use of gate.creole.annic.apache.lucene.search.Query in project gate-core by GateNLP.
the class QueryParser method parse.
/**
* Given a query, this method parses it to convert it into one or more
* lucene queries.
* @throws gate.creole.ir.SearchException
*/
public Query[] parse(String field, String query, String baseTokenAnnotationType, String corpusID, String annotationSetToSearchIn) throws gate.creole.ir.SearchException {
this.field = field;
this.baseTokenAnnotationType = baseTokenAnnotationType;
this.position = 0;
// at the moment this supports only | operator
// it also support klene operators * and +
// implicit operator is &
// It supports simple String queries
// it supports eight kinds of tokens
// 1. String (without quotes)
// 2. "String" (with quotes)
// 3. {AnnotationType}
// 4. {AnnotationType==String}
// 5. {AnnotationType=="String"}
// 7. {AnnotationType.feature==string}
// 8. {AnnotationType.feature=="string"}
// Steps
// The query would we searched from left to right order
// returned arraylist contains queries where each query is required
// to
// be converted into the Phrase query
queries = SubQueryParser.parseQuery(query);
Query[] q = new Query[queries.size()];
for (int i = 0; i < queries.size(); i++) {
Query phraseQuery = createPhraseQuery(queries.get(i));
// boolean query
if (corpusID == null && annotationSetToSearchIn == null) {
BooleanQuery booleanQuery = new BooleanQuery();
Term t = new Term(Constants.ANNOTATION_SET_ID, Constants.COMBINED_SET);
TermQuery tQuery = new TermQuery(t);
booleanQuery.add(tQuery, false, true);
booleanQuery.add(phraseQuery, true, false);
q[i] = booleanQuery;
} else {
BooleanQuery booleanQuery = new BooleanQuery();
booleanQuery.add(phraseQuery, true, false);
if (corpusID != null) {
Term t = new Term(Constants.CORPUS_ID, corpusID);
TermQuery tQuery = new TermQuery(t);
booleanQuery.add(tQuery, true, false);
}
if (annotationSetToSearchIn != null) {
Term t = new Term(Constants.ANNOTATION_SET_ID, annotationSetToSearchIn);
TermQuery tQuery = new TermQuery(t);
booleanQuery.add(tQuery, true, false);
} else {
Term t = new Term(Constants.ANNOTATION_SET_ID, Constants.COMBINED_SET);
TermQuery tQuery = new TermQuery(t);
booleanQuery.add(tQuery, false, true);
}
q[i] = booleanQuery;
}
}
return q;
}
use of gate.creole.annic.apache.lucene.search.Query in project gate-core by GateNLP.
the class LuceneSearchThread method search.
/**
* This method collects the necessary information from lucene and uses
* it when the next method is called
*
* @param query query supplied by the user
* @param patternWindow number of tokens to refer on left and right
* context
* @param indexLocation location of the index the searcher should
* search in
* @param luceneSearcher an instance of lucene search from where the
* instance of SearchThread is invoked
* @return true iff search was successful false otherwise
*/
@SuppressWarnings("unchecked")
public boolean search(String query, int patternWindow, String indexLocation, String corpusToSearchIn, String annotationSetToSearchIn, LuceneSearcher luceneSearcher) throws SearchException {
this.query = query;
this.contextWindow = patternWindow;
this.indexLocation = indexLocation;
this.queryParser = new QueryParser();
this.luceneSearcher = luceneSearcher;
/*
* reset all parameters that keep track of where we are in our
* searching. These parameters are used mostly to keep track of
* where to start fetching the next results from
*/
searchResultInfoMap = new HashMap<String, List<QueryItem>>();
serializedFileIDIndex = 0;
queryItemIndex = 0;
serializedFilesIDsList = new ArrayList<String>();
ftpIndex = -1;
success = false;
fwdIterationEnded = false;
try {
// first find out the location of Index
// TODO does this just replace \ with / if so we should do this better
StringBuilder temp = new StringBuilder();
for (int i = 0; i < indexLocation.length(); i++) {
if (indexLocation.charAt(i) == '\\') {
temp.append("/");
} else {
temp.append(indexLocation.charAt(i));
}
}
indexLocation = temp.toString();
/*
* for each different location there can be different
* baseTokenAnnotationType each index will have their index
* Definition file stored under the index directory so first see
* if given location is a valid directory
*/
File locationFile = new File(indexLocation);
if (!locationFile.isDirectory()) {
System.out.println("Skipping the invalid Index Location :" + indexLocation);
return false;
}
if (!indexLocation.endsWith("/")) {
indexLocation += "/";
}
// otherwise let us read the index definition file
locationFile = new File(indexLocation + "LuceneIndexDefinition.xml");
// check if this file is available
if (!locationFile.exists()) {
System.out.println("Index Definition file not found - Skipping the invalid Index Location :" + indexLocation + "LuceneIndexDefinition.xml");
return false;
}
Map<String, Object> indexInformation = null;
// other wise read this file
XStream xstream = new XStream(new StaxDriver());
try (FileReader fileReader = new FileReader(indexLocation + "LuceneIndexDefinition.xml")) {
// Saving was accomplished by using XML serialization of the map.
