[issue23401] Add pickle support of Mapping views

[Serhiy Storchaka]

Changes by Serhiy Storchaka :


--
resolution:  -> rejected
stage: patch review -> resolved
status: open -> closed

___
Python tracker 

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

[issue23401] Add pickle support of Mapping views

[Guido van Rossum]

Guido van Rossum added the comment:

Please close.

--

___
Python tracker 

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

[issue23401] Add pickle support of Mapping views

[Serhiy Storchaka]

Serhiy Storchaka added the comment:

Can this issue be closed?

--

___
Python tracker 

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

[issue23401] Add pickle support of Mapping views

[Serhiy Storchaka]

Serhiy Storchaka added the comment:

OK, then I withdraw my proposition.

Thank you for your attention Raymond and Guido.

--

___
Python tracker 

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

[issue23401] Add pickle support of Mapping views

[Guido van Rossum]

Guido van Rossum added the comment:

No, it should just be up to the implementation. Though you might add a note
to the docs explaining that it's probably a misguided idea.

On Tue, Jun 7, 2016 at 2:23 AM, Serhiy Storchaka 
wrote:

>
> Serhiy Storchaka added the comment:
>
> Should we forbid pickling MappingView?
>
> And what about copying and deepcopying?
>
> --
>
> ___
> Python tracker 
> 
> ___
>

--

___
Python tracker 

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

[issue23401] Add pickle support of Mapping views

[Serhiy Storchaka]

Serhiy Storchaka added the comment:

Should we forbid pickling MappingView?

And what about copying and deepcopying?

--

___
Python tracker 

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

[issue23401] Add pickle support of Mapping views

[Raymond Hettinger]

Raymond Hettinger added the comment:

> Raymond, what is your opinion?

Whether and how to pickle should be at the discretion of a concrete class 
rather than a requirement for being a mapping.  For example, it may not make 
any sense for a persistent dictionary such as a FileDict or an SQLDict.  Also, 
my mental model of a mapping view is something that is transparent, a window to 
the world rather than the world itself.

--

___
Python tracker 

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

[issue23401] Add pickle support of Mapping views

[Guido van Rossum]

Guido van Rossum added the comment:

I find it extremely suspicious that if you pickle the keys of a large object it 
also pickles that object. These are views, not copied data, for a reason. I 
also take back that this is okay for concrete dict.

As for your reason (http://bugs.python.org/issue23401#msg267317), that doesn't 
explain the use case for pickling keys views. If you forbade pickling them 
would any user code break? I seriously doubt it. Who on earth would want to 
pickle a keys view on an OrderedDict? It doesn't contain any information other 
than the underlying object. These views exist as alternate APIs to dicts, not 
as objects by themselves.

--

___
Python tracker 

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

[issue23401] Add pickle support of Mapping views

[Serhiy Storchaka]

Serhiy Storchaka added the comment:

In this case we need to implement pickle support separately in OrderedDict and 
any Mapping subclass.

--

___
Python tracker 

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

[issue23401] Add pickle support of Mapping views

[Guido van Rossum]

Guido van Rossum added the comment:

I think it's fine to do this for concrete classes, I just don't think it
should be added to ABCs.

--Guido (mobile)

--

___
Python tracker 

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

[issue23401] Add pickle support of Mapping views

[Serhiy Storchaka]

Serhiy Storchaka added the comment:

Because I can! I just want to ensure that all basic types (for which it makes 
sense) are correctly and portable pickleable, especially collections and 
related classes. Note that currently Mapping views are pickleable (as most pure 
Python classes), but pickle exposes implementation detail (the _mapping 
attribute). For example pickled keys view of Python implementation of 
OrderedDict is not fully compatible with C implementation. I want to provide 
implementation agnostic and more efficient pickling.

>>> keys = collections.abc.KeysView({1: 2, 3: 4})
>>> keys.__reduce_ex__(3)
(, (,), 
(None, {'_mapping': {1: 2, 3: 4}}), None, None)

There are many ways to do this (msg235477, msg235493). I hesitate with choosing,

--

___
Python tracker 

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

[issue23401] Add pickle support of Mapping views

[Guido van Rossum]

Guido van Rossum added the comment:

Serhiy, why would you want to separately pickle such a view? It's a trivial 
wrapper around a dict/mapping.

What's the use case?

