protected void doPrefetch(ResponseBuilder rb) throws IOException {
SolrQueryRequest req = rb.req;
SolrQueryResponse rsp = rb.rsp;
// pre-fetch returned documents
if (!req.getParams().getBool(ShardParams.IS_SHARD, false)
&& rb.getResults().docList != null
&& rb.getResults().docList.size() <= 50) {
// TODO: this may depend on the highlighter component (or other components?)
SolrPluginUtils.optimizePreFetchDocs(rb.getResults().docList, rb.getQuery(), req, rsp);
}
}
private void returnFields(ResponseBuilder rb, ShardRequest sreq) {
// Keep in mind that this could also be a shard in a multi-tiered system.
// TODO: if a multi-tiered system, it seems like some requests
// could/should bypass middlemen (like retrieving stored fields)
// TODO: merge fsv to if requested
if ((sreq.purpose & ShardRequest.PURPOSE_GET_FIELDS) != 0) {
boolean returnScores = (rb.getFieldFlags() & SolrIndexSearcher.GET_SCORES) != 0;
assert (sreq.responses.size() == 1);
ShardResponse srsp = sreq.responses.get(0);
SolrDocumentList docs =
(SolrDocumentList) srsp.getSolrResponse().getResponse().get("response");
String keyFieldName = rb.req.getSchema().getUniqueKeyField().getName();
for (SolrDocument doc : docs) {
Object id = doc.getFieldValue(keyFieldName);
ShardDoc sdoc = rb.resultIds.get(id.toString());
if (sdoc != null) {
if (returnScores && sdoc.score != null) {
doc.setField("score", sdoc.score);
}
rb._responseDocs.set(sdoc.positionInResponse, doc);
}
}
}
}
private void createRetrieveDocs(ResponseBuilder rb) {
// TODO: in a system with nTiers > 2, we could be passed "ids" here
// unless those requests always go to the final destination shard
// for each shard, collect the documents for that shard.
HashMap<String, Collection<ShardDoc>> shardMap = new HashMap<String, Collection<ShardDoc>>();
for (ShardDoc sdoc : rb.resultIds.values()) {
Collection<ShardDoc> shardDocs = shardMap.get(sdoc.shard);
if (shardDocs == null) {
shardDocs = new ArrayList<ShardDoc>();
shardMap.put(sdoc.shard, shardDocs);
}
shardDocs.add(sdoc);
}
SchemaField uniqueField = rb.req.getSchema().getUniqueKeyField();
// Now create a request for each shard to retrieve the stored fields
for (Collection<ShardDoc> shardDocs : shardMap.values()) {
ShardRequest sreq = new ShardRequest();
sreq.purpose = ShardRequest.PURPOSE_GET_FIELDS;
sreq.shards = new String[] {shardDocs.iterator().next().shard};
sreq.params = new ModifiableSolrParams();
// add original params
sreq.params.add(rb.req.getParams());
// no need for a sort, we already have order
sreq.params.remove(CommonParams.SORT);
// we already have the field sort values
sreq.params.remove(ResponseBuilder.FIELD_SORT_VALUES);
// make sure that the id is returned for correlation.
String fl = sreq.params.get(CommonParams.FL);
if (fl != null) {
fl = fl.trim();
// currently, "score" is synonymous with "*,score" so
// don't add "id" if the fl is empty or "score" or it would change the meaning.
if (fl.length() != 0 && !"score".equals(fl) && !"*".equals(fl)) {
sreq.params.set(CommonParams.FL, fl + ',' + uniqueField.getName());
}
}
ArrayList<String> ids = new ArrayList<String>(shardDocs.size());
for (ShardDoc shardDoc : shardDocs) {
// TODO: depending on the type, we may need more tha a simple toString()?
ids.add(shardDoc.id.toString());
}
sreq.params.add(ShardParams.IDS, StrUtils.join(ids, ','));
rb.addRequest(this, sreq);
}
}
private void createMainQuery(ResponseBuilder rb) {
ShardRequest sreq = new ShardRequest();
sreq.purpose = ShardRequest.PURPOSE_GET_TOP_IDS;
sreq.params = new ModifiableSolrParams(rb.req.getParams());
// TODO: base on current params or original params?
