polymorphism w/out signatures?

[Laughlin, Joseph V]

Knuth.

Joe Laughlin


> -----Original Message-----
> From: Roy Smith [mailto:roy at panix.com] 
> Sent: Monday, May 10, 2004 11:12 AM
> To: python-list at python.org
> Subject: Re: polymorphism w/out signatures?
> 
> 
> Peter Hansen <peter at engcorp.com> wrote:
> > Or is there some way that the following code can
> > possibly fail with a TypeError, for any reason
> > whatsoever?
> > 
> >    someName = [someName]
> 
> It's a bit contrived (and also not exactly what you asked for), but 
> consider something like:
> 
> class foo:
>     def __getattr__ (self, name):
>         return 1 + ""
> 
>     def barf (self):
>         self.someName = [self.someName]
> 
> f = foo ()
> f.barf ()
> 
> which produces this when you run it:
> 
> Roy-Smiths-Computer:tmp$ ./try.py 
> Traceback (most recent call last):
>   File "./try.py", line 11, in ?
>     f.barf ()
>   File "./try.py", line 8, in barf
>     self.someName = [self.someName]
>   File "./try.py", line 5, in __getattr__
>     return 1 + ""
> TypeError: unsupported operand type(s) for +: 'int' and 'str'
> 
> The general point is that when catching exceptions, it's 
> usually best to 
> include as little code in the try block as you can, to make 
> sure you're 
> catching the exception you think you are.  Especially in a highly 
> dynamic language like Python, it's sometimes difficult to 
> think of all 
> the possible ways code can fail.
> 
> Was it Knuth who said, "I've only proven that my program is 
> correct, I 
> havn't actually tested it"?  Or was that Dijkstra?
> -- 
> http://mail.python.org/mailman/listinfo/python-list
>