[issue27598] Add SizedIterable to collections.abc and typing

[Guido van Rossum]

Guido van Rossum added the comment:

ReIterable and Collection are different concepts.
- ReIterable just implements __iter__ but promises certain semantics.
- Collection implements __iter__, __contains__ and __len__.

If we had ReIterable I would probably decree that Collection inherits from it 
(practicality beats purity).

But I don't think ReIterable by itself would do much good; apart from a few 
"show-off" hacks, any reasonable ReIterable would also be able to implement 
__contains__ and __len__ easily.

And I would certainly call range() a Collection.

--

___
Python tracker 

___
___
Python-bugs-list mailing list
Unsubscribe: 
https://mail.python.org/mailman/options/python-bugs-list/archive%40mail-archive.com

[issue27598] Add SizedIterable to collections.abc and typing

[Bar Harel]

Bar Harel added the comment:

I still believe "Reiterable" better demonstrates the concept.

When you request a Reiterable as a function parameter or assert if something is 
a Reiterable the other side knows exactly what you mean.

A "Collection" is way more ambiguous - if you create an object that acts like 
range() but you can cycle over it more than once I wouldn't exactly call it a 
collection. I would though call it a Reiterable and it would be clear for any 
Python programmer familiar with the concept of iterators.

I believe this is a funny case in which the naming is more important than the 
implementation as it will turn into a term or concept that will be further used 
in many places to come.

--
nosy: +bar.harel

___
Python tracker 

___
___
Python-bugs-list mailing list
Unsubscribe: 
https://mail.python.org/mailman/options/python-bugs-list/archive%40mail-archive.com

[issue27598] Add SizedIterable to collections.abc and typing

[Guido van Rossum]

Guido van Rossum added the comment:

No on adding __eq__.

If another core dev is available please go ahead and commit.

--
title: Add Collection to collections.abc and typing -> Add SizedIterable to 
collections.abc and typing

___
Python tracker 

___
___
Python-bugs-list mailing list
Unsubscribe: 
https://mail.python.org/mailman/options/python-bugs-list/archive%40mail-archive.com

[issue27598] Add SizedIterable to collections.abc and typing

[Neil Girdhar]

Neil Girdhar added the comment:

Given issue http://bugs.python.org/issue27802, it might be worth considering 
that all Collections implement __eq__ and __ne__, so maybe these should be 
abstract methods on Collection?

--

___
Python tracker 

___
___
Python-bugs-list mailing list
Unsubscribe: 
https://mail.python.org/mailman/options/python-bugs-list/archive%40mail-archive.com

[issue27598] Add SizedIterable to collections.abc and typing

[Ivan Levkivskyi]

Ivan Levkivskyi added the comment:

Thank you Neil, I agree that Sequence is a reversible collection.

I combined you documentation patch with updated _collections_abc and updated 
test_functools (it tests MRO) into a single patch.

--
Added file: http://bugs.python.org/file44148/collection_neil_combined.diff

___
Python tracker 

___
___
Python-bugs-list mailing list
Unsubscribe: 
https://mail.python.org/mailman/options/python-bugs-list/archive%40mail-archive.com

[issue27598] Add SizedIterable to collections.abc and typing

[Neil Girdhar]

Neil Girdhar added the comment:

(added the documentation changes)

--
Added file: http://bugs.python.org/file44146/doc_changes.diff

___
Python tracker 

___
___
Python-bugs-list mailing list
Unsubscribe: 
https://mail.python.org/mailman/options/python-bugs-list/archive%40mail-archive.com

[issue27598] Add SizedIterable to collections.abc and typing

[Neil Girdhar]

Neil Girdhar added the comment:

Great patch.  Shouldn't Sequence be a "Reversible, Collection"?

--

___
Python tracker 

___
___
Python-bugs-list mailing list
Unsubscribe: 
https://mail.python.org/mailman/options/python-bugs-list/archive%40mail-archive.com

[issue27598] Add SizedIterable to collections.abc and typing

[Guido van Rossum]

Guido van Rossum added the comment:

LGTM, but I'm on vacation through Monday.

--

___
Python tracker 

___
___
Python-bugs-list mailing list
Unsubscribe: 
https://mail.python.org/mailman/options/python-bugs-list/archive%40mail-archive.com

[issue27598] Add SizedIterable to collections.abc and typing

[Ivan Levkivskyi]

Ivan Levkivskyi added the comment:

I submit a simple solution in line with __subclasshook__ of other ABCs.
I added several tests, it looks like it is doing everything right.

