[Python-ideas] Adding collections.abc.Ordered

[Wes Turner]

* collections.abc.Ordered
* collections.abc.Reversible
* collections.abc.Infinite [...]

* collections.abc.Sorted ?
* collections.abc.Recursive ?

Rationale:

These are all attributes of collections that would allow us to reason about
[complexity, types, runtime]?

[While someone is at it, annotating functions and methods with complexity
class URI fragments accessible at runtime could also be useful for [dynamic
programming].]

* Ordered is justified by this thread
* Reversible is distinct from Ordered (because Infinite sequences)
* Infinite is the distinction between Ordered and Reversible

* collections.abc.Sorted would be a useful property to keep track of
(because then you don't have to do element-wise comparisons for each
collection member)

...

* collections.abc.Recursive could also be a useful property to mixin [again
for dynamic programming]

https://en.wikipedia.org/wiki/Dynamic_programming
On Dec 26, 2015 7:34 PM, "Wes Turner" <wes.turner at gmail.com> wrote:

> * collections.abc.Ordered
> * collections.abc.Reversible
> * collections.abc.Infinite [...]
>
> * collections.abc.Sorted ?
> On Dec 26, 2015 7:24 PM, "Andrew Barnert via Python-ideas" <
> python-ideas at python.org> wrote:
>
>> On Dec 26, 2015, at 16:09, Serhiy Storchaka <storchaka at gmail.com> wrote:
>> >
>> >> On 27.12.15 01:05, Guido van Rossum wrote:
>> >> There is a precedent for declaring that a method isn't implemented:
>> >> __hash__. The convention is to set it to None in the subclass that
>> >> explicitly doesn't want to implement it. The __subclasshook__ in
>> >> collections.Hashable checks for this. The pattern is also used for
>> >> __await__.
>> >
>> > Yes, this was the first thing that I tried, but it doesn't work, as
>> shown in my example in issue25864. This is yet one thing that should be
>> fixed in Reversible.
>>
>> As Guido pointed out in your typehinting #170, that isn't a bug with
>> typing.Reversible. You can't use typing types in runtime type tests with
>> issubclass. That isn't supposed to work, and the fact that it often kind of
>> does work is actually a bug that he's fixing. So the fact that it doesn't
>> work in this case is correct.
>>
>> That also means your attempted solution to this thread is wrong;
>> typing.Reversible cannot be used as a substitute for
>> collections.abc.Ordered.
>>
>> > May be we have to use this idiom more widely, and specially handle
>> assigning special methods to None. The error message "'sometype' can't be
>> reverted" looks better than "'NoneType' is not callable".
>>
>> I agree with this. A new collections.abc.Reversible (interposed between
>> Iterable and Sequence) would be a potential substitute for Ordered, and
>> would have this problem, which would be solvable by treating __reversed__ =
>> None specially, just like __hash__ = None. And I'm pretty sure it would
>> come up in practice (see issue 25864). And once we've got two or three
>> special methods doing this instead of one, making it more general does
>> sound like a good idea. So, if we need Reversible as a substitute for
>> Ordered, then I think we want the general "is None" test.
>>
>> But I'm still not sure Reversible is a good substitute for Ordered
>> (again, consider an infinitely long collection, or just a lazy proxy that
>> doesn't compute values until needed and doesn't know its length in
>> advance--they're clearly ordered, and just as clearly not reversible), and
>> I'm not sure we actually need either Reversible or Ordered in the first
>> place.
>>
>> _______________________________________________
>> Python-ideas mailing list
>> Python-ideas at python.org
>> https://mail.python.org/mailman/listinfo/python-ideas
>> Code of Conduct: http://python.org/psf/codeofconduct/
>>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.python.org/pipermail/python-ideas/attachments/20151226/55d47ba5/attachment-0001.html>