// don't pass through any shards param
sreq.params.remove(ShardParams.SHARDS);
// set the start (offset) to 0 for each shard request so we can properly merge
// results from the start.
if (rb.shards_start > -1) {
// if the client set shards.start set this explicitly
sreq.params.set(CommonParams.START, rb.shards_start);
} else {
sreq.params.set(CommonParams.START, "0");
}
// TODO: should we even use the SortSpec? That's obtained from the QParser, and
// perhaps we shouldn't attempt to parse the query at this level?
// Alternate Idea: instead of specifying all these things at the upper level,
// we could just specify that this is a shard request.
if (rb.shards_rows > -1) {
// if the client set shards.rows set this explicity
sreq.params.set(CommonParams.ROWS, rb.shards_rows);
} else {
sreq.params.set(
CommonParams.ROWS, rb.getSortSpec().getOffset() + rb.getSortSpec().getCount());
}
// in this first phase, request only the unique key field
// and any fields needed for merging.
sreq.params.set(ResponseBuilder.FIELD_SORT_VALUES, "true");
if ((rb.getFieldFlags() & SolrIndexSearcher.GET_SCORES) != 0
|| rb.getSortSpec().includesScore()) {
sreq.params.set(CommonParams.FL, rb.req.getSchema().getUniqueKeyField().getName() + ",score");
} else {
sreq.params.set(CommonParams.FL, rb.req.getSchema().getUniqueKeyField().getName());
}
rb.addRequest(this, sreq);
}
@Override
public void prepare(ResponseBuilder rb) throws IOException {
SolrQueryRequest req = rb.req;
SolrParams params = req.getParams();
if (!params.getBool(COMPONENT_NAME, true)) {
return;
}
SolrQueryResponse rsp = rb.rsp;
// Set field flags
String fl = params.get(CommonParams.FL);
int fieldFlags = 0;
if (fl != null) {
fieldFlags |= SolrPluginUtils.setReturnFields(fl, rsp);
}
rb.setFieldFlags(fieldFlags);
String defType = params.get(QueryParsing.DEFTYPE, QParserPlugin.DEFAULT_QTYPE);
// get it from the response builder to give a different component a chance
// to set it.
String queryString = rb.getQueryString();
if (queryString == null) {
// this is the normal way it's set.
queryString = params.get(CommonParams.Q);
rb.setQueryString(queryString);
}
try {
QParser parser = QParser.getParser(rb.getQueryString(), defType, req);
Query q = parser.getQuery();
if (q == null) {
// normalize a null query to a query that matches nothing
q = new BooleanQuery();
}
rb.setQuery(q);
rb.setSortSpec(parser.getSort(true));
rb.setQparser(parser);
String[] fqs = req.getParams().getParams(CommonParams.FQ);
if (fqs != null && fqs.length != 0) {
List<Query> filters = rb.getFilters();
if (filters == null) {
filters = new ArrayList<Query>();
rb.setFilters(filters);
}
for (String fq : fqs) {
if (fq != null && fq.trim().length() != 0) {
QParser fqp = QParser.getParser(fq, null, req);
filters.add(fqp.getQuery());
}
}
}
} catch (ParseException e) {
throw new SolrException(SolrException.ErrorCode.BAD_REQUEST, e);
}
// TODO: temporary... this should go in a different component.
