`NpathMetric`'s testability impeded by intertwining `mvn` process execution with output analysis

[ivanovmg]

It seems reasonable to refactor NpathMetric towards better testability.

The problem is that this class runs mvn process under the hood
and then analyzed the output produced.
Consider the option of splitting the running mvn part
from the actual analysis of the process output.

Then we can test thoroughly the part related to the analysis of the output file,
while the part related to running the process will be a part of a simple imperative shell.

@ivanovmg ideally, we should change the implementation of npath/main.py and remove the dependency on Maven. It should not be difficult to implement it in Python.

Originally posted by @yegor256 in #702

[ivanovmg]

@yegor256 please consider this issue. Xref: #702

[yegor256]

@ivanovmg I would rather get rid of mvn entirely. It shouldn't be hard to implement NPath calculation with pure Python, since we have the AST available.

[0crat]

@ivanovmg 🎉 Awesome bug report! You've earned +12 points, boosting your balance to +112. Your keen eye helps elevate our project's quality. Keep spotting things that don't look quite right and don't hesitate to report them. Your contributions are invaluable!

[ivanovmg]

@yegor256 I could start with implementing NPathMetric without use of mvn. But we will have to split this task into several ones.
I will start with if and if-else statements. I suppose, I should create a separate ticket on this.

[yegor256]

@ivanovmg I agree, let's move forward incrementally

[kodsurfer]

@ivanovmg @yegor256 hello, is this issue completed or need some help?

[ivanovmg]

@kodsurfer there is quite some stuff to do.
From the top of my head, currently MvnFreeNPathMetric works fine when it comes to finding a total NPath for all methods in a class (https://github.com/cqfn/aibolit/blob/master/aibolit/metrics/npath/main.py#L93). So, when it comes to computing NPath, it works OK, but the resulting output is probably not what we ultimately expect.

It seems reasonable to provide the report on NPath metric for a given file in the following format, maybe:

- filename
  - Class1
    - method1 - 1
    - method2 - 5
    - method3 - 2
  - Class2
    - method1 - 2
    - method2 - 3
    - method3 - 1

@yegor256 what's your take on the output format?