Implemented skip violations

[dbalabka]

Initial implementation of #199

Scope

Implement possibility to skip class dependency violation in depfile.yml. There will be new configuration section skip_violations (maybe skipped_violations better?). The developer can configure it in the following way:

skip_violations:
  Library\LibClass:
    - Core\CoreClass
    - Core\AnotherClass

skip_violations section contains associative array where key (Library\LibClass) is the name of dependant class and values (Core\CoreClass, Core\AnotherClass) are dependency classes.

Future scope

There might be implemented possibility for PHP functions dependency as well. In this case, we can specify functions names instead of class names. PHP allows to create functions with names which can be equal to class names in the same namespace:

<?php

namespace Foo;

class bar {
    function __invoke()
    {
        echo __CLASS__ . PHP_EOL;
    }
}

function bar() {
    echo __FUNCTION__ . PHP_EOL;
}

bar();
(new bar())();

output:

Foo\bar
Foo\bar

So function names should be identified differently from clasess. We can sufix each function name with ():

skip_violations:
  Library\LibClass:
    - Foo\bar()
    - Foo\bar

TODO

• Implement class violations skipping possibility
• Unit tests
• Update documentation

[timglabisch]

thanks for trying to make deptrac better,

i am unsure about this PR. I would agree that we need some way to skip violations but there are a bunch of issues with this approach.

when i wrote deptrac i had the idea that a dependency is something abstract. in fact we (mostly) use deptrac on the "class"-level but if we think further dependencies has nothing to do with classes.

for example we could write a npath collector that collects every function that is "too complex". than you could write a rule to prevent some parts of your code to prevent on too complex code. You would end up with a dependency between (may) 2 functions, not classes.
skip_violations is only build on class level.

atm. i am unsure about how to solve this issue but i guess the approach is not generic enough.

[dbalabka]

@timglabisch agree with you.
Talking about classes and functions. I can change the configuration format to support both functions and classes to avoid configuration BC breaks. Or it might be improved in future releases until \SensioLabs\Deptrac\DependencyResult\DependencyInterface does support only classes.

When you integrate DepTrac in CI/CD with large code base (our case) all dependencies cannot be easily fixed, because it would take a lot of developer-hour. IMO DepTrac definitely should support violation skip possibility otherwise integration process might not be so smooth.

[dbalabka]

@timglabisch as you can see in #199 we have implemented a workaround using existing functionality, but not the best option. We would like to see jUnit support as well to see and control the amount of existing skipped violations. Also, an approach described in #199 provides a possibility to skip some class completely from an analysis, but in the scope of this PR I have implemented a possibility to skip exact violation which minimizes probability to introduce new violations.

[dbalabka]

@timglabisch any ideas? Should I continue work on PR or you will just close it?

[smoench]

I think this would be good start having a skipping mechanism. If there is somebody having the use case skipping functions/npath then they/we should start thinking about abstracting the mechanism.

@torinaki 👍 feel free to continue

[smoench]

@torinaki Some documentation in the README.md would be great :)

[dbalabka]

@smoench , @timglabisch it seems that PR is ready for review.

[DavidBadura]

except for the little things 👍

[smoench]

@torinaki Would be great if you could rebase on actual master. I extended the travis ci config to enforce static analysis. As soon as the pipeline gets green I will merge your PR :)

[smoench]

@torinaki It seems you have just a few coding style issues. They could be easily fixed with php-cs-fixer. Just download the latest phar-file and run following command php php-cs-fixer.phar fix. Thanks :)

[dbalabka]

@smoench CS is fixed

[smoench]

Thank you @torinaki