Csaba Ringhofer created IMPALA-10781:
----------------------------------------
Summary: Avoid nested loop join when there is OR in the join
condition
Key: IMPALA-10781
URL: https://issues.apache.org/jira/browse/IMPALA-10781
Project: IMPALA
Issue Type: Improvement
Components: Backend, Frontend
Reporter: Csaba Ringhofer
The following query becomes a nested loop join in Impala:
{code}
SELECT * FROM t1 JOIN t2 ON t1_col1 = t2_col1 OR t1_col2 = t2_col2;
{code}
A possible solution is to rewrite the join into an union of two joins where
each join becomes an equi join. Currently this has to be done by hand.
It is possible to create a more efficient solution that doesn't need to reread
the right side of the join by adding an operator that duplicates rows and adds
an extra column that identifies the join condition.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
