[Question] rationale for PLW1514 autofix / doc suggesting "locale" instead of "utf-8" ?

[mikaelarguedas]

Although it's been around for a long time, I discovered PLW1514 with the latest ruff release and have been surprised by the fix which made me curious about it.

ruff docs state the following:

Instead, consider using the encoding parameter to enforce a specific encoding. [PEP 597](https://peps.python.org/pep-0597/) recommends using locale.getpreferredencoding(False) as the default encoding on versions earlier than Python 3.10, and encoding="locale" on Python 3.10 and later.

And the autofix implemented does the following:

-open("test.txt")
+open("test.txt", encoding="locale")

Reading the PEP:
It looks like (from my reading of it):

• the source of the PEP is that without specifying encoding people would assume it is UTF-8 and would get bugs as UTF-8 is not default on Windows
• Python would like to default to UTF-8 for file opening in the future
• The encoding="locale" is recommended only when opening text files that have been saved using locale encoding but the overall recommendation (from my understanding) is to use encoding="utf-8"

Finally the example given by ruff differs from the autofix suggested as it gives utf-8 encoding as an example

I understand that encoding="locale" is the functional equivalent of not specifying the encoding for now, but from the pep it seems most users would want and potentially should using utf-8 instead.
So I was wondering if there is a specific rationale for steering users towards using "locale" over "utf-8" ?

Thanks! 🙏

@Avasam @qdegraaf @charliermarsh

[Avasam]

The short answer:
Fixing to "locale" is safer and will keep the code working as-is / as expected. Whilst resolving the lint issue about forward compatibility and implicit encoding.

That being said, I agree that in the documentation, it should steer towards trying to use utf-8 wherever possible.
The autofixed encoding could even be configurable 🤷‍♂️💡

Also after doing this migration in setuptools, I realised that the current suggestion of using locale.getpreferredencoding(False) on older python doesn't help at all resolving EncodingWarning as you just end up with a different warning. Imo it should instead be "locale" if sys.version >=(3,10) else None