Python-Swift 互操作 (PythonKit)：更改默认标准输入编码

Question

I've imported a custom Python package into Swift using PythonKit, within a Swift Package (SPM) using XCode. The Python code at various points performs f = open("somefile","r") , and then f.readline() .

When invoked from Swift, the following error is raised from within Python :

File "/Users/user/[...]/lib/python3.8/site-packages/module/unihan_variants.py", line 115, in unihan_variants_dict
    line = f.readline()

...

Python exception: 'ascii' codec can't decode byte 0xc2 in position 121: ordinal not in range(128)

Investigating further, I see that the default stdin decoding is ascii instead of utf-8 :

let sys = Python.import("sys")
print(sys.stdin.encoding) // asciii
print(sys.stdout.encoding) // ascii

When I go to my original Python module, I find I can avoid the error in PythonKit by specifying f = open("somefile","r",encoding="utf-8") , but unfortunately I've just assumed utf-8 throughout all of my projects to date.

Is there a way to change default encoding for stdin and stdout from XCode or PythonKit?

Or is it necessary/advised to go back and specify utf-8 throughout all my Python code?

(This doesn't work) :

sys.stdin.encoding = "utf-8" //Python exception: readonly attribute: file /Users/brianparker/Library/Developer/Xcode/DerivedData/Morphology-efqracfjfhxtguetmduhdszyzezb/SourcePackages/checkouts/PythonKit/PythonKit/Python.swift, line 540

Further notes :

• print(platform.python_version()) returns 3.8.5
• My method for importing my Python package was by adding the path to site-packages within the virtualenv to sys.path in Swift/PythonKit
• print(sys.getdefaultencoding()) indicates 'utf-8'. It seems weird that this differs from stdin/stdout

---

Update :

I've managed to get sys.stdin.encoding to indicate "utf-8" by setting the PYTHONIOENCODING environment variable. However the same error gets thrown at readline()

Answer 1

Python's TextIOWrapper open 's default encoding isn't determined by the sys properties or environment variables proposed in my question. Instead, it defaults to locale.getpreferredencoding(False) .

My Python environment shows the following, which explains why I typically don't need to specify an encoding when opening a file :

>>> locale.getlocale()
('en_US', 'UTF-8')
>>> locale.getpreferredencoding(do_setlocale=False)
'UTF-8'

Python within Swift however does not have a default locale (at least when run on my machine):

let locale = Python.import("locale")
print(locale.getlocale()) // (None, None)

The following code will set the desired locale for Python within the Swift app, allowing TextIOWrapper to default to utf-8 when opening files :

let locale = Python.import("locale")
if locale.getlocale().tuple2 == (Python.None,Python.None) {
   locale.setlocale(locale.LC_ALL, locale: PythonObject(["en_US","UTF-8"]))
}

Note that there's a Python proposal from 2019 to make utf-8 TextIOWrapper's default encoding without checking locale , though I'm not sure if it will be adopted.