Hi Michael,On Jan 5, 2006, at 7:38 AM, Michael Bouschen wrote: Hi Craig,
[...]
CLR: And another negative test for HAVING that uses a term that's not an aggregate. SELECT department, AVG(weeklyhours) FROM Employee GROUP BY department HAVING middlename != NULL
MBO: OK, then we are back to the original negative query which had a HAVING clause: HAVING firstname = 'emp1first'. But I can add this, too.
Yes, please. It's a different negative test from others because it's a boolean _expression_ (ok) using non-aggregate terms (bad).
Just for my understanding: it is ok to use a non-aggregate _expression_ as long as it is a grouping _expression_, correct?
Yes. Here's my latest attempt at the having _expression_: A14.6.10-2 [When “having” is specified, the “having” _expression_ consists of arithmetic and boolean expressions containing expressions that are either aggregate expressions or contained in a grouping _expression_.]
Since the having _expression_ is applied *after* the grouping is complete, this restriction is required. "Having" expressions without aggregates doesn't make sense; these expressions should be part of the where clause.
I added a new negative test to class Having: SELECT department, AVG(weeklyhours) FROM Employee GROUP BY department HAVING firstname == 'emp1First' JPOX does not catch this error.
I propose the following: - I check in the change with the new negative test.
ok - I reopen issue JDO-245.
ok - I really think the title of JDO-245 talking about "having clause refering fields which are not part of the result clause" is misleading, because the result clause does not matter here. I would like to change the title.
If you can edit the JIRA please do so. I think you have karma. - I would like to move the test class from the result package to jdoql.
ok
Regards,
Craig
What do you think?
Regards Michael -- Michael Bouschen [EMAIL PROTECTED] Engineering GmbH Tel.:++49/30/235 520-33 Buelowstr. 66 Fax.:++49/30/2175 2012 D-10783 Berlin
|
smime.p7s
Description: S/MIME cryptographic signature
