On Thu, Feb 17, 2011 at 6:27 PM, Eric Deplagne <[email protected]> wrote:
> On Thu, 17 Feb 2011 10:10:38 +0100, Arve Knudsen wrote: > > On Thu, Feb 17, 2011 at 9:53 AM, Sylvain Thénault < > > [email protected]> wrote: > > > > > On 17 février 09:51, Arve Knudsen wrote: > > > > While this is a sound guideline, it should not be applied with > religious > > > > zeal. In contrast to a purely stylistic matter such as the naming of > > > > classes/variables etc., local imports can be motivated by technical > > > > concerns. When you perform local imports, it does not only affect the > > > > program's readability, but also its *behaviour*. I will in the > general > > > case > > > > import globally at the beginning of my modules, but in certain cases > I > > > want > > > > to defer module loading, maybe depending on a user option, and > delegate > > > it > > > > to a function. Therefore, I think pylint (for instance) should leave > this > > > to > > > > the programmer's discretion, and not try to be too smart about it. > It's > > > > complicated enough in my experience to define style rules that don't > get > > > too > > > > much in your way, in real life applications. > > > > > > That is expected pylint behaviour, for now. And I personnaly don't wish > to > > > change it. Though we could add a special warning message for 'local' > > > imports > > > (fairly easy to implement). If someone wish this behaviour, he can file > a > > > ticket :) > > > > > > I'd prefer to see it as off by default at least (please), were it to be > > added, the way gcc doesn't get too fascist about things until you specify > > -Wall :) My code is riddled enough with pylint suppressions where it > happens > > to do the wrong thing, I find. > > > > Arve > > This is exactly something I hate int gcc... > > The tool has to be fascist by default, and you turn off what you want to > turn off, > with *you* being responsible... > It has to? Why? For religious reasons? Arve
_______________________________________________ Python-Projects mailing list [email protected] http://lists.logilab.org/mailman/listinfo/python-projects
