Search in sources :

Example 1 with FieldDoc

use of org.apache.lucene.search.FieldDoc in project crate by crate.

the class ScoreDocRowFunction method apply.

@Nullable
@Override
public Row apply(@Nullable ScoreDoc input) {
    if (input == null) {
        return null;
    }
    FieldDoc fieldDoc = (FieldDoc) input;
    scorer.score(fieldDoc.score);
    for (OrderByCollectorExpression orderByCollectorExpression : orderByCollectorExpressions) {
        orderByCollectorExpression.setNextFieldDoc(fieldDoc);
    }
    List<LeafReaderContext> leaves = indexReader.leaves();
    int readerIndex = ReaderUtil.subIndex(fieldDoc.doc, leaves);
    LeafReaderContext subReaderContext = leaves.get(readerIndex);
    int subDoc = fieldDoc.doc - subReaderContext.docBase;
    for (LuceneCollectorExpression<?> expression : expressions) {
        try {
            expression.setNextReader(subReaderContext);
        } catch (IOException e) {
            throw Throwables.propagate(e);
        }
        expression.setNextDocId(subDoc);
    }
    return inputRow;
}
Also used : FieldDoc(org.apache.lucene.search.FieldDoc) OrderByCollectorExpression(io.crate.operation.reference.doc.lucene.OrderByCollectorExpression) LeafReaderContext(org.apache.lucene.index.LeafReaderContext) IOException(java.io.IOException) Nullable(javax.annotation.Nullable)

Example 2 with FieldDoc

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

the class AbstractFieldDataImplTestCase method testSortMultiValuesFields.

public void testSortMultiValuesFields() throws Exception {
    fillExtendedMvSet();
    IndexFieldData indexFieldData = getForField("value");
    IndexSearcher searcher = new IndexSearcher(DirectoryReader.open(writer));
    TopFieldDocs topDocs = searcher.search(new MatchAllDocsQuery(), 10, new Sort(new SortField("value", indexFieldData.comparatorSource(null, MultiValueMode.MIN, null))));
    assertThat(topDocs.totalHits, equalTo(8));
    assertThat(topDocs.scoreDocs.length, equalTo(8));
    assertThat(topDocs.scoreDocs[0].doc, equalTo(7));
    assertThat(((BytesRef) ((FieldDoc) topDocs.scoreDocs[0]).fields[0]).utf8ToString(), equalTo("!08"));
    assertThat(topDocs.scoreDocs[1].doc, equalTo(0));
    assertThat(((BytesRef) ((FieldDoc) topDocs.scoreDocs[1]).fields[0]).utf8ToString(), equalTo("02"));
    assertThat(topDocs.scoreDocs[2].doc, equalTo(2));
    assertThat(((BytesRef) ((FieldDoc) topDocs.scoreDocs[2]).fields[0]).utf8ToString(), equalTo("03"));
    assertThat(topDocs.scoreDocs[3].doc, equalTo(3));
    assertThat(((BytesRef) ((FieldDoc) topDocs.scoreDocs[3]).fields[0]).utf8ToString(), equalTo("04"));
    assertThat(topDocs.scoreDocs[4].doc, equalTo(4));
    assertThat(((BytesRef) ((FieldDoc) topDocs.scoreDocs[4]).fields[0]).utf8ToString(), equalTo("06"));
    assertThat(topDocs.scoreDocs[5].doc, equalTo(6));
    assertThat(((BytesRef) ((FieldDoc) topDocs.scoreDocs[5]).fields[0]).utf8ToString(), equalTo("08"));
    assertThat(topDocs.scoreDocs[6].doc, equalTo(1));
    assertThat((BytesRef) ((FieldDoc) topDocs.scoreDocs[6]).fields[0], equalTo(null));
    assertThat(topDocs.scoreDocs[7].doc, equalTo(5));
    assertThat((BytesRef) ((FieldDoc) topDocs.scoreDocs[7]).fields[0], equalTo(null));
    topDocs = searcher.search(new MatchAllDocsQuery(), 10, new Sort(new SortField("value", indexFieldData.comparatorSource(null, MultiValueMode.MAX, null), true)));
    assertThat(topDocs.totalHits, equalTo(8));
    assertThat(topDocs.scoreDocs.length, equalTo(8));
    assertThat(topDocs.scoreDocs[0].doc, equalTo(6));
    assertThat(((BytesRef) ((FieldDoc) topDocs.scoreDocs[0]).fields[0]).utf8ToString(), equalTo("10"));
    assertThat(topDocs.scoreDocs[1].doc, equalTo(4));
    assertThat(((BytesRef) ((FieldDoc) topDocs.scoreDocs[1]).fields[0]).utf8ToString(), equalTo("08"));
    assertThat(topDocs.scoreDocs[2].doc, equalTo(3));
    assertThat(((BytesRef) ((FieldDoc) topDocs.scoreDocs[2]).fields[0]).utf8ToString(), equalTo("06"));
    assertThat(topDocs.scoreDocs[3].doc, equalTo(0));
    assertThat(((BytesRef) ((FieldDoc) topDocs.scoreDocs[3]).fields[0]).utf8ToString(), equalTo("04"));
    assertThat(topDocs.scoreDocs[4].doc, equalTo(2));
    assertThat(((BytesRef) ((FieldDoc) topDocs.scoreDocs[4]).fields[0]).utf8ToString(), equalTo("03"));
    assertThat(topDocs.scoreDocs[5].doc, equalTo(7));
    assertThat(((BytesRef) ((FieldDoc) topDocs.scoreDocs[5]).fields[0]).utf8ToString(), equalTo("!10"));
    assertThat(topDocs.scoreDocs[6].doc, equalTo(1));
    assertThat(((FieldDoc) topDocs.scoreDocs[6]).fields[0], equalTo(null));
    assertThat(topDocs.scoreDocs[7].doc, equalTo(5));
    assertThat(((FieldDoc) topDocs.scoreDocs[7]).fields[0], equalTo(null));
}
Also used : IndexSearcher(org.apache.lucene.search.IndexSearcher) FieldDoc(org.apache.lucene.search.FieldDoc) TopFieldDocs(org.apache.lucene.search.TopFieldDocs) Sort(org.apache.lucene.search.Sort) SortField(org.apache.lucene.search.SortField) MatchAllDocsQuery(org.apache.lucene.search.MatchAllDocsQuery)

