MNT Remove encoding declarations: # -*- coding: utf-8 -*-
#21260
Conversation
In Python 3, the default source file encoding is UTF-8.
571474c to
86e5679
Compare
# -*- coding: utf-8 -*-
|
While for Python it's indeed not necessary, editors/IDE will also need an indication what encoding it is (and I guess otherwise could use the system encoding). https://stackoverflow.com/a/14083123 For instance on Windows the default encoding is not UTF-8, So I think it's probably safer to keep those headers. |
|
The default encoding of Python source files and the default encoding associated to the system locale are two different things.
Do you know of a broken IDE that would understand the encoding declaration Currently, some pure ASCII files do have the encoding declaration and some UTF-8 files lack it. No one complains about it, this is a strong indication the encoding declaration is probably NOT important. Also the current situation is inconsistent. Either all source files should start with the encoding declaration, or all UTF-8 files (and only them) should start with the encoding declaration - whatever you prefer. |
|
A majority of the files I have modified seem to be ASCII files: |
|
Here are the UTF-8 And here are the UTF-8 |
Well the question is more how confident are you that IDE will rely on the file extension to determine the encoding :) For instance in https://www.jetbrains.com/help/pycharm/encoding.html I see nothing about it being a Python or a non Python file. I guess it depends on the default settings. One could hope they are reasonable, but personally I don't know. I just would rather be careful here. So a confirmation that UTF-8 is used by default, particularly on Windows for some of the more popular browsers would be useful. In VS Code the default is indeed UTF-8. It's probably fine, but doesn't hurt to double check. |
|
I don't have Windows to double-check. Instead I can add Alternatively we can close this PR. |
|
Note that VS saves into UTF8-BOM by default, but we probably want to the avoid the BOM in a multi-platform context. |
Well the problem is that a file that is currently ASCII will become non ASCII as soon as one adds a non ASCII character. Let's keep it open. Maybe someone else has other opinions or could provide feedback on their IDE configuration. |
Then we need to add |
|
Note that large projects like NumPy live without the encoding declarations. It seems to be working well for them. |
Maybe they not not have authorship lines with "Gaël", "Müller" or "Loïc" or docstrings with "Schölkopf". |
|
+1 for keeping UTF-8 markers to avoid problems with editors on Windows and add them when needed on a case by case basis. |
The authorship lines are not relevant as such. The number of UTF-8 files and total files is similar in both projects, but I agree Scikit-learn has twice the NumPy proportion of UTF-8 files. Still, it's similar. In NumPy: In Scikit-learn: |
Just to be clear:
|
|
I don't know. Can someone with a windows machine try to edit a On Linux and macOS I am pretty sure that UTF-8 is always use by default nowadays. |
|
If you really want to take into account Windows editors, I think |
I checked with these 4 editors. VSCode, PyCharm, notepad++ use UTF-8 by default. Spider seems to be using ASCII by default but automatically switches to UTF-8 as soon as you add a non-ASCII character. In all cases, reading the file shows the characters correctly. I think we can safely remove the UTF-8 markers. |
|
(please don't ask me more experiments with these editors, I'm uninstalling 3 of them as I speak 😄) |
Co-authored-by: Jérémie du Boisberranger <34657725+jeremiedbb@users.noreply.github.com>
What does this implement/fix? Explain your changes.
In Python 3, the default source file encoding is UTF-8.
Any other comments?
This is a follow-up of #21246 which was limited to examples.