Martijn van Oosterhout wrote:
> On Tue, May 16, 2006 at 10:29:27AM -0700, Don Y wrote:
>> Given a user defined type foo...
>> I've created several casts to/from foo and built-in types.
>> I had adopted a naming convention of:
>> baz foo_to_baz(foo);
>> foo foo_from_baz(baz);
>>
>> But:
>
> <snip>
>
>> I don't see how I can do this in my declarations. E.g.,
>> if I have
>> baz = {int4, text, float8, ...}
>> then I end up with several (C) functions all named foo()
>> but each taking a different argument type (baz). Since
>> C doesn't support more than a single namespace for functions,
>> this just won't work.
>>
>> What am I failing to see, here?
>
> That the name of the function in C doesn't have to be the same as the
> name of the function in SQL. You can even define many SQL functions
> that all refer to the same C function.
>
> So in your C file yo call them:
>
> cast_foo_to_baz()
>
> and in the SQL you declare as just:
>
> baz()
But what *binds* my C declaration to the corresponding SQL
"CREATE CAST"?
E.g.,
CREATE FUNCTION foo_from_baz(baz)
RETURNS foo
AS '...'
LANGUAGE 'C' IMMUTABLE STRICT;
yet, to support the foo(baz) syntax, I would then need (?)
to continue with:
CREATE CAST (baz AS foo)
WITH FUNCTION foo(baz);
or, do I use:
CREATE CAST (baz as foo)
WITH FUNCTION foo_from_baz(baz);
but then how does the "foo(baz)" syntax originate?
or, is this a built in "feature" of the interpreter?
(In which case, why all the commentary <snip>ed that
tells me to name my functions in this way?)
--don