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

Example 41 with ScoreDoc

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

the class BaseGeoPointTestCase method doRandomDistanceTest.

private void doRandomDistanceTest(int numDocs, int numQueries) throws IOException {
    Directory dir = newDirectory();
    IndexWriterConfig iwc = newIndexWriterConfig();
    // Else seeds may not reproduce:
    iwc.setMergeScheduler(new SerialMergeScheduler());
    int pointsInLeaf = 2 + random().nextInt(4);
    iwc.setCodec(new FilterCodec("Lucene70", TestUtil.getDefaultCodec()) {

        @Override
        public PointsFormat pointsFormat() {
            return new PointsFormat() {

                @Override
                public PointsWriter fieldsWriter(SegmentWriteState writeState) throws IOException {
                    return new Lucene60PointsWriter(writeState, pointsInLeaf, BKDWriter.DEFAULT_MAX_MB_SORT_IN_HEAP);
                }

                @Override
                public PointsReader fieldsReader(SegmentReadState readState) throws IOException {
                    return new Lucene60PointsReader(readState);
                }
            };
        }
    });
    RandomIndexWriter writer = new RandomIndexWriter(random(), dir, iwc);
    for (int i = 0; i < numDocs; i++) {
        double latRaw = nextLatitude();
        double lonRaw = nextLongitude();
        // pre-normalize up front, so we can just use quantized value for testing and do simple exact comparisons
        double lat = quantizeLat(latRaw);
        double lon = quantizeLon(lonRaw);
        Document doc = new Document();
        addPointToDoc("field", doc, lat, lon);
        doc.add(new StoredField("lat", lat));
        doc.add(new StoredField("lon", lon));
        writer.addDocument(doc);
    }
    IndexReader reader = writer.getReader();
    IndexSearcher searcher = newSearcher(reader);
    for (int i = 0; i < numQueries; i++) {
        double lat = nextLatitude();
        double lon = nextLongitude();
        double radius = 50000000D * random().nextDouble();
        BitSet expected = new BitSet();
        for (int doc = 0; doc < reader.maxDoc(); doc++) {
            double docLatitude = reader.document(doc).getField("lat").numericValue().doubleValue();
            double docLongitude = reader.document(doc).getField("lon").numericValue().doubleValue();
            double distance = SloppyMath.haversinMeters(lat, lon, docLatitude, docLongitude);
            if (distance <= radius) {
                expected.set(doc);
            }
        }
        TopDocs topDocs = searcher.search(newDistanceQuery("field", lat, lon, radius), reader.maxDoc(), Sort.INDEXORDER);
        BitSet actual = new BitSet();
        for (ScoreDoc doc : topDocs.scoreDocs) {
            actual.set(doc.doc);
        }
        try {
            assertEquals(expected, actual);
        } catch (AssertionError e) {
            System.out.println("center: (" + lat + "," + lon + "), radius=" + radius);
            for (int doc = 0; doc < reader.maxDoc(); doc++) {
                double docLatitude = reader.document(doc).getField("lat").numericValue().doubleValue();
                double docLongitude = reader.document(doc).getField("lon").numericValue().doubleValue();
                double distance = SloppyMath.haversinMeters(lat, lon, docLatitude, docLongitude);
                System.out.println("" + doc + ": (" + docLatitude + "," + docLongitude + "), distance=" + distance);
            }
            throw e;
        }
    }
    reader.close();
    writer.close();
    dir.close();
}

Example 42 with ScoreDoc

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

the class LTRRescorer method heapAdjust.

private void heapAdjust(ScoreDoc[] hits, int size, int root) {
    final ScoreDoc doc = hits[root];
    final float score = doc.score;
    int i = root;
    while (i <= ((size >> 1) - 1)) {
        final int lchild = (i << 1) + 1;
        final ScoreDoc ldoc = hits[lchild];
        final float lscore = ldoc.score;
        float rscore = Float.MAX_VALUE;
        final int rchild = (i << 1) + 2;
        ScoreDoc rdoc = null;
        if (rchild < size) {
            rdoc = hits[rchild];
            rscore = rdoc.score;
        }
        if (lscore < score) {
            if (rscore < lscore) {
                hits[i] = rdoc;
                hits[rchild] = doc;
                i = rchild;
            } else {
                hits[i] = ldoc;
                hits[lchild] = doc;
                i = lchild;
            }
        } else if (rscore < score) {
            hits[i] = rdoc;
            hits[rchild] = doc;
            i = rchild;
        } else {
            return;
        }
    }
}

Example 43 with ScoreDoc

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

the class TestLTRReRankingPipeline method testDifferentTopN.

