On Fri, Jul 02, 2004 at 10:27:35 -0400,
Tom Lane <tgl@sss.pgh.pa.us> wrote:
>
> IIRC there were objections the last time it was seriously proposed,
> basically of the form "but that will break my application which relies
> on writing so-and-so without a cast". We did get as far as removing all
> the implicit cross-category coercions except these casts to text:
>
> bigint | text
> smallint | text
> integer | text
> real | text
> double precision | text
> numeric | text
> oid | text
> date | text
> time without time zone | text
> time with time zone | text
> timestamp without time zone | text
> timestamp with time zone | text
> interval | text
>
> but these seem to have a constituency :-(
I think it is reasoable to expect people to use an explicit cast when
doing these conversions. These are not conversions that you would normally
want to do. And even if you do, you need to be very concerned about how
the converted text looks for different ranges of source data values.
Allowing these makes it harder to notice mistakes. And I expect that
overall the hiding mistakes case occurs a lot more than the I meant to
do that but didn't want to type ::text case.