[Python-Dev] PEP 469: Restoring the iterkeys/values/items() methods

[Steven D'Aprano]

On Sun, Apr 20, 2014 at 03:07:39PM +0000, Kristján Valur Jónsson wrote:

> Does one ever use iteritems() et al without first invoking iter() on 
> it?  

I can't speak for others, but I never invoke iteritems *with* iter(). 
What would be the point? iteritems is documented as returning an 
interator.

# never this
for key, value in iter(mydict.iteritems()): ...

# but this
for key, value in mydict.iteritems(): ...


> I.e. is it important that it is an iterator, rather than an 
> iterable? I think we could easily relax that requirement in the pep 
> and solve 99% of the use cases.

And the other 1% of cases would be a land-mine waiting to blow the 
user's code up.

Would it actually solve 99% of the use cases? Or only 90%? Or 50%? How 
do you know?

In Python 2.7 iteritems() etc is documented as returning an iterator. 
That's a promise of the language, and people will rely on it. But they 
won't be able to rely on that promise in polygot 2+3 code -- exactly the 
use-case this PEP is trying to satisfy -- because the promise to return 
an iterator will be broken in 3.

It would be actively misleading, since Python 3's iteritems() would 
return a view, not an iter, and it would fail at solving the backwards 
compatibility issue since views and iterators are not interchangeable 
except for the most basic use of iteration.