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

Example 6 with TwoPhaseIterator

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

the class TestDrillSideways method testRandom.

public void testRandom() throws Exception {
    while (aChance == 0.0) {
        aChance = random().nextDouble();
    }
    while (bChance == 0.0) {
        bChance = random().nextDouble();
    }
    while (cChance == 0.0) {
        cChance = random().nextDouble();
    }
    //aChance = .01;
    //bChance = 0.5;
    //cChance = 1.0;
    double sum = aChance + bChance + cChance;
    aChance /= sum;
    bChance /= sum;
    cChance /= sum;
    int numDims = TestUtil.nextInt(random(), 2, 5);
    //int numDims = 3;
    int numDocs = atLeast(3000);
    //int numDocs = 20;
    if (VERBOSE) {
        System.out.println("numDims=" + numDims + " numDocs=" + numDocs + " aChance=" + aChance + " bChance=" + bChance + " cChance=" + cChance);
    }
    String[][] dimValues = new String[numDims][];
    int valueCount = 2;
    for (int dim = 0; dim < numDims; dim++) {
        Set<String> values = new HashSet<>();
        while (values.size() < valueCount) {
            String s = TestUtil.randomRealisticUnicodeString(random());
            //String s = _TestUtil.randomString(random());
            if (s.length() > 0) {
                values.add(s);
            }
        }
        dimValues[dim] = values.toArray(new String[values.size()]);
        valueCount *= 2;
    }
    List<Doc> docs = new ArrayList<>();
    for (int i = 0; i < numDocs; i++) {
        Doc doc = new Doc();
        doc.id = "" + i;
        doc.contentToken = randomContentToken(false);
        doc.dims = new int[numDims];
        doc.dims2 = new int[numDims];
        for (int dim = 0; dim < numDims; dim++) {
            if (random().nextInt(5) == 3) {
                // This doc is missing this dim:
                doc.dims[dim] = -1;
            } else if (dimValues[dim].length <= 4) {
                int dimUpto = 0;
                doc.dims[dim] = dimValues[dim].length - 1;
                while (dimUpto < dimValues[dim].length) {
                    if (random().nextBoolean()) {
                        doc.dims[dim] = dimUpto;
                        break;
                    }
                    dimUpto++;
                }
            } else {
                doc.dims[dim] = random().nextInt(dimValues[dim].length);
            }
            if (random().nextInt(5) == 3) {
                // 2nd value:
                doc.dims2[dim] = random().nextInt(dimValues[dim].length);
            } else {
                doc.dims2[dim] = -1;
            }
        }
        docs.add(doc);
    }
    Directory d = newDirectory();
    Directory td = newDirectory();
    IndexWriterConfig iwc = newIndexWriterConfig(new MockAnalyzer(random()));
    iwc.setInfoStream(InfoStream.NO_OUTPUT);
    RandomIndexWriter w = new RandomIndexWriter(random(), d, iwc);
    DirectoryTaxonomyWriter tw = new DirectoryTaxonomyWriter(td, IndexWriterConfig.OpenMode.CREATE);
    FacetsConfig config = new FacetsConfig();
    for (int i = 0; i < numDims; i++) {
        config.setMultiValued("dim" + i, true);
    }
    boolean doUseDV = random().nextBoolean();
    for (Doc rawDoc : docs) {
        Document doc = new Document();
        doc.add(newStringField("id", rawDoc.id, Field.Store.YES));
        doc.add(new SortedDocValuesField("id", new BytesRef(rawDoc.id)));
        doc.add(newStringField("content", rawDoc.contentToken, Field.Store.NO));
        if (VERBOSE) {
            System.out.println("  doc id=" + rawDoc.id + " token=" + rawDoc.contentToken);
        }
        for (int dim = 0; dim < numDims; dim++) {
            int dimValue = rawDoc.dims[dim];
            if (dimValue != -1) {
                if (doUseDV) {
                    doc.add(new SortedSetDocValuesFacetField("dim" + dim, dimValues[dim][dimValue]));
                } else {
                    doc.add(new FacetField("dim" + dim, dimValues[dim][dimValue]));
                }
                doc.add(new StringField("dim" + dim, dimValues[dim][dimValue], Field.Store.YES));
                if (VERBOSE) {
                    System.out.println("    dim" + dim + "=" + new BytesRef(dimValues[dim][dimValue]));
                }
            }
            int dimValue2 = rawDoc.dims2[dim];
            if (dimValue2 != -1) {
                if (doUseDV) {
                    doc.add(new SortedSetDocValuesFacetField("dim" + dim, dimValues[dim][dimValue2]));
                } else {
                    doc.add(new FacetField("dim" + dim, dimValues[dim][dimValue2]));
