Add option to force color output

[chadrik]

mypy ignores color_output=true if there is no tty, but in a CI context color is often supported and desired even though there's no tty. pytest solves this with a tri-state flag: --color=yes/no/auto, where color=auto is equivalent to mypy's color_output=true, and color=yes forces color on.

chad$ pytest --help | grep color
  --color=color         color terminal output (yes/no/auto).

I'm happy to submit a PR for this if you think it makes sense.

There is the issue of backward compatibility. We have a few options

• introduce a new flag --color=yes/no/auto and deprecate the old boolean one. This would let us maintain backward compatibility for a few versions while people switch over
• change the type of --color-output/--no-color-output to enum string. This would break backward compatibility, at least for the CLI. We might be able to hack something to work for config files.

[msullivan]

There is an undocumented environment variable MYPY_FORCE_COLOR that will force color.

At the very least that should get documented but probably we should do the true/false/auto thing.

Probably I'd want to maintain compatibility at least for a few versions..
Hm, is it possible with argparse to support both --color and --color=yes syntaxes?

[chadrik]

There is an undocumented environment variable MYPY_FORCE_COLOR that will force color.

Thanks for the heads up!

Probably I'd want to maintain compatibility at least for a few versions..
Hm, is it possible with argparse to support both --color and --color=yes syntaxes?

I don't think there's anything natively supported for it, so it's probably a question of how difficult it is to hack.

[ilevkivskyi]

I agree the three-state (with default set to auto) will be better than two state plus an env variable.

About backwards compatibility, it is unlikely that many people use --color-output right now because it is a default. It is more likely that they use --no-color-output.

Another thing that I didn't try but might work is to have four valid choices for --color-output: 'yes', 'no', 'auto', and '' (empty string). Maybe it is possible in argparse, and then we can make '' an alias to 'auto', and --no-color-output an alias to --color-output=no?

[ilevkivskyi]

(edit: fixed current flag name)

[wkschwartz]

Workaround for those arriving here because they're looking for color output from Mypy when run via Tox: setting passenv = TERM solved it for me.

[PhilippSelenium]

Any updates on this?

When I run mypy with tox I get no colour output regardless of which environment variables I pass:

passenv =
    MYPY_FORCE_COLOR
    PY_COLORS

@wkschwartz passenv = TERM had some unwanted side effects in my case.

[yowsa]

Any news on this? Color is working perfectly when I run mypy manually in PyCharm but doesn't show up in my pre-commit hooks when using iTerm. I tried setting MYPY_FORCE_COLOR=1 but it had no effect.

[abravalheri]

On #7771 (comment)

... passenv = TERM had some unwanted side effects in my case.

Hi @PhilippSelenium, in my setup tox seems to be working with passenv, but I am still unsure if I should include this workaround in the CI environment or not... Would you mind to share which unwanted side effects were you facing?

[PhilippSelenium]

@abravalheri I cannot remember. I just updated tox and mypy and added TERM and had no problems ...

[b0g3r]

@yowsa in my case it fixed with explicit TERM=xterm-color as environment variable in pycharm run configuration