Turn contextlib.{redirect_stdout,suppress} into ContextDecorators

[anntzer]

BPO	22531
Nosy	@ncoghlan, @bitdancer, @anntzer

^{Note: these values reflect the state of the issue at the time it was migrated and might not reflect the current state.}

Show more details

GitHub fields:

assignee = None
closed_at = <Date 2014-10-01.14:55:08.557>
created_at = <Date 2014-10-01.09:50:23.843>
labels = ['library']
title = 'Turn contextlib.{redirect_stdout,suppress} into ContextDecorators'
updated_at = <Date 2014-10-01.14:55:08.555>
user = 'https://github.com/anntzer'

bugs.python.org fields:

activity = <Date 2014-10-01.14:55:08.555>
actor = 'ncoghlan'
assignee = 'none'
closed = True
closed_date = <Date 2014-10-01.14:55:08.557>
closer = 'ncoghlan'
components = ['Library (Lib)']
creation = <Date 2014-10-01.09:50:23.843>
creator = 'Antony.Lee'
dependencies = []
files = []
hgrepos = []
issue_num = 22531
keywords = []
message_count = 3.0
messages = ['228065', '228075', '228078']
nosy_count = 3.0
nosy_names = ['ncoghlan', 'r.david.murray', 'Antony.Lee']
pr_nums = []
priority = 'normal'
resolution = 'rejected'
stage = 'resolved'
status = 'closed'
superseder = None
type = None
url = 'https://bugs.python.org/issue22531'
versions = ['Python 3.4', 'Python 3.5']

[anntzer]

A small lib improvement suggestion could be to make contextlib.redirect_stdout and contextlib.suppress inherit from ContextDecorator.

As a side note, the source of contextlib has some classes inheriting explicitly from object while others don't, so perhaps this can be harmonized at the same time.

[bitdancer]

-100 on doing this for suppress. That would be exactly the kind of wrong-headed code that I was worried that context manager would invite.

[ncoghlan]

Aye, suppress is only intended for use around a single line of code. Using it for an entire function would be "OnError Resume Next" levels of poor coding style.

I'm also -1 on implicitly wrapping redirect_stdout around functions due to the immediate thread safety problem doing so introduces. The stdout redirection really only makes sense in a single-threaded scripting context, and using it as a decorator rather than inline makes it far too easy to inadvertently violate that constraint.

As far as the "explicitly inherit from object or not" goes, it wouldn't surprise me if that's just a "migrated from Python 2" vs "first introduced in Python 3" distinction. It's not something I would change solely for the sake of consistency.