@Override
public void transform(SolrDocument doc, int docid) {
// This is only good for random-access functions
try {
// TODO: calculate this stuff just once across diff functions
int idx = ReaderUtil.subIndex(docid, readerContexts);
LeafReaderContext rcontext = readerContexts.get(idx);
FunctionValues values = docValuesArr[idx];
if (values == null) {
docValuesArr[idx] = values = valueSource.getValues(fcontext, rcontext);
}
int localId = docid - rcontext.docBase;
setValue(doc, values.objectVal(localId));
} catch (IOException e) {
throw new SolrException(
SolrException.ErrorCode.SERVER_ERROR,
"exception at docid " + docid + " for valuesource " + valueSource,
e);
}
}
@Override
public void execute(SearchContext context) {
FieldsVisitor fieldsVisitor;
Set<String> fieldNames = null;
List<String> extractFieldNames = null;
boolean loadAllStored = false;
if (!context.hasFieldNames()) {
// no fields specified, default to return source if no explicit indication
if (!context.hasScriptFields() && !context.hasFetchSourceContext()) {
context.fetchSourceContext(new FetchSourceContext(true));
}
fieldsVisitor = new FieldsVisitor(context.sourceRequested());
} else if (context.fieldNames().isEmpty()) {
fieldsVisitor = new FieldsVisitor(context.sourceRequested());
} else {
for (String fieldName : context.fieldNames()) {
if (fieldName.equals("*")) {
loadAllStored = true;
continue;
}
if (fieldName.equals(SourceFieldMapper.NAME)) {
if (context.hasFetchSourceContext()) {
context.fetchSourceContext().fetchSource(true);
} else {
context.fetchSourceContext(new FetchSourceContext(true));
}
continue;
}
MappedFieldType fieldType = context.smartNameFieldType(fieldName);
if (fieldType == null) {
// Only fail if we know it is a object field, missing paths / fields shouldn't fail.
if (context.getObjectMapper(fieldName) != null) {
throw new IllegalArgumentException("field [" + fieldName + "] isn't a leaf field");
}
} else if (fieldType.stored()) {
if (fieldNames == null) {
fieldNames = new HashSet<>();
}
fieldNames.add(fieldType.names().indexName());
} else {
if (extractFieldNames == null) {
extractFieldNames = newArrayList();
}
extractFieldNames.add(fieldName);
}
}
if (loadAllStored) {
fieldsVisitor = new AllFieldsVisitor(); // load everything, including _source
} else if (fieldNames != null) {
boolean loadSource = extractFieldNames != null || context.sourceRequested();
fieldsVisitor = new CustomFieldsVisitor(fieldNames, loadSource);
} else {
fieldsVisitor = new FieldsVisitor(extractFieldNames != null || context.sourceRequested());
}
}
InternalSearchHit[] hits = new InternalSearchHit[context.docIdsToLoadSize()];
FetchSubPhase.HitContext hitContext = new FetchSubPhase.HitContext();
for (int index = 0; index < context.docIdsToLoadSize(); index++) {
int docId = context.docIdsToLoad()[context.docIdsToLoadFrom() + index];
int readerIndex = ReaderUtil.subIndex(docId, context.searcher().getIndexReader().leaves());
LeafReaderContext subReaderContext =
context.searcher().getIndexReader().leaves().get(readerIndex);
int subDocId = docId - subReaderContext.docBase;
final InternalSearchHit searchHit;
try {
int rootDocId = findRootDocumentIfNested(context, subReaderContext, subDocId);
if (rootDocId != -1) {
searchHit =
createNestedSearchHit(
context,
docId,
subDocId,
rootDocId,
extractFieldNames,
loadAllStored,
fieldNames,
subReaderContext);
} else {
searchHit =
createSearchHit(
context, fieldsVisitor, docId, subDocId, extractFieldNames, subReaderContext);
}
} catch (IOException e) {
throw ExceptionsHelper.convertToElastic(e);
}
hits[index] = searchHit;
hitContext.reset(searchHit, subReaderContext, subDocId, context.searcher());
for (FetchSubPhase fetchSubPhase : fetchSubPhases) {
if (fetchSubPhase.hitExecutionNeeded(context)) {
fetchSubPhase.hitExecute(context, hitContext);
