@Override
protected void doSetNextReader(LeafReaderContext context) throws IOException {
if (segmentFacetCounts != null) {
segmentResults.add(createSegmentResult());
}
groupFieldTermsIndex = DocValues.getSorted(context.reader(), groupField);
facetFieldTermsIndex = DocValues.getSorted(context.reader(), facetField);
// 1+ to allow for the -1 "not set":
segmentFacetCounts = new int[facetFieldTermsIndex.getValueCount() + 1];
segmentTotalCount = 0;
segmentGroupedFacetHits.clear();
for (GroupedFacetHit groupedFacetHit : groupedFacetHits) {
int facetOrd =
groupedFacetHit.facetValue == null
? -1
: facetFieldTermsIndex.lookupTerm(groupedFacetHit.facetValue);
if (groupedFacetHit.facetValue != null && facetOrd < 0) {
continue;
}
int groupOrd =
groupedFacetHit.groupValue == null
? -1
: groupFieldTermsIndex.lookupTerm(groupedFacetHit.groupValue);
if (groupedFacetHit.groupValue != null && groupOrd < 0) {
continue;
}
int segmentGroupedFacetsIndex =
groupOrd * (facetFieldTermsIndex.getValueCount() + 1) + facetOrd;
segmentGroupedFacetHits.put(segmentGroupedFacetsIndex);
}
if (facetPrefix != null) {
startFacetOrd = facetFieldTermsIndex.lookupTerm(facetPrefix);
if (startFacetOrd < 0) {
// Points to the ord one higher than facetPrefix
startFacetOrd = -startFacetOrd - 1;
}
BytesRefBuilder facetEndPrefix = new BytesRefBuilder();
facetEndPrefix.append(facetPrefix);
facetEndPrefix.append(UnicodeUtil.BIG_TERM);
endFacetOrd = facetFieldTermsIndex.lookupTerm(facetEndPrefix.get());
assert endFacetOrd < 0;
endFacetOrd = -endFacetOrd - 1; // Points to the ord one higher than facetEndPrefix
} else {
startFacetOrd = -1;
endFacetOrd = facetFieldTermsIndex.getValueCount();
}
}
@Test
public void testSvValues() throws IOException {
int numDocs = 1000000;
int numOrdinals = numDocs / 4;
Map<Integer, Long> controlDocToOrdinal = new HashMap<>();
OrdinalsBuilder builder = new OrdinalsBuilder(numDocs);
long ordinal = builder.currentOrdinal();
for (int doc = 0; doc < numDocs; doc++) {
if (doc % numOrdinals == 0) {
ordinal = builder.nextOrdinal();
}
controlDocToOrdinal.put(doc, ordinal);
builder.addDoc(doc);
}
Ordinals ords = builder.build(ImmutableSettings.EMPTY);
assertThat(ords, instanceOf(SinglePackedOrdinals.class));
RandomAccessOrds docs = ords.ordinals();
final SortedDocValues singleOrds = DocValues.unwrapSingleton(docs);
assertNotNull(singleOrds);
for (Map.Entry<Integer, Long> entry : controlDocToOrdinal.entrySet()) {
assertThat(entry.getValue(), equalTo((long) singleOrds.getOrd(entry.getKey())));
}
}
@Override
public DocIdSet getDocIdSet(LeafReaderContext context, Bits acceptDocs) throws IOException {
final Bits docsWithField = DocValues.getDocsWithField(context.reader(), field);
if (negate) {
if (docsWithField instanceof MatchAllBits) {
return null;
}
return new DocValuesDocIdSet(context.reader().maxDoc(), acceptDocs) {
@Override
protected final boolean matchDoc(int doc) {
return !docsWithField.get(doc);
}
};
} else {
if (docsWithField instanceof MatchNoBits) {
return null;
}
if (docsWithField instanceof BitSet) {
// UweSays: this is always the case for our current impl - but who knows
// :-)
return BitsFilteredDocIdSet.wrap(new BitDocIdSet((BitSet) docsWithField), acceptDocs);
}
return new DocValuesDocIdSet(context.reader().maxDoc(), acceptDocs) {
@Override
protected final boolean matchDoc(int doc) {
return docsWithField.get(doc);
}
};
}
}
/** Returns a SortedSetDocValues view of this instance */
public SortedSetDocValues iterator(LeafReader reader) throws IOException {
if (isEmpty()) {
return DocValues.emptySortedSet();
} else {
return new Iterator(reader);
}
}
private static SortedNumericDocValues sparseSingles(
final NumericDocValues deltas,
final long minValue,
final long missingValue,
final int maxDoc) {
final NumericDocValues values =
new NumericDocValues() {
@Override
public long get(int docID) {
final long delta = deltas.get(docID);