--

___
Python tracker 

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

[issue23401] Add pickle support of Mapping views

[Guido van Rossum]

Guido van Rossum added the comment:

Raymond, what is your opinion?

--

___
Python tracker 

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

[issue23401] Add pickle support of Mapping views

[Serhiy Storchaka]

Serhiy Storchaka added the comment:

The option 3 is now available with pickle protocols >= 2 (but less efficient 
with protocols < 4).

Adding pickle support of Mapping views doesn't add a requirement for all 
classes that purport to be mappings.  But if you implement pickle support of 
your mapping, its views become automatically pickleable.

--
versions: +Python 3.6 -Python 3.5

___
Python tracker 

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

[issue23401] Add pickle support of Mapping views

[Raymond Hettinger]

Raymond Hettinger added the comment:

Guido, is it your intention that the ABCs grow pickle support?  

I presume if this were implemented that it would become a requirement for all 
classes (and their subclasses) that purport to be mappings.  I thought that you 
had designed to the ABCs to stay focused only on the essential characteristics 
of mappings.

--
assignee:  -> gvanrossum
nosy: +gvanrossum

___
Python tracker 

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

[issue23401] Add pickle support of Mapping views

[Serhiy Storchaka]

Serhiy Storchaka added the comment:

Oh, and yet one option:

5) As a list or a tuple. `copyreg.__newobj__, (list,), None, 
iter(self._mapping.keys())` or `copyreg.__newobj__, (tuple, 
tuple(self._mapping))`.

--

___
Python tracker rep...@bugs.python.org
http://bugs.python.org/issue23401
___
___
Python-bugs-list mailing list
Unsubscribe: 
https://mail.python.org/mailman/options/python-bugs-list/archive%40mail-archive.com

[issue23401] Add pickle support of Mapping views

[Serhiy Storchaka]

Changes by Serhiy Storchaka storch...@gmail.com:


--
keywords: +patch
Added file: http://bugs.python.org/file38024/pickle_mapping_views.patch

___
Python tracker rep...@bugs.python.org
http://bugs.python.org/issue23401
___
___
Python-bugs-list mailing list
Unsubscribe: 
https://mail.python.org/mailman/options/python-bugs-list/archive%40mail-archive.com

[issue23401] Add pickle support of Mapping views

[Serhiy Storchaka]

New submission from Serhiy Storchaka:

Proposed patch adds pickle support of general Mapping views. See also 
issue23264.

--
components: Library (Lib)
messages: 235475
nosy: alexandre.vassalotti, pitrou, rhettinger, serhiy.storchaka, stutzbach
priority: normal
severity: normal
stage: patch review
status: open
title: Add pickle support of Mapping views
type: enhancement
versions: Python 3.5

___
Python tracker rep...@bugs.python.org
http://bugs.python.org/issue23401
___
___
Python-bugs-list mailing list
Unsubscribe: 
https://mail.python.org/mailman/options/python-bugs-list/archive%40mail-archive.com

[issue23401] Add pickle support of Mapping views

[Serhiy Storchaka]

Serhiy Storchaka added the comment:

There are many ways to pickle e.g. Mapping keys:

1) Default implementation (as an instances of Python class MappingKey). This is 
implementation depended and exposes private variable _mapping. Changing 
implementation of the keys() method will break compatibility.

2) As `self.__class__, (self._mapping,)`. It depends on implementation of 
keys() as MappingKey or MappingKey subclass. It doesn't very subclass friendly 
if keys() is more complex or MappingKey subclass requires additional parameters.

3) As `self._mapping.__class__.keys, (self._mapping,)`. Implementation 
agnostic, perhaps more efficient than option 4 because 
self._mapping.__class__.keys may be shared. But needs protocol 4.

4) As `self._mapping.keys, ()`. Implementation agnostic.

pickle_mapping_views.patch implements option 4, the patch in issue23264 
implements option 3 (it works for dict with all protocols, but needs protocol 4 
for dict subclasses). Perhaps implementations should be unified (use option 3 
with protocol 4 and option 3 otherwise).

--

___
Python tracker rep...@bugs.python.org
http://bugs.python.org/issue23401
___
___
Python-bugs-list mailing list
Unsubscribe: 
https://mail.python.org/mailman/options/python-bugs-list/archive%40mail-archive.com