On Tuesday 09 March 2010 17:59:02 Giovanni Gherdovich wrote:

hello,

sorry for the late answer, I was not here last week. 

I came across your problem a long time ago;
actually, I believe that it worked the right way when I started working on 
Pyreverse last year, but it got somewhere broken while merging it to 
Pylint...
However we don't see it in our tests, for some reason I had no time to 
understand: take a look at pylint/test/data : 
classes_No_Name.dot   __init__.py           suppliermodule_test.py
clientmodule_test.py  packages_No_Name.dot
where the expected output classes_No_Name.dot is correctly produced during 
the test, but not when we work directly with pyreverse.

> Sylvain Thénault wrote:
>  > I think you've spotted it: relative import. IIRC by default
>  > pyreverse only display on the diagram classes which are coming
>  > from the same package. The pb is probably that due to the relative
>  > import (and potentially a bad python path), pyreverse think both
>  > classes don't belong to the same package.
>
> I don't know how to check this.

I think, you should look in astng/manager.py to the 'project_from_files' 
method, where I can read (l.261) :

        # insert current working directory to the python path to have a 
correct
        # behaviour
        sys.path.insert(0, os.getcwd())

This doesn't seem to work anymore ... 

or is it line 297 in the same method :
        finally:
            sys.path.pop(0)
?

I hope this helps :) don't go into _ast ! The problem should be somewhere 
around astng vs pyreverse .

>  > Are you launching pyreverse from withing the package directory
>  > you're asking pyreverse to analyze?
>
> Yes.
>
>  > Also you should probably use absolute import
>  > anyway, pylint should have told you relative import are bad ;)
>
Relative import does not seem to be the problem: when I go to pylint/test 
and do 

$ pylint data

I get no 'relative import' complain.
However, Pyreverse still does not get the 'associated' and 'ancestors' 
relations.

So, don't hesitate if you have further questions.
-- 

Emile Anclin <emile.anc...@logilab.fr>
http://www.logilab.fr/   http://www.logilab.org/ 
Informatique scientifique & et gestion de connaissances
_______________________________________________
Python-Projects mailing list
Python-Projects@lists.logilab.org
http://lists.logilab.org/mailman/listinfo/python-projects

Reply via email to