String shards = params.get(ShardParams.SHARDS);
if (shards != null) {
List<String> lst = StrUtils.splitSmart(shards, ",", true);
rb.shards = lst.toArray(new String[lst.size()]);
}
String shards_rows = params.get(ShardParams.SHARDS_ROWS);
if (shards_rows != null) {
rb.shards_rows = Integer.parseInt(shards_rows);
}
String shards_start = params.get(ShardParams.SHARDS_START);
if (shards_start != null) {
rb.shards_start = Integer.parseInt(shards_start);
}
}
private void mergeIds(ResponseBuilder rb, ShardRequest sreq) {
SortSpec ss = rb.getSortSpec();
Sort sort = ss.getSort();
SortField[] sortFields = null;
if (sort != null) sortFields = sort.getSort();
else {
sortFields = new SortField[] {SortField.FIELD_SCORE};
}
SchemaField uniqueKeyField = rb.req.getSchema().getUniqueKeyField();
// id to shard mapping, to eliminate any accidental dups
HashMap<Object, String> uniqueDoc = new HashMap<Object, String>();
// Merge the docs via a priority queue so we don't have to sort *all* of the
// documents... we only need to order the top (rows+start)
ShardFieldSortedHitQueue queue;
queue = new ShardFieldSortedHitQueue(sortFields, ss.getOffset() + ss.getCount());
long numFound = 0;
Float maxScore = null;
for (ShardResponse srsp : sreq.responses) {
SolrDocumentList docs =
(SolrDocumentList) srsp.getSolrResponse().getResponse().get("response");
// calculate global maxScore and numDocsFound
if (docs.getMaxScore() != null) {
maxScore = maxScore == null ? docs.getMaxScore() : Math.max(maxScore, docs.getMaxScore());
}
numFound += docs.getNumFound();
NamedList sortFieldValues =
(NamedList) (srsp.getSolrResponse().getResponse().get("sort_values"));
// go through every doc in this response, construct a ShardDoc, and
// put it in the priority queue so it can be ordered.
for (int i = 0; i < docs.size(); i++) {
SolrDocument doc = docs.get(i);
Object id = doc.getFieldValue(uniqueKeyField.getName());
String prevShard = uniqueDoc.put(id, srsp.getShard());
if (prevShard != null) {
// duplicate detected
numFound--;
// For now, just always use the first encountered since we can't currently
// remove the previous one added to the priority queue. If we switched
// to the Java5 PriorityQueue, this would be easier.
continue;
// make which duplicate is used deterministic based on shard
// if (prevShard.compareTo(srsp.shard) >= 0) {
// TODO: remove previous from priority queue
// continue;
// }
}
ShardDoc shardDoc = new ShardDoc();
shardDoc.id = id;
shardDoc.shard = srsp.getShard();
shardDoc.orderInShard = i;
Object scoreObj = doc.getFieldValue("score");
if (scoreObj != null) {
if (scoreObj instanceof String) {
shardDoc.score = Float.parseFloat((String) scoreObj);
} else {
shardDoc.score = (Float) scoreObj;
}
}
shardDoc.sortFieldValues = sortFieldValues;
queue.insertWithOverflow(shardDoc);
} // end for-each-doc-in-response
} // end for-each-response
// The queue now has 0 -> queuesize docs, where queuesize <= start + rows
// So we want to pop the last documents off the queue to get
// the docs offset -> queuesize
int resultSize = queue.size() - ss.getOffset();
resultSize = Math.max(0, resultSize); // there may not be any docs in range
Map<Object, ShardDoc> resultIds = new HashMap<Object, ShardDoc>();
for (int i = resultSize - 1; i >= 0; i--) {
ShardDoc shardDoc = (ShardDoc) queue.pop();
shardDoc.positionInResponse = i;
// Need the toString() for correlation with other lists that must
// be strings (like keys in highlighting, explain, etc)
resultIds.put(shardDoc.id.toString(), shardDoc);
}
SolrDocumentList responseDocs = new SolrDocumentList();
if (maxScore != null) responseDocs.setMaxScore(maxScore);
responseDocs.setNumFound(numFound);
responseDocs.setStart(ss.getOffset());
// size appropriately
for (int i = 0; i < resultSize; i++) responseDocs.add(null);
// save these results in a private area so we can access them
// again when retrieving stored fields.