if (delta == missingValue) {
return 0;
}
return minValue + delta;
}
};
final Bits docsWithFields =
new Bits() {
@Override
public boolean get(int index) {
return deltas.get(index) != missingValue;
}
@Override
public int length() {
return maxDoc;
}
};
return DocValues.singleton(values, docsWithFields);
}
@Override
public Bits getDocsWithField(FieldInfo field) throws IOException {
if (field.getDocValuesType() == DocValuesType.SORTED_SET) {
return DocValues.docsWithValue(getSortedSet(field), maxDoc);
} else {
return new Bits.MatchAllBits(maxDoc);
}
}
@Override
public FunctionValues getValues(Map context, LeafReaderContext readerContext) throws IOException {
final NumericDocValues arr = DocValues.getNumeric(readerContext.reader(), field);
final Bits valid = DocValues.getDocsWithField(readerContext.reader(), field);
return new LongDocValues(this) {
@Override
public long longVal(int doc) {
return arr.get(doc);
}
@Override
public boolean exists(int doc) {
return valid.get(doc);
}
@Override
public Object objectVal(int doc) {
return exists(doc) ? longToObject(arr.get(doc)) : null;
}
@Override
public String strVal(int doc) {
return exists(doc) ? longToString(arr.get(doc)) : null;
}
@Override
public ValueFiller getValueFiller() {
return new ValueFiller() {
private final MutableValueDate mval = new MutableValueDate();
@Override
public MutableValue getValue() {
return mval;
}
@Override
public void fillValue(int doc) {
mval.value = arr.get(doc);
mval.exists = exists(doc);
}
};
}
};
}
@Override
public LeafFieldComparator getLeafComparator(LeafReaderContext context) throws IOException {
LeafReader reader = context.reader();
FieldInfo info = reader.getFieldInfos().fieldInfo(field);
if (info != null) {
Geo3DDocValuesField.checkCompatible(info);
}
currentDocs = DocValues.getSortedNumeric(reader, field);
return this;
}
@Override
public MultiGeoPointValues getGeoPointValues() {
final RandomAccessOrds ords = ordinals.ordinals();
final SortedDocValues singleOrds = DocValues.unwrapSingleton(ords);
if (singleOrds != null) {
final GeoPoint point = new GeoPoint();
final GeoPointValues values =
new GeoPointValues() {
@Override
public GeoPoint get(int docID) {
final int ord = singleOrds.getOrd(docID);
if (ord >= 0) {
return point.reset(lat.get(ord), lon.get(ord));
}
return point.reset(Double.NaN, Double.NaN);
}
};
return FieldData.singleton(values, DocValues.docsWithValue(singleOrds, maxDoc));
} else {
final GeoPoint point = new GeoPoint();
return new MultiGeoPointValues() {
@Override
public GeoPoint valueAt(int index) {
final long ord = ords.ordAt(index);
if (ord >= 0) {
return point.reset(lat.get(ord), lon.get(ord));
}
return point.reset(Double.NaN, Double.NaN);
}
@Override
public void setDocument(int docId) {
ords.setDocument(docId);
}
@Override
public int count() {
return ords.cardinality();
}
};
}
}
@Override
public Explanation explain(LeafReaderContext context, int doc) throws IOException {
SortedDocValues values = DocValues.getSorted(context.reader(), joinField);
if (values != null) {
int segmentOrd = values.getOrd(doc);
if (segmentOrd != -1) {
BytesRef joinValue = values.lookupOrd(segmentOrd);
return Explanation.match(
queryNorm, "Score based on join value " + joinValue.utf8ToString());
}
}
return Explanation.noMatch("Not a match");
}
private static SortedNumericDocValues withOrdinals(
Ordinals ordinals, final LongValues values, int maxDoc) {
final RandomAccessOrds ords = ordinals.ordinals();
final SortedDocValues singleOrds = DocValues.unwrapSingleton(ords);
if (singleOrds != null) {
final NumericDocValues singleValues =
new NumericDocValues() {
@Override
public long get(int docID) {
final int ord = singleOrds.getOrd(docID);
if (ord >= 0) {
return values.get(singleOrds.getOrd(docID));
} else {
return 0;
}
}
};
return DocValues.singleton(singleValues, DocValues.docsWithValue(ords, maxDoc));
} else {