}
}
}
for (FetchSubPhase fetchSubPhase : fetchSubPhases) {
if (fetchSubPhase.hitsExecutionNeeded(context)) {
fetchSubPhase.hitsExecute(context, hits);
}
}
context
.fetchResult()
.hits(
new InternalSearchHits(
hits,
context.queryResult().topDocs().totalHits,
context.queryResult().topDocs().getMaxScore()));
}
@Override
public PercolateShardResponse doPercolate(
PercolateShardRequest request, PercolateContext context) {
Engine.Searcher percolatorSearcher = context.indexShard().acquireSearcher("percolate");
try {
MatchAndSort matchAndSort = QueryCollector.matchAndSort(logger, context);
queryBasedPercolating(percolatorSearcher, context, matchAndSort);
TopDocs topDocs = matchAndSort.topDocs();
long count = topDocs.totalHits;
List<BytesRef> matches = new ArrayList<BytesRef>(topDocs.scoreDocs.length);
float[] scores = new float[topDocs.scoreDocs.length];
List<Map<String, HighlightField>> hls = null;
if (context.highlight() != null) {
hls = new ArrayList<Map<String, HighlightField>>(topDocs.scoreDocs.length);
}
final FieldMapper<?> idMapper =
context.mapperService().smartNameFieldMapper(IdFieldMapper.NAME);
final IndexFieldData<?> idFieldData = context.fieldData().getForField(idMapper);
int i = 0;
final HashedBytesRef spare = new HashedBytesRef(new BytesRef());
for (ScoreDoc scoreDoc : topDocs.scoreDocs) {
int segmentIdx =
ReaderUtil.subIndex(scoreDoc.doc, percolatorSearcher.reader().leaves());
AtomicReaderContext atomicReaderContext =
percolatorSearcher.reader().leaves().get(segmentIdx);
BytesValues values = idFieldData.load(atomicReaderContext).getBytesValues(true);
final int localDocId = scoreDoc.doc - atomicReaderContext.docBase;
final int numValues = values.setDocument(localDocId);
assert numValues == 1;
spare.bytes = values.nextValue();
spare.hash = values.currentValueHash();
matches.add(values.copyShared());
if (hls != null) {
Query query = context.percolateQueries().get(spare);
context.parsedQuery(new ParsedQuery(query, ImmutableMap.<String, Filter>of()));
context.hitContext().cache().clear();
highlightPhase.hitExecute(context, context.hitContext());
hls.add(i, context.hitContext().hit().getHighlightFields());
}
scores[i++] = scoreDoc.score;
}
if (hls != null) {
return new PercolateShardResponse(
matches.toArray(new BytesRef[matches.size()]),
hls,
count,
scores,
context,
request.index(),
request.shardId());
} else {
return new PercolateShardResponse(
matches.toArray(new BytesRef[matches.size()]),
count,
scores,
context,
request.index(),
request.shardId());
}
} catch (Throwable e) {
logger.debug("failed to execute", e);
throw new PercolateException(context.indexShard().shardId(), "failed to execute", e);
} finally {
percolatorSearcher.release();
}
}
private Map<Integer, Object> highlightField(
String field,
String contents[],
BreakIterator bi,
BytesRef terms[],
int[] docids,
List<AtomicReaderContext> leaves,
int maxPassages,
Query query)
throws IOException {
Map<Integer, Object> highlights = new HashMap<>();
PassageFormatter fieldFormatter = getFormatter(field);
if (fieldFormatter == null) {
throw new NullPointerException("PassageFormatter cannot be null");
}
// check if we should do any multiterm processing
Analyzer analyzer = getIndexAnalyzer(field);
CharacterRunAutomaton automata[] = new CharacterRunAutomaton[0];
if (analyzer != null) {
automata = MultiTermHighlighting.extractAutomata(query, field);
}
// resize 'terms', where the last term is the multiterm matcher
if (automata.length > 0) {
BytesRef newTerms[] = new BytesRef[terms.length + 1];
System.arraycopy(terms, 0, newTerms, 0, terms.length);
terms = newTerms;
}
// we are processing in increasing docid order, so we only need to reinitialize stuff on segment
// changes
// otherwise, we will just advance() existing enums to the new document in the same segment.
