"Kevin Grittner" <Kevin.Grittner@wicourts.gov> writes:
> Tom Lane <tgl@sss.pgh.pa.us> wrote:
>> Here's a proposed patch, sans documentation as yet.
> I see you took the surgical approach -- only a cast from a record to
> a character string type is affected. I agree that will fix the
> complaints I've seen, and I imagine you're keeping the change narrow
> to minimize the risk of breaking existing code, but this still looks
> weird to me:
> test=# select ('2010-11-05'::date).text;
> text
> ------------
> 2010-11-05
> (1 row)
Perhaps, but it's been accepted since 7.3, with few complaints.
I think we should only remove the behavior that was added in 8.4.
regards, tom lane