Example 3 with FieldDoc

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

the class NestedSortingTests method testDuel.

public void testDuel() throws Exception {
    final int numDocs = scaledRandomIntBetween(100, 1000);
    for (int i = 0; i < numDocs; ++i) {
        final int numChildren = randomInt(2);
        List<Document> docs = new ArrayList<>(numChildren + 1);
        for (int j = 0; j < numChildren; ++j) {
            Document doc = new Document();
            doc.add(new StringField("f", TestUtil.randomSimpleString(random(), 2), Field.Store.NO));
            doc.add(new StringField("__type", "child", Field.Store.NO));
            docs.add(doc);
        }
        if (randomBoolean()) {
            docs.add(new Document());
        }
        Document parent = new Document();
        parent.add(new StringField("__type", "parent", Field.Store.NO));
        docs.add(parent);
        writer.addDocuments(docs);
        if (rarely()) {
            // we need to have a bit more segments than what RandomIndexWriter would do by default
            DirectoryReader.open(writer).close();
        }
    }
    writer.commit();
    MultiValueMode sortMode = randomFrom(Arrays.asList(MultiValueMode.MIN, MultiValueMode.MAX));
    DirectoryReader reader = DirectoryReader.open(writer);
    reader = ElasticsearchDirectoryReader.wrap(reader, new ShardId(indexService.index(), 0));
    IndexSearcher searcher = new IndexSearcher(reader);
    PagedBytesIndexFieldData indexFieldData1 = getForField("f");
    IndexFieldData<?> indexFieldData2 = NoOrdinalsStringFieldDataTests.hideOrdinals(indexFieldData1);
    final String missingValue = randomBoolean() ? null : TestUtil.randomSimpleString(random(), 2);
    final int n = randomIntBetween(1, numDocs + 2);
    final boolean reverse = randomBoolean();
    final TopDocs topDocs1 = getTopDocs(searcher, indexFieldData1, missingValue, sortMode, n, reverse);
    final TopDocs topDocs2 = getTopDocs(searcher, indexFieldData2, missingValue, sortMode, n, reverse);
    for (int i = 0; i < topDocs1.scoreDocs.length; ++i) {
        final FieldDoc fieldDoc1 = (FieldDoc) topDocs1.scoreDocs[i];
        final FieldDoc fieldDoc2 = (FieldDoc) topDocs2.scoreDocs[i];
        assertEquals(fieldDoc1.doc, fieldDoc2.doc);
        assertArrayEquals(fieldDoc1.fields, fieldDoc2.fields);
    }
    searcher.getIndexReader().close();
}
Also used : IndexSearcher(org.apache.lucene.search.IndexSearcher) PagedBytesIndexFieldData(org.elasticsearch.index.fielddata.plain.PagedBytesIndexFieldData) FieldDoc(org.apache.lucene.search.FieldDoc) ElasticsearchDirectoryReader(org.elasticsearch.common.lucene.index.ElasticsearchDirectoryReader) DirectoryReader(org.apache.lucene.index.DirectoryReader) ArrayList(java.util.ArrayList) Document(org.apache.lucene.document.Document) MultiValueMode(org.elasticsearch.search.MultiValueMode) ShardId(org.elasticsearch.index.shard.ShardId) TopDocs(org.apache.lucene.search.TopDocs) StringField(org.apache.lucene.document.StringField)

