sumit <[EMAIL PROTECTED]> writes:
>       Can someone inform us of the developments regarding our code on 
> DATACUBE? We would like to know if there are somethings which we might 
> have overlooked. If possible, can we also know if we can expect it to be 
> added to this release?

I'm sorry, but that patch will never be accepted in anything close to
its present form.  You've taken what is fundamentally a grouping
construct and implemented it in an altogether inappropriate place,
namely the top level of the system.  You mention that it doesn't work in
subselects, but there are a bunch of other ways that it fails to work,
too.  (For example, SQL99 expects that a CUBE list can be used in
combination with other grouping columns.)  If the patch had a growth
path that would eventually address such concerns, we would consider it
as a partial implementation, but as far as I can see the only growth
path is to throw it out and start over.  We've accepted patches in the
past that did things at the wrong level (INTERSECT/EXCEPT comes to mind),
and have learned painfully that we don't want to do that again.

An implementation that would actually meet the spec would probably
involve extending the code in executor/nodeGroup.c and related files,
and perhaps tweaking the planner if CUBE requires any changes in the
structure of the plan tree (offhand I can't tell if it does or not).
You'd also need to extend the parsetree representation of GROUP BY
lists so that a CUBE operation could be distinguished from ordinary

                        regards, tom lane

---------------------------(end of broadcast)---------------------------
TIP 2: you can get off all lists at once with the unregister command
    (send "unregister YourEmailAddressHere" to [EMAIL PROTECTED])

Reply via email to