On Fri, 2 Jun 2000, Tom Lane wrote:
> >> None works. I tried. There is no type that can be cast to inet. And
> >> inet_in has a different calling sequence, and you can't really use that.
>
> > Added to TODO.
>
> If we had "C string" (or something like that) as a genuine type in the
> type system, then it would be possible/reasonable for the parser to
> understand
> textvariable::cstring::inet
> as a request to invoke text_out followed by inet_in.
>
> I think it'd be a real bad idea to invoke such conversions silently,
> since then we'd essentially have no type system at all (you could get
> from anything to anything else via cstring, so how can the thing check
> for errors?). But it'd be awful darn handy to be able to invoke the
> type i/o routines explicitly...
I think its a great idea. Is it just a matter of altering the catalog? Or
apparently we'd need postgres engine to know that cstring is a special
type and instead of looking for conversion routines it should use xxx_in
and yyy_out
I like it in any case. Last ditch do-not-use-unless-you-sure-its-what you
want thingy...
-alex