                }
                doc.add(new StringField("dim" + dim, dimValues[dim][dimValue2], Field.Store.YES));
                if (VERBOSE) {
                    System.out.println("      dim" + dim + "=" + new BytesRef(dimValues[dim][dimValue2]));
                }
            }
        }
        w.addDocument(config.build(tw, doc));
    }
    if (random().nextBoolean()) {
        // Randomly delete a few docs:
        int numDel = TestUtil.nextInt(random(), 1, (int) (numDocs * 0.05));
        if (VERBOSE) {
            System.out.println("delete " + numDel);
        }
        int delCount = 0;
        while (delCount < numDel) {
            Doc doc = docs.get(random().nextInt(docs.size()));
            if (!doc.deleted) {
                if (VERBOSE) {
                    System.out.println("  delete id=" + doc.id);
                }
                doc.deleted = true;
                w.deleteDocuments(new Term("id", doc.id));
                delCount++;
            }
        }
    }
    if (random().nextBoolean()) {
        if (VERBOSE) {
            System.out.println("TEST: forceMerge(1)...");
        }
        w.forceMerge(1);
    }
    IndexReader r = w.getReader();
    final SortedSetDocValuesReaderState sortedSetDVState;
    IndexSearcher s = newSearcher(r);
    if (doUseDV) {
        sortedSetDVState = new DefaultSortedSetDocValuesReaderState(s.getIndexReader());
    } else {
        sortedSetDVState = null;
    }
    if (VERBOSE) {
        System.out.println("r.numDocs() = " + r.numDocs());
    }
    // NRT open
    TaxonomyReader tr = new DirectoryTaxonomyReader(tw);
    int numIters = atLeast(10);
    for (int iter = 0; iter < numIters; iter++) {
        String contentToken = random().nextInt(30) == 17 ? null : randomContentToken(true);
        int numDrillDown = TestUtil.nextInt(random(), 1, Math.min(4, numDims));
        if (VERBOSE) {
            System.out.println("\nTEST: iter=" + iter + " baseQuery=" + contentToken + " numDrillDown=" + numDrillDown + " useSortedSetDV=" + doUseDV);
        }
        String[][] drillDowns = new String[numDims][];
        int count = 0;
        boolean anyMultiValuedDrillDowns = false;
        while (count < numDrillDown) {
            int dim = random().nextInt(numDims);
            if (drillDowns[dim] == null) {
                if (random().nextBoolean()) {
                    // Drill down on one value:
                    drillDowns[dim] = new String[] { dimValues[dim][random().nextInt(dimValues[dim].length)] };
                } else {
                    int orCount = TestUtil.nextInt(random(), 1, Math.min(5, dimValues[dim].length));
                    drillDowns[dim] = new String[orCount];
                    anyMultiValuedDrillDowns |= orCount > 1;
                    for (int i = 0; i < orCount; i++) {
                        while (true) {
                            String value = dimValues[dim][random().nextInt(dimValues[dim].length)];
                            for (int j = 0; j < i; j++) {
                                if (value.equals(drillDowns[dim][j])) {
                                    value = null;
                                    break;
                                }
                            }
                            if (value != null) {
                                drillDowns[dim][i] = value;
                                break;
                            }
                        }
                    }
                }
                if (VERBOSE) {
                    BytesRef[] values = new BytesRef[drillDowns[dim].length];
                    for (int i = 0; i < values.length; i++) {
                        values[i] = new BytesRef(drillDowns[dim][i]);
                    }
                    System.out.println("  dim" + dim + "=" + Arrays.toString(values));
                }
                count++;
            }
        }
        Query baseQuery;
        if (contentToken == null) {
            baseQuery = new MatchAllDocsQuery();
        } else {
            baseQuery = new TermQuery(new Term("content", contentToken));
        }
        DrillDownQuery ddq = new DrillDownQuery(config, baseQuery);
        for (int dim = 0; dim < numDims; dim++) {
            if (drillDowns[dim] != null) {
                for (String value : drillDowns[dim]) {
                    ddq.add("dim" + dim, value);
                }
            }
        }
        Query filter;
        if (random().nextInt(7) == 6) {
            if (VERBOSE) {
                System.out.println("  only-even filter");
            }
            filter = new Query() {

