I wrote a collision detection class for VTK – vtkCollisionDectionFilter – back in 2003, as a simple way to simulate joint articulation. Its been open sourced since 2006 in my vtkbioeng project on sourceforge.
The package has been downloaded a little over 1000 times and checked-out of cvs a little under a 1000 times. Judging by emailed questions most people are after the collision detection source code. (Btw, the highest percent of downloads are from China at 25%).
After an emailed nudge from David Doria, it was time to try and submit the class for inclusion in the official repo, so I’ll be creating a branch on github and putting an updated class, examples, and test up there. Gerrit also seems like a good submission review process, so I’ll have a go with that too.
Why did it take so long to submit? Firstly, the class was open sourced in a cmake ready package – developers could easily compile it themselves if they wanted the class. Unfortunately this meant that users of precompiled binaries of VTK couldn’t access the class. Secondly, there is an implicit promise to support a submitted class far into the future – scary stuff and not to be taken lightly. As against this, I co-wrote a submitted class vtkGL2PSExporter with Prabhu Ramachandran and while checking up on it the odd time, it has not given me any grief.
As a version 2.0 of the proposed branch, I’ll be looking at an abstract superclass for collision detection with subclasses implementing the current class, a wrapper class around the Bullet physics library, and perhaps a multi-body subclass too (in version 3.0)
If all goes to plan, an article to the VTK journal would be a good endpoint.