Am Dienstag, 27. Mai 2008 21:32:44 schrieb Donal K. Fellows:
>  Arnulf Wiedemann wrote:
>  > Do I understand correctly, that is the case when a class method is called
>  > without the "my" in front? In that case there is a problem, if the same
>  > method exists in global namespace, as the object unknown method is called
>  > after looking up in global namespace.

> 
>  Umm, I'm not sure what the problem is. Methods exist in objects and
>  classes, not namespaces. That unknown handler is the handler for methods
>  that are unknown, i.e. it will be called when you do:
> 
>     theobject foobarbooftangftangbiscuitbarrel
> 

Yes, you are right, I had itcl-ng implementation in mind, where it is 
different, there the methods do exist as commands in the namespace.

>  (Sorry, been listening to Monty Python recordings...)
> 
>  If you call a command in a namespace, it is first resolved according to
>  the namespace's resolver policy (including its path and a look in the
>  global namespace) and only if that search turns up a blank is the
>  unknown handler for the namespace called (which might or might not be
>  the same as the global namespace's unknown handler). But that handler is
>  inherently disjoint from the namespace unknown handler; they've nothing
>  in common unless you set that to be otherwise. Note that the resolution
>  to the unknown handler does not go through the mechanisms you set up at
>  the scripted level; it's an inherent characteristic of the code that
>  constructs call chains.
> 
>  FWIW, if a command isn't in a namespace, the namespace's command
>  resolver gets first bite at the cherry, even before looking in the
>  global namespace, and I *think* it can inhibit the global lookup.

you are right here too.

>  I 
>  could be wrong there; the resolver API is only poorly documented at
>  best, a consequence of (or maybe reason for) never being made public. If
>  you are doing things to avoid 'my', you're in the space of commands and
>  not method names.
> 

Yes that is what I had in mind (how it is working in itcl-ng)

>  I don't know if this resolves your concerns. :-)
> 
>  (I suppose that the code that sparked this discussion doesn't need an
>  unknown handler for instance objects; the code to extract the list of
>  methods could also install forwardings during subclass creation, which
>  would be faster overall. Designing a good base-class system for widgets
>  is not the easiest thing ever; there are some tricky nuances.)

have to think about that, if I perhaps can use what you describe here for 
itcl-ng

Arnulf
> 
>  Donal.



-------------------------------------------------------------------------
This SF.net email is sponsored by: Microsoft
Defy all challenges. Microsoft(R) Visual Studio 2008.
http://clk.atdmt.com/MRT/go/vse0120000070mrt/direct/01/
_______________________________________________
Tcl-Core mailing list
Tcl-Core@[...].net
https://lists.sourceforge.net/lists/listinfo/tcl-core