Wang Han created LUCENE-5766:
--------------------------------
Summary: bug in rewrite function of boolean query
Key: LUCENE-5766
URL: https://issues.apache.org/jira/browse/LUCENE-5766
Project: Lucene - Core
Issue Type: Bug
Reporter: Wang Han
when optimize boolean query. When the only query sub clause has a boost with
1.0, it should be cloned too.
if (minNrShouldMatch == 0 && clauses.size() == 1) { //
optimize 1-clause queries
BooleanClause c = clauses.get(0);
if (!c.isProhibited()) { // just return clause
Query query = c.getQuery().rewrite(reader); // rewrite first
if (getBoost() != 1.0f) { // incorporate boost
if (query == c.getQuery()) { // if rewrite was no-op
query = query.clone(); // then clone before boost
}
// Since the BooleanQuery only has 1 clause, the BooleanQuery will be
// written out. Therefore the rewritten Query's boost must
incorporate both
// the clause's boost, and the boost of the BooleanQuery itself
query.setBoost(getBoost() * query.getBoost());
}
return query;
}
}
when boolean query nested in a disjunction query, rewrite function return the
original term query without clone and may cause bad boost value in upper
queries. Delete the if statement and always do
if (query == c.getQuery()) { // if rewrite was no-op
query = query.clone(); // then clone before boost
}
can fix this bug.
sample query may cause this bug: ((+((title:at)~1^2.0)))^4.5 after rewrite
the query will be changed to ((+((title:at)~1^9.0)))^4.5 and Query.clone()
won't work for this problem because your lazy clone strategy.
--
This message was sent by Atlassian JIRA
(v6.2#6252)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