                @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 {
                            DocIdSetIterator approximation = DocIdSetIterator.all(context.reader().maxDoc());
                            return new ConstantScoreScorer(this, score(), new TwoPhaseIterator(approximation) {

                                @Override
                                public boolean matches() throws IOException {
                                    int docID = approximation.docID();
                                    return (Integer.parseInt(context.reader().document(docID).get("id")) & 1) == 0;
                                }

                                @Override
                                public float matchCost() {
                                    return 1000f;
                                }
                            });
                        }
                    };
                }

                @Override
                public String toString(String field) {
                    return "drillSidewaysTestFilter";
                }

                @Override
                public boolean equals(Object o) {
                    return o == this;
                }

                @Override
                public int hashCode() {
                    return System.identityHashCode(this);
                }
            };
        } else {
            filter = null;
        }
        // Verify docs are always collected in order.  If we
        // had an AssertingScorer it could catch it when
        // Weight.scoresDocsOutOfOrder lies!:
        getNewDrillSideways(s, config, tr).search(ddq, new SimpleCollector() {

            int lastDocID;

            @Override
            public void collect(int doc) {
                assert doc > lastDocID;
                lastDocID = doc;
            }

            @Override
            protected void doSetNextReader(LeafReaderContext context) throws IOException {
                lastDocID = -1;
            }

            @Override
            public boolean needsScores() {
                return false;
            }
        });
        // subScorers are on the same docID:
        if (!anyMultiValuedDrillDowns) {
            // Can only do this test when there are no OR'd
            // drill-down values, because in that case it's
            // easily possible for one of the DD terms to be on
            // a future docID:
            getNewDrillSidewaysScoreSubdocsAtOnce(s, config, tr).search(ddq, new AssertingSubDocsAtOnceCollector());
        }
        TestFacetResult expected = slowDrillSidewaysSearch(s, docs, contentToken, drillDowns, dimValues, filter);
        Sort sort = new Sort(new SortField("id", SortField.Type.STRING));
        DrillSideways ds;
        if (doUseDV) {
            ds = getNewDrillSideways(s, config, sortedSetDVState);
        } else {
            ds = getNewDrillSidewaysBuildFacetsResult(s, config, tr);
        }
        // Retrieve all facets:
        DrillSidewaysResult actual = ds.search(ddq, filter, null, numDocs, sort, true, true);
        TopDocs hits = s.search(baseQuery, numDocs);
        Map<String, Float> scores = new HashMap<>();
        for (ScoreDoc sd : hits.scoreDocs) {
            scores.put(s.doc(sd.doc).get("id"), sd.score);
        }
        if (VERBOSE) {
            System.out.println("  verify all facets");
        }
        verifyEquals(dimValues, s, expected, actual, scores, doUseDV);
        // Make sure drill down doesn't change score:
        Query q = ddq;
        if (filter != null) {
            q = new BooleanQuery.Builder().add(q, Occur.MUST).add(filter, Occur.FILTER).build();
        }
        TopDocs ddqHits = s.search(q, numDocs);
        assertEquals(expected.hits.size(), ddqHits.totalHits);
        for (int i = 0; i < expected.hits.size(); i++) {
            // Score should be IDENTICAL:
            assertEquals(scores.get(expected.hits.get(i).id), ddqHits.scoreDocs[i].score, 0.0f);
        }
    }
    w.close();
    IOUtils.close(r, tr, tw, d, td);
}

Example 7 with TwoPhaseIterator

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

the class PhraseHelper method getTermToSpans.