return new SortedNumericDocValues() {
@Override
public long valueAt(int index) {
return values.get(ords.ordAt(index));
}
@Override
public void setDocument(int doc) {
ords.setDocument(doc);
}
@Override
public int count() {
return ords.cardinality();
}
};
}
}
private static SortedNumericDocValues pagedSingles(
final PackedLongValues values, final Bits docsWithValue) {
return DocValues.singleton(
new NumericDocValues() {
// we need to wrap since NumericDocValues must return 0 when a doc has no value
@Override
public long get(int docID) {
if (docsWithValue == null || docsWithValue.get(docID)) {
return values.get(docID);
} else {
return 0;
}
}
},
docsWithValue);
}
private static SortedNumericDocValues singles(
final NumericDocValues deltas, final long minValue) {
final NumericDocValues values;
if (minValue == 0) {
values = deltas;
} else {
values =
new NumericDocValues() {
@Override
public long get(int docID) {
return minValue + deltas.get(docID);
}
};
}
return DocValues.singleton(values, null);
}
@Override
public Scorer scorer(LeafReaderContext context) throws IOException {
SortedDocValues values = DocValues.getSorted(context.reader(), joinField);
if (values == null) {
return null;
}
Scorer approximationScorer = approximationWeight.scorer(context);
if (approximationScorer == null) {
return null;
} else if (globalOrds != null) {
return new OrdinalMapScorer(
this, collector, values, approximationScorer, globalOrds.getGlobalOrds(context.ord));
} else {
return new SegmentOrdinalScorer(this, collector, values, approximationScorer);
}
}
@Override
protected void doSetNextReader(LeafReaderContext context) throws IOException {
index = DocValues.getSorted(context.reader(), groupField);
// Clear ordSet and fill it with previous encountered groups that can occur in the current
// segment.
ordSet.clear();
for (BytesRef countedGroup : groups) {
if (countedGroup == null) {
ordSet.put(-1);
} else {
int ord = index.lookupTerm(countedGroup);
if (ord >= 0) {
ordSet.put(ord);
}
}
}
}
@Override
public LeafCollector getLeafCollector(LeafReaderContext context) throws IOException {
final int base = context.docBase;
final NumericDocValues values = DocValues.getNumeric(context.reader(), "sort_i");
return new LeafCollector() {
@Override
public void setScorer(Scorer scorer) throws IOException {}
public boolean acceptsDocsOutOfOrder() {
return false;
}
public void collect(int doc) {
list.add(new ScoreDoc(doc + base, (float) values.get(doc)));
}
};
}
@Override
public Explanation explain(LeafReaderContext context, int doc) throws IOException {
SortedDocValues values = DocValues.getSorted(context.reader(), joinField);
if (values != null) {
int segmentOrd = values.getOrd(doc);
if (segmentOrd != -1) {
final float score;
if (globalOrds != null) {
long globalOrd = globalOrds.getGlobalOrds(context.ord).get(segmentOrd);
score = collector.score((int) globalOrd);
} else {
score = collector.score(segmentOrd);
}
BytesRef joinValue = values.lookupOrd(segmentOrd);
return Explanation.match(score, "Score based on join value " + joinValue.utf8ToString());
}
}
return Explanation.noMatch("Not a match");
}
@Override
public AtomicNumericFieldData load(LeafReaderContext context) {
try {
final BinaryDocValues values = DocValues.getBinary(context.reader(), fieldNames.indexName());
if (numericType.isFloatingPoint()) {
return new AtomicDoubleFieldData(-1) {
@Override
public SortedNumericDoubleValues getDoubleValues() {
switch (numericType) {
case FLOAT:
return new BinaryAsSortedNumericFloatValues(values);
case DOUBLE:
return new BinaryAsSortedNumericDoubleValues(values);
default:
throw new ElasticsearchIllegalArgumentException("" + numericType);
}
}
@Override
public Collection<Accountable> getChildResources() {
return Collections.emptyList();
}
};
} else {
return new AtomicLongFieldData(0) {
@Override
public SortedNumericDocValues getLongValues() {
