationOperator

Returns:

a map of name to columns for the result table

startTrackingPrevValues

public void startTrackingPrevValues()

Description copied from interface: IterativeChunkedAggregationOperator

Called after initialization; when the operator's result columns must have previous tracking enabled.

Specified by:

startTrackingPrevValues in interface IterativeChunkedAggregationOperator

initializeRefreshing

public UnaryOperator<ModifiedColumnSet> initializeRefreshing(@NotNull
 @NotNull QueryTable resultTable,
 @NotNull
 @NotNull LivenessReferent aggregationUpdateListener)

Description copied from interface: IterativeChunkedAggregationOperator

Initialize refreshing result support for this operator. As a side effect, make a factory method for converting upstream modified column sets to result modified column sets, to be invoked whenever this operator reports a modification in order to determine the operator's contribution to the final result modified column set.

Specified by:

initializeRefreshing in interface IterativeChunkedAggregationOperator

Parameters:

resultTable - The result QueryTable after initialization

aggregationUpdateListener - The aggregation update listener, which may be needed for referential integrity

Returns:

A factory that produces a result modified column set from the upstream modified column set

getInputResultColumns

public Map<String,? extends ColumnSource<?>> getInputResultColumns()

Description copied from interface: GroupByOperator

Get a map from input column names to the corresponding output ColumnSource.

Specified by:

getInputResultColumns in interface GroupByOperator

hasModifications

public boolean hasModifications(boolean columnsModified)

Description copied from interface: GroupByOperator

Determine whether to propagate changes when input columns have been modified.

Specified by:

hasModifications in interface GroupByOperator

Parameters:

columnsModified - have any of the input columns been modified (as per the MCS)?

Returns:

true if we have modified our output (e.g., because of additions or modifications).

resetForStep

public boolean resetForStep(@NotNull
 @NotNull TableUpdate upstream,
 int startingDestinationsCount)

Description copied from interface: IterativeChunkedAggregationOperator

Reset any per-step internal state. Note that the arguments to this method should not be mutated in any way.

Specified by:

resetForStep in interface IterativeChunkedAggregationOperator

Parameters:

upstream - The upstream ShiftAwareListener.Update

startingDestinationsCount - The number of used destinations at the beginning of this step

Returns:

true if this operator must generate modifications on this cycle; typically an operator returns false and depends on the actual add/remove/modify/shift calls to determine modifications

propagateInitialState

public void propagateInitialState(@NotNull
 @NotNull QueryTable resultTable,
 int startingDestinationsCount)

Description copied from interface: IterativeChunkedAggregationOperator

Perform any internal state keeping needed for destinations that were added during initialization.

Specified by:

propagateInitialState in interface IterativeChunkedAggregationOperator

Parameters:

resultTable - The result QueryTable after initialization

startingDestinationsCount - The number of used destinations at the beginning of this step

propagateUpdates

public void propagateUpdates(@NotNull
 @NotNull TableUpdate downstream,
 @NotNull
 @NotNull RowSet newDestinations)

Description copied from interface: IterativeChunkedAggregationOperator

Perform any internal state keeping needed for destinations that were added (went from 0 keys to > 0), removed (went from > 0 keys to 0), or modified (keys added or removed, or keys modified) by this iteration. Note that the arguments to this method should not be mutated in any way.

Specified by:

propagateUpdates in interface IterativeChunkedAggregationOperator

Parameters:

downstream - The downstream TableUpdate (which does not have its ModifiedColumnSet finalized yet)

newDestinations - New destinations added on this update

getExposedRowSetsAs

public String getExposedRowSetsAs()

getAggregatedColumnPairs

public MatchPair[] getAggregatedColumnPairs()

getHiddenResults

public List<String> getHiddenResults()

resultExtractor

@NotNull
public @NotNull IterativeChunkedAggregationOperator resultExtractor(List<Pair> resultPairs)