Please review.

--
keywords: +patch
Added file: http://bugs.python.org/file44144/collection.diff

___
Python tracker 

___
___
Python-bugs-list mailing list
Unsubscribe: 
https://mail.python.org/mailman/options/python-bugs-list/archive%40mail-archive.com

[issue27598] Add SizedIterable to collections.abc and typing

[Guido van Rossum]

Guido van Rossum added the comment:

I thought about this some more. It won't work because when a class method
is called via super(), 'cls' is still set to the derived class (and in
general for class methods that's the right thing to do, just as it is for
regular methods that self is still the actual object). If you read my
previous long comment you'll understand that all the __subclasshook__
methods carefully return NotImplemented immediately if their 'cls' argument
is not the class where they are defined. So it won't work.

I also don't see the use case (you're not supposed to do this at home,
basically).

--

___
Python tracker 

___
___
Python-bugs-list mailing list
Unsubscribe: 
https://mail.python.org/mailman/options/python-bugs-list/archive%40mail-archive.com

[issue27598] Add SizedIterable to collections.abc and typing

[Guido van Rossum]

Guido van Rossum added the comment:

Why don't you give it a try? I'd be happy to review a patch from you.

--

___
Python tracker 

___
___
Python-bugs-list mailing list
Unsubscribe: 
https://mail.python.org/mailman/options/python-bugs-list/archive%40mail-archive.com

[issue27598] Add SizedIterable to collections.abc and typing

[Neil Girdhar]

Neil Girdhar added the comment:

@gvanrossum is there any reason that subclasshook is implemented by overriding 
instead of cooperation?  E.g.,:

class Sized(metaclass=ABCMeta):

@classmethod
def __subclasshook__(cls, C):
return (super().__subclasshook__(C) and 
any("__len__" in B.__dict__ for B in C.__mro__))


etc.  And then Collection does not need to implement subclasshook since its 
base classes cooperate?

--
nosy: +neil.g

___
Python tracker 

___
___
Python-bugs-list mailing list
Unsubscribe: 
https://mail.python.org/mailman/options/python-bugs-list/archive%40mail-archive.com

[issue27598] Add SizedIterable to collections.abc and typing

[Guido van Rossum]

Guido van Rossum added the comment:

Ping -- I'd really like to see this happening, with "Collection" as the name.

--

___
Python tracker 

___
___
Python-bugs-list mailing list
Unsubscribe: 
https://mail.python.org/mailman/options/python-bugs-list/archive%40mail-archive.com

[issue27598] Add SizedIterable to collections.abc and typing

[Sven R. Kunze]

Sven R. Kunze added the comment:

I am way too late to the discussion but I also support the term "collections". 
I think it also helps developers coming from a C# background:

https://msdn.microsoft.com/library/92t2ye13(v=vs.110).aspx

--
nosy: +srkunze

___
Python tracker 

___
___
Python-bugs-list mailing list
Unsubscribe: 
https://mail.python.org/mailman/options/python-bugs-list/archive%40mail-archive.com

[issue27598] Add SizedIterable to collections.abc and typing

[Ivan Levkivskyi]

Changes by Ivan Levkivskyi :


--
nosy: +levkivskyi

___
Python tracker 

___
___
Python-bugs-list mailing list
Unsubscribe: 
https://mail.python.org/mailman/options/python-bugs-list/archive%40mail-archive.com

[issue27598] Add SizedIterable to collections.abc and typing

[Guido van Rossum]

Guido van Rossum added the comment:

Regarding the design of __subclasshook__, these two flaws kind of cancel each 
other out. The idea is that if you have a "one-trick pony" class whose 
issubclass() check must verify that the purported sublass implements a specific 
method (e.g. __hash__), you don't want that subclass check to be inherited by a 
subclass. I suppose perhaps this ought to have been done by making 
ABCMeta.__subclasscheck__ check for the __subclasshook__ in the class dict, but 
(for reasons I've forgotten) it's not doing it that way, instead just calling 
cls.__subclasscheck__. All known __subclasshook__ implementations (those in 
collections.abc anyway :-) compensate for this by returning NotImplemented if 
the class for which they are being called isn't exactly the class in which they 
are defined. The final piece of the puzzle is object.__subclasshook__(), which 
always returns NotImplemented.

(NOTE: When reading the above paragraph, be careful to distinguish between 
__subclasscheck__, which is invoked by issubclass() and isinstance(), and 
__subclasshook__, which is only invoked by ABCMeta.__subclasscheck__().)