indexInformation = (Map<String, Object>) xstream.fromXML(fileReader);
}
// find out if the current index was indexed by annicIndexPR
String indexedWithANNICIndexPR = (String) indexInformation.get(Constants.CORPUS_INDEX_FEATURE);
if (indexedWithANNICIndexPR == null || !indexedWithANNICIndexPR.equals(Constants.CORPUS_INDEX_FEATURE_VALUE)) {
System.out.println("This corpus was not indexed by Annic Index PR - Skipping the invalid Index");
return false;
}
// find out the baseTokenAnnotationType name
baseTokenAnnotationType = ((String) indexInformation.get(Constants.BASE_TOKEN_ANNOTATION_TYPE)).trim();
int separatorIndex = baseTokenAnnotationType.lastIndexOf('.');
if (separatorIndex >= 0) {
baseTokenAnnotationType = baseTokenAnnotationType.substring(separatorIndex + 1);
}
// create various Queries from the user's query
Query[] luceneQueries = queryParser.parse("contents", query, baseTokenAnnotationType, corpusToSearchIn, annotationSetToSearchIn);
if (queryParser.needValidation()) {
if (DEBUG)
System.out.println("Validation enabled!");
} else {
if (DEBUG)
System.out.println("Validation disabled!");
}
// create an instance of Index Searcher
LuceneIndexSearcher searcher = new LuceneIndexSearcher(indexLocation);
try {
// we need to iterate through one query at a time
for (int luceneQueryIndex = 0; luceneQueryIndex < luceneQueries.length; luceneQueryIndex++) {
/*
* this call reinitializes the first Term positions arraylists
* which are being used to store the results
*/
searcher.initializeTermPositions();
/*
* and now execute the query result of which will be stored in
* hits
*/
Hits hits = searcher.search(luceneQueries[luceneQueryIndex]);
/*
* and so now find out the positions of the first terms in the
* returned results. first term position is the position of the
* first term in the found pattern
*/
List<?>[] firstTermPositions = searcher.getFirstTermPositions();
// if no result available, set null to our scores
if (firstTermPositions[0].size() == 0) {
// do nothing
continue;
}
// information
for (int hitIndex = 0; hitIndex < hits.length(); hitIndex++) {
int index = firstTermPositions[0].indexOf(Integer.valueOf(hits.id(hitIndex)));
// we fetch all the first term positions for the query
// issued
List<?> ftp = (List<?>) firstTermPositions[1].get(index);
/*
* pattern length (in terms of total number of annotations
* following one other)
*/
int patLen = ((Integer) firstTermPositions[2].get(index)).intValue();
/*
* and the type of query (if it has only one annotation in it,
* or multiple terms following them)
*/
int qType = ((Integer) firstTermPositions[3].get(index)).intValue();
// find out the documentID
String serializedFileID = hits.doc(hitIndex).get(Constants.DOCUMENT_ID_FOR_SERIALIZED_FILE);
QueryItem queryItem = new QueryItem();
queryItem.annotationSetName = hits.doc(hitIndex).get(Constants.ANNOTATION_SET_ID).intern();
queryItem.id = hits.id(hitIndex);
queryItem.documentID = hits.doc(hitIndex).get(Constants.DOCUMENT_ID).intern();
queryItem.ftp = ftp;
queryItem.patLen = patLen;
queryItem.qType = qType;
queryItem.query = luceneQueries[luceneQueryIndex];
queryItem.queryString = queryParser.getQueryString(luceneQueryIndex).intern();
/*
* all these information go in the top level arrayList. we
* create separate arrayList for each individual document
* where each element in the arrayList provides information
* about different query issued over it
*/
List<QueryItem> queryItemsList = searchResultInfoMap.get(serializedFileID);
if (queryItemsList == null) {
queryItemsList = new ArrayList<QueryItem>();
queryItemsList.add(queryItem);
searchResultInfoMap.put(serializedFileID, queryItemsList);
serializedFilesIDsList.add(serializedFileID);
} else {
// // before inserting we check if it is already added
// if(!doesAlreadyExist(queryItem, queryItemsList)) {
queryItemsList.add(queryItem);
// }
}
}
}
} finally {
searcher.close();
}
// if any result possible, return true
if (searchResultInfoMap.size() > 0)
success = true;
else
success = false;
} catch (IOException | gate.creole.ir.SearchException e) {
throw new SearchException(e);
}
return success;
}
Aggregations