[SciPy-dev] try_umfpack.py

Robert Cimrman cimrman3@ntc.zcu...
Mon Feb 16 11:16:00 CST 2009


Nils Wagner wrote:
> Hi Robert C.,
> 
> 
> I have updated the URL in try_umfpack.py
> 
> #defaultURL = 'http://www.cise.ufl.edu/research/sparse/HBformat/' 
> defaultURL = 'http://www.cise.ufl.edu/research/sparse/RB/HB/'
> 
> 
> ./try_umfpack.py -d  bcsstm21.tar.gz 
> ************************************************** url: 
> http://www.cise.ufl.edu/research/sparse/RB/HB/bcsstm21.tar.gz file:
> /tmp/tmpDwKACc.gz format: triplet reading... Traceback (most recent
> call last): File "./try_umfpack.py", line 222, in <module> main() 
> File "./try_umfpack.py", line 138, in main mtx = readMatrix(
> matrixName, options ) File "./try_umfpack.py", line 98, in readMatrix
>  mtx = readMatrix( fd ) File "./try_umfpack.py", line 37, in
> read_triplet nRow, nCol = map( int, fd.readline().split() ) 
> ValueError: invalid literal for int() with base 10: 
> 'bcsstm21/bcsstm21.rb'
> 
> Any idea ?

Hi Nils,

thanks for your import fixes.

It seems that the matrices are no longer stored simply in a gzipped
files in the triplet (coordinate) format, but instead are stored in a
tar.gz archive with .txt info files and .rb file in the
Rutherford-Boeing format. try_umfpack.py cannot read this format yet -
what is the status of the ticket
http://projects.scipy.org/scipy/scipy/ticket/354 ? As soon as that gets
into scipy, fixing try_umfpack.py is trivial.

r.


More information about the Scipy-dev mailing list