Does MegaLinter support flake8-cognitive-complexity?

[kcfedun-fincad]

Hello there,

Does MegaLinter support flake8-cognitive-complexity? If not, can it be extended to do so? perhaps a new PYTHON_FLAKE8_ variable could be used to enable/disable these checks?

https://github.com/Melevir/flake8-cognitive-complexity

Many thanks.

[kcfedun-fincad]

I've applied a bit of a workaround for now using the following code, but a more sustainable solution would be ideal.

.mega-linter.yml
---
...

PYTHON_FLAKE8_PRE_COMMANDS:
  - command: |-
      echo $PWD
      cd /venvs/flake8 \
      && source bin/activate \
      && PYTHONDONTWRITEBYTECODE=1 pip3 install --no-cache-dir flake8-cognitive-complexity \
      && deactivate

[nvuillam]

If we add it by default with flake8 installation, will it enforce new rules by default for all repos ?

[Kurt-von-Laven]

Can you elaborate on what is unsustainable about this approach? If you hadn't provided it, using PYTHON_FLAKE8_PRE_COMMANDS to install the desired flake8 plugins is what I would have recommended (and what we have been doing). For my education, did you encounter issues with a simple pip install flake8-cognitive-complexity? For the sake of folks who may copy and paste this snippet in the future, some may prefer a standalone shell script that is called from PYTHON_FLAKE8_PRE_COMMANDS. I am a big fan of MegaLinter's value of minimizing the configuration burden on its users, but it feels quite aggressive for MegaLinter to decide which of the many available flake8 plugins to run on our users' source code. This particular plugin looks quite appealing to me personally, but at face value it doesn't appear to be very mature or regularly maintained. It also may be viewed as redundant with and/or contradictory to the more popular mccabe plugin, for example.

[Kurt-von-Laven]

@nvuillam, there are some exceptions, but yes, most flake8 plugins will run by default if both they and flake8 are installed.

[kcfedun-fincad]

Can you elaborate on what is unsustainable about this approach? If you hadn't provided it, using PYTHON_FLAKE8_PRE_COMMANDS to install the desired flake8 plugins is what I would have recommended (and what we have been doing). For my education, did you encounter issues with a simple pip install flake8-cognitive-complexity? For the sake of folks who may copy and paste this snippet in the future, some may prefer a standalone shell script that is called from PYTHON_FLAKE8_PRE_COMMANDS. I am a big fan of MegaLinter's value of minimizing the configuration burden on its users, but it feels quite aggressive for MegaLinter to decide which of the many available flake8 plugins to run on our users' source code. This particular plugin looks quite appealing to me personally, but at face value it doesn't appear to be very mature or regularly maintained. It also may be viewed as redundant with and/or contradictory to the more popular mccabe plugin, for example.

My sustainability concerns arise in cases where MegaLinter could decide to change the /venvs/flake8 directory or how it manages pip installations. That would have a trickle-down effect on my custom installs.

I did try running pip3 install --no-cache-dir flake8-cognitive-complexity in the PYTHON_FLAKE8_PRE_COMMANDS (without cd-ing into the /venvs/flake8 dir and activating the virtualenv. This resulted in none of the cognitive-complexity check being run. This implies that the pip package must be installed in the flake8 venv.

I wouldn't want other users to be forced to use the checks provided by the pip extensions, be it mccabe, cognitive-complexity, or others. Ideally there would be a switch/variable that could be used to enable them if the user so chooses.

[nvuillam]

It depends if we want plugins rules to be considered as standard in megalinter:)
A good test could be to create a pull request, and as megalinter job does own ml code linting, wewould see the gap in ml code with such plugin :)

[nvuillam]

(basically, if it triggers 657 errors within MegaLinter own code, I'll prefer to let it as an option and just add in the doc about how to add such plugin instead of adding/activating it by default ^^
The best would be to have a ML config var to enable additional rules (which would be disabled by default ^^)

[Kurt-von-Laven]

I did try running pip3 install --no-cache-dir flake8-cognitive-complexity in the PYTHON_FLAKE8_PRE_COMMANDS (without cd-ing into the /venvs/flake8 dir and activating the virtualenv. This resulted in none of the cognitive-complexity check being run. This implies that the pip package must be installed in the flake8 venv.

I am unable to reproduce this. Installing flake8 plugins into the virtualenv might be better practice, but at least I do observe that flake8-bugbear runs when installed via a simple pip install. If you want to use the virtualenv, it may at least be possible to remove the cd by specifying the absolute rather than relative filepath? You can ask flake8 which plugins it picked up via PYTHON_FLAKE8_ARGUMENTS: --version.

[nvuillam]

@kcfedun-fincad @Kurt-von-Laven I simplified the way to add pip plugins in venvs :)

PRE_COMMANDS:
  - command: pip install flake8-cognitive-complexity
    venv: flake8 # Will be run within flake8 python virtualenv. There is one virtualenv per python-based linter, with the same name

https://github.com/oxsecurity/megalinter/actions/runs/3370888552/jobs/5592370481

Already available in beta, and soon in next minor release :)

[Kurt-von-Laven]

That is quite a nice solution 👍.