DocsAndPositionsEnum postings[] = null;
TermsEnum termsEnum = null;
int lastLeaf = -1;
for (int i = 0; i < docids.length; i++) {
String content = contents[i];
if (content.length() == 0) {
continue; // nothing to do
}
bi.setText(content);
int doc = docids[i];
int leaf = ReaderUtil.subIndex(doc, leaves);
AtomicReaderContext subContext = leaves.get(leaf);
AtomicReader r = subContext.reader();
assert leaf >= lastLeaf; // increasing order
// if the segment has changed, we must initialize new enums.
if (leaf != lastLeaf) {
Terms t = r.terms(field);
if (t != null) {
termsEnum = t.iterator(null);
postings = new DocsAndPositionsEnum[terms.length];
}
}
if (termsEnum == null) {
continue; // no terms for this field, nothing to do
}
// if there are multi-term matches, we have to initialize the "fake" enum for each document
if (automata.length > 0) {
DocsAndPositionsEnum dp =
MultiTermHighlighting.getDocsEnum(analyzer.tokenStream(field, content), automata);
dp.advance(doc - subContext.docBase);
postings[terms.length - 1] = dp; // last term is the multiterm matcher
}
Passage passages[] =
highlightDoc(
field,
terms,
content.length(),
bi,
doc - subContext.docBase,
termsEnum,
postings,
maxPassages);
if (passages.length == 0) {
// no passages were returned, so ask for a default summary
passages = getEmptyHighlight(field, bi, maxPassages);
}
if (passages.length > 0) {
highlights.put(doc, fieldFormatter.format(passages, content));
}
lastLeaf = leaf;
}
return highlights;
}
int resolveParentDocuments(
TopDocs topDocs,
SearchContext context,
Recycler.V<ObjectObjectOpenHashMap<Object, ParentDoc[]>> parentDocs) {
int parentHitsResolved = 0;
Recycler.V<ObjectObjectOpenHashMap<Object, Recycler.V<IntObjectOpenHashMap<ParentDoc>>>>
parentDocsPerReader =
cacheRecycler.hashMap(context.searcher().getIndexReader().leaves().size());
for (ScoreDoc scoreDoc : topDocs.scoreDocs) {
int readerIndex =
ReaderUtil.subIndex(scoreDoc.doc, context.searcher().getIndexReader().leaves());
AtomicReaderContext subContext =
context.searcher().getIndexReader().leaves().get(readerIndex);
int subDoc = scoreDoc.doc - subContext.docBase;
// find the parent id
HashedBytesArray parentId =
context.idCache().reader(subContext.reader()).parentIdByDoc(parentType, subDoc);
if (parentId == null) {
// no parent found
continue;
}
// now go over and find the parent doc Id and reader tuple
for (AtomicReaderContext atomicReaderContext : context.searcher().getIndexReader().leaves()) {
AtomicReader indexReader = atomicReaderContext.reader();
int parentDocId = context.idCache().reader(indexReader).docById(parentType, parentId);
Bits liveDocs = indexReader.getLiveDocs();
if (parentDocId != -1 && (liveDocs == null || liveDocs.get(parentDocId))) {
// we found a match, add it and break
Recycler.V<IntObjectOpenHashMap<ParentDoc>> readerParentDocs =
parentDocsPerReader.v().get(indexReader.getCoreCacheKey());
if (readerParentDocs == null) {
readerParentDocs = cacheRecycler.intObjectMap(indexReader.maxDoc());
parentDocsPerReader.v().put(indexReader.getCoreCacheKey(), readerParentDocs);
}
ParentDoc parentDoc = readerParentDocs.v().get(parentDocId);
if (parentDoc == null) {
parentHitsResolved++; // we have a hit on a parent
parentDoc = new ParentDoc();
parentDoc.docId = parentDocId;
parentDoc.count = 1;
parentDoc.maxScore = scoreDoc.score;
parentDoc.sumScores = scoreDoc.score;
readerParentDocs.v().put(parentDocId, parentDoc);
} else {
parentDoc.count++;
parentDoc.sumScores += scoreDoc.score;
if (scoreDoc.score > parentDoc.maxScore) {
parentDoc.maxScore = scoreDoc.score;
}
}
}
}
}
boolean[] states = parentDocsPerReader.v().allocated;
Object[] keys = parentDocsPerReader.v().keys;
Object[] values = parentDocsPerReader.v().values;
for (int i = 0; i < states.length; i++) {
if (states[i]) {
Recycler.V<IntObjectOpenHashMap<ParentDoc>> value =
(Recycler.V<IntObjectOpenHashMap<ParentDoc>>) values[i];
ParentDoc[] _parentDocs = value.v().values().toArray(ParentDoc.class);
Arrays.sort(_parentDocs, PARENT_DOC_COMP);
parentDocs.v().put(keys[i], _parentDocs);
Releasables.release(value);
}
}
Releasables.release(parentDocsPerReader);
return parentHitsResolved;
}
public void handleMergeFields(ResponseBuilder rb, SolrIndexSearcher searcher)
throws IOException {
SolrQueryRequest req = rb.req;
SolrQueryResponse rsp = rb.rsp;
// The query cache doesn't currently store sort field values, and SolrIndexSearcher doesn't
// currently have an option to return sort field values. Because of this, we
// take the documents given and re-derive the sort values.
