[Distutils] setup.cfg new format proposal

Floris Bruynooghe floris.bruynooghe at gmail.com
Wed Sep 23 14:51:17 CEST 2009

On Fri, Sep 18, 2009 at 11:17:40AM +0100, Chris Withers wrote:
> Floris Bruynooghe wrote:
> >On Fri, Sep 11, 2009 at 07:02:53PM +0200, Tarek Ziadé wrote:
> >>On Fri, Sep 11, 2009 at 6:32 PM, Floris Bruynooghe
> >><floris.bruynooghe at gmail.com> wrote:
> >>>On Fri, Sep 11, 2009 at 05:14:58PM +0200, Tarek Ziadé wrote:
> >>>[...]
> >>>>That can be dropped since we have '==' and it's all strings.
> >>>'2.6' < '2.10'?
> >>>
> >>>Although not close you must be wary of this.  Bring on PEP 386 ;-)
> >>Right. Maybe we can drop < >, >= and >=, completely.
> >>
> >>In real world use cases, we can use "or" to make a section work for
> >>2.6, 3.1, etc.. for python versions.
> >>
> >>I can't think of a use case where "in" and "==" is not enough.
> >
> >Version comparison would be nice though.  Taking your example from
> >your blog post: if I where to try to install that on Python 2.4 I'd
> >have to edit the condition of the [py25] section to "in ('2.4',
> >'2.5')".  Maybe this is what you want though, in case you explicitly
> >don't support for 2.4.  OTOH it's nice if things would just work in
> >case of developers simply not having tried earlier versions.
> >
> >This also shows how "in" implies a tuple or a list as data type, your
> >mini-language was trying to avoid that.  Or did I miss something?
> A lot of these problems would go away if we could just treat
> "Python" as a package like every other package when it comes to
> dependencies.

How would that help in specifying dependencies based on a python
version?  The conditional syntax as discussed so far does not allow
for conditions based on versions of other packages detected.  And
doing so would seem like a lot more complicated then what's being
discussed so far (at first sight anyway, haven't tought too deeply
about this yet).


Debian GNU/Linux -- The Power of Freedom
www.debian.org | www.gnu.org | www.kernel.org

More information about the Distutils-SIG mailing list