[Python-3000] Iterators for dict keys, values, and items == annoying :)

[Nick Coghlan]

Adam DePrince wrote:
> On Thu, 2006-03-30 at 23:37 -0800, Neal Norwitz wrote:
>> On 3/30/06, Terry Reedy <tjreedy at udel.edu> wrote:
>>> "Aahz" <aahz at pythoncraft.com> wrote in message
>>> news:20060331012037.GA7960 at panix.com...
>>>> What do we want to tell people who have code like this:
>>>>
>>>> keys = d.keys()
>>>> keys.sort()
>>> Could a good-enough code analyzer detect such, even if separated by
>>> intervening lines?  If so, it could suggest sorted() as a fix.  I wonder if
>>> the pypy analyzer could be adapted for 2.x to 3.0 warning and upgrade
>>> purposes.  Or do pylint or pychecker gather enough information?
>> pychecker is supposed to have this info, but only if d is known to be
>> a dict.  It could be extended to assume any method keys() (and
>> friends) should return iterators.  In which case, it would say that an
>> iterator doesn't have a sort method.  Below is the output of the
>> current version.
> 
> With the views that we were talking about before, with d.keys() all you
> have is the view, neither an iter nor list have been instantiated.  
> 
> I'd almost say that 
> 
> keys = d.keys()
> foo = keys.sort() 
> 
> could fit quite well into the view framework.

Not a good idea, since the long history of "list.sort()" encourages people to 
think of the sort() method as an inplace operation, which it wouldn't be on a 
view. "sorted()", on the other hand, already creates a new object. The only 
downside is that Py2.x & Py3k compatible code would look like:

   keys = sorted(d.keys())

which is likely to create the list *twice* in Py2.x.

Cheers,
Nick.

-- 
Nick Coghlan   |   ncoghlan at gmail.com   |   Brisbane, Australia
---------------------------------------------------------------
             http://www.boredomandlaziness.org