Here's an example showing why we don't want __subclasshook__ to be inherited. 
Suppose we have a class SupportsInt, like this:

class SupportsInt(ABC):
@classmethod
def __subclasshook__(cls, C):
return hasattr(C, '__int__')

Now suppose we had a concrete class inheriting from this:

class MyInteger(SupportsInt):
def __int__(self):
return 0

Now, alas, everything with an __int__ method is considered to be a MyInteger, 
for example isinstance(0, MyInteger) returns True.

So what should Collection.__subclasshook__ do? It could do something like this:

class Collection(Set, Iterable, Container):
@classmethod
def __subclasshook__(cls, C):
if cls is not Collection:
return NotImplemented
for base in Set, Iterable, Container:
ok = base.__subclasshook__(C)
if ok != True: return False
return True

(Untested, hopefully you get the idea. Hope this helps.)

--

___
Python tracker 

___
___
Python-bugs-list mailing list
Unsubscribe: 
https://mail.python.org/mailman/options/python-bugs-list/archive%40mail-archive.com

[issue27598] Add SizedIterable to collections.abc and typing

[Raymond Hettinger]

Raymond Hettinger added the comment:

In putting together a patch for this, I noticed that the __subclasshook__ 
classmethods don't compose well.   I'm unclear about the semantics about what 
these hooks are supposed to mean for subclasses and whether there is a bug in 
the existing code for Set, Sequence, and Mapping.  Each of those inherits from 
Sized, Iterable, and Container, but only the Sized.__subclasshook__ is visible 
to the subclass.  Also, the code for Sized.__subclasshook__ has an identity 
check for Sized which doesn't seem correct from the point-of-view of the 
subclasses.

--

___
Python tracker 

___
___
Python-bugs-list mailing list
Unsubscribe: 
https://mail.python.org/mailman/options/python-bugs-list/archive%40mail-archive.com

[issue27598] Add SizedIterable to collections.abc and typing

[Guido van Rossum]

Guido van Rossum added the comment:

It would be a deviation from existing practice for a collection ABC to
exist in typing but not in collections.abc, and it would mean the
inheritance tree for the types in typing.py would be different from
that in collections.abc.

These ABCs, while not so useful by themselves, are building blocks for
the more useful ABCs. They also give people the right terminology.
That arguably is the biggest problem here -- it's clear that people
desperately want to be able to talk about the common API for sets and
sequences, which happens to be __len__, __iter__ and __contains__. But
it's also clear that people often believe that that common API is
called Iterable, which it isn't.

--

___
Python tracker 

___
___
Python-bugs-list mailing list
Unsubscribe: 
https://mail.python.org/mailman/options/python-bugs-list/archive%40mail-archive.com

[issue27598] Add SizedIterable to collections.abc and typing

[Raymond Hettinger]

Raymond Hettinger added the comment:

Does all of the recognition patterns in typing.py also need to be in 
collections.abc? 

AFAICT, the current collections.abc classes for Sized, Iterable, and Container 
almost never get used (possibly because they don't provide any useful mixin 
methods).  In the interest of maximizing the ratio of useful ABCs (like 
MutableMapping), could we defer expanding collections.abc until we've seen some 
real use cases?

--
nosy: +rhettinger

___
Python tracker 

___
___
Python-bugs-list mailing list
Unsubscribe: 
https://mail.python.org/mailman/options/python-bugs-list/archive%40mail-archive.com

[issue27598] Add SizedIterable to collections.abc and typing

[Guido van Rossum]

Guido van Rossum added the comment:

I'm thinking that it should be called Collection after all.

--Guido (mobile)

--

___
Python tracker 

___
___
Python-bugs-list mailing list
Unsubscribe: 
https://mail.python.org/mailman/options/python-bugs-list/archive%40mail-archive.com

[issue27598] Add SizedIterable to collections.abc and typing

[Brett Cannon]

New submission from Brett Cannon:

See the discussion on python-ideas entitled "An ABC representing "Iterable, 
Sized, Container"".

--
components: Library (Lib)
messages: 271088
nosy: brett.cannon, gvanrossum
priority: normal
severity: normal
stage: test needed
status: open
title: Add SizedIterable to collections.abc and typing
versions: Python 3.6

___
Python tracker 

___
___
Python-bugs-list mailing list
Unsubscribe: 
https://mail.python.org/mailman/options/python-bugs-list/archive%40mail-archive.com