org.apache.lucene.facet.search
Class TopKInEachNodeHandler

java.lang.Object
  org.apache.lucene.facet.search.FacetResultsHandler
      org.apache.lucene.facet.search.TopKInEachNodeHandler

public class TopKInEachNodeHandler
extends FacetResultsHandler

Generates FacetResult from the count arrays aggregated for a particular FacetRequest. The generated FacetResult is a subtree of the taxonomy tree. Its root node, FacetResult.getFacetResultNode(), is the facet specified by FacetRequest.getCategoryPath(), and the enumerated children, FacetResultNode.getSubResults(), of each node in that FacetResult are the top K ( = FacetRequest.getNumResults()) among its children in the taxonomy. Top in the sense FacetRequest.getSortBy(), which can be by the values aggregated in the count arrays, or by ordinal numbers; also specified is the sort order, FacetRequest.getSortOrder(), ascending or descending, of these values or ordinals before their top K are selected. The depth (number of levels excluding the root) of the FacetResult tree is specified by FacetRequest.getDepth().

Because the number of selected children of each node is restricted, and not the overall number of nodes in the FacetResult, facets not selected into FacetResult might have better values, or ordinals, (typically, higher counts), than facets that are selected into the FacetResult.

The generated FacetResult also provides with FacetResult.getNumValidDescendants(), which returns the total number of facets that are descendants of the root node, no deeper than FacetRequest.getDepth(), and which have valid value. The rootnode itself is not counted here. Valid value is determined by the FacetResultsHandler. TopKInEachNodeHandler defines valid as != 0.

NOTE: this code relies on the assumption that TaxonomyReader.INVALID_ORDINAL == -1, a smaller value than any valid ordinal.

WARNING: This API is experimental and might change in incompatible ways in the next release.

Nested Class Summary

static class

TopKInEachNodeHandler.IntermediateFacetResultWithHash Intermediate result to hold counts from one or more partitions processed thus far.

Field Summary

Fields inherited from class org.apache.lucene.facet.search.FacetResultsHandler

facetRequest, taxonomyReader

Constructor Summary

TopKInEachNodeHandler(TaxonomyReader taxonomyReader, FacetRequest facetRequest)

Method Summary

IntermediateFacetResult	fetchPartitionResult(FacetArrays arrays, int offset) Recursively explore all facets that can be potentially included in the FacetResult to be generated, and that belong to the given partition, so that values can be examined and collected.
void	labelResult(FacetResult facetResult) Label results according to settings in FacetRequest, such as FacetRequest.getNumLabel().
IntermediateFacetResult	mergeResults(IntermediateFacetResult... tmpResults) Merge several partitions' IntermediateFacetResult-s into one of the same format
FacetResult	rearrangeFacetResult(FacetResult facetResult) Perform any rearrangement as required on a facet result that has changed after it was rendered.
FacetResult	renderFacetResult(IntermediateFacetResult tmpResult) Create a facet result from the temporary result.

Methods inherited from class org.apache.lucene.facet.search.FacetResultsHandler

getFacetRequest, getTaxonomyReader, isSelfPartition

Methods inherited from class java.lang.Object

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

Constructor Detail

TopKInEachNodeHandler

public TopKInEachNodeHandler(TaxonomyReader taxonomyReader,
                             FacetRequest facetRequest)

Method Detail

fetchPartitionResult

public IntermediateFacetResult fetchPartitionResult(FacetArrays arrays,
                                                    int offset)
                                             throws IOException

Recursively explore all facets that can be potentially included in the FacetResult to be generated, and that belong to the given partition, so that values can be examined and collected. For each such node, gather its top K (FacetRequest.getNumResults()) children among its children that are encountered in the given particular partition (aka current counting list).

Specified by:

fetchPartitionResult in class FacetResultsHandler

Parameters:

arrays - the already filled in count array, potentially only covering one partition: the ordinals ranging from

offset - to offset + the length of the count arrays within arrays (exclusive)

Returns:

IntermediateFacetResult consisting of IntToObjectMap that maps potential FacetResult nodes to their top K children encountered in the current partition. Note that the mapped potential tree nodes need not belong to the given partition, only the top K children mapped to. The aim is to identify nodes that are certainly excluded from the FacetResult to be eventually (after going through all the partitions) returned by this handler, because they have K better siblings, already identified in this partition. For the identified excluded nodes, we only count number of their descendants in the subtree (to be included in FacetResult.getNumValidDescendants()), but not bother with selecting top K in these generations, which, by definition, are, too, excluded from the FacetResult tree.

Throws:

IOException - in case TaxonomyReader.getOrdinal(org.apache.lucene.facet.taxonomy.CategoryPath) does.

See Also:

FacetResultsHandler.fetchPartitionResult(FacetArrays, int)

mergeResults

public IntermediateFacetResult mergeResults(IntermediateFacetResult... tmpResults)
                                     throws ClassCastException,
                                            IllegalArgumentException

Merge several partitions' IntermediateFacetResult-s into one of the same format

Specified by:

mergeResults in class FacetResultsHandler

Parameters:

tmpResults - one or more temporary results created by this handler.

Returns:

temporary facet result that represents to union, as specified by this handler, of the input temporary facet results.

Throws:

ClassCastException - if the temporary result passed was not created by this handler

IllegalArgumentException - if passed facetResults do not have the same FacetRequest

See Also:

FacetResultsHandler.mergeResults(IntermediateFacetResult...)

labelResult

public void labelResult(FacetResult facetResult)
                 throws IOException

Description copied from class: FacetResultsHandler

Label results according to settings in FacetRequest, such as FacetRequest.getNumLabel(). Usually invoked by FacetsAccumulator.accumulate(ScoredDocIDs)

Specified by:

labelResult in class FacetResultsHandler

Parameters:

facetResult - facet result to be labeled.

Throws:

IOException - on error

rearrangeFacetResult

public FacetResult rearrangeFacetResult(FacetResult facetResult)

Description copied from class: FacetResultsHandler

Perform any rearrangement as required on a facet result that has changed after it was rendered.

Possible use case: a sampling facets accumulator invoked another other facets accumulator on a sample set of documents, obtained rendered facet results, fixed their counts, and now it is needed to sort the results differently according to the fixed counts.

Specified by:

rearrangeFacetResult in class FacetResultsHandler

Parameters:

facetResult - result to be rearranged.

See Also:

FacetResultNode.setValue(double)

renderFacetResult

public FacetResult renderFacetResult(IntermediateFacetResult tmpResult)
                              throws IOException

Description copied from class: FacetResultsHandler

Create a facet result from the temporary result.

Specified by:

renderFacetResult in class FacetResultsHandler

Parameters:

tmpResult - temporary result to be rendered as a FacetResult

Throws:

IOException - on error.