Example 4 with FieldDoc

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

the class SearchPhaseController method getHits.

private SearchHits getHits(ReducedQueryPhase reducedQueryPhase, boolean ignoreFrom, ScoreDoc[] sortedDocs, AtomicArray<? extends QuerySearchResultProvider> fetchResultsArr) {
    List<? extends AtomicArray.Entry<? extends QuerySearchResultProvider>> fetchResults = fetchResultsArr.asList();
    boolean sorted = false;
    int sortScoreIndex = -1;
    if (reducedQueryPhase.oneResult.topDocs() instanceof TopFieldDocs) {
        TopFieldDocs fieldDocs = (TopFieldDocs) reducedQueryPhase.oneResult.queryResult().topDocs();
        if (fieldDocs instanceof CollapseTopFieldDocs && fieldDocs.fields.length == 1 && fieldDocs.fields[0].getType() == SortField.Type.SCORE) {
            sorted = false;
        } else {
            sorted = true;
            for (int i = 0; i < fieldDocs.fields.length; i++) {
                if (fieldDocs.fields[i].getType() == SortField.Type.SCORE) {
                    sortScoreIndex = i;
                }
            }
        }
    }
    // clean the fetch counter
    for (AtomicArray.Entry<? extends QuerySearchResultProvider> entry : fetchResults) {
        entry.value.fetchResult().initCounter();
    }
    int from = ignoreFrom ? 0 : reducedQueryPhase.oneResult.queryResult().from();
    int numSearchHits = (int) Math.min(reducedQueryPhase.fetchHits - from, reducedQueryPhase.oneResult.size());
    // with collapsing we can have more fetch hits than sorted docs
    numSearchHits = Math.min(sortedDocs.length, numSearchHits);
    // merge hits
    List<SearchHit> hits = new ArrayList<>();
    if (!fetchResults.isEmpty()) {
        for (int i = 0; i < numSearchHits; i++) {
            ScoreDoc shardDoc = sortedDocs[i];
            QuerySearchResultProvider fetchResultProvider = fetchResultsArr.get(shardDoc.shardIndex);
            if (fetchResultProvider == null) {
                continue;
            }
            FetchSearchResult fetchResult = fetchResultProvider.fetchResult();
            int index = fetchResult.counterGetAndIncrement();
            if (index < fetchResult.hits().internalHits().length) {
                SearchHit searchHit = fetchResult.hits().internalHits()[index];
                searchHit.score(shardDoc.score);
                searchHit.shard(fetchResult.shardTarget());
                if (sorted) {
                    FieldDoc fieldDoc = (FieldDoc) shardDoc;
                    searchHit.sortValues(fieldDoc.fields, reducedQueryPhase.oneResult.sortValueFormats());
                    if (sortScoreIndex != -1) {
                        searchHit.score(((Number) fieldDoc.fields[sortScoreIndex]).floatValue());
                    }
                }
                hits.add(searchHit);
            }
        }
    }
    return new SearchHits(hits.toArray(new SearchHit[hits.size()]), reducedQueryPhase.totalHits, reducedQueryPhase.maxScore);
}
Also used : AtomicArray(org.elasticsearch.common.util.concurrent.AtomicArray) QuerySearchResultProvider(org.elasticsearch.search.query.QuerySearchResultProvider) SearchHit(org.elasticsearch.search.SearchHit) FieldDoc(org.apache.lucene.search.FieldDoc) FetchSearchResult(org.elasticsearch.search.fetch.FetchSearchResult) ArrayList(java.util.ArrayList) IntArrayList(com.carrotsearch.hppc.IntArrayList) CollapseTopFieldDocs(org.apache.lucene.search.grouping.CollapseTopFieldDocs) TopFieldDocs(org.apache.lucene.search.TopFieldDocs) ScoreDoc(org.apache.lucene.search.ScoreDoc) SearchHits(org.elasticsearch.search.SearchHits) CollapseTopFieldDocs(org.apache.lucene.search.grouping.CollapseTopFieldDocs)

