On Tuesday 27 July 2010 16:15:25 Edward K. Ream wrote:
> Hello all,
hello,
>
> I've just published a design for an incremental pylint on my blog:
> http://edreamleo.blogspot.com/2010/07/design-for-inc-lint-incremental-p
>ylint.html
I read your thoughts about Pylint with much of interest even if could not
understand all steps.
> Anyway, some of you may be interested. I would appreciate your
> comments. Thanks.
I have some thoughts and questions :
* Remark about Inference: I once tried to cache the *direct* inferences,
i.e. calling infer() without the context. There, I discovered that:
- only 5 % of the nodes are infered more than one time
- we basicly only infer Name and Getattr nodes directly
- however, there should be a lot of double-infering with regard
to the Context / Callcontext management; I think there are some
low hanging fruits there (like there were discovered some on our
last ASTNG refactoring) for a simple inference cache.
* Concerning your *Inc-lint*: What is quite obvious is, that Pylint looses
a lot of time of parsing almost always basicly all builtin and all
the Python standard library.
So the question is, how to store all the data about modified and
unmodified files, including a relevant representation of the astng tree?
And how reading all this information can be done faster than just astng?
Then, I am not sure if we need to go deeper than the level of modified
modules. (However I am not sure if I understood everything...) Or, at
least if we could just store the informations about unmodified modules,
it would probably be a huge improvement, and I wonder if going below
would result in another important improvement: the _ast-parsing of a
single file is quite fast.
thx anyway!
--
Emile Anclin <[email protected]>
http://www.logilab.fr/ http://www.logilab.org/
Informatique scientifique & et gestion de connaissances
_______________________________________________
Python-Projects mailing list
[email protected]
http://lists.logilab.org/mailman/listinfo/python-projects
