private ParsedDocument parseFetchedDoc(
PercolateContext context,
BytesReference fetchedDoc,
IndexService documentIndexService,
String type) {
ParsedDocument doc = null;
XContentParser parser = null;
try {
parser = XContentFactory.xContent(fetchedDoc).createParser(fetchedDoc);
MapperService mapperService = documentIndexService.mapperService();
DocumentMapper docMapper = mapperService.documentMapperWithAutoCreate(type);
doc = docMapper.parse(source(parser).type(type).flyweight(true));
if (context.highlight() != null) {
doc.setSource(fetchedDoc);
}
} catch (Throwable e) {
throw new ElasticsearchParseException("failed to parse request", e);
} finally {
if (parser != null) {
parser.close();
}
}
if (doc == null) {
throw new ElasticsearchParseException("No doc to percolate in the request");
}
return doc;
}
@Override
public PercolateShardResponse doPercolate(
PercolateShardRequest request, PercolateContext context) {
long count = 0;
List<BytesRef> matches = new ArrayList<>();
List<Map<String, HighlightField>> hls = new ArrayList<>();
Lucene.ExistsCollector collector = new Lucene.ExistsCollector();
for (Map.Entry<HashedBytesRef, Query> entry : context.percolateQueries().entrySet()) {
collector.reset();
if (context.highlight() != null) {
context.parsedQuery(
new ParsedQuery(entry.getValue(), ImmutableMap.<String, Filter>of()));
context.hitContext().cache().clear();
}
try {
context.docSearcher().search(entry.getValue(), collector);
} catch (Throwable e) {
logger.debug("[" + entry.getKey() + "] failed to execute query", e);
throw new PercolateException(context.indexShard().shardId(), "failed to execute", e);
}
if (collector.exists()) {
if (!context.limit || count < context.size()) {
matches.add(entry.getKey().bytes);
if (context.highlight() != null) {
highlightPhase.hitExecute(context, context.hitContext());
hls.add(context.hitContext().hit().getHighlightFields());
}
}
count++;
}
}
BytesRef[] finalMatches = matches.toArray(new BytesRef[matches.size()]);
return new PercolateShardResponse(
finalMatches, hls, count, context, request.index(), request.shardId());
}
@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 ParsedDocument parseRequest(
IndexService documentIndexService, PercolateShardRequest request, PercolateContext context)
throws ElasticsearchException {
BytesReference source = request.source();
if (source == null || source.length() == 0) {
return null;
}
// TODO: combine all feature parse elements into one map
Map<String, ? extends SearchParseElement> hlElements = highlightPhase.parseElements();
Map<String, ? extends SearchParseElement> facetElements = facetPhase.parseElements();
Map<String, ? extends SearchParseElement> aggregationElements =
aggregationPhase.parseElements();
ParsedDocument doc = null;
XContentParser parser = null;
// Some queries (function_score query when for decay functions) rely on a SearchContext being
// set:
// We switch types because this context needs to be in the context of the percolate queries in
// the shard and
// not the in memory percolate doc
String[] previousTypes = context.types();
context.types(new String[] {TYPE_NAME});
SearchContext.setCurrent(context);
try {
parser = XContentFactory.xContent(source).createParser(source);
String currentFieldName = null;
XContentParser.Token token;
while ((token = parser.nextToken()) != XContentParser.Token.END_OBJECT) {
if (token == XContentParser.Token.FIELD_NAME) {
currentFieldName = parser.currentName();
// we need to check the "doc" here, so the next token will be START_OBJECT which is
// the actual document starting
if ("doc".equals(currentFieldName)) {
if (doc != null) {
throw new ElasticsearchParseException("Either specify doc or get, not both");
}
MapperService mapperService = documentIndexService.mapperService();
DocumentMapper docMapper =
mapperService.documentMapperWithAutoCreate(request.documentType());
doc = docMapper.parse(source(parser).type(request.documentType()).flyweight(true));
// the document parsing exists the "doc" object, so we need to set the new current
// field.