// TODO: use ResponseBuilder (w/ comments) or the request context?
rb.resultIds = resultIds;
rb._responseDocs = responseDocs;
}
protected void doFieldSortValues(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.
boolean fsv = req.getParams().getBool(ResponseBuilder.FIELD_SORT_VALUES, false);
if (fsv) {
Sort sort = rb.getSortSpec().getSort();
SortField[] sortFields =
sort == null ? new SortField[] {SortField.FIELD_SCORE} : sort.getSort();
NamedList sortVals = new NamedList(); // order is important for the sort fields
Field field = new Field("dummy", "", Field.Store.YES, Field.Index.NO); // a dummy Field
SolrIndexReader reader = searcher.getReader();
SolrIndexReader[] readers = reader.getLeafReaders();
SolrIndexReader subReader = reader;
if (readers.length == 1) {
// if there is a single segment, use that subReader and avoid looking up each time
subReader = readers[0];
readers = null;
}
int[] offsets = reader.getLeafOffsets();
for (SortField sortField : sortFields) {
int type = sortField.getType();
if (type == SortField.SCORE || type == SortField.DOC) continue;
FieldComparator comparator = null;
FieldComparator comparators[] =
(readers == null) ? null : new FieldComparator[readers.length];
String fieldname = sortField.getField();
FieldType ft = fieldname == null ? null : req.getSchema().getFieldTypeNoEx(fieldname);
DocList docList = rb.getResults().docList;
ArrayList<Object> vals = new ArrayList<Object>(docList.size());
DocIterator it = rb.getResults().docList.iterator();
int offset = 0;
int idx = 0;
while (it.hasNext()) {
int doc = it.nextDoc();
if (readers != null) {
idx = SolrIndexReader.readerIndex(doc, offsets);
subReader = readers[idx];
offset = offsets[idx];
comparator = comparators[idx];
}
if (comparator == null) {
comparator = sortField.getComparator(1, 0);
comparator = comparator.setNextReader(subReader, offset);
if (comparators != null) comparators[idx] = comparator;
}
doc -= offset; // adjust for what segment this is in
comparator.copy(0, doc);
Object val = comparator.value(0);
// Sortable float, double, int, long types all just use a string
// comparator. For these, we need to put the type into a readable
// format. One reason for this is that XML can't represent all
// string values (or even all unicode code points).
// indexedToReadable() should be a no-op and should
// thus be harmless anyway (for all current ways anyway)
if (val instanceof String) {
field.setValue((String) val);
val = ft.toObject(field);
}
// Must do the same conversion when sorting by a
// String field in Lucene, which returns the terms
// data as BytesRef:
if (val instanceof BytesRef) {
field.setValue(((BytesRef) val).utf8ToString());
val = ft.toObject(field);
}
vals.add(val);
}
sortVals.add(fieldname, vals);
}
rsp.add("sort_values", sortVals);
}
}
/** Actually run the query */
@Override
public void process(ResponseBuilder rb) throws IOException {
SolrQueryRequest req = rb.req;
SolrQueryResponse rsp = rb.rsp;
SolrParams params = req.getParams();
if (!params.getBool(COMPONENT_NAME, true)) {
return;
}
SolrIndexSearcher searcher = req.getSearcher();
if (rb.getQueryCommand().getOffset() < 0) {
throw new SolrException(
SolrException.ErrorCode.BAD_REQUEST, "'start' parameter cannot be negative");
}
// -1 as flag if not set.
long timeAllowed = (long) params.getInt(CommonParams.TIME_ALLOWED, -1);
// Optional: This could also be implemented by the top-level searcher sending
// a filter that lists the ids... that would be transparent to
// the request handler, but would be more expensive (and would preserve score
// too if desired).