return new BinaryAsSortedNumericDocValues(values);
}
@Override
public Collection<Accountable> getChildResources() {
return Collections.emptyList();
}
};
}
} catch (IOException e) {
throw new ElasticsearchIllegalStateException("Cannot load doc values", e);
}
}
@Override
public FunctionValues getValues(Map context, LeafReaderContext readerContext) throws IOException {
final int off = readerContext.docBase;
final LeafReader r;
Object o = context.get("searcher");
if (o instanceof SolrIndexSearcher) {
SolrIndexSearcher is = (SolrIndexSearcher) o;
SchemaField sf = is.getSchema().getFieldOrNull(field);
if (sf != null
&& sf.hasDocValues() == false
&& sf.multiValued() == false
&& sf.getType().getNumericType() != null) {
// it's a single-valued numeric field: we must currently create insanity :(
List<LeafReaderContext> leaves = is.getIndexReader().leaves();
LeafReader insaneLeaves[] = new LeafReader[leaves.size()];
int upto = 0;
for (LeafReaderContext raw : leaves) {
insaneLeaves[upto++] = Insanity.wrapInsanity(raw.reader(), field);
}
r = SlowCompositeReaderWrapper.wrap(new MultiReader(insaneLeaves));
} else {
// reuse ordinalmap
r = ((SolrIndexSearcher) o).getLeafReader();
}
} else {
IndexReader topReader = ReaderUtil.getTopLevelContext(readerContext).reader();
r = SlowCompositeReaderWrapper.wrap(topReader);
}
// if it's e.g. tokenized/multivalued, emulate old behavior of single-valued fc
final SortedDocValues sindex =
SortedSetSelector.wrap(DocValues.getSortedSet(r, field), SortedSetSelector.Type.MIN);
final int end = sindex.getValueCount();
return new IntDocValues(this) {
@Override
public int intVal(int doc) {
return (end - sindex.getOrd(doc + off) - 1);
}
};
}
@Override
public Scorer scorer(LeafReaderContext context, Bits acceptDocs) throws IOException {
SortedDocValues values = DocValues.getSorted(context.reader(), joinField);
if (values == null) {
return null;
}
Scorer approximationScorer = approximationWeight.scorer(context, acceptDocs);
if (approximationScorer == null) {
return null;
}
if (globalOrds != null) {
return new OrdinalMapScorer(
this,
queryNorm,
foundOrds,
values,
approximationScorer,
globalOrds.getGlobalOrds(context.ord));
}
{
return new SegmentOrdinalScorer(this, queryNorm, foundOrds, values, approximationScorer);
}
}
public void setNextReader(LeafReaderContext context) throws IOException {
this.vals = DocValues.getNumeric(context.reader(), this.field);
}
public DelegatingCollector getFilterCollector(IndexSearcher indexSearcher) {
try {
SolrIndexSearcher searcher = (SolrIndexSearcher) indexSearcher;
SortedDocValues docValues = null;
FunctionQuery funcQuery = null;
docValues = DocValues.getSorted(searcher.getLeafReader(), this.field);
FieldType fieldType = null;
if (this.max != null) {
if (this.max.indexOf("(") == -1) {
fieldType = searcher.getSchema().getField(this.max).getType();
} else {
LocalSolrQueryRequest request = null;
try {
SolrParams params = new ModifiableSolrParams();
request = new LocalSolrQueryRequest(searcher.getCore(), params);
FunctionQParser functionQParser = new FunctionQParser(this.max, null, null, request);
funcQuery = (FunctionQuery) functionQParser.parse();
} catch (Exception e) {
throw new IOException(e);
} finally {
request.close();
}
}
}
if (this.min != null) {
if (this.min.indexOf("(") == -1) {
fieldType = searcher.getSchema().getField(this.min).getType();
} else {
LocalSolrQueryRequest request = null;
try {
SolrParams params = new ModifiableSolrParams();
request = new LocalSolrQueryRequest(searcher.getCore(), params);
FunctionQParser functionQParser = new FunctionQParser(this.min, null, null, request);
funcQuery = (FunctionQuery) functionQParser.parse();
} catch (Exception e) {
throw new IOException(e);
} finally {
request.close();
}
}
}
int maxDoc = searcher.maxDoc();
int leafCount = searcher.getTopReaderContext().leaves().size();
// Deal with boosted docs.
