[core] How to wrap CPD as a rule

[rsoesemann]

This is more a discussion item that an issue, but I post it here to mention the experts (e.g. @adangel @jsotuyod ) to get some feedback.

I am the creator of the Apex language module and of ApexMetrics which wraps PMD and makes it available to the Code Climate platform for code analysis. PMD and CodeClimate are a perfect match as both work on a per file basis and know the concept of issues of a certain severity.

Besides PMD I am a heavy user of the Copy Paste Detector that is part of PMD. But only as a local standalone tool. I would like to create a CopyPasteRule instead and wonder if that is doable?

As nobody has done it in the last 10 years I fear its just not doable because of tthe way PMD works. I mean Visitor pattern, issues checked per file.

What do you think?

[jsotuyod]

This one is tough.

There are 2 main differences on how PMD and CPD work that makes this hard:

• CPD works on tokens, PMD on the AST. The AST is a higher level of abstraction, and the codebases for each are not fully shared (that's why you can support CPD without having PMD rules).
• CPD checks token sequences even from different files. For PMD each file can be analyzed on it's own. Several parts of PMD take this for granted and make use of it. That's why we could build an analysis cache to speed up PMD, but not for CPD. This is also used when running multithreaded analysis on PMD.

There are ideas around to bridge this gap (see wiki) but they are early and will probably not do so fully. Having CPD as a PMD rule would require forcefully disabling the analysis cache, and is probably not much to be gained there.

What may be possible, is having a run option for running CPD as part of PMD, and have found violations redirected to a PMD report rather to the standard CPD one. To do so one would have to add required flags to CLI / Ant, and at the end of PMD.doPMD() call CPD.go(), retrieve the found matches, and manually adding them to the reporter.... possibly the RuleViolation interface will require some modification to fit CPD violations (which in turn may require updating renderers).

It's not the cleanest solution, but the only one I can see actually working.

On the other hand, you could simply add support for different renderers in CPD, just like PMD has (bonus points for reusing existing codebase), move the current rendering code to an XMLRenderer, and add your own Code Climate renderer, to be used just as you currently use PMD.

[rsoesemann]

Thanks @jsotuyod fro providing valuable feedback here. I guess your second idea is the solution. Make CDP run seperately and redirect / convert the ouput so CodeClimate can understand it.