Memory leak during repeated lucene query searches?

Question

Hi all,

Basically, I simply want to do many searches on a given lucene index.

Therefore, I made a class Data with final 'analyzer', 'reader', 'searcher' and 'parser' fields, (all properly initialized in the constructor). The class also provides a 'search' method to search the index. This is all shown in the code below.

The problem is however that memory gradually becomes filled after many calls to 'search' (with different queries). I did not check what happens when always the same query is used. I looked around already for possible answers, and it seems to be best practice to keep the searcher etc open across different searches, so I guess that is not the problem. Any other ideas?

Thanks, Joachim.

(example) code:

public class Data {

private final Analyzer analyzer;
private final IndexReader reader;
private final IndexSearcher searcher;   
private final QueryParser parser;

public Data(String indexPath, Analyzer analyzer) throws IOException {
    this.analyzer = analyzer;
    Directory directory =  FSDirectory.open(indexPath);
    reader = new FilterIndexReader(IndexReader.open(directory, true));
            directory.close();
    searcher = new IndexSearcher(reader);
    parser =  new QueryParser(Version.LUCENE_CURRENT, 
                                           FieldName.CONTENT, analyzer);
}    

public TopDocs search(String line, Integer maxHits) throws ... { 
    Query query = parser.parse(QueryParser.escape(line));
    return searcher.search(query, maxHits);
}

}

Answer 1

The garbage collector will probably free up memory at some point before the app runs out of memory.

I don't see why instances would be held in memory by the search method, but you could use a profiling tool such as JProfiler to confirm.