@Override
public void flush(SegmentWriteState state, Sorter.DocMap sortMap, DocValuesConsumer dvConsumer)
throws IOException {
final int valueCount = hash.size();
final PackedLongValues ords;
final int[] sortedValues;
final int[] ordMap;
if (finalOrds == null) {
sortedValues = hash.sort();
ords = pending.build();
ordMap = new int[valueCount];
for (int ord = 0; ord < valueCount; ord++) {
ordMap[sortedValues[ord]] = ord;
}
} else {
sortedValues = finalSortedValues;
ords = finalOrds;
ordMap = finalOrdMap;
}
final int[] sorted;
if (sortMap != null) {
sorted =
sortDocValues(
state.segmentInfo.maxDoc(),
sortMap,
new BufferedSortedDocValues(
hash, valueCount, ords, sortedValues, ordMap, docsWithField.iterator()));
} else {
sorted = null;
}
dvConsumer.addSortedField(
fieldInfo,
new EmptyDocValuesProducer() {
@Override
public SortedDocValues getSorted(FieldInfo fieldInfoIn) {
if (fieldInfoIn != fieldInfo) {
throw new IllegalArgumentException("wrong fieldInfo");
}
final SortedDocValues buf =
new BufferedSortedDocValues(
hash, valueCount, ords, sortedValues, ordMap, docsWithField.iterator());
if (sorted == null) {
return buf;
}
return new SortingLeafReader.SortingSortedDocValues(buf, sorted);
}
});
}
/**
* Merges the sorted docvalues from <code>toMerge</code>.
*
* <p>The default implementation calls {@link #addSortedField}, passing an Iterable that merges
* ordinals and values and filters deleted documents .
*/
public void mergeSortedField(
FieldInfo fieldInfo, final MergeState mergeState, List<SortedDocValues> toMerge)
throws IOException {
mergeState.checkAbort.work(mergeState.segmentInfo.getDocCount());
final AtomicReader readers[] = mergeState.readers.toArray(new AtomicReader[toMerge.size()]);
final SortedDocValues dvs[] = toMerge.toArray(new SortedDocValues[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];
SortedDocValues 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)) {
int ord = dv.getOrd(i);
if (ord >= 0) {
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
addSortedField(
fieldInfo,
// ord -> value
new Iterable<BytesRef>() {
@Override
public Iterator<BytesRef> iterator() {
return new Iterator<BytesRef>() {
int currentOrd;
@Override
public boolean hasNext() {
return currentOrd < map.getValueCount();
}
@Override
public BytesRef next() {
if (!hasNext()) {
throw new NoSuchElementException();
}
int segmentNumber = map.getFirstSegmentNumber(currentOrd);
int segmentOrd = (int) map.getFirstSegmentOrd(currentOrd);
final BytesRef term = dvs[segmentNumber].lookupOrd(segmentOrd);
currentOrd++;
return term;
}
@Override
public void remove() {
throw new UnsupportedOperationException();
}
};
}
},
// doc -> ord
new Iterable<Number>() {
@Override
public Iterator<Number> iterator() {
return new Iterator<Number>() {
int readerUpto = -1;
int docIDUpto;
int nextValue;
AtomicReader currentReader;
Bits currentLiveDocs;
LongValues currentMap;
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();
currentMap = map.getGlobalOrds(readerUpto);
}
docIDUpto = 0;
continue;
}
if (currentLiveDocs == null || currentLiveDocs.get(docIDUpto)) {
nextIsSet = true;
int segOrd = dvs[readerUpto].getOrd(docIDUpto);
nextValue = segOrd == -1 ? -1 : (int) currentMap.get(segOrd);
docIDUpto++;
return true;
}
docIDUpto++;
}
}
};
}
});
}