From: Robert Klemme <bob.news@...>
Date: 2004-03-19T17:39:43+09:00
Subject: Re: `#catch'ing many symbols


"George Ogata" <g_ogata@optushome.com.au> schrieb im Newsbeitrag
news:878yhy0xwl.fsf@optushome.com.au...
> "Robert Klemme" <bob.news@gmx.net> writes:
>
> > I don't think so.  Problem is that catch returns the instance
associated
> > with the throw and when catching multiple symbols you can't
distinguish
> > what happened.  In that case a single symbol is enough because you can
use
> > the returned element to carry information about the event:
>
> Thanks, I'd forgotten about the second throw arg.  I guess that
> satisfies the uncondtional case.  Conditional was really the one I was
> interested in, though.
>
> > Conditional catch does not make sense IMHO.  What do you want to
> > achieve with that?
>
> I think it can be useful in recursive scenarios.  To unwind the stack
> to the first call, you catch the symbol only if it's the first
> invocation.

Well, that's not too difficult, since one often has an entry method to
recursion:

def enter(*args)
  catch :quit do
    catch :done do
      recur_impl(*args)
    end
  end
end

def recur(*args)
  catch :done do
    recur_impl(*args)
  end
end

def recur_impl(*args)
  if ...
     recur x
  else
      ...
  end
end

> In my case, I have a stackable command interpreter interface
> ("stackable" in the sense that you can recursively invoke one while
> one is running, like the way you sometimes nest shells).  One command
> exits just the topmost layer to return to the one below.  Another
> exits the whole stack.

Is that really a case for recursion?

> To implement this, the former command throws :done, the latter :quit.
> Only the bottom-most interpreter catches :quit, the others just catch
> :done.
>
> I suppose there're few uses for multi-catching though, so fair enough.
> If it's a harmless extension, though...

As I said, I don't think it's a reasonable extension...

Regards

    robert