String ids = params.get(ShardParams.IDS);
if (ids != null) {
SchemaField idField = req.getSchema().getUniqueKeyField();
List<String> idArr = StrUtils.splitSmart(ids, ",", true);
int[] luceneIds = new int[idArr.size()];
int docs = 0;
for (int i = 0; i < idArr.size(); i++) {
int id =
req.getSearcher()
.getFirstMatch(
new Term(idField.getName(), idField.getType().toInternal(idArr.get(i))));
if (id >= 0) luceneIds[docs++] = id;
}
DocListAndSet res = new DocListAndSet();
res.docList = new DocSlice(0, docs, luceneIds, null, docs, 0);
if (rb.isNeedDocSet()) {
List<Query> queries = new ArrayList<Query>();
queries.add(rb.getQuery());
List<Query> filters = rb.getFilters();
if (filters != null) queries.addAll(filters);
res.docSet = searcher.getDocSet(queries);
}
rb.setResults(res);
rsp.add("response", rb.getResults().docList);
return;
}
SolrIndexSearcher.QueryCommand cmd = rb.getQueryCommand();
cmd.setTimeAllowed(timeAllowed);
SolrIndexSearcher.QueryResult result = new SolrIndexSearcher.QueryResult();
//
// grouping / field collapsing
//
boolean doGroup = params.getBool(GroupParams.GROUP, false);
if (doGroup) {
try {
cmd.groupCommands = new ArrayList<Grouping.Command>();
String[] fields = params.getParams(GroupParams.GROUP_FIELD);
String[] funcs = params.getParams(GroupParams.GROUP_FUNC);
String[] queries = params.getParams(GroupParams.GROUP_QUERY);
String groupSortStr = params.get(GroupParams.GROUP_SORT);
Sort groupSort =
groupSortStr != null ? QueryParsing.parseSort(groupSortStr, req.getSchema()) : null;
int limitDefault = cmd.getLen(); // this is normally from "rows"
int docsPerGroupDefault = params.getInt(GroupParams.GROUP_LIMIT, 1);
// temporary: implement all group-by-field as group-by-func
if (funcs == null) {
funcs = fields;
} else if (fields != null) {
// catenate functions and fields
String[] both = new String[fields.length + funcs.length];
System.arraycopy(fields, 0, both, 0, fields.length);
System.arraycopy(funcs, 0, both, fields.length, funcs.length);
funcs = both;
}
if (funcs != null) {
for (String groupByStr : funcs) {
QParser parser = QParser.getParser(groupByStr, "func", rb.req);
Query q = parser.getQuery();
Grouping.CommandFunc gc = new Grouping.CommandFunc();
gc.groupSort = groupSort;
if (q instanceof FunctionQuery) {
gc.groupBy = ((FunctionQuery) q).getValueSource();
} else {
gc.groupBy = new QueryValueSource(q, 0.0f);
}
gc.key = groupByStr;
gc.groupLimit = limitDefault;
gc.docsPerGroup = docsPerGroupDefault;
cmd.groupCommands.add(gc);
}
}
if (cmd.groupCommands.size() == 0) cmd.groupCommands = null;
if (cmd.groupCommands != null) {
if (rb.doHighlights || rb.isDebug()) {
// we need a single list of the returned docs
cmd.setFlags(SolrIndexSearcher.GET_DOCLIST);
}
searcher.search(result, cmd);
rb.setResult(result);
rsp.add("grouped", result.groupedResults);
// TODO: get "hits" a different way to log
return;
}
} catch (ParseException e) {
throw new SolrException(SolrException.ErrorCode.BAD_REQUEST, e);
}
}
// normal search result
searcher.search(result, cmd);
rb.setResult(result);
rsp.add("response", rb.getResults().docList);
rsp.getToLog().add("hits", rb.getResults().docList.matches());
doFieldSortValues(rb, searcher);
doPrefetch(rb);
}
