public override Explanation Explain(IndexSearcher searcher, Explanation firstPassExplanation, int docID)
{
Explanation result = base.Explain(searcher, firstPassExplanation, docID);
IList <AtomicReaderContext> leaves = searcher.IndexReader.Leaves;
int subReader = ReaderUtil.SubIndex(docID, leaves);
AtomicReaderContext readerContext = leaves[subReader];
int docIDInSegment = docID - readerContext.DocBase;
var context = new Dictionary <string, object>();
var fakeScorer = new FakeScorer {
score = firstPassExplanation.Value, doc = docIDInSegment
};
context["scorer"] = fakeScorer;
foreach (string variable in expression.variables)
{
result.AddDetail(new Explanation((float)bindings.GetValueSource(variable).GetValues
(context, readerContext).DoubleVal(docIDInSegment), "variable \"" + variable + "\""
));
}
return(result);
}
private void SumValues(IList <MatchingDocs> matchingDocs, bool keepScores, ValueSource valueSource)
{
FakeScorer scorer = new FakeScorer();
IDictionary context = new Dictionary <string, Scorer>();
if (keepScores)
{
context["scorer"] = scorer;
}
Int32sRef scratch = new Int32sRef();
foreach (MatchingDocs hits in matchingDocs)
{
OrdinalsReader.OrdinalsSegmentReader ords = ordinalsReader.GetReader(hits.Context);
int scoresIdx = 0;
float[] scores = hits.Scores;
FunctionValues functionValues = valueSource.GetValues(context, hits.Context);
DocIdSetIterator docs = hits.Bits.GetIterator();
int doc;
while ((doc = docs.NextDoc()) != DocIdSetIterator.NO_MORE_DOCS)
{
ords.Get(doc, scratch);
if (keepScores)
{
scorer.docID = doc;
scorer.score = scores[scoresIdx++];
}
float value = (float)functionValues.DoubleVal(doc);
for (int i = 0; i < scratch.Length; i++)
{
m_values[scratch.Int32s[i]] += value;
}
}
}
Rollup();
}
/// <summary>
/// Returns the grouped results. Returns null if the
/// number of groups collected is <= groupOffset.
///
/// <para>
/// <b>NOTE</b>: This collector is unable to compute
/// the groupValue per group so it will always be null.
/// This is normally not a problem, as you can obtain the
/// value just like you obtain other values for each
/// matching document (eg, via stored fields, via
/// FieldCache, etc.)
/// </para>
/// </summary>
/// <typeparam name="TGroupValue">The expected return type for group value</typeparam>
/// <param name="withinGroupSort">
/// The <see cref="Sort"/> used to sort
/// documents within each group. Passing null is
/// allowed, to sort by relevance.
/// </param>
/// <param name="groupOffset">Which group to start from</param>
/// <param name="withinGroupOffset">
/// Which document to start from within each group
/// </param>
/// <param name="maxDocsPerGroup">
/// How many top documents to keep within each group.
/// </param>
/// <param name="fillSortFields">
/// If true then the Comparable values for the sort fields will be set
/// </param>
public virtual ITopGroups <TGroupValue> GetTopGroups <TGroupValue>(Sort withinGroupSort, int groupOffset, int withinGroupOffset, int maxDocsPerGroup, bool fillSortFields)
{
//if (queueFull) {
//System.out.println("getTopGroups groupOffset=" + groupOffset + " topNGroups=" + topNGroups);
//}
if (subDocUpto != 0)
{
ProcessGroup();
}
if (groupOffset >= groupQueue.Count)
{
return(null);
}
int totalGroupedHitCount = 0;
FakeScorer fakeScorer = new FakeScorer();
float maxScore = float.Epsilon; // LUCENENET: Epsilon in .NET is the same as MIN_VALUE in Java
GroupDocs <TGroupValue>[] groups = new GroupDocs <TGroupValue> [groupQueue.Count - groupOffset];
for (int downTo = groupQueue.Count - groupOffset - 1; downTo >= 0; downTo--)
{
OneGroup og = groupQueue.Pop();
// At this point we hold all docs w/ in each group,
// unsorted; we now sort them:
ITopDocsCollector collector;
if (withinGroupSort == null)
{
// Sort by score
if (!needsScores)
{
throw new ArgumentException("cannot sort by relevance within group: needsScores=false");
}
collector = TopScoreDocCollector.Create(maxDocsPerGroup, true);
}
else
{
// Sort by fields
collector = TopFieldCollector.Create(withinGroupSort, maxDocsPerGroup, fillSortFields, needsScores, needsScores, true);
}
collector.SetScorer(fakeScorer);
collector.SetNextReader(og.readerContext);
for (int docIDX = 0; docIDX < og.count; docIDX++)
{
int doc = og.docs[docIDX];
fakeScorer.doc = doc;
if (needsScores)
{
fakeScorer.score = og.scores[docIDX];
}
collector.Collect(doc);
}
totalGroupedHitCount += og.count;
object[] groupSortValues;
if (fillSortFields)
{
groupSortValues = new IComparable[comparers.Length];
for (int sortFieldIDX = 0; sortFieldIDX < comparers.Length; sortFieldIDX++)
{
groupSortValues[sortFieldIDX] = comparers[sortFieldIDX][og.comparerSlot];
}
}
else
{
groupSortValues = null;
}
TopDocs topDocs = collector.GetTopDocs(withinGroupOffset, maxDocsPerGroup);
// TODO: we could aggregate scores across children
// by Sum/Avg instead of passing NaN:
groups[downTo] = new GroupDocs <TGroupValue>(float.NaN,
topDocs.MaxScore,
og.count,
topDocs.ScoreDocs,
/// <summary>
/// Accumulates groups for the BlockJoinQuery specified by its slot.
