capfd in binary mode?

[asottile]

I'd like something like this to work:

def test_thing(capfd_bytes):
    subprocess.check_call(('echo', '-e', '\\x98\\xfe'))
    out, err = capfd_bytes.readouterr()
    assert out == b'\x98\xfe'

My actual usecase is running tar as stdout and then inspecting the output

I'm hacking around with this (which works, but is undeniably fragile):

def test(capfd):
    ...
    capfd._capture.out.tmpfile.encoding = None
    capfd._capture.err.tmpfile.encoding = None
    out, err = capfd.readouterr()
    capfd._capture.out.tmpfile.encoding = 'UTF-8'
    capfd._capture.err.tmpfile.encoding = 'UTF-8'
    # ...

would pytest be amenable to a PR which enables something like this without hax?

[nicoddemus]

Hi @asottile!

I think so, yes, thanks for the offer. Do you have an API in mind?

(I could swear that this has been brought up before, but I can't find the issue...)

[asottile]

It was probably me before 😆

• Add readouterr_bytes for retrieving captured binary output #2279
• Make capsys more like stdio streams in python3. Resolves #1407. #2266

I'm not sure I necessarily like the api I wrote in #2279 (too easy to mix and match bytes / text and get unicode errors when mixing and matching). I think it would be better to add new separate capfdbinary and capsysbinary (feel free to improve the names here) which have the same interface as capfd and capsys?

[nicoddemus]

It was probably me before

Rá I knew this has been discussed before. 😆

I think it would be better to add new separate capfdbinary and capsysbinary (feel free to improve the names here) which have the same interface as capfd and capsys?

I think this might be the way to go, it might not be easy to change the stream types after it has been created, and having separate methods is clumsy as you noted.

From my POV please go ahead. 👍

[asottile]

via #2925 (I'll open another issue for capsysbinary so I can changelog)