Examples with DocIdSet org.apache.lucene.search.DocIdSet used on opensource projects

Example 16 with DocIdSet

use of org.apache.lucene.search.DocIdSet in project lucene-solr by apache.

the class TestDocSet method doFilterTest.

public void doFilterTest(IndexReader reader) throws IOException {
    IndexReaderContext topLevelContext = reader.getContext();
    FixedBitSet bs = getRandomSet(reader.maxDoc(), rand.nextInt(reader.maxDoc() + 1));
    DocSet a = new BitDocSet(bs);
    DocSet b = getIntDocSet(bs);
    Filter fa = a.getTopFilter();
    Filter fb = b.getTopFilter();
    /* top level filters are no longer supported
    // test top-level
    DocIdSet da = fa.getDocIdSet(topLevelContext);
    DocIdSet db = fb.getDocIdSet(topLevelContext);
    doTestIteratorEqual(da, db);
    ***/
    DocIdSet da;
    DocIdSet db;
    List<LeafReaderContext> leaves = topLevelContext.leaves();
    // first test in-sequence sub readers
    for (LeafReaderContext readerContext : leaves) {
        da = fa.getDocIdSet(readerContext, null);
        db = fb.getDocIdSet(readerContext, null);
        doTestIteratorEqual(da, db);
    }
    int nReaders = leaves.size();
    // now test out-of-sequence sub readers
    for (int i = 0; i < nReaders; i++) {
        LeafReaderContext readerContext = leaves.get(rand.nextInt(nReaders));
        da = fa.getDocIdSet(readerContext, null);
        db = fb.getDocIdSet(readerContext, null);
        doTestIteratorEqual(da, db);
    }
}

Example 17 with DocIdSet

use of org.apache.lucene.search.DocIdSet in project greplin-lucene-utils by Cue.

the class HasFieldFilterTest method testBasics.

@Test
public void testBasics() throws Exception {
    Filter field1 = new HasFieldFilter("field1");
    Filter field2 = new HasFieldFilter("field2");
    DocIdSet hasField1 = field1.getDocIdSet(this.reader);
    Assert.assertTrue(hasField1.isCacheable());
    DocIdSet hasField2 = field2.getDocIdSet(this.reader);
    Assert.assertTrue(hasField2.isCacheable());
    assertDocIds(hasField1, true, false);
    assertDocIds(hasField2, true, true);
}

Example 18 with DocIdSet

use of org.apache.lucene.search.DocIdSet in project elasticsearch by elastic.

the class TermsSliceQuery method createWeight.

@Override
public Weight createWeight(IndexSearcher searcher, boolean needsScores) throws IOException {
    return new ConstantScoreWeight(this) {

        @Override
        public Scorer scorer(LeafReaderContext context) throws IOException {
            final DocIdSet disi = build(context.reader());
            final DocIdSetIterator leafIt = disi.iterator();
            return new ConstantScoreScorer(this, score(), leafIt);
        }
    };
}

Example 19 with DocIdSet

use of org.apache.lucene.search.DocIdSet in project lucene-solr by apache.

the class FacetFieldProcessorByArrayDV method collectDocs.

@Override
protected void collectDocs() throws IOException {
    int domainSize = fcontext.base.size();
    if (nTerms <= 0 || domainSize < effectiveMincount) {
        // TODO: what about allBuckets? missing bucket?
        return;
    }
    // TODO: refactor some of this logic into a base class
    boolean countOnly = collectAcc == null && allBucketsAcc == null;
    boolean fullRange = startTermIndex == 0 && endTermIndex == si.getValueCount();
    // Are we expecting many hits per bucket?
    // FUTURE: pro-rate for nTerms?
    // FUTURE: better take into account number of values in multi-valued fields.  This info is available for indexed fields.
    // FUTURE: take into account that bigger ord maps are more expensive than smaller ones
    // One test: 5M doc index, faceting on a single-valued field with almost 1M unique values, crossover point where global counting was slower
    // than per-segment counting was a domain of 658k docs.  At that point, top 10 buckets had 6-7 matches each.
    // this was for heap docvalues produced by UninvertingReader
    // Since these values were randomly distributed, lets round our domain multiplier up to account for less random real world data.
    long domainMultiplier = multiValuedField ? 4L : 2L;
    // +3 to increase test coverage with small tests
    boolean manyHitsPerBucket = domainSize * domainMultiplier > (si.getValueCount() + 3);
    // If we're only calculating counts, we're not prefixing, and we expect to collect many documents per unique value,
    // then collect per-segment before mapping to global ords at the end.  This will save redundant seg->global ord mappings.
    // FUTURE: there are probably some other non "countOnly" cases where we can use this as well (i.e. those where
    // the docid is not used)
    boolean canDoPerSeg = countOnly && fullRange;
    boolean accumSeg = manyHitsPerBucket && canDoPerSeg;
    // internal - override perSeg heuristic
    if (freq.perSeg != null)
        accumSeg = canDoPerSeg && freq.perSeg;
    final List<LeafReaderContext> leaves = fcontext.searcher.getIndexReader().leaves();
    Filter filter = fcontext.base.getTopFilter();
    for (int subIdx = 0; subIdx < leaves.size(); subIdx++) {
        LeafReaderContext subCtx = leaves.get(subIdx);
        setNextReaderFirstPhase(subCtx);
        // solr docsets already exclude any deleted docs
        DocIdSet dis = filter.getDocIdSet(subCtx, null);
        DocIdSetIterator disi = dis.iterator();
        SortedDocValues singleDv = null;
        SortedSetDocValues multiDv = null;
        if (multiValuedField) {
            // TODO: get sub from multi?
            multiDv = subCtx.reader().getSortedSetDocValues(sf.getName());
            if (multiDv == null) {
                multiDv = DocValues.emptySortedSet();
            }
            // this will be null if this is not a wrapped single valued docvalues.
            if (unwrap_singleValued_multiDv) {
                singleDv = DocValues.unwrapSingleton(multiDv);
            }
        } else {
            singleDv = subCtx.reader().getSortedDocValues(sf.getName());
            if (singleDv == null) {
                singleDv = DocValues.emptySorted();
            }
        }
        LongValues toGlobal = ordinalMap == null ? null : ordinalMap.getGlobalOrds(subIdx);
        if (singleDv != null) {
            if (accumSeg) {
                collectPerSeg(singleDv, disi, toGlobal);
            } else {
                if (canDoPerSeg && toGlobal != null) {
                    collectCounts(singleDv, disi, toGlobal);
                } else {
                    collectDocs(singleDv, disi, toGlobal);
                }
            }
        } else {
            if (accumSeg) {
                collectPerSeg(multiDv, disi, toGlobal);
            } else {
                if (canDoPerSeg && toGlobal != null) {
                    collectCounts(multiDv, disi, toGlobal);
                } else {
                    collectDocs(multiDv, disi, toGlobal);
                }
            }
        }
    }
    // better GC
    reuse = null;
}

Example 20 with DocIdSet

use of org.apache.lucene.search.DocIdSet in project lucene-solr by apache.

the class AbstractPrefixTreeQuery method createWeight.

@Override
public Weight createWeight(IndexSearcher searcher, boolean needsScores, float boost) throws IOException {
    return new ConstantScoreWeight(this, boost) {

        @Override
        public Scorer scorer(LeafReaderContext context) throws IOException {
            DocIdSet docSet = getDocIdSet(context);
            if (docSet == null) {
                return null;
            }
            DocIdSetIterator disi = docSet.iterator();
            if (disi == null) {
                return null;
            }
            return new ConstantScoreScorer(this, score(), disi);
        }
    };
}