/// </summary>
/// <param name="slot"> Search query's slot </param>
/// <param name="offset"> Parent docs offset </param>
/// <param name="maxDocsPerGroup"> Upper bound of documents per group number </param>
/// <param name="withinGroupOffset"> Offset within each group of child docs </param>
/// <param name="withinGroupSort"> Sort criteria within groups </param>
/// <param name="fillSortFields"> Specifies whether to add sort fields or not </param>
/// <returns> TopGroups for the query specified by slot </returns>
/// <exception cref="IOException"> if there is a low-level I/O error </exception>
private TopGroups<int> AccumulateGroups(int slot, int offset, int maxDocsPerGroup, int withinGroupOffset, Sort withinGroupSort, bool fillSortFields)
{
var groups = new GroupDocs<int>[sortedGroups.Length - offset];
var fakeScorer = new FakeScorer();
int totalGroupedHitCount = 0;
//System.out.println("slot=" + slot);
for (int groupIdx = offset; groupIdx < sortedGroups.Length; groupIdx++)
{
OneGroup og = sortedGroups[groupIdx];
int numChildDocs;
if (slot == -1 || slot >= og.counts.Length)
{
numChildDocs = 0;
}
else
{
numChildDocs = og.counts[slot];
}
// Number of documents in group should be bounded to prevent redundant memory allocation
int numDocsInGroup = Math.Max(1, Math.Min(numChildDocs, maxDocsPerGroup));
//System.out.println("parent doc=" + og.doc + " numChildDocs=" + numChildDocs + " maxDocsPG=" + maxDocsPerGroup);
// At this point we hold all docs w/ in each group, unsorted; we now sort them:
Collector collector;
if (withinGroupSort == null)
{
//System.out.println("sort by score");
// Sort by score
if (!trackScores)
{
throw new ArgumentException("cannot sort by relevance within group: trackScores=false");
}
collector = TopScoreDocCollector.Create(numDocsInGroup, true);
}
else
{
// Sort by fields
collector = TopFieldCollector.Create(withinGroupSort, numDocsInGroup, fillSortFields, trackScores, trackMaxScore, true);
}
collector.Scorer = fakeScorer;
collector.NextReader = og.readerContext;
for (int docIdx = 0; docIdx < numChildDocs; docIdx++)
{
//System.out.println("docIDX=" + docIDX + " vs " + og.docs[slot].length);
int doc = og.docs[slot][docIdx];
fakeScorer.doc = doc;
if (trackScores)
{
fakeScorer._score = og.scores[slot][docIdx];
}
collector.Collect(doc);
}
totalGroupedHitCount += numChildDocs;
object[] groupSortValues;
if (fillSortFields)
{
groupSortValues = new object[comparators.Length];
for (int sortFieldIdx = 0; sortFieldIdx < comparators.Length; sortFieldIdx++)
{
groupSortValues[sortFieldIdx] = comparators[sortFieldIdx].Value(og.Slot);
}
}
else
{
groupSortValues = null;
}
TopDocs topDocs;
if (withinGroupSort == null)
{
var tempCollector = (TopScoreDocCollector) collector;
topDocs = tempCollector.TopDocs(withinGroupOffset, numDocsInGroup);
}
else
{
var tempCollector = (TopFieldCollector) collector;
topDocs = tempCollector.TopDocs(withinGroupOffset, numDocsInGroup);
}
groups[groupIdx - offset] = new GroupDocs<int>(og.Score, topDocs.MaxScore, numChildDocs, topDocs.ScoreDocs, og.Doc, groupSortValues);
}
return new TopGroups<int>(new TopGroups<int>(sort.GetSort(), withinGroupSort == null ? null : withinGroupSort.GetSort(), 0, totalGroupedHitCount, groups, maxScore), totalHitCount);
}
public override bool Score(ICollector collector, int maxDoc)
{
if (maxDoc != int.MaxValue)
{
throw new ArgumentException("maxDoc must be System.Int32.MaxValue");
}
//if (DEBUG) {
// System.out.println("\nscore: reader=" + context.reader());
//}
//System.out.println("score r=" + context.reader());
FakeScorer scorer = new FakeScorer(this);
collector.SetScorer(scorer);
if (drillDownCollector != null)
{
drillDownCollector.SetScorer(scorer);
drillDownCollector.SetNextReader(context);
}
foreach (DocsAndCost dim in dims)
{
dim.sidewaysCollector.SetScorer(scorer);
dim.sidewaysCollector.SetNextReader(context);
}
