On Wed, Aug 17, 2016 at 10:37 PM, Patrik Iselind
Thanks, i got it to work now.
I have some feedback on the page you referred to.
I was missing the implements line as well. This was not highlighted at all.
Bullet two concerning priority. How checkers are ordered (internally?) doesn't help me to decide if i should set a negative priority close to zero or very far from zero in order for the checker to be first or at least among the first.
I interpret this page as a resource you'd give programmers new to pylint checkers to get them started, awesome start! I think it would be really helpful if the page gave a overview of how the AST is buid and how i'm meant to access for example function variables in a function or the identifiers defined in a module. Where can i find a list of all the visit_* methods that i can use? I cannot find any class defining them all, are they dynamically called somehow?
In examples directory referred to, looking at custom_raw.py. It implements process_module() instead of visit_*. What is the difference? When should i choose one over the other?
Thanks a lot for your assistance in getting my first checker running.
// Patrik
Hi Patrik, These are all really good questions! Would you mind adding your feedback over this issue (https://github.com/PyCQA/pylint/issues/1071), where I am tracking the documentation effort for writing a new checker? In the mean time, some answers for your questions: - probably you shouldn't care that much about the priority, I don't know what its initial purpose was, so I presume it could be removed at some point. For what I care, each checker should not care about its order of execution, especially when considering calling them in a parallel environment. - sure, we need to document these as well. Basically, any node from the builtin ast module should be available in pylint as well, under a visit_<node name> method. - the process_module method is called whenever the checker is also a raw checker. The checkers can be AST checkers, operating on trees, token checkers, operating on tokens and raw checkers, operating on the file content itself. A checker can be all of these at the same time, if you declare the right __implements__. process_module is usually used when you need the entire content of the file for some reason. Of course, this can also be implemented with a visit_module() and getting the string representation of the module node, although we have some problems now with that, since we can't respect the invariant that ast.to_string() == file content.