Package org.elasticsearch.search.aggregations

Class LeafBucketCollector

java.lang.Object

org.elasticsearch.search.aggregations.LeafBucketCollector

All Implemented Interfaces:

org.apache.lucene.search.LeafCollector

Direct Known Subclasses:

LeafBucketCollectorBase, ProfilingLeafBucketCollector

public abstract class LeafBucketCollector
extends Object
implements org.apache.lucene.search.LeafCollector

Collects results for a particular segment. See the docs for collect(int, long) for how to do the collecting.

Field Summary

Fields

Modifier and Type	Field	Description
static LeafBucketCollector	NO_OP_COLLECTOR	A LeafBucketCollector that doesn't collect anything.

Constructor Summary

Constructors

Constructor	Description
LeafBucketCollector()

Method Summary

Modifier and Type	Method	Description
void	collect(int doc)
abstract void	collect(int doc, long owningBucketOrd)	Collect the given doc in the bucket owned by owningBucketOrd.
boolean	isNoop()	Does this collector collect anything? If this returns true we can safely just never call collect(int, long).
void	setScorer(org.apache.lucene.search.Scorable scorer)

Methods inherited from class java.lang.Object

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

Methods inherited from interface org.apache.lucene.search.LeafCollector

competitiveIterator

Field Details

NO_OP_COLLECTOR

public static final LeafBucketCollector NO_OP_COLLECTOR

A LeafBucketCollector that doesn't collect anything. Aggregators will return this if they've already collected their results and don't need to hook into the primary search. It is always safe to skip calling this calling setScorer(org.apache.lucene.search.Scorable) and collect(int, long) on this collector.

Constructor Details

LeafBucketCollector

public LeafBucketCollector()

Method Details

collect

public abstract void collect(int doc,
 long owningBucketOrd)
                      throws IOException

Collect the given doc in the bucket owned by owningBucketOrd.

The implementation of this method metric aggregations is generally something along the lines of

 array[owningBucketOrd] += loadValueFromDoc(doc)
 

Bucket aggregations have more trouble because their job is to make new ordinals. So their implementation generally looks kind of like

 long myBucketOrd = mapOwningBucketAndValueToMyOrd(owningBucketOrd, loadValueFromDoc(doc));
 collectBucket(doc, myBucketOrd);
 

Some bucket aggregations "know" how many ordinals each owning ordinal needs so they can map "densely". The range aggregation, for example, can perform this mapping with something like:

 return rangeCount * owningBucketOrd + matchingRange(value);
 

Other aggregations don't know how many buckets will fall into any particular owning bucket. The terms aggregation, for example, uses LongKeyedBucketOrds which amounts to a hash lookup.

Throws:

IOException

isNoop

public boolean isNoop()

Does this collector collect anything? If this returns true we can safely just never call collect(int, long).

collect

public final void collect(int doc)
                   throws IOException

Specified by:

collect in interface org.apache.lucene.search.LeafCollector

Throws:

IOException

setScorer

public void setScorer(org.apache.lucene.search.Scorable scorer)
               throws IOException

Specified by:

setScorer in interface org.apache.lucene.search.LeafCollector

Throws:

IOException