[Bf-committers] Collada importer/exporter kickout

[Brecht Van Lommel]

Hi,

> +1 for immediate creation of Collada tracker and move collada bugs
> there (can help with this, it worked well for BMesh).

Also agree on creating a Collada tracker.

> Other topics can be postponed closer to release, though I'd expect
> 2.62 would include collada, perhaps we could set some target for the
> Collada tracker & call for developers to help out, or disable by
> default (actual code removal can be checked on much later if we want
> to phase out or move to something else).

Right, I don't think we should throw it out or disable it for release just yet.

> My own experience with trying to fix collada bugs has not been great -
> the few issues I ran into were either very hard to debug or bugs in
> opencollada (which I forwarded to the collada tracker) - just to say
> that ~5 or so times I seriously tried to fix some of our collada bugs
> and didn't get far.

I've had the same experience, tried to fix bugs but found it very
difficult. I looked into armature bugs mostly, and it seems the
Collada and Blender models of those are quite different, but the code
doesn't properly address this issue and seems to need major
refactoring. I think good animation/armature support is critical,
otherwise you might as well use .obj/.3ds.

Now, as for the path forward, it's apparently a lot of work to get a
complete importer/exporter working, not only with basic test cases but
also compatible with other applications. Especially the importer for
different Collada versions seems difficult to implement well.

>From the point of view of attracting more developers, a python
importer seems the best solution, also since there's good python
utility code already there from other IO addons. However from what I
can see the bpycollada importer is still at prototype level. The c++
importer is further along but needs difficult code refactoring to fix
bugs.

Brecht.