[Python-ideas] Introduce collections.Reiterable

[Neil Girdhar]

On Sat, Sep 21, 2013 at 2:25 AM, Stephen J. Turnbull <stephen at xemacs.org>wrote:

> Neil Girdhar writes:
>
>  > You're right that you should go ahead and use something however you
>  > want to.  However, there are plenty of times where you can't do that,
>  > e.g., you want to know if something is callable before calling it,
>  > and similarly if something is reiterable before iterating it and
>  > exhausting.  That is the purpose of collections.abc,
>
> I don't think so.  It's documented that way:
>
>     This module provides abstract base classes that can be used to
>     test whether a class provides a particular interface; for example,
>     whether it is hashable or whether it is a mapping.
>
> But I wouldn't do explicit testing with isinstance, but rather use
> implicit assertions (at instantiation time) by deriving from the ABC.
> I don't see how Reiterable could be adapted to this style of
> programming because the API of iterables is basically fixed (support
> __iter__ or __getitem__).
>

Wouldn't you need to define a new ABC to do this?

Here's one possibility with Tim's iterable and not iterator:

class Reiterable(collections.Iterable,
                 metaclass=collections.abc.ABCMeta):
    @classmethod
    def __subclasshook__(cls, subclass):
        if (collections.Iterable.__subclasshook__(subclass)
            and not issubclass(subclass, collections.Iterator)):
            return True
        return NotImplemented

for obj in [list(), tuple(), dict(), set(), range(4),
            (x * x for x in range(4))]:
    print(type(obj),
          isinstance(obj, Reiterable))

Another possibility would be to explicitly register Views and so on using

Reiterable.register(...)



>
>  > and that's what I thought we were discussing.
>
> You were, I agree.  But you proposed a new API, which pretty well
> guarantees many discussants will take a more global view, like "do the
> use cases justify this addition?"
>

It's a good point.  I think if I'm the only with this problem, then the
answer is clearly no.  I will just cast to list and so what if it's a
little bit slower in some cases. How could I know that I was the only one
with this problem?


> Another such question is "what exactly is the specification?"  Tim
> Delany, for example, AIUI doesn't have a problem with saying that any
> iterable is reiterable, because it won't raise an exception if the
> program iterates it after exhaustion.  It simply does nothing, but in
> some cases that's perfectly acceptable.  I know you disagree, and I
> don't think that's a useful definition.  Still it demonstrates the
> wide range of opinions on what "reiterable" can or should guarantee.
>

Yes, agreed that there are a wide range of reasonable opinions.

Best,

Neil
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.python.org/pipermail/python-ideas/attachments/20130921/b05a4f95/attachment-0001.html>