stylistic question -- optional return value

[Paddy McCarthy]

Erik Max Francis <max at alcyone.com> wrote in message news:<3D6D3F1D.42BD6B41 at alcyone.com>...
> Andrew Koenig wrote:
> 
> > Paddy> I would suggest something like (True, x, y) and (False, x, y)
> > Paddy> where False means disregard any y value. You could then
> > Paddy> consistently assign the return value of your function to three
> > Paddy> variables and examine the first to see if y contains useful
> > Paddy> data.
> > 
> > Hmmm... I think I like the (x, y) or (x, sentinel) idea better,
> > assuming that (x, y) or (x, None) isn't acceptable.
> 
> I agree.  Adding the additional element that acts as a "key" as to what
> follows seems rather distasteful.  I've done this a few times, mostly
> for quick hacks, but I've never been particularly proud of it.  (In all
> cases it was not part of some important API that external people would
> have to use.

If there IS no sentinel value i.e. you want to allow x and y to take any 
value then you either have to supply a third flag value as above, or maybe 
encode the return value if return is a tuple of two values then x and y 
are those values, otherwise the return value is a tuple whose first 
element is x.
The later way of doing things is less clear to me than the Paddy approach 
above (posted from my home account).