[SciPy-dev] Test case for integrate/ode.py: banded systems

David M. Cooke cookedm at physics.mcmaster.ca
Wed Mar 8 16:46:15 CST 2006


Jesper Friis <jesper.friis at material.ntnu.no> writes:

> Andrew Straw wrote:
>> And now, upon reading the comments in the code of that page, it seems
>> you've uncovered what may be a bug in the VODE wrapper. Would you say it
>> is? If so, we should report the bug, hopefully fix it either ourselves
>> or have a developer do it, and fix up the page appropriately.
>>
>> Cheers!
>> Andrew
>
>
> I have to admit that I just started last week to use
> python instead of Fortran, so I think that you are in a better
> position to say if it should be considered a bug or not that the
> nrowpd argument is missing in the python interface.
>
> Searching through the function integrate/odepack/vode.f revealed that
> the Jacobian function is only called in two places. For full systems
> it is called with nrowpd=neq while for banded systems it is called
> with nrowpd=ml*2+mu+1. So, as long as these values for nrowpd are
> provided in the documentation there should be no problems using the
> python interface.
>
> Note however that the lower and upper bandwidths, ml and mu are
> interchanged compared to what I wrote in the comment in the example
> script. The reason for this turned out to be another little miss in
> integrate/ode.py, which is corrected with the patch below.

Done.

> It might also be interesting (at least for me) to add a way to obtain
> information about the solution, like the number of steps that was
> needed to reach the solution. What is the best way to do this?  Maybe
> one could add a method called infodict() that returns a dictionary
> containing all optional output from the solver?

Come up with a good solution and a patch :-)

> In the end of the ode module there are two simple tests for non-banded
> systems. Maybe they should be added to the Cookbook together with my
> banded example. What do you think?

Works for me.

-- 
|>|\/|<
/--------------------------------------------------------------------------\
|David M. Cooke                      http://arbutus.physics.mcmaster.ca/dmc/
|cookedm at physics.mcmaster.ca




More information about the Scipy-dev mailing list