search

okfn-brasil / serenata-toolbox

📦 pip module containing code shared across Serenata de Amor's projects | ** Este repositório não recebe atualizações frequentes **
MIT License
154 stars 69 forks source link

Better config Coveralls for accuracy #75

Closed yrachid closed 7 years ago

yrachid commented 7 years ago

Hello, I hope it is not a problem to open an issue out of the blue like this.

Lately, @lipemorais was showing me this project and, as a way to try to start contributing to it, I thought it would be interesting to replace Coveralls by another quality measuring tool. My reasons behind such suggestion are in the fact that the Coveralls UI is very confusing, I can't quickly get a picture of what is actually going on in the project when looking at its dashboard. Besides, it is showing a quite imprecise metric at this moment, since it's also considering third-party libraries in its analysis, lowering the metrics to only 25%.

I believe that CodeClimate would be a better option to solve this problem. I've forked this project and configured its forked copy to be analyzed by CodeClimate. Below, it's possible to see a result of the first analysis:

screen shot 2017-05-24 at 12 57 19 am

Besides the code styling, it's also possible to harvest some metrics about testing coverage:

screen shot 2017-05-24 at 2 25 43 am screen shot 2017-05-24 at 2 25 59 am

I've also noticed that Jarbas makes use of both Coveralls and CodeClimate. Maybe, CodeClimate could be configured to be used as a test coverage reporter, replacing Coveralls there as well.

I would like to apologize in advance if this is not the right way to open a new issue.

cuducos commented 7 years ago

Hi @othman853,

Many thanks for you suggestion. Unfortunately I tend to disagree.

Coveralls UI is very confusing

This is a quite subjective matter — I use it a lot so I feel familiar in their dashboard.

it is showing a quite imprecise metric at this moment, since it's also considering third-party libraries

This is not a Coveralls problem, it's lack of configuration. We added it only 6 days ago and so we haven't fine tuned it yet. If we configure it in a similar way we did in Jarbas, this doesn't happen.

In spite of that Code Climate doesn't offer an analysis as complete as Landscape.io in terms of style, code smell and errors — specially when using it with veryhigh strictness: it help us with docstrings (important for #23 as I noted elsewhere), unused imports, thing accessed before creation, line breaks etc.

Anyway. I'm really glad you pointed is toward to some flaws in the configuration of our CI services, but I do not agree that changing would be better: after all the same settings needed now for Coveralls will also be needed with for Code Climate.

I'll let this issue open for debate for a while, but feel free to open an issue or PR fine tuning or Coveralls ; )

yrachid commented 7 years ago

Hello! Thanks for the answer! I believe that a simple Coveralls tuning would end up being more appropriate then. I'll check how it can be fixed and open a PR. Sounds good? Do you believe that a new issue should be open or we could open a PR to solve this current issue?

Once again, thank you very much! :)

cuducos commented 7 years ago

I'll check how it can be fixed and open a PR. Sounds good?

That's awesome ; )

Do you believe that a new issue should be open or we could open a PR to solve this current issue?

If you don't mind I can edit the title of this one — I think keeping the new issue togather with this discussion could be helpful. May I?

yrachid commented 7 years ago

Absolutely! Thank you for all the help!