// We have to deal with it here rather then the constructor because
// because the QueryElevationComponent runs after the Queries are constructed.
IntIntOpenHashMap boostDocs = null;
Map context = null;
SolrRequestInfo info = SolrRequestInfo.getRequestInfo();
if (info != null) {
context = info.getReq().getContext();
}
if (this.boosted == null && context != null) {
this.boosted =
(Map<BytesRef, Integer>) context.get(QueryElevationComponent.BOOSTED_PRIORITY);
}
boostDocs = getBoostDocs(searcher, this.boosted, context);
if (this.min != null || this.max != null) {
return new CollapsingFieldValueCollector(
maxDoc,
leafCount,
docValues,
this.nullPolicy,
max != null ? this.max : this.min,
max != null,
this.needsScores,
fieldType,
boostDocs,
funcQuery,
searcher);
} else {
return new CollapsingScoreCollector(
maxDoc, leafCount, docValues, this.nullPolicy, boostDocs);
}
} catch (Exception e) {
throw new RuntimeException(e);
}
}
@Override
public NumericDocValues getNorms(FieldInfo field) throws IOException {
final NormsEntry entry = norms.get(field.number);
if (entry.docsWithFieldOffset == -2) {
// empty
return DocValues.emptyNumeric();
} else if (entry.docsWithFieldOffset == -1) {
// dense
final LongValues normValues = getNormValues(entry);
return new NumericDocValues() {
int doc = -1;
@Override
public long longValue() throws IOException {
return normValues.get(doc);
}
@Override
public int docID() {
return doc;
}
@Override
public int nextDoc() throws IOException {
return advance(doc + 1);
}
@Override
public int advance(int target) throws IOException {
if (target >= maxDoc) {
return doc = NO_MORE_DOCS;
}
return doc = target;
}
@Override
public long cost() {
return maxDoc;
}
};
} else {
// sparse
final LongValues normValues = getNormValues(entry);
final SparseDISI disi;
synchronized (data) {
disi = new SparseDISI(maxDoc, data, entry.docsWithFieldOffset, entry.numDocsWithField);
}
return new NumericDocValues() {
@Override
public int advance(int target) throws IOException {
return disi.advance(target);
}
@Override
public int nextDoc() throws IOException {
return disi.nextDoc();
}
@Override
public int docID() {
return disi.docID();
}
@Override
public long cost() {
return entry.numDocsWithField;
}
@Override
public long longValue() throws IOException {
return normValues.get(disi.index());
}
};
}
}
public static NamedList<Integer> getCounts(
SolrIndexSearcher searcher,
DocSet docs,
String fieldName,
int offset,
int limit,
int mincount,
boolean missing,
String sort,
String prefix,
String contains,
boolean ignoreCase)
throws IOException {
SchemaField schemaField = searcher.getSchema().getField(fieldName);
FieldType ft = schemaField.getType();
NamedList<Integer> res = new NamedList<>();
// TODO: remove multiValuedFieldCache(), check dv type / uninversion type?
