[SciPy-User] scipy.sparse.linalg.factorized

[David Kershaw]

Christopher Mutel <cmutel <at> gmail.com> writes:

> 
> Hi David-
> 
> UMFPACK needs to be wrapped to work with scipy, and the wrapper was
> removed because of incompatible licenses (SuiteSparse, of which
> UMFPACK is a part, is GPL-licensed). The wrapper code was also
> available as a scikit, which is now maintained by Robert Cimrman on
> github: https://github.com/rc/scikit-umfpack
> 
> I have built scikit-umfpack as a wheel for OS X, but don't have enough
> expertise to build it as a wheel for Windows. To install it on your
> machine you will need to compile SuiteSparse, and then install
> scikit-umfpack with a suitable config that shows where your UMFPACK
> and other SuiteSparse libraries are.
> 
> It is possible that scikit-umfpack could use the library in cvxopt; I
> don't know anything about this, but I guess it is not trivial.
> 
> If someone wanted to build scikit-umfpack as a wheel (which would
> include the UMFPACK libraries, so no one else would have to figure out
> how to compile them), that would be super awesome.
> 
> You can test if UMFPACK is being used by scipy with:
> 
> from scipy.sparse.linalg.dsolve.linsolve import useUmfpack
> 
> useUmfpack will be either True or False.
> 
> Yours,
> -Chris

Thanks for your help Chris.

I just found at:
https://pypi.python.org/pypi/scikit-umfpack

a file for download:
File    Type    Py Version    Uploaded on    Size
scikit_umfpack-0.1-cp27-none-macosx_10_6_intel.whl (md5)   Python Wheel  
cp27  2015-01-21 875KB 

Is this "scikit-umfpack as a wheel" for python 2.7.9 and how do I add it to 
my existing python(x,y) installation? I don't have a clue how to install a 
python wheel.

Best regards,
David 


> 
> On 2 February 2015 at 09:19, David Kershaw <dskgair <at> gmail.com> wrote:
> > I have the Python(x,y)-2.7.9.0.exe installation of python(x,y) running 
on
> > windows-7 OS. I want to use scipy.sparse.linalg.factorized to solve 
linear
> > systems using Umfpack. As far as I can tell, the python(x,y) 
installation
> > does not include the necessary Umfpack code. How do I install Umfpack 
and
> > how can I tell that I'm getting Umfpack and not superLU when I call
> > factorized?  The Microsoft Visual C++ Compiler Package for Python 2.7 
was
> > installed when I ran Python(x,y)-2.7.9.0.exe.
> >
> > _______________________________________________
> > SciPy-User mailing list
> > SciPy-User <at> scipy.org
> > http://mail.scipy.org/mailman/listinfo/scipy-user
>