Re: set of valid characters for type names

[Tim Janik <timj gtk org>]

On Wed, 9 May 2001, James Henstridge wrote:

> On Tue, 8 May 2001, Tim Janik wrote:
> 
> > "." is a normal struct/class operator i nmany languages.
> > how are those bindings supposed to interface to
> > type names that contain "."?
> 
> Well, in pygtk, typecodes are just treated as strings.  In the cases where
> I want to use dots in type names, I doubt that other languages would be
> involved.  Allowing dots would mean I could use more meaningful type
> names, rather than having to mangle them.
> 
> > i'm somewhat in doubt about "+" already, "-"/"_" should at least
> > normally be exchangable, and is just added convenience for scheme.
> 
> Well, neither `+' or `-' are valid characters as part of an identifier in
> C or python.

well, for '-' i'd usually expect people to eitehr use '_' or '-' so
you can always s/-/_/ in the binding. for '+' i'm not so sure.

>  Is it really worth going for lowest common denominator for
> this?

that's a good question, and i'm not sure i have the best answer to it.
if we allow arbitrary characters, some LBs (and remember there can be
a multitude of bindings in a process that use types interchangably)
may have no chance but to s/[^a-z]/_/, and if types are not anymore
uniquely identifiable after this substitution things go to hell.

> 
> James.
> 

---
ciaoTJ