final boolean multiValued = schemaField.multiValued() || ft.multiValuedFieldCache();
final SortedSetDocValues si; // for term lookups only
OrdinalMap ordinalMap = null; // for mapping per-segment ords to global ones
if (multiValued) {
si = searcher.getLeafReader().getSortedSetDocValues(fieldName);
if (si instanceof MultiSortedSetDocValues) {
ordinalMap = ((MultiSortedSetDocValues) si).mapping;
}
} else {
SortedDocValues single = searcher.getLeafReader().getSortedDocValues(fieldName);
si = single == null ? null : DocValues.singleton(single);
if (single instanceof MultiSortedDocValues) {
ordinalMap = ((MultiSortedDocValues) single).mapping;
}
}
if (si == null) {
return finalize(res, searcher, schemaField, docs, -1, missing);
}
if (si.getValueCount() >= Integer.MAX_VALUE) {
throw new UnsupportedOperationException(
"Currently this faceting method is limited to " + Integer.MAX_VALUE + " unique terms");
}
final BytesRefBuilder prefixRef;
if (prefix == null) {
prefixRef = null;
} else if (prefix.length() == 0) {
prefix = null;
prefixRef = null;
} else {
prefixRef = new BytesRefBuilder();
prefixRef.copyChars(prefix);
}
int startTermIndex, endTermIndex;
if (prefix != null) {
startTermIndex = (int) si.lookupTerm(prefixRef.get());
if (startTermIndex < 0) startTermIndex = -startTermIndex - 1;
prefixRef.append(UnicodeUtil.BIG_TERM);
endTermIndex = (int) si.lookupTerm(prefixRef.get());
assert endTermIndex < 0;
endTermIndex = -endTermIndex - 1;
} else {
startTermIndex = -1;
endTermIndex = (int) si.getValueCount();
}
final int nTerms = endTermIndex - startTermIndex;
int missingCount = -1;
final CharsRefBuilder charsRef = new CharsRefBuilder();
if (nTerms > 0 && docs.size() >= mincount) {
// count collection array only needs to be as big as the number of terms we are
// going to collect counts for.
final int[] counts = new int[nTerms];
Filter filter = docs.getTopFilter();
List<LeafReaderContext> leaves = searcher.getTopReaderContext().leaves();
for (int subIndex = 0; subIndex < leaves.size(); subIndex++) {
LeafReaderContext leaf = leaves.get(subIndex);
DocIdSet dis =
filter.getDocIdSet(leaf, null); // solr docsets already exclude any deleted docs
DocIdSetIterator disi = null;
if (dis != null) {
disi = dis.iterator();
}
if (disi != null) {
if (multiValued) {
SortedSetDocValues sub = leaf.reader().getSortedSetDocValues(fieldName);
if (sub == null) {
sub = DocValues.emptySortedSet();
}
final SortedDocValues singleton = DocValues.unwrapSingleton(sub);
if (singleton != null) {
// some codecs may optimize SORTED_SET storage for single-valued fields
accumSingle(counts, startTermIndex, singleton, disi, subIndex, ordinalMap);
} else {
accumMulti(counts, startTermIndex, sub, disi, subIndex, ordinalMap);
}
} else {
SortedDocValues sub = leaf.reader().getSortedDocValues(fieldName);
if (sub == null) {
sub = DocValues.emptySorted();
}
accumSingle(counts, startTermIndex, sub, disi, subIndex, ordinalMap);
}
}
}
if (startTermIndex == -1) {
missingCount = counts[0];
}
// IDEA: we could also maintain a count of "other"... everything that fell outside
// of the top 'N'
int off = offset;
int lim = limit >= 0 ? limit : Integer.MAX_VALUE;
if (sort.equals(FacetParams.FACET_SORT_COUNT)
|| sort.equals(FacetParams.FACET_SORT_COUNT_LEGACY)) {
int maxsize = limit > 0 ? offset + limit : Integer.MAX_VALUE - 1;
maxsize = Math.min(maxsize, nTerms);
LongPriorityQueue queue =
new LongPriorityQueue(Math.min(maxsize, 1000), maxsize, Long.MIN_VALUE);
int min = mincount - 1; // the smallest value in the top 'N' values
for (int i = (startTermIndex == -1) ? 1 : 0; i < nTerms; i++) {
int c = counts[i];
if (contains != null) {
final BytesRef term = si.lookupOrd(startTermIndex + i);
if (!SimpleFacets.contains(term.utf8ToString(), contains, ignoreCase)) {
continue;
}
}
if (c > min) {
// NOTE: we use c>min rather than c>=min as an optimization because we are going in
// index order, so we already know that the keys are ordered. This can be very
// important if a lot of the counts are repeated (like zero counts would be).
