Email Re: PyLint (fwd)

Sylvain Thénault schrieb am 10.05.2004 11:52:

Date: Wed, 05 May 2004 09:05:51 +0200
From: Andreas Amoroso <a.amoroso@freenet.de>
To: contact@logilab.fr
Subject: PyLint

Dear reader,

please find attached a zip-archive containing to files (with diffs) from
the latest PyLint distribution with modifications. These modifications:
- fix buglets (added checks for attribute presence before these
attributes are accessed)
    

this one has already been fixed in the CVS.

  

- add file output via two new options (useful e.g. when checking console
applications that use curses or something similar)
    

I've checked in your patch with some slight modifications (only add one
option to tell that we want output to files, format is specified with
the usual --html and --parseable options). It will be included in the
soon to come pylint's next release.

One thing I'm not sure about : did you want to have one file per checked 
module or one file per module/package specified on the command line ?
(currently it's the latter, plus one file for statistic reports if any)

  

Basically both options would be OK as long as file output is available at all.

I think these snippets might contribute to the stability and usability
of PyLint. 
    

yes, thanks for the patch !

  

Additionally, please consider to include an empty __init__.py
in the logilab directory of the common installer as the absence of this
file prevents PyLint from starting.
    

hum, the setup.py script of both pylint and common should create the
file if it doesn't exists. Do you have installed them using the regular
"python setup.py install" command ?

  

To be honest, I do not remember. I think I used a binary installer, but I might have created it myself. Anyway, there seems to be something missing as I know that somebody else encountered the same problem (and was not interested enough to solve it himself, apparently).

BTW, I am currently using PyLint to analyze a little program that uses "relative imports" (imports from files in the same dir as the main module). I have two questions related to this:
1. Why are these relative imports considered to be bad (flagged with warnings)
2. Is there (and if not should there be) an option to disable specific warnings?
These are not too urgent, of course, but if the answer to the second question should be no I think the proposed option would be worth considering for future releases. And while I am at it, I would consider it a plus if one could be more specific about naming conventions for identifiers (e.g. separate rules for global functions and member functions and the like). This all leads to the customization issues you probably have been discussing all the way, but still I could not resist ;-)
However, not to leave the wrong impression, I like PyLint even in its present state and consider it very useful.

Regards

regards