Add pathlib.chown method

[vajrasky]

BPO	20779
Nosy	@pitrou, @vajrasky, @sedrubal, @eamanu, @y0urself
PRs	gh-64978: Add chown() to pathlib.Path #31212
Files	add_chown_to_pathlib.patch

^{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 = None
created_at = <Date 2014-02-26.09:51:50.192>
labels = ['type-feature', 'library', '3.11']
title = 'Add pathlib.chown method'
updated_at = <Date 2022-02-08.11:17:25.947>
user = 'https://github.com/vajrasky'

bugs.python.org fields:

activity = <Date 2022-02-08.11:17:25.947>
actor = 'y0urself'
assignee = 'none'
closed = False
closed_date = None
closer = None
components = ['Library (Lib)']
creation = <Date 2014-02-26.09:51:50.192>
creator = 'vajrasky'
dependencies = []
files = ['34227']
hgrepos = []
issue_num = 20779
keywords = ['patch']
message_count = 7.0
messages = ['212245', '212378', '212411', '212425', '396629', '398695', '412787']
nosy_count = 8.0
nosy_names = ['pitrou', 'cvrebert', 'neologix', 'vajrasky', 'sedrubal', 'eamanu', 'Tom Cook', 'y0urself']
pr_nums = ['31212']
priority = 'normal'
resolution = None
stage = 'patch review'
status = 'open'
superseder = None
type = 'enhancement'
url = 'https://bugs.python.org/issue20779'
versions = ['Python 3.11']

Linked PRs

• GH-64978: Add pathlib.Path.chown() #104183

[vajrasky]

For pragmatic and philosophical reasons, I would argue that we should add chown to pathlib library.

[pitrou]

I don't know about philosophical reasons ;-), but I don't think chown is very often used. Do you know of a context where it is?

[vajrasky]

Pragmatic reasons:

I use chown moderately often. Usually as root, I want to change the uid of the files after generating a bunch of files. But I almost never changed gid.

I am in two mind whether I should add lchown as well or not (just like lchmod) and make the signature of pathlib.chown same as os.chown (with third parameter, follow_symbolic_links).

If you close this as won't fix, I will not hold grudge against you, though. ;)

[pitrou]

Well, I don't know yet, but if we ever want chown() in pathlib, I think it should be higher-level and accept symbolic uids as well.

[TomCook]

+1 this.

I have a program that opens a UNIX socket as root for other processes to communicate with it. I need to set the permissions to 0o775 and set the owner gid to a specific group so that members of that group can communicate with the process. It's annoying to have to drop back to os.chown(str(path), ...) for this.

[eamanu]

Hello!

I also use it frequently, but looking this issue I can agree with pitrou that there aren't lot of people that use chown.

If I'm not wrong, there's a trend to move os to pathlib, perhaps chown should be in pathlib too :)

I can work on this.

[y0urself]

I would love to use chown for a pathlib Path, too.

It may not be used often, but if anyone want's to change all the os references from a file, it would be cool to have it on pathlib!

[barneygale]

I'm -1 on this: I don't think pathlib should start accumulating further functionality from os unless its used to implement high-level path behaviour (like walking directory trees).

[CAM-Gerlach]

To note, I can't confirm myself since I'm on a Windows machine at the moment, but os.chown is documented as natively supporting pathlib.Paths since Python 3.6. There's also the higher level shutil.chown which has a nicer API with more Pythonic argument handling and accepts user/group names as well as IDs, unlike the os version. As it simply passes path through to the underlying os.chown, it in turn supports pathlib.Paths as well.

This seems to address the major concrete issue expressed by users here (e.g. "It's annoying to have to drop back to os.chown(str(path), ...)", "I would love to use chown for a pathlib Path, too."), as well as exposing a nicer higher-level interface without the need to add yet another chown in pathlib.

(Sidenote: Perhaps, if there's interest, it's worth allowing passing through dir_fd or particularly follow_symlinks in shutil.chown to os.chown so the latter is a proper superset of the former, but that's an entirely separate issue of course and unrelated to pathlib.)