Our application requires the push-down sql from named/dynamic query be the same
each time a same JPQL is executed. In the following JPQL example,
query="UPDATE BasicA t set t.name= ?1, t.age = ?2 WHERE t.id = ?3"
we observe two different push-down sql could be generated:
UPDATE PDQBasicA t0 SET name = ?, age = ? WHERE (t0.id = ?)
UPDATE PDQBasicA t0 SET age = ?, name = ? WHERE (t0.id = ?)
This unpredictable behavior breaks our application. The
indeterministic ordering of the update list is due to the indeterminsitic
ordering provided by HashMap and HashSet in QueryExpressions and
JPQLExpressionBuilder, respectively.
When the HashMap is changed to LinkedHashMap and HashSet to
LinkedHashSet, the access order based on insertion will be preserved and
the generated push-down sql will have predictable ordering of update list. Is
there an alternative way to address this issue?
Should a JIRA issue be open to fix this problem? Any comment is
appreciated.
In QueryExpressions:
/**
* Add an update.
*/
public void putUpdate(Path path, Value val) {
if (updates == Collections.EMPTY_MAP)
updates = new HashMap(); <== change HashMap to LinkedHashMap
updates.put(path, val);
}
In JPQLExpressionBuilder:
JPQLNode[] findChildrenByID(int id) {
Collection set = new HashSet(); <== change HashSet to LinkedHashSet
findChildrenByID(id, set);
return (JPQLNode[]) set.toArray(new JPQLNode[set.size()]);
}
Regards,
Fay
____________________________________________________________________________________
Be a better friend, newshound, and
know-it-all with Yahoo! Mobile. Try it now.
http://mobile.yahoo.com/;_ylt=Ahu06i62sR8HDtDypao8Wcj9tAcJ
