On 29 June 2010 18:37, Adrien Grand <[email protected]> wrote:
> Hello Hyrum,
>
> On Mon, Jun 28, 2010 at 1:20 PM, Hyrum K. Wright
> <[email protected]> wrote:
>>> But, beg your pardon, what's so "heavy-weight" with RAT? I can't see
>>> any advantages of a Python implementation, other than it would be
>>> preferrable to Python developers.
>>
>> In our discussion, Robert and I observed that improving and extending
>> RAT as it is today can be quite a chore.  He explained the origins of
>> RAT as "and experiment in software architecture" and that it had
>> horribly gone wrong.  The architecture is way over-engineered for the
>> core mission of RAT, but because the tool proved useful, it was
>> adopted anyway.  mouse is an attempt to greatly reduce the amount of
>> code needed to accomplish that mission, and be much easier to extend
>> down the road.
>>
>> Another goal of mouse is to be easily incorporated into a post-commit
>> hook, so that it could audit the files being modified during each
>> commit.
>
> I'm not sure using Python as a programming language would make it
> simpler : Many people from the Java world enjoy audit tools such as
> CPD, Findbugs, Cobertura and so on, but they will only use a tool like
> RAT if it integrates with their build and integration tools such as
> Ant, Maven and Hudson. I have the feeling that you may waste much more
> time developing Maven/Ant plug-ins that call Python code (need to fork
> a process, what if Python is not installed?) than what you will gain
> by writing the core in Python. Another solution would be to rely on
> Jython to seemlessly integrate your code inside a Maven/Ant plug-in,
> but it looks more like a magic trick than the right thing to do, in my
> opinion.
>
> I admit that my arguments may sound a little bit Java-centric, but I
> think that there are many potential users of RAT among Java developers
> because of their love for audit tools. :-)

I'd just add that RAT currently has Ant and Maven interfaces which are
likely to have been used by lots of people/projects.

So Mouse needs to support any such interfaces.

Furthermore any change to the API which requires users to change their
scripts will not be popular and will hinder takeup.

> --
> Adrien Grand
>

Reply via email to