@Override
public TermsEnum iterator() throws IOException {
return values.termsEnum();
}
/**
* Merges the sortedset docvalues from <code>toMerge</code>.
*
* <p>The default implementation calls {@link #addSortedSetField}, passing an Iterable that merges
* ordinals and values and filters deleted documents .
*/
public void mergeSortedSetField(
FieldInfo fieldInfo, final MergeState mergeState, List<SortedSetDocValues> toMerge)
throws IOException {
mergeState.checkAbort.work(mergeState.segmentInfo.getDocCount());
final AtomicReader readers[] = mergeState.readers.toArray(new AtomicReader[toMerge.size()]);
final SortedSetDocValues dvs[] = toMerge.toArray(new SortedSetDocValues[toMerge.size()]);
// step 1: iterate thru each sub and mark terms still in use
TermsEnum liveTerms[] = new TermsEnum[dvs.length];
long[] weights = new long[liveTerms.length];
for (int sub = 0; sub < liveTerms.length; sub++) {
AtomicReader reader = readers[sub];
SortedSetDocValues dv = dvs[sub];
Bits liveDocs = reader.getLiveDocs();
if (liveDocs == null) {
liveTerms[sub] = dv.termsEnum();
weights[sub] = dv.getValueCount();
} else {
LongBitSet bitset = new LongBitSet(dv.getValueCount());
for (int i = 0; i < reader.maxDoc(); i++) {
if (liveDocs.get(i)) {
dv.setDocument(i);
long ord;
while ((ord = dv.nextOrd()) != SortedSetDocValues.NO_MORE_ORDS) {
bitset.set(ord);
}
}
}
liveTerms[sub] = new BitsFilteredTermsEnum(dv.termsEnum(), bitset);
weights[sub] = bitset.cardinality();
}
}
// step 2: create ordinal map (this conceptually does the "merging")
final OrdinalMap map = OrdinalMap.build(this, liveTerms, weights, PackedInts.COMPACT);
// step 3: add field
addSortedSetField(
fieldInfo,
// ord -> value
new Iterable<BytesRef>() {
@Override
public Iterator<BytesRef> iterator() {
return new Iterator<BytesRef>() {
long currentOrd;
@Override
public boolean hasNext() {
return currentOrd < map.getValueCount();
}
@Override
public BytesRef next() {
if (!hasNext()) {
throw new NoSuchElementException();
}
int segmentNumber = map.getFirstSegmentNumber(currentOrd);
long segmentOrd = map.getFirstSegmentOrd(currentOrd);
final BytesRef term = dvs[segmentNumber].lookupOrd(segmentOrd);
currentOrd++;
return term;
}
@Override
public void remove() {
throw new UnsupportedOperationException();
}
};
}
},
// doc -> ord count
new Iterable<Number>() {
@Override
public Iterator<Number> iterator() {
return new Iterator<Number>() {
int readerUpto = -1;
int docIDUpto;
int nextValue;
AtomicReader currentReader;
Bits currentLiveDocs;
boolean nextIsSet;
@Override
public boolean hasNext() {
return nextIsSet || setNext();
}
@Override
public void remove() {
throw new UnsupportedOperationException();
}
@Override
public Number next() {
if (!hasNext()) {
throw new NoSuchElementException();
}
assert nextIsSet;
nextIsSet = false;
// TODO make a mutable number
return nextValue;
}
private boolean setNext() {
while (true) {
if (readerUpto == readers.length) {
return false;
}
if (currentReader == null || docIDUpto == currentReader.maxDoc()) {
readerUpto++;
if (readerUpto < readers.length) {
currentReader = readers[readerUpto];
currentLiveDocs = currentReader.getLiveDocs();
}
docIDUpto = 0;
continue;
}
if (currentLiveDocs == null || currentLiveDocs.get(docIDUpto)) {
nextIsSet = true;
SortedSetDocValues dv = dvs[readerUpto];
dv.setDocument(docIDUpto);
nextValue = 0;
while (dv.nextOrd() != SortedSetDocValues.NO_MORE_ORDS) {
nextValue++;
}
docIDUpto++;
return true;
}
docIDUpto++;
}
}
};
}
},
// ords
new Iterable<Number>() {
@Override
public Iterator<Number> iterator() {
return new Iterator<Number>() {
int readerUpto = -1;
int docIDUpto;
long nextValue;
AtomicReader currentReader;
Bits currentLiveDocs;
LongValues currentMap;
boolean nextIsSet;
long ords[] = new long[8];
int ordUpto;
int ordLength;
@Override
public boolean hasNext() {
return nextIsSet || setNext();
}
@Override
public void remove() {
throw new UnsupportedOperationException();
}
@Override
public Number next() {
if (!hasNext()) {
throw new NoSuchElementException();
}
assert nextIsSet;
nextIsSet = false;
// TODO make a mutable number
return nextValue;
}
private boolean setNext() {
while (true) {
if (readerUpto == readers.length) {
return false;
}
if (ordUpto < ordLength) {
nextValue = ords[ordUpto];
ordUpto++;
nextIsSet = true;
return true;
}
if (currentReader == null || docIDUpto == currentReader.maxDoc()) {
readerUpto++;
if (readerUpto < readers.length) {
currentReader = readers[readerUpto];
currentLiveDocs = currentReader.getLiveDocs();
currentMap = map.getGlobalOrds(readerUpto);
}
docIDUpto = 0;
continue;
}
if (currentLiveDocs == null || currentLiveDocs.get(docIDUpto)) {
assert docIDUpto < currentReader.maxDoc();
SortedSetDocValues dv = dvs[readerUpto];
dv.setDocument(docIDUpto);
ordUpto = ordLength = 0;
long ord;
while ((ord = dv.nextOrd()) != SortedSetDocValues.NO_MORE_ORDS) {
if (ordLength == ords.length) {
ords = ArrayUtil.grow(ords, ordLength + 1);
}
ords[ordLength] = currentMap.get(ord);
ordLength++;
}
docIDUpto++;
continue;
}
docIDUpto++;
}
}
};
}
});
}
@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;
}
}
