[SciPy-user] OpenOpt svn broken on Windows

Matthieu Brucher matthieu.brucher@gmail....
Thu Jul 17 02:02:53 CDT 2008


>> What is the purpose of ooMisc.__solverPaths__ ? If it should be a
>> dictionary once scikits.openopt is imported, you have an error in your
>> program. If it is not, you have some junk code that can be deleted.
>>
> Why it should be a dictionary once scikits.openopt is imported? What's
> wrong now? scikits.openopt doesn't contain __solverPaths__, it's
> auxiliary variable used deeply in Kernel only, and anyone don't have to
> care about this variable value.

There are two parts in what I'm saying:
- the first is that there is a langage (Python) misunderstanding, see
the link Alan provided sooner in the discussion, probably leading to
mistakes in the future,
- the second is that you have to expose it if you want people to help you.

Let me explain this.
You are complaining that you have to connect several optimizers
yourself to OO. This is sad, because, as you've stated, you don't have
much time, you have to work on the GSoC. So your best interest and the
best interest of the OO users is that additional people show up and
help you.
So what is the best course of action? Not everyone wants to have
access to the SVN. Enabling adding solvers on the fly solves this (and
you already stated before that you don't like people committing
changes to the SVN without your approval, so this IS the only way of
properly doing things). People install OO from the latest tarball.
Then they can create their own solver, add some tests to check if it
is working and then come back to you. This way, you will have feedback
from the community as well, and you won't feel alone developing the
package.

So if you want people to help you, please advertise __solverPaths__
and write a small tutorial.
If you don't want people helping you, please stop saying that it is
not your job to connect solver X or solver Y. It is, as you are the
only developer of this specific part of the package.

> Currently I removed import  __solverPaths__, add __solverPaths__= None
> in rPS.py file and it works ok with my Python 2.5 but I got a error in
> 2.4, that's why I had moved __solverPaths__ to ooMisc. I intend to
> remove it (from svn) later, when Python 2.4 will be too obsolete.

This is strange. I'll try to find out what is happening and I'll let you know ;)

Matthieu
-- 
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


More information about the SciPy-user mailing list