//
// TODO: See SOLR-5595
boolean fsv = req.getParams().getBool(ResponseBuilder.FIELD_SORT_VALUES, false);
if (fsv) {
NamedList<Object[]> sortVals = new NamedList<>(); // order is important for the sort fields
IndexReaderContext topReaderContext = searcher.getTopReaderContext();
List<LeafReaderContext> leaves = topReaderContext.leaves();
LeafReaderContext currentLeaf = null;
if (leaves.size() == 1) {
// if there is a single segment, use that subReader and avoid looking up each time
currentLeaf = leaves.get(0);
leaves = null;
}
DocList docList = rb.getResults().docList;
// sort ids from lowest to highest so we can access them in order
int nDocs = docList.size();
final long[] sortedIds = new long[nDocs];
final float[] scores = new float[nDocs]; // doc scores, parallel to sortedIds
DocList docs = rb.getResults().docList;
DocIterator it = docs.iterator();
for (int i = 0; i < nDocs; i++) {
sortedIds[i] = (((long) it.nextDoc()) << 32) | i;
scores[i] = docs.hasScores() ? it.score() : Float.NaN;
}
// sort ids and scores together
new InPlaceMergeSorter() {
@Override
protected void swap(int i, int j) {
long tmpId = sortedIds[i];
float tmpScore = scores[i];
sortedIds[i] = sortedIds[j];
scores[i] = scores[j];
sortedIds[j] = tmpId;
scores[j] = tmpScore;
}
@Override
protected int compare(int i, int j) {
return Long.compare(sortedIds[i], sortedIds[j]);
}
}.sort(0, sortedIds.length);
SortSpec sortSpec = rb.getSortSpec();
Sort sort = searcher.weightSort(sortSpec.getSort());
SortField[] sortFields =
sort == null ? new SortField[] {SortField.FIELD_SCORE} : sort.getSort();
List<SchemaField> schemaFields = sortSpec.getSchemaFields();
for (int fld = 0; fld < schemaFields.size(); fld++) {
SchemaField schemaField = schemaFields.get(fld);
FieldType ft = null == schemaField ? null : schemaField.getType();
SortField sortField = sortFields[fld];
SortField.Type type = sortField.getType();
// :TODO: would be simpler to always serialize every position of SortField[]
if (type == SortField.Type.SCORE || type == SortField.Type.DOC) continue;
FieldComparator<?> comparator = null;
LeafFieldComparator leafComparator = null;
Object[] vals = new Object[nDocs];
int lastIdx = -1;
int idx = 0;
for (int i = 0; i < sortedIds.length; ++i) {
long idAndPos = sortedIds[i];
float score = scores[i];
int doc = (int) (idAndPos >>> 32);
int position = (int) idAndPos;
if (leaves != null) {
idx = ReaderUtil.subIndex(doc, leaves);
currentLeaf = leaves.get(idx);
if (idx != lastIdx) {
// we switched segments. invalidate comparator.
comparator = null;
}
}
if (comparator == null) {
comparator = sortField.getComparator(1, 0);
leafComparator = comparator.getLeafComparator(currentLeaf);
}
doc -= currentLeaf.docBase; // adjust for what segment this is in
leafComparator.setScorer(new FakeScorer(doc, score));
leafComparator.copy(0, doc);
Object val = comparator.value(0);
if (null != ft) val = ft.marshalSortValue(val);
vals[position] = val;
}
sortVals.add(sortField.getField(), vals);
}
rsp.add("merge_values", sortVals);
}
}