For GROUP BY queries, PostgreSQL require all columns in the
unaggregated part of the select list to also be in the
GROUP BY list.
To fix this, add user_name and user_id to the GROUP BY list along side
qcc_title.
This addition cannot change the query results, because user_name and
user_id are functionally dependent on qcc_title: user_name because
of the WHERE clause for this query, and user_id due to the 1 to 1
relationship between user_name and user_id enforced by their unique
and not null constraints on the underlying table.
This has been tested on PostgreSQL and MySQL.
Bug: 68087
Change-Id: I3aed715b40ff6e2290220122acbda0f0e74b5b36
) . ')';
}
+ if ( $dbr->implicitGroupby() ) {
+ $options = array( 'GROUP BY' => array( 'qcc_title' ) );
+ } else {
+ $options = array( 'GROUP BY' => array( 'user_name', 'user_id', 'qcc_title' ) );
+ }
+
return array(
'tables' => array( 'querycachetwo', 'user', 'recentchanges' ),
'fields' => array( 'user_name', 'user_id', 'recentedits' => 'COUNT(*)', 'qcc_title' ),
- 'options' => array( 'GROUP BY' => array( 'qcc_title' ) ),
+ 'options' => $options,
'conds' => $conds
);
}