Example 5 with FieldDoc

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

the class CollapsingTopDocsCollectorTests method assertSearchCollapse.

private <T extends Comparable> void assertSearchCollapse(CollapsingDocValuesProducer<T> dvProducers, boolean numeric, boolean multivalued) throws IOException {
    final int numDocs = randomIntBetween(1000, 2000);
    int maxGroup = randomIntBetween(2, 500);
    final Directory dir = newDirectory();
    final RandomIndexWriter w = new RandomIndexWriter(random(), dir);
    Set<T> values = new HashSet<>();
    int totalHits = 0;
    for (int i = 0; i < numDocs; i++) {
        final T value = dvProducers.randomGroup(maxGroup);
        values.add(value);
        Document doc = new Document();
        dvProducers.add(doc, value, multivalued);
        doc.add(new NumericDocValuesField("sort1", randomIntBetween(0, 10)));
        doc.add(new NumericDocValuesField("sort2", randomLong()));
        w.addDocument(doc);
        totalHits++;
    }
    List<T> valueList = new ArrayList<>(values);
    Collections.sort(valueList);
    final IndexReader reader = w.getReader();
    final IndexSearcher searcher = newSearcher(reader);
    final SortField collapseField = dvProducers.sortField(multivalued);
    final SortField sort1 = new SortField("sort1", SortField.Type.INT);
    final SortField sort2 = new SortField("sort2", SortField.Type.LONG);
    Sort sort = new Sort(sort1, sort2, collapseField);
    int expectedNumGroups = values.size();
    final CollapsingTopDocsCollector collapsingCollector;
    if (numeric) {
        collapsingCollector = CollapsingTopDocsCollector.createNumeric(collapseField.getField(), sort, expectedNumGroups, false);
    } else {
        collapsingCollector = CollapsingTopDocsCollector.createKeyword(collapseField.getField(), sort, expectedNumGroups, false);
    }
    TopFieldCollector topFieldCollector = TopFieldCollector.create(sort, totalHits, true, false, false);
    searcher.search(new MatchAllDocsQuery(), collapsingCollector);
    searcher.search(new MatchAllDocsQuery(), topFieldCollector);
    CollapseTopFieldDocs collapseTopFieldDocs = collapsingCollector.getTopDocs();
    TopFieldDocs topDocs = topFieldCollector.topDocs();
    assertEquals(collapseField.getField(), collapseTopFieldDocs.field);
    assertEquals(expectedNumGroups, collapseTopFieldDocs.scoreDocs.length);
    assertEquals(totalHits, collapseTopFieldDocs.totalHits);
    assertEquals(totalHits, topDocs.scoreDocs.length);
    assertEquals(totalHits, topDocs.totalHits);
    Set<Object> seen = new HashSet<>();
    // collapse field is the last sort
    int collapseIndex = sort.getSort().length - 1;
    int topDocsIndex = 0;
    for (int i = 0; i < expectedNumGroups; i++) {
        FieldDoc fieldDoc = null;
        for (; topDocsIndex < totalHits; topDocsIndex++) {
            fieldDoc = (FieldDoc) topDocs.scoreDocs[topDocsIndex];
            if (seen.contains(fieldDoc.fields[collapseIndex]) == false) {
                break;
            }
        }
        FieldDoc collapseFieldDoc = (FieldDoc) collapseTopFieldDocs.scoreDocs[i];
        assertNotNull(fieldDoc);
        assertEquals(collapseFieldDoc.doc, fieldDoc.doc);
        assertArrayEquals(collapseFieldDoc.fields, fieldDoc.fields);
        seen.add(fieldDoc.fields[fieldDoc.fields.length - 1]);
    }
    for (; topDocsIndex < totalHits; topDocsIndex++) {
        FieldDoc fieldDoc = (FieldDoc) topDocs.scoreDocs[topDocsIndex];
        assertTrue(seen.contains(fieldDoc.fields[collapseIndex]));
    }
    // check merge
    final IndexReaderContext ctx = searcher.getTopReaderContext();
    final SegmentSearcher[] subSearchers;
    final int[] docStarts;
    if (ctx instanceof LeafReaderContext) {
        subSearchers = new SegmentSearcher[1];
        docStarts = new int[1];
        subSearchers[0] = new SegmentSearcher((LeafReaderContext) ctx, ctx);
        docStarts[0] = 0;