// smaller term numbers sort higher, so subtract the term number instead
long pair = (((long) c) << 32) + (Integer.MAX_VALUE - i);
boolean displaced = queue.insert(pair);
if (displaced) min = (int) (queue.top() >>> 32);
}
}
// if we are deep paging, we don't have to order the highest "offset" counts.
int collectCount = Math.max(0, queue.size() - off);
assert collectCount <= lim;
// the start and end indexes of our list "sorted" (starting with the highest value)
int sortedIdxStart = queue.size() - (collectCount - 1);
int sortedIdxEnd = queue.size() + 1;
final long[] sorted = queue.sort(collectCount);
for (int i = sortedIdxStart; i < sortedIdxEnd; i++) {
long pair = sorted[i];
int c = (int) (pair >>> 32);
int tnum = Integer.MAX_VALUE - (int) pair;
final BytesRef term = si.lookupOrd(startTermIndex + tnum);
ft.indexedToReadable(term, charsRef);
res.add(charsRef.toString(), c);
}
} else {
// add results in index order
int i = (startTermIndex == -1) ? 1 : 0;
if (mincount <= 0 && contains == null) {
// if mincount<=0 and we're not examining the values for contains, then
// we won't discard any terms and we know exactly where to start.
i += off;
off = 0;
}
for (; i < nTerms; i++) {
int c = counts[i];
if (c < mincount) continue;
BytesRef term = null;
if (contains != null) {
term = si.lookupOrd(startTermIndex + i);
if (!SimpleFacets.contains(term.utf8ToString(), contains, ignoreCase)) {
continue;
}
}
if (--off >= 0) continue;
if (--lim < 0) break;
if (term == null) {
term = si.lookupOrd(startTermIndex + i);
}
ft.indexedToReadable(term, charsRef);
res.add(charsRef.toString(), c);
}
}
}
return finalize(res, searcher, schemaField, docs, missingCount, missing);
}
@Override
public SortedSetDocValues getSortedSet(FieldInfo field) throws IOException {
final FSTEntry entry = fsts.get(field.name);
if (entry.numOrds == 0) {
return DocValues.emptySortedSet(); // empty FST!
}
FST<Long> instance;
synchronized (this) {
instance = fstInstances.get(field.name);
if (instance == null) {
data.seek(entry.offset);
instance = new FST<>(data, PositiveIntOutputs.getSingleton());
if (!merging) {
ramBytesUsed.addAndGet(instance.ramBytesUsed());
fstInstances.put(field.name, instance);
}
}
}
final BinaryDocValues docToOrds = getBinary(field);
final FST<Long> fst = instance;
// per-thread resources
final BytesReader in = fst.getBytesReader();
final Arc<Long> firstArc = new Arc<>();
final Arc<Long> scratchArc = new Arc<>();
final IntsRefBuilder scratchInts = new IntsRefBuilder();
final BytesRefFSTEnum<Long> fstEnum = new BytesRefFSTEnum<>(fst);
final ByteArrayDataInput input = new ByteArrayDataInput();
return new SortedSetDocValues() {
final BytesRefBuilder term = new BytesRefBuilder();
BytesRef ordsRef;
long currentOrd;
@Override
public long nextOrd() {
if (input.eof()) {
return NO_MORE_ORDS;
} else {
currentOrd += input.readVLong();
return currentOrd;
}
}
@Override
public void setDocument(int docID) {
ordsRef = docToOrds.get(docID);
input.reset(ordsRef.bytes, ordsRef.offset, ordsRef.length);
currentOrd = 0;
}
@Override
public BytesRef lookupOrd(long ord) {
try {
in.setPosition(0);
fst.getFirstArc(firstArc);