// TODO: if we ever allow null baseScorer ... it will
// mean we DO score docs out of order ... hmm, or if we
// change up the order of the conjuntions below
if (Debugging.AssertsEnabled)
{
Debugging.Assert(baseScorer != null);
}
// some scorers, eg ReqExlScorer, can hit NPE if cost is called after nextDoc
long baseQueryCost = baseScorer.GetCost();
int numDims = dims.Length;
long drillDownCost = 0;
for (int dim = 0; dim < numDims; dim++)
{
DocIdSetIterator disi = dims[dim].disi;
if (dims[dim].bits == null && disi != null)
{
drillDownCost += disi.GetCost();
}
}
long drillDownAdvancedCost = 0;
if (numDims > 1 && dims[1].disi != null)
{
drillDownAdvancedCost = dims[1].disi.GetCost();
}
// Position all scorers to their first matching doc:
baseScorer.NextDoc();
int numBits = 0;
foreach (DocsAndCost dim in dims)
{
if (dim.disi != null)
{
dim.disi.NextDoc();
}
else if (dim.bits != null)
{
numBits++;
}
}
IBits[] bits = new IBits[numBits];
ICollector[] bitsSidewaysCollectors = new ICollector[numBits];
DocIdSetIterator[] disis = new DocIdSetIterator[numDims - numBits];
ICollector[] sidewaysCollectors = new ICollector[numDims - numBits];
int disiUpto = 0;
int bitsUpto = 0;
for (int dim = 0; dim < numDims; dim++)
{
DocIdSetIterator disi = dims[dim].disi;
if (dims[dim].bits == null)
{
disis[disiUpto] = disi;
sidewaysCollectors[disiUpto] = dims[dim].sidewaysCollector;
disiUpto++;
}
else
{
bits[bitsUpto] = dims[dim].bits;
bitsSidewaysCollectors[bitsUpto] = dims[dim].sidewaysCollector;
bitsUpto++;
}
}
/*
* System.out.println("\nbaseDocID=" + baseScorer.docID() + " est=" + estBaseHitCount);
* System.out.println(" maxDoc=" + context.reader().maxDoc());
* System.out.println(" maxCost=" + maxCost);
* System.out.println(" dims[0].freq=" + dims[0].freq);
* if (numDims > 1) {
* System.out.println(" dims[1].freq=" + dims[1].freq);
* }
*/
if (bitsUpto > 0 || scoreSubDocsAtOnce || baseQueryCost < drillDownCost / 10)
{
//System.out.println("queryFirst: baseScorer=" + baseScorer + " disis.length=" + disis.length + " bits.length=" + bits.length);
DoQueryFirstScoring(collector, disis, sidewaysCollectors, bits, bitsSidewaysCollectors);
}
else if (numDims > 1 && (dims[1].disi == null || drillDownAdvancedCost < baseQueryCost / 10))
{
//System.out.println("drillDownAdvance");
DoDrillDownAdvanceScoring(collector, disis, sidewaysCollectors);
}
else
{
//System.out.println("union");
DoUnionScoring(collector, disis, sidewaysCollectors);
}
return(false);
}
private void SumValues(IList<MatchingDocs> matchingDocs, bool keepScores, ValueSource valueSource)
{
FakeScorer scorer = new FakeScorer();
IDictionary context = new Dictionary<string, Scorer>();
if (keepScores)
{
context["scorer"] = scorer;
}
IntsRef scratch = new IntsRef();
foreach (MatchingDocs hits in matchingDocs)
{
OrdinalsReader.OrdinalsSegmentReader ords = ordinalsReader.GetReader(hits.Context);
int scoresIdx = 0;
float[] scores = hits.Scores;
FunctionValues functionValues = valueSource.GetValues(context, hits.Context);
DocIdSetIterator docs = hits.Bits.GetIterator();
int doc;
while ((doc = docs.NextDoc()) != DocIdSetIterator.NO_MORE_DOCS)
{
ords.Get(doc, scratch);
if (keepScores)
{
scorer.docID_Renamed = doc;
scorer.score_Renamed = scores[scoresIdx++];
}
float value = (float)functionValues.DoubleVal(doc);
for (int i = 0; i < scratch.Length; i++)
{
values[scratch.Ints[i]] += value;
}
}
}
Rollup();
}
public override bool Score(Collector collector, int max)
{
FakeScorer fakeScorer = new FakeScorer();
collector.Scorer = fakeScorer;
if (_doc == -1)
{
_doc = NextDocOutOfOrder();
}
while (_doc < max)
{
fakeScorer.doc = _doc;
fakeScorer._score = outerInstance._scores[outerInstance._ords[_scoreUpto]];
collector.Collect(_doc);
_doc = NextDocOutOfOrder();
}
return _doc != DocIdSetIterator.NO_MORE_DOCS;
}
/// <summary>
/// Returns the grouped results. Returns null if the
/// number of groups collected is <= groupOffset.
