CodeDiscussion

Yesterday I posted a new version of dopri (0.7.050921b) that appears to fix crashes/segfaults on delete as well as the occassional double catching of events bug. The integration wrapper and event finding code is now being used with the radau integration code; I hope to have radau available for release by tomorrow.

-- ErikSherwood 2005-09-22 14:39:23

FYI: If you run the dopri integrator and get a Python error like: SystemError: error return without exception set

this means that the C code in the dopri module returned a C NULL value instead of a PyObject*. I have had this error pop up while testing the latest dopri version (not yet available to everyone), but I can't seem to reproduce it. Just be on the look out for it now that I've changed how memory initializations are being done by the C code. If you could record the circumstances under which it happens and let me know, that would be great.

-- ErikSherwood 2005-09-17 22:14:30

I have finished a version of the dopri integrator (location listed on the main code topics page) that incorporates the changes regarding eventDelay and eventInterval.

I have also implemented most of the changes necessary to be able to evaluate the vector field, jacobians, etc. at given points/times without calling the integration routine. However, I will need to make significant changes to the calling sequence from python in order to make vf, etc. evaluation work properly. That is, we will need to do something like call integ.Init(...) before integrating via integ.Integrate() or say integ.Vfield(). We will also have to call integ.Cleanup() when finished with the integration object. It looks like I will need to have a lot more python scaffolding, so we should discuss how this will look Thursday.

-- ErikSherwood 2005-09-06 21:38:05

last edited 2005-09-22 14:39:29 by ErikSherwood