[SciPy-dev] [scikits] openopt SVN instable for the moment
Wed Apr 9 11:40:35 CDT 2008
2008/4/9, Alan Isaac <firstname.lastname@example.org>:
> OK, it seems that things are largely getting worked out
> It looks to me like Matthieu implemented some useful
> changes, and one of the good things that happened is that
> a bug was uncovered. As far as I can tell, there has been
> an increase in functionality and ease of use with no loss of
> functionality or ease of use. Dmitrey, please correct me if
> I misunderstand (by stating in simple terms what
> functionality is lost).
Well, the only drawback with this implementation is that Dmitrey will have
to explicitely add the new wrappers he's writting to the registry. Aside
from this "issue", there is only more ease of use and of modularity.
I have been encouraging Dmitrey to rely more on test-driven
> development. This event highlights its importance, and it
> is one of the things I will be enouraging if I end up
> mentoring him for the GSoC (as I hope I will).
> I think that aside from the design issues, another kind of
> issue was uncovered that is quite natural. Dmitrey and
> Matthieu are working on code that sometimes overlaps in
> terms of needs and tests. But of course it is natural for
> each to feel protective of his primary area of work, since
> they are not actually working jointly on a single project.
We work on different part of the package, but I have had this issue for
several months (not being able to use the generic framework). This change
was not discussed here, but was hinted on several other ML some months ago
(python-list for instance).
This means that good tests become all the more
> crucial---tests that should be passed in advance of commits.
> But in addition, I suggest a 2-stage procedure in areas of
> design disagreement:
> 1. discuss with each other any large or medium size
> changes in design
> 2. if there are unresolvable disagreements, ask the
> developers list for feedback on the best resolution
This seems natural, if everyone plays by the same rule and pays attention
not to cripple functionalities.
French PhD student
Website : http://matthieu-brucher.developpez.com/
Blogs : http://matt.eifelle.com and http://blog.developpez.com/?blog=92
LinkedIn : http://www.linkedin.com/in/matthieubrucher
-------------- next part --------------
An HTML attachment was scrubbed...
More information about the Scipy-dev