[SciPy-user] Would anyone connect fortran constrained linear least squares solver to Python?

[Dominique Orban]

Hi Dmitrey,

On 1/7/08, dmitrey <dmitrey.kroshko at scipy.org> wrote:

> Unfortunately, NLPy still lacks precise convenient documentation. Would
> you provide something like the one:
> http://projects.scipy.org/scipy/scikits/browser/trunk/openopt/scikits/openopt/examples/llsp_1.py
> then I could provide openopt binding for the solver.

In a sense, my situation is not unlike yours since I am not often able
to spend much time writing documentation. However, a text document is
not the only medium to communicate documentation. NLPy has demo code
for almost all features in the Examples subfolder. Moreover, many
modules can be executed (to run a basic test). Regarding LSQR, the
default demo is

http://nlpy.svn.sourceforge.net/viewvc/nlpy/trunk/nlpy/Examples/demo_lsqr.py?revision=68&view=markup

and shows how to use the code. The function aprod() could be any
function that computes a matrix-vector product. The module lsqr.py
itself has a docstring explaining what problem it is trying to solve
and how. It also gives references to papers, for those who are that
motivated.

Cheers,
Dominique