///
/// <para>
/// <b>NOTE</b>: This collector is unable to compute
/// the groupValue per group so it will always be null.
/// This is normally not a problem, as you can obtain the
/// value just like you obtain other values for each
/// matching document (eg, via stored fields, via
/// FieldCache, etc.)
/// </para>
/// </summary>
/// <typeparam name="TGroupValue">The expected return type for group value</typeparam>
/// <<param name="withinGroupSort">
/// The <see cref="Sort"/> used to sort
/// documents within each group. Passing null is
/// allowed, to sort by relevance.
/// </param>
/// <param name="groupOffset">Which group to start from</param>
/// <param name="withinGroupOffset">
/// Which document to start from within each group
/// </param>
/// <param name="maxDocsPerGroup">
/// How many top documents to keep within each group.
/// </param>
/// <param name="fillSortFields">
/// If true then the Comparable values for the sort fields will be set
/// </param>
public virtual ITopGroups <TGroupValue> GetTopGroups <TGroupValue>(Sort withinGroupSort, int groupOffset, int withinGroupOffset, int maxDocsPerGroup, bool fillSortFields)
{
//if (queueFull) {
//System.out.println("getTopGroups groupOffset=" + groupOffset + " topNGroups=" + topNGroups);
//}
if (subDocUpto != 0)
{
ProcessGroup();
}
if (groupOffset >= groupQueue.Size())
{
return(null);
}
int totalGroupedHitCount = 0;
FakeScorer fakeScorer = new FakeScorer();
float maxScore = float.MinValue;
GroupDocs <TGroupValue>[] groups = new GroupDocs <TGroupValue> [groupQueue.Size() - groupOffset];
for (int downTo = groupQueue.Size() - groupOffset - 1; downTo >= 0; downTo--)
{
OneGroup og = groupQueue.Pop();
// At this point we hold all docs w/ in each group,
// unsorted; we now sort them:
ITopDocsCollector collector;
if (withinGroupSort == null)
{
// Sort by score
if (!needsScores)
{
throw new ArgumentException("cannot sort by relevance within group: needsScores=false");
}
collector = TopScoreDocCollector.Create(maxDocsPerGroup, true);
}
else
{
// Sort by fields
collector = TopFieldCollector.Create(withinGroupSort, maxDocsPerGroup, fillSortFields, needsScores, needsScores, true);
}
collector.Scorer = fakeScorer;
collector.NextReader = og.readerContext;
for (int docIDX = 0; docIDX < og.count; docIDX++)
{
int doc = og.docs[docIDX];
fakeScorer.doc = doc;
if (needsScores)
{
fakeScorer.score = og.scores[docIDX];
}
collector.Collect(doc);
}
totalGroupedHitCount += og.count;
object[] groupSortValues;
if (fillSortFields)
{
groupSortValues = new IComparable[comparators.Length];
for (int sortFieldIDX = 0; sortFieldIDX < comparators.Length; sortFieldIDX++)
{
groupSortValues[sortFieldIDX] = comparators[sortFieldIDX].Value(og.comparatorSlot);
}
}
else
{
groupSortValues = null;
}
TopDocs topDocs = collector.TopDocs(withinGroupOffset, maxDocsPerGroup);
// TODO: we could aggregate scores across children
// by Sum/Avg instead of passing NaN:
groups[downTo] = new GroupDocs <TGroupValue>(float.NaN,
topDocs.MaxScore,
og.count,
topDocs.ScoreDocs,
default(TGroupValue),
groupSortValues);
maxScore = Math.Max(maxScore, topDocs.MaxScore);
}
/*
* while (groupQueue.size() != 0) {
* final OneGroup og = groupQueue.pop();
* //System.out.println(" leftover: og ord=" + og.groupOrd + " count=" + og.count);
* totalGroupedHitCount += og.count;
* }
*/
return(new TopGroups <TGroupValue>(new TopGroups <TGroupValue>(groupSort.GetSort(),
withinGroupSort == null ? null : withinGroupSort.GetSort(),
totalHitCount, totalGroupedHitCount, groups, maxScore),
totalGroupCount));
}
