@Override
public LeafBucketCollector getLeafCollector(LeafReaderContext ctx, final LeafBucketCollector sub)
throws IOException {
if (valuesSource == null) {
return LeafBucketCollector.NO_OP_COLLECTOR;
}
final SortedNumericDocValues values = valuesSource.longValues(ctx);
return new LeafBucketCollectorBase(sub, values) {
@Override
public void collect(int doc, long bucket) throws IOException {
assert bucket == 0;
values.setDocument(doc);
final int valuesCount = values.count();
long previousKey = Long.MIN_VALUE;
for (int i = 0; i < valuesCount; ++i) {
long value = values.valueAt(i);
long key = rounding.roundKey(value);
assert key >= previousKey;
if (key == previousKey) {
continue;
}
long bucketOrd = bucketOrds.add(key);
if (bucketOrd < 0) { // already seen
bucketOrd = -1 - bucketOrd;
collectExistingBucket(sub, doc, bucketOrd);
} else {
collectBucket(sub, doc, bucketOrd);
}
previousKey = key;
}
}
};
}
private Collector createCollector(AtomicReaderContext reader) {
// if rehash is false then the value source is either already hashed, or the user explicitly
// requested not to hash the values (perhaps they already hashed the values themselves before
// indexing the doc)
// so we can just work with the original value source as is
if (!rehash) {
MurmurHash3Values hashValues =
MurmurHash3Values.cast(((ValuesSource.Numeric) valuesSource).longValues());
return new DirectCollector(counts, hashValues);
}
if (valuesSource instanceof ValuesSource.Numeric) {
ValuesSource.Numeric source = (ValuesSource.Numeric) valuesSource;
MurmurHash3Values hashValues =
source.isFloatingPoint()
? MurmurHash3Values.hash(source.doubleValues())
: MurmurHash3Values.hash(source.longValues());
return new DirectCollector(counts, hashValues);
}
if (valuesSource instanceof ValuesSource.Bytes.WithOrdinals) {
ValuesSource.Bytes.WithOrdinals source = (ValuesSource.Bytes.WithOrdinals) valuesSource;
final RandomAccessOrds ordinalValues = source.ordinalsValues();
final long maxOrd = ordinalValues.getValueCount();
if (maxOrd == 0) {
return new EmptyCollector();
}
final long ordinalsMemoryUsage = OrdinalsCollector.memoryOverhead(maxOrd);
final long countsMemoryUsage = HyperLogLogPlusPlus.memoryUsage(precision);
// only use ordinals if they don't increase memory usage by more than 25%
if (ordinalsMemoryUsage < countsMemoryUsage / 4) {
return new OrdinalsCollector(counts, ordinalValues, bigArrays);
}
}
return new DirectCollector(counts, MurmurHash3Values.hash(valuesSource.bytesValues()));
}
@Override
public boolean needsScores() {
return (valuesSource != null && valuesSource.needsScores()) || super.needsScores();
}