    } else {
        final CompositeReaderContext compCTX = (CompositeReaderContext) ctx;
        final int size = compCTX.leaves().size();
        subSearchers = new SegmentSearcher[size];
        docStarts = new int[size];
        int docBase = 0;
        for (int searcherIDX = 0; searcherIDX < subSearchers.length; searcherIDX++) {
            final LeafReaderContext leave = compCTX.leaves().get(searcherIDX);
            subSearchers[searcherIDX] = new SegmentSearcher(leave, compCTX);
            docStarts[searcherIDX] = docBase;
            docBase += leave.reader().maxDoc();
        }
    }
    final CollapseTopFieldDocs[] shardHits = new CollapseTopFieldDocs[subSearchers.length];
    final Weight weight = searcher.createNormalizedWeight(new MatchAllDocsQuery(), false);
    for (int shardIDX = 0; shardIDX < subSearchers.length; shardIDX++) {
        final SegmentSearcher subSearcher = subSearchers[shardIDX];
        final CollapsingTopDocsCollector c;
        if (numeric) {
            c = CollapsingTopDocsCollector.createNumeric(collapseField.getField(), sort, expectedNumGroups, false);
        } else {
            c = CollapsingTopDocsCollector.createKeyword(collapseField.getField(), sort, expectedNumGroups, false);
        }
        subSearcher.search(weight, c);
        shardHits[shardIDX] = c.getTopDocs();
    }
    CollapseTopFieldDocs mergedFieldDocs = CollapseTopFieldDocs.merge(sort, 0, expectedNumGroups, shardHits);
    assertTopDocsEquals(mergedFieldDocs, collapseTopFieldDocs);
    w.close();
    reader.close();
    dir.close();
}
Also used : IndexSearcher(org.apache.lucene.search.IndexSearcher) FieldDoc(org.apache.lucene.search.FieldDoc) ArrayList(java.util.ArrayList) CollapseTopFieldDocs(org.apache.lucene.search.grouping.CollapseTopFieldDocs) TopFieldDocs(org.apache.lucene.search.TopFieldDocs) SortField(org.apache.lucene.search.SortField) SortedSetSortField(org.apache.lucene.search.SortedSetSortField) SortedNumericSortField(org.apache.lucene.search.SortedNumericSortField) Document(org.apache.lucene.document.Document) SortedNumericDocValuesField(org.apache.lucene.document.SortedNumericDocValuesField) NumericDocValuesField(org.apache.lucene.document.NumericDocValuesField) Sort(org.apache.lucene.search.Sort) LeafReaderContext(org.apache.lucene.index.LeafReaderContext) TopFieldCollector(org.apache.lucene.search.TopFieldCollector) CollapseTopFieldDocs(org.apache.lucene.search.grouping.CollapseTopFieldDocs) Directory(org.apache.lucene.store.Directory) HashSet(java.util.HashSet) MatchAllDocsQuery(org.apache.lucene.search.MatchAllDocsQuery) IndexReaderContext(org.apache.lucene.index.IndexReaderContext) Weight(org.apache.lucene.search.Weight) CompositeReaderContext(org.apache.lucene.index.CompositeReaderContext) IndexReader(org.apache.lucene.index.IndexReader) CollapsingTopDocsCollector(org.apache.lucene.search.grouping.CollapsingTopDocsCollector) RandomIndexWriter(org.apache.lucene.index.RandomIndexWriter)

Aggregations

FieldDoc (org.apache.lucene.search.FieldDoc)50 Sort (org.apache.lucene.search.Sort)26 TopFieldDocs (org.apache.lucene.search.TopFieldDocs)23 SortField (org.apache.lucene.search.SortField)21 ScoreDoc (org.apache.lucene.search.ScoreDoc)20 MatchAllDocsQuery (org.apache.lucene.search.MatchAllDocsQuery)18 IndexSearcher (org.apache.lucene.search.IndexSearcher)17 ArrayList (java.util.ArrayList)14 Directory (org.apache.lucene.store.Directory)14 TopDocs (org.apache.lucene.search.TopDocs)12 RandomIndexWriter (org.apache.lucene.index.RandomIndexWriter)10 TermQuery (org.apache.lucene.search.TermQuery)10 Document (org.apache.lucene.document.Document)9 Term (org.apache.lucene.index.Term)9 BytesRef (org.apache.lucene.util.BytesRef)9 HashMap (java.util.HashMap)8 Query (org.apache.lucene.search.Query)8 IndexReader (org.apache.lucene.index.IndexReader)7 HashSet (java.util.HashSet)5 DirectoryReader (org.apache.lucene.index.DirectoryReader)5