IntsRef output = Util.getByOutput(fst, ord, in, firstArc, scratchArc, scratchInts);
term.grow(output.length);
term.clear();
return Util.toBytesRef(output, term);
} catch (IOException bogus) {
throw new RuntimeException(bogus);
}
}
@Override
public long lookupTerm(BytesRef key) {
try {
InputOutput<Long> o = fstEnum.seekCeil(key);
if (o == null) {
return -getValueCount() - 1;
} else if (o.input.equals(key)) {
return o.output.intValue();
} else {
return -o.output - 1;
}
} catch (IOException bogus) {
throw new RuntimeException(bogus);
}
}
@Override
public long getValueCount() {
return entry.numOrds;
}
@Override
public TermsEnum termsEnum() {
return new FSTTermsEnum(fst);
}
};
}
@Override
protected boolean tryBulkMerge(DocValues docValues) {
// only bulk merge if value type is the same otherwise size differs
return super.tryBulkMerge(docValues) && docValues.getType() == template.getType();
}
@Override
protected void doSetNextReader(LeafReaderContext context) throws IOException {
if (segmentFacetCounts != null) {
segmentResults.add(createSegmentResult());
}
groupFieldTermsIndex = DocValues.getSorted(context.reader(), groupField);
facetFieldDocTermOrds = DocValues.getSortedSet(context.reader(), facetField);
facetFieldNumTerms = (int) facetFieldDocTermOrds.getValueCount();
if (facetFieldNumTerms == 0) {
facetOrdTermsEnum = null;
} else {
facetOrdTermsEnum = facetFieldDocTermOrds.termsEnum();
}
// [facetFieldNumTerms() + 1] for all possible facet values and docs not containing facet
// field
segmentFacetCounts = new int[facetFieldNumTerms + 1];
segmentTotalCount = 0;
segmentGroupedFacetHits.clear();
for (GroupedFacetHit groupedFacetHit : groupedFacetHits) {
int groupOrd =
groupedFacetHit.groupValue == null
? -1
: groupFieldTermsIndex.lookupTerm(groupedFacetHit.groupValue);
if (groupedFacetHit.groupValue != null && groupOrd < 0) {
continue;
}
int facetOrd;
if (groupedFacetHit.facetValue != null) {
if (facetOrdTermsEnum == null
|| !facetOrdTermsEnum.seekExact(groupedFacetHit.facetValue)) {
continue;
}
facetOrd = (int) facetOrdTermsEnum.ord();
} else {
facetOrd = facetFieldNumTerms;
}
// (facetFieldDocTermOrds.numTerms() + 1) for all possible facet values and docs not
// containing facet field
int segmentGroupedFacetsIndex = groupOrd * (facetFieldNumTerms + 1) + facetOrd;
segmentGroupedFacetHits.put(segmentGroupedFacetsIndex);
}
if (facetPrefix != null) {
TermsEnum.SeekStatus seekStatus;
if (facetOrdTermsEnum != null) {
seekStatus = facetOrdTermsEnum.seekCeil(facetPrefix);
} else {
seekStatus = TermsEnum.SeekStatus.END;
}
if (seekStatus != TermsEnum.SeekStatus.END) {
startFacetOrd = (int) facetOrdTermsEnum.ord();
} else {
startFacetOrd = 0;
endFacetOrd = 0;
return;
}
BytesRefBuilder facetEndPrefix = new BytesRefBuilder();
facetEndPrefix.append(facetPrefix);
facetEndPrefix.append(UnicodeUtil.BIG_TERM);
seekStatus = facetOrdTermsEnum.seekCeil(facetEndPrefix.get());
if (seekStatus != TermsEnum.SeekStatus.END) {
endFacetOrd = (int) facetOrdTermsEnum.ord();
} else {
endFacetOrd = facetFieldNumTerms; // Don't include null...
}
} else {
startFacetOrd = 0;
endFacetOrd = facetFieldNumTerms + 1;
}
}