@Ignore
@Test
public void testDifferentTopN() throws IOException {
    final Directory dir = newDirectory();
    final RandomIndexWriter w = new RandomIndexWriter(random(), dir);
    Document doc = new Document();
    doc.add(newStringField("id", "0", Field.Store.YES));
    doc.add(newTextField("field", "wizard oz oz oz oz oz", Field.Store.NO));
    doc.add(new FloatDocValuesField("final-score", 1.0f));
    w.addDocument(doc);
    doc = new Document();
    doc.add(newStringField("id", "1", Field.Store.YES));
    doc.add(newTextField("field", "wizard oz oz oz oz the", Field.Store.NO));
    doc.add(new FloatDocValuesField("final-score", 2.0f));
    w.addDocument(doc);
    doc = new Document();
    doc.add(newStringField("id", "2", Field.Store.YES));
    doc.add(newTextField("field", "wizard oz oz oz the the ", Field.Store.NO));
    doc.add(new FloatDocValuesField("final-score", 3.0f));
    w.addDocument(doc);
    doc = new Document();
    doc.add(newStringField("id", "3", Field.Store.YES));
    doc.add(newTextField("field", "wizard oz oz the the the the ", Field.Store.NO));
    doc.add(new FloatDocValuesField("final-score", 4.0f));
    w.addDocument(doc);
    doc = new Document();
    doc.add(newStringField("id", "4", Field.Store.YES));
    doc.add(newTextField("field", "wizard oz the the the the the the", Field.Store.NO));
    doc.add(new FloatDocValuesField("final-score", 5.0f));
    w.addDocument(doc);
    final IndexReader r = w.getReader();
    w.close();
    // Do ordinary BooleanQuery:
    final BooleanQuery.Builder bqBuilder = new BooleanQuery.Builder();
    bqBuilder.add(new TermQuery(new Term("field", "wizard")), BooleanClause.Occur.SHOULD);
    bqBuilder.add(new TermQuery(new Term("field", "oz")), BooleanClause.Occur.SHOULD);
    final IndexSearcher searcher = getSearcher(r);
    // first run the standard query
    TopDocs hits = searcher.search(bqBuilder.build(), 10);
    assertEquals(5, hits.totalHits);
    assertEquals("0", searcher.doc(hits.scoreDocs[0].doc).get("id"));
    assertEquals("1", searcher.doc(hits.scoreDocs[1].doc).get("id"));
    assertEquals("2", searcher.doc(hits.scoreDocs[2].doc).get("id"));
    assertEquals("3", searcher.doc(hits.scoreDocs[3].doc).get("id"));
    assertEquals("4", searcher.doc(hits.scoreDocs[4].doc).get("id"));
    final List<Feature> features = makeFieldValueFeatures(new int[] { 0, 1, 2 }, "final-score");
    final List<Normalizer> norms = new ArrayList<Normalizer>(Collections.nCopies(features.size(), IdentityNormalizer.INSTANCE));
    final List<Feature> allFeatures = makeFieldValueFeatures(new int[] { 0, 1, 2, 3, 4, 5, 6, 7, 8, 9 }, "final-score");
    final LTRScoringModel ltrScoringModel = TestLinearModel.createLinearModel("test", features, norms, "test", allFeatures, null);
    final LTRRescorer rescorer = new LTRRescorer(new LTRScoringQuery(ltrScoringModel));
    // rerank @ 0 should not change the order
    hits = rescorer.rescore(searcher, hits, 0);
    assertEquals("0", searcher.doc(hits.scoreDocs[0].doc).get("id"));
    assertEquals("1", searcher.doc(hits.scoreDocs[1].doc).get("id"));
    assertEquals("2", searcher.doc(hits.scoreDocs[2].doc).get("id"));
    assertEquals("3", searcher.doc(hits.scoreDocs[3].doc).get("id"));
    assertEquals("4", searcher.doc(hits.scoreDocs[4].doc).get("id"));
    for (int topN = 1; topN <= 5; topN++) {
        log.info("rerank {} documents ", topN);
        hits = searcher.search(bqBuilder.build(), 10);
        final ScoreDoc[] slice = new ScoreDoc[topN];
        System.arraycopy(hits.scoreDocs, 0, slice, 0, topN);
        hits = new TopDocs(hits.totalHits, slice, hits.getMaxScore());
        hits = rescorer.rescore(searcher, hits, topN);
        for (int i = topN - 1, j = 0; i >= 0; i--, j++) {
            log.info("doc {} in pos {}", searcher.doc(hits.scoreDocs[j].doc).get("id"), j);
            assertEquals(i, Integer.parseInt(searcher.doc(hits.scoreDocs[j].doc).get("id")));
            assertEquals(i + 1, hits.scoreDocs[j].score, 0.00001);
        }
    }
    r.close();
    dir.close();
}

Example 44 with ScoreDoc

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

the class StoredFieldsShardRequestFactory method mapShardToDocs.

private void mapShardToDocs(HashMap<String, Set<ShardDoc>> shardMap, ScoreDoc[] scoreDocs) {
    for (ScoreDoc scoreDoc : scoreDocs) {
        ShardDoc solrDoc = (ShardDoc) scoreDoc;
        Set<ShardDoc> shardDocs = shardMap.get(solrDoc.shard);
        if (shardDocs == null) {
            shardMap.put(solrDoc.shard, shardDocs = new HashSet<>());
        }
        shardDocs.add(solrDoc);
    }
}

Example 45 with ScoreDoc

use of org.apache.lucene.search.ScoreDoc 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);
}