// code extracted & refactored from WSTE.extractWeightedSpanTerms()
private void getTermToSpans(SpanQuery spanQuery, LeafReaderContext readerContext, int doc, Map<BytesRef, Spans> result) throws IOException {
    // note: in WSTE there was some field specific looping that seemed pointless so that isn't here.
    final IndexSearcher searcher = new IndexSearcher(readerContext.reader());
    searcher.setQueryCache(null);
    if (willRewrite) {
        // searcher.rewrite loops till done
        spanQuery = (SpanQuery) searcher.rewrite(spanQuery);
    }
    // Get the underlying query terms
    // sorted so we can loop over results in order shortly...
    TreeSet<Term> termSet = new FieldFilteringTermSet();
    //needsScores==false
    searcher.createWeight(spanQuery, false, 1.0f).extractTerms(termSet);
    // Get Spans by running the query against the reader
    // TODO it might make sense to re-use/cache the Spans instance, to advance forward between docs
    SpanWeight spanWeight = (SpanWeight) searcher.createNormalizedWeight(spanQuery, false);
    Spans spans = spanWeight.getSpans(readerContext, SpanWeight.Postings.POSITIONS);
    if (spans == null) {
        return;
    }
    TwoPhaseIterator twoPhaseIterator = spans.asTwoPhaseIterator();
    if (twoPhaseIterator != null) {
        if (twoPhaseIterator.approximation().advance(doc) != doc || !twoPhaseIterator.matches()) {
            return;
        }
    } else if (spans.advance(doc) != doc) {
        // preposition, and return doing nothing if find none
        return;
    }
    // Consume the Spans into a cache.  This instance is used as a source for multiple cloned copies.
    // It's important we do this and not re-use the same original Spans instance since these will be iterated
    // independently later on; sometimes in ways that prevents sharing the original Spans.
    // consumes spans for this doc only and caches
    CachedSpans cachedSpansSource = new CachedSpans(spans);
    // we don't use it below
    spans = null;
    // Map terms to a Spans instance (aggregate if necessary)
    for (final Term queryTerm : termSet) {
        //   spanQuery list were already filtered by these conditions.
        if (positionInsensitiveTerms.contains(queryTerm)) {
            continue;
        }
        // copy-constructor refers to same data (shallow) but has iteration state from the beginning
        CachedSpans cachedSpans = new CachedSpans(cachedSpansSource);
        // Add the span to whatever span may or may not exist
        Spans existingSpans = result.get(queryTerm.bytes());
        if (existingSpans != null) {
            if (existingSpans instanceof MultiSpans) {
                ((MultiSpans) existingSpans).addSpans(cachedSpans);
            } else {
                // upgrade to MultiSpans
                MultiSpans multiSpans = new MultiSpans();
                multiSpans.addSpans(existingSpans);
                multiSpans.addSpans(cachedSpans);
                result.put(queryTerm.bytes(), multiSpans);
            }
        } else {
            result.put(queryTerm.bytes(), cachedSpans);
        }
    }
}

Example 8 with TwoPhaseIterator

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

the class IntersectsRPTVerifyQuery method createWeight.

