[SciPy-Dev] Adding new wrappers (dassl, daspk version 2, cvode version 2.7.0) to scipy.integrate?

Geoff Oxberry goxberry@gmail....
Tue Jun 19 11:32:02 CDT 2012

I posted a feature request ticket (#1678, http://projects.scipy.org/scipy/ticket/1678) about this on the SciPy Dev Wiki, and Ralf Gommers suggested that I follow up on the scipy-dev mailing list.

To recap the gist of my ticket, there are problems for which DVODE (currently wrapped in scipy.integrate.ode) are inadequate, such as in computational combustion. It would be helpful to researchers in combustion to wrap updated numerical integration libraries (such as DASSL, DASPK, and CVODE); DAE solver wrappers would also be very helpful (technically, DASSL and DASPK are DAE solvers, but one thing at a time here).

Since some of these wrappers already exist in scikits.odes, would it be possible to merge them into scipy.integrate.ode? If so, in the process, could the Sundials wrappers also be updated to operate with Sundials version 2.5.0 instead of Sundials version 2.4.0? The newest version of Sundials contains only minor changes to the API (a few integer arguments are changed to long integers to accommodate larger problem sizes), and several bugs are fixed.

There have been many e-mails on the Sundials-users mailing list requesting a good, well-maintained and supported Python interface to Sundials with a reputable institution behind it. It looks like scikits.odes is the best-maintained Python interface out there, since pysundials and python-sundials are no longer maintained, and the Assimulo documentation uses some nomenclature that is nonstandard in numerical analysis (which makes me doubt the software). Furthermore, I know that users in the combustion community, frustrated with this situation, have implemented their own wrappers to such solvers (see for instance, https://github.com/jwallen/PyDAS, written by a colleague of mine).

I am willing to help with this effort after I graduate (I am in the process of finishing my PhD) in a few months.


More information about the SciPy-Dev mailing list