OT: Re: Just took a look in the perl newsgroup....

[Bengt Richter]

On Thu, 29 May 2003 15:46:34 -0600, Steven Taschuk <staschuk at telusplanet.net> wrote:

>Quoth Bengt Richter:
>> On Wed, 28 May 2003 23:10:57 -0600, Steven Taschuk <staschuk at telusplanet.net> wrote:
[...]
>  [...]
>> ====< caseswitch.py >=============================================
>> class Switch(object):
>>     def __init__(self, *caseNames):
>>         self.halias = {}
>>         for name in caseNames:
>>             try:
>>                 hname = '_%X' % hash(name)
>>             except TypeError:
>>                 raise ValueError, 'Switch requires hashable case values'
>>             exec ('class %s(Exception):pass'%hname) in self.__dict__
>>             self.halias[name] = getattr(self, hname)
>>         del self.__dict__['__builtins__']
>
>One point: why not just this?
>
>    def __init__(self, *caseNames):
>        self.halias = {}
>        for name in caseNames:
>            class AwfulHack(Exception):
>                pass
>            self.halias[name] = AwfulHack

Good point. Forgot that just differing ids of generated classes would suffice ;-/
>
>The hash code name is not used except in error messages, where it
>is not helpful in general.  The proposed name at least tells the
>user that it's the programmer's fault.  (A better name could be
>done with a metaclass __str__, I imagine.)
Good idea too. Are you sure you want to contribute towards making this
that slick ;-)
>
>  [...]
>> Note that I had to make a choice re the tuple ambiguity. [...]
>
>Yes, that's slightly unfortunate, though it doesn't deserve a
>yikes.
>
>It's still abuse, though.
Well, at least I didn't include slice notation in specifying multiple cases ;-)

Regards,
Bengt Richter