@Override
public Weight createWeight(IndexSearcher searcher, boolean needsScores, float boost) throws IOException {
    final Map valueSourceContext = ValueSource.newContext(searcher);
    return new ConstantScoreWeight(this, boost) {

        @Override
        public Scorer scorer(LeafReaderContext context) throws IOException {
            // Compute approx & exact
            final IntersectsDifferentiatingQuery.IntersectsDifferentiatingVisitor result = intersectsDiffQuery.compute(context);
            if (result.approxDocIdSet == null) {
                return null;
            }
            final DocIdSetIterator approxDISI = result.approxDocIdSet.iterator();
            if (approxDISI == null) {
                return null;
            }
            final DocIdSetIterator exactIterator;
            if (result.exactDocIdSet != null) {
                // If both sets are the same, there's nothing to verify; we needn't return a TwoPhaseIterator
                if (result.approxDocIdSet == result.exactDocIdSet) {
                    return new ConstantScoreScorer(this, score(), approxDISI);
                }
                exactIterator = result.exactDocIdSet.iterator();
                assert exactIterator != null;
            } else {
                exactIterator = null;
            }
            final FunctionValues predFuncValues = predicateValueSource.getValues(valueSourceContext, context);
            final TwoPhaseIterator twoPhaseIterator = new TwoPhaseIterator(approxDISI) {

                @Override
                public boolean matches() throws IOException {
                    final int doc = approxDISI.docID();
                    if (exactIterator != null) {
                        if (exactIterator.docID() < doc) {
                            exactIterator.advance(doc);
                        }
                        if (exactIterator.docID() == doc) {
                            return true;
                        }
                    }
                    return predFuncValues.boolVal(doc);
                }

                @Override
                public float matchCost() {
                    // TODO: use cost of exactIterator.advance() and predFuncValues.boolVal()
                    return 100;
                }
            };
            return new ConstantScoreScorer(this, score(), twoPhaseIterator);
        }
    };
}

Example 9 with TwoPhaseIterator

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

the class SortedNumericDocValuesRangeQuery 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 {
            SortedNumericDocValues values = getValues(context.reader(), field);
            if (values == null) {
                return null;
            }
            final NumericDocValues singleton = DocValues.unwrapSingleton(values);
            final TwoPhaseIterator iterator;
            if (singleton != null) {
                iterator = new TwoPhaseIterator(singleton) {

                    @Override
                    public boolean matches() throws IOException {
                        final long value = singleton.longValue();
                        return value >= lowerValue && value <= upperValue;
                    }

                    @Override
                    public float matchCost() {
                        // 2 comparisons
                        return 2;
                    }
                };
            } else {
                iterator = new TwoPhaseIterator(values) {

                    @Override
                    public boolean matches() throws IOException {
                        for (int i = 0, count = values.docValueCount(); i < count; ++i) {
                            final long value = values.nextValue();
                            if (value < lowerValue) {
                                continue;
                            }
                            // Values are sorted, so the first value that is >= lowerValue is our best candidate
                            return value <= upperValue;
                        }
                        // all values were < lowerValue
                        return false;
                    }

                    @Override
                    public float matchCost() {
                        // 2 comparisons
                        return 2;
                    }
                };
            }
            return new ConstantScoreScorer(this, score(), iterator);
        }
    };
}

Example 10 with TwoPhaseIterator

use of org.apache.lucene.search.TwoPhaseIterator in project pyramid by cheng-li.

the class CustomConjunctionSpans method asTwoPhaseIterator.

/**
 * Return a {@link TwoPhaseIterator} view of this ConjunctionSpans.
 */
@Override
public TwoPhaseIterator asTwoPhaseIterator() {
    float totalMatchCost = 0;
    // Compute the matchCost as the total matchCost/positionsCostant of the sub spans.
    for (Spans spans : subSpans) {
        TwoPhaseIterator tpi = spans.asTwoPhaseIterator();
        if (tpi != null) {
            totalMatchCost += tpi.matchCost();
        } else {
            totalMatchCost += spans.positionsCost();
        }
    }
    final float matchCost = totalMatchCost;
    return new TwoPhaseIterator(conjunction) {

        @Override
        public boolean matches() throws IOException {
            return twoPhaseCurrentDocMatches();
        }

        @Override
        public float matchCost() {
            return matchCost;
        }
    };
}