Repository navigation
Add pathlib.chown method #64978
Description
| BPO | 20779 |
|---|---|
| Nosy | @pitrou, @vajrasky, @sedrubal, @eamanu, @y0urself |
| PRs | chown() to pathlib.Path #31212 |
| Files |
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']
Activity
vajrasky commentedon Feb 26, 2014
For pragmatic and philosophical reasons, I would argue that we should add chown to pathlib library.
pitrou commentedon Feb 27, 2014
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 commentedon Feb 28, 2014
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 commentedon Feb 28, 2014
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 commentedon Jun 28, 2021
+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 commentedon Aug 1, 2021
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 commentedon Feb 7, 2022
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 commentedon Jan 8, 2023
I'm -1 on this: I don't think pathlib should start accumulating further functionality from
osunless its used to implement high-level path behaviour (like walking directory trees).CAM-Gerlach commentedon Jan 8, 2023
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 levelshutil.chownwhich has a nicer API with more Pythonic argument handling and accepts user/group names as well as IDs, unlike theosversion. As it simply passespaththrough to the underlyingos.chown, it in turn supportspathlib.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 anotherchowninpathlib.(Sidenote: Perhaps, if there's interest, it's worth allowing passing through
dir_fdor particularlyfollow_symlinksinshutil.chowntoos.chownso the latter is a proper superset of the former, but that's an entirely separate issue of course and unrelated topathlib.)AlexWaygood commentedon Jan 8, 2023
I agree with @barneygale, @CAM-Gerlach, and @brettcannon (Brett voiced his opinions on the PR here: #31212 (comment)). There seems to be little point adding this method, if it's just a thin wrapper around
os.chown().24 remaining items
barneygale commentedon May 4, 2023
PR available:
pathlib.Path.chown()#104183Merge branch 'main' into pythongh-64978-add-pathlib-path-chown
nineteendo commentedon Jun 2, 2024
Barney, what's the status here?
barneygale commentedon Jun 2, 2024
I'm still undecided as to whether this is a good idea.
IIUC we can't make
shutilusepathlibbecause pathlib strips trailing slashes, and soshutil.chown('not_a_dir/', ...)would succeed where previously it failed withNotADirectoryError. This is fixable if we introduce a private superclass ofPaththat skips normalization, but there's little appetite for that right now.Consequently we'd either need to make
pathlibuseshutil, which feels wrong to some core devs and I'm trying to avoid if I can, or we'd just copy-paste the shutil implementation into pathlib, which I feel pretty "meh" about too.barneygale commentedon Jun 2, 2024
#73991 is the more important shutil/pathlib issue. I hope to solve that first, and by doing so, shed some light on whether this request is worth doing or not.
barneygale commentedon Jul 3, 2024
I'm going to close this as "won't do". It turns out that
Path.chown()won't be helpful for copying file metadata in #73991, and I can see no other compelling reason to add it. Plus it's POSIX-only and usually needs root. Useos.chown()orshutil.chown()instead please! And feel free to leave a comment if you disagree and would like to see this re-opened.