Hi Dmitry,
You are right this is not a sorting, this is a ranking by two parameters
where both of them equal in priority.
For such task in math, we have an idea of "dominating subset".
It simply too understand by example:
P1, P2
E1 2, 2
E2 1, 2
E3 2, 1
E4 1, 1
Here element E1 dominate all other elements. E2 and E3 dominate E4, and E4
is a loser.
So we can do a ranking like that: E1 has rank 1, E2 and E3 have rank 2, and
E4 has rank 3.
I have done the same with JoCoCo results.
Where less coverage is better, and greate complexity is better (because we
are looking for classes with the problems).
In my list, there are all classes with rank 1 (except well covered, small,
and with a strange result of 0% coverage).
Finally, I have sorted this list by one value "coverage/complexity" -- just
for better viewing.
I have written R script for this, and if you think about some other
parameters or filtering -- let me know.
2018-05-14 20:54 GMT+03:00 Dmitry Pavlov <dpavlov....@gmail.com>:
> Hi Alexander,
>
> I did not quite understand how this list of top tests was obtained. It does
> not look like sorted.
>
> Can you clarify?
>
> Sincerely,
> Dmitriy Pavlov
>
> пн, 14 мая 2018 г. в 20:48, Александр Меньшиков <sharple...@gmail.com>:
>
> > Vyacheslav,
> > I have made a simple ranking for your data for highlighting classes which
> > more than other needs testing.
> >
> > Please take a look at the list which I have done (36 classes):
> >
> > https://drive.google.com/file/d/1RJrtnN77MeEWLTShFzqaLInCdOSb9
> WdC/view?usp=sharing
> >
> >
> > Below you can read how did it in details:
> >
> > First I have removed all small classes (less than 200 lines), classes
> with
> > more than 80% coverage, and with 0% (too strange result).
> > After that, I have extracted subset which dominates other classes by two
> > parameters: coverage and complexity.
> > It means all other classes have not less coverage and not greater
> > complexity, and also better at least at one of these parameters.
> >
> > Finally, I have sorted this list by value "coverage/complexity".
> >
>
