Re: Casting INT4 to BOOL...

>>> Can you add some regression tests, please?
>
>> Given the simplicity of the casts, does this really need a
>> require a regression test?
>
> That request seems quite over-the-top to me too. The real problem here
> is just whether we want to be accepting a feature addition, small
> though
> it be, at this stage of the beta cycle. I've got mixed emotions about
> that myself. The odds of breaking anything with this patch seem
> essentially zero, and we have certainly seen this requested multiple
> times before, so one part of me says "sure, push it in". But from a
> project-management standpoint it's hard to justify not saying "it's got
> to wait for 8.1".

*puts on patch advocate's hat*

From a feature/usability standpoint, it'd be "more convenient" to have
booleans behave consistently starting with the 8.X branch as opposed to
saying, "starting with 8.1, booleans can be explicitly cast to integers
and visa versa." From a patch maintainer stand point, given the work
of the original patch, it's hardly worth the effort to maintain and
given the possibility of OIDs being used, waiting is more work than
committing it now. :)

>>> The patch changes the system catalog; it probably ought to also bump
>>> the catalog version number.
>
>> System catalog bumps have been coming through with some degree of
>> regularity so I wasn't worried about providing the patch to bump the
>> catalog date. -sc
>
> I think the agreed protocol is that the descriptive text should
> remind the committer to bump the catversion upon application. Just
> to make sure he doesn't forget ;-)

Fair enough... though with this discussion it would seem like a rather
unnecessary cudgel to the head. Next time I'll bump it when adding a
built-in function, however. Thanks for the protocol FYI. -sc

--
Sean Chittenden