currentFieldName = parser.currentName();
}
} else if (token == XContentParser.Token.START_OBJECT) {
SearchParseElement element = hlElements.get(currentFieldName);
if (element == null) {
element = facetElements.get(currentFieldName);
if (element == null) {
element = aggregationElements.get(currentFieldName);
}
}
if ("query".equals(currentFieldName)) {
if (context.percolateQuery() != null) {
throw new ElasticsearchParseException("Either specify query or filter, not both");
}
context.percolateQuery(documentIndexService.queryParserService().parse(parser).query());
} else if ("filter".equals(currentFieldName)) {
if (context.percolateQuery() != null) {
throw new ElasticsearchParseException("Either specify query or filter, not both");
}
Filter filter =
documentIndexService.queryParserService().parseInnerFilter(parser).filter();
context.percolateQuery(new XConstantScoreQuery(filter));
} else if ("sort".equals(currentFieldName)) {
parseSort(parser, context);
} else if (element != null) {
element.parse(parser, context);
}
} else if (token == XContentParser.Token.START_ARRAY) {
if ("sort".equals(currentFieldName)) {
parseSort(parser, context);
}
} else if (token == null) {
break;
} else if (token.isValue()) {
if ("size".equals(currentFieldName)) {
context.size(parser.intValue());
if (context.size() < 0) {
throw new ElasticsearchParseException(
"size is set to ["
+ context.size()
+ "] and is expected to be higher or equal to 0");
}
} else if ("sort".equals(currentFieldName)) {
parseSort(parser, context);
} else if ("track_scores".equals(currentFieldName)
|| "trackScores".equals(currentFieldName)) {
context.trackScores(parser.booleanValue());
}
}
}
// We need to get the actual source from the request body for highlighting, so parse the
// request body again
// and only get the doc source.
if (context.highlight() != null) {
parser.close();
currentFieldName = null;
parser = XContentFactory.xContent(source).createParser(source);
token = parser.nextToken();
assert token == XContentParser.Token.START_OBJECT;
while ((token = parser.nextToken()) != XContentParser.Token.END_OBJECT) {
if (token == XContentParser.Token.FIELD_NAME) {
currentFieldName = parser.currentName();
} else if (token == XContentParser.Token.START_OBJECT) {
if ("doc".equals(currentFieldName)) {
BytesStreamOutput bStream = new BytesStreamOutput();
XContentBuilder builder = XContentFactory.contentBuilder(XContentType.SMILE, bStream);
builder.copyCurrentStructure(parser);
builder.close();
doc.setSource(bStream.bytes());
break;
} else {
parser.skipChildren();
}
} else if (token == null) {
break;
}
}
}
} catch (Throwable e) {
throw new ElasticsearchParseException("failed to parse request", e);
} finally {
context.types(previousTypes);
SearchContext.removeCurrent();
if (parser != null) {
parser.close();
}
}
return doc;
}
public PercolateShardResponse percolate(PercolateShardRequest request) {
IndexService percolateIndexService = indicesService.indexServiceSafe(request.index());
IndexShard indexShard = percolateIndexService.shardSafe(request.shardId());
ShardPercolateService shardPercolateService = indexShard.shardPercolateService();
shardPercolateService.prePercolate();
long startTime = System.nanoTime();
SearchShardTarget searchShardTarget =
new SearchShardTarget(clusterService.localNode().id(), request.index(), request.shardId());
final PercolateContext context =
new PercolateContext(
request,
searchShardTarget,
indexShard,
percolateIndexService,
cacheRecycler,
pageCacheRecycler,
bigArrays,
scriptService);
try {
ParsedDocument parsedDocument = parseRequest(percolateIndexService, request, context);
if (context.percolateQueries().isEmpty()) {
return new PercolateShardResponse(context, request.index(), request.shardId());
}
if (request.docSource() != null && request.docSource().length() != 0) {
parsedDocument =
parseFetchedDoc(
context, request.docSource(), percolateIndexService, request.documentType());
} else if (parsedDocument == null) {
throw new ElasticsearchIllegalArgumentException("Nothing to percolate");
}
if (context.percolateQuery() == null
&& (context.trackScores()
|| context.doSort
|| context.facets() != null
|| context.aggregations() != null)) {
context.percolateQuery(new MatchAllDocsQuery());
}
if (context.doSort && !context.limit) {
throw new ElasticsearchIllegalArgumentException("Can't sort if size isn't specified");
}
if (context.highlight() != null && !context.limit) {
throw new ElasticsearchIllegalArgumentException("Can't highlight if size isn't specified");
}
if (context.size() < 0) {
context.size(0);
}
// parse the source either into one MemoryIndex, if it is a single document or index multiple
// docs if nested
PercolatorIndex percolatorIndex;
if (indexShard.mapperService().documentMapper(request.documentType()).hasNestedObjects()) {
percolatorIndex = multi;
} else {
percolatorIndex = single;
}
PercolatorType action;
if (request.onlyCount()) {
action = context.percolateQuery() != null ? queryCountPercolator : countPercolator;
} else {
if (context.doSort) {
action = topMatchingPercolator;
} else if (context.percolateQuery() != null) {
action = context.trackScores() ? scoringPercolator : queryPercolator;
} else {
action = matchPercolator;
}
}
context.percolatorTypeId = action.id();
percolatorIndex.prepare(context, parsedDocument);
indexShard.readAllowed();
return action.doPercolate(request, context);
} finally {
context.release();
shardPercolateService.postPercolate(System.nanoTime() - startTime);
}
}