[Bf-committers] Need ideas on COLLADA improvement for GSoC

[Jass]

I mentioned one idea a few weeks ago regarding organization of
exporter/importer development. Here is an update:

1.) Create a "Blender Collada Core" which is built in c and which
builds on top of the openCollada libraries. This core would support all
features that Blender WANTS to expose to Collada (allows to export).
Ideally the importer/exporter pair should allow lossless round trip.
I believe that it CAN be done because the Collada specifications allow
insertion of tool dependent data. Of course the lossless roundtrip
would only work reliably for blender -> collada -> blender round trips.

2.) Add the ability to "inject" preparation scripts and post processing
scripts into the export or import. These scripts could keep target specific
adjustments. The script injection would be doable via a python API,
such that...

3.) Now each target (such as Second Life, Maya, 3DS max, ...) can be
supported by adding target specific exporters/importer addons.
These addons can do anything starting from just presetting or fixing
the exporter options (that's what i would do for the hypothetical
Second Life exporter right now), or add more elaborate changes
like adding an XSL Transformation to the exported .dae or whatever.

in my opinion such an approach would help a lot to keep target
dependencies out of the Blender Collada implementation. It would
also help less skilled programmers to keep hands off blender code :)
and separate target specific development into python scripts.

cheers,
Gaia

On 21.03.2012 12:00, Daniel Balster wrote:
> Hi Arystanbek,
>
> ...I just started a new COLLADA feature at home a few weeks ago:
> profile_GLES. Since it goes beyond a small set of bug fixes it should go
> into a review/approval process.
>
> Problem description:
>
> profile_COMMON does not make me happy. Normally you write your "cool"
> shaders manually and feed them with parameters - but in profile_COMMON
> everything is hard-wired:
> - it's hard to match blender's current material+textures into a
> profile_COMMON effect
> - if you want to use COLLADA with custom shaders, you're missing a lot of
> parameters
> - just have a look at the<bump>  hack
> - normally you have a shader (ie "skinning") that can be used for multiple
> different models that only differ in a specific material setting (ie
> textures). Currently it's not possible to create a material/effect with
> exchangable samplers.
>
> My current COLLADA feature:
>
> profile_GLES (or Cg, GLSL,...) is much more versatile since it lets you
> spit out a lot of generic parameters, passes and even the actual shader
> sources.
>
> - via #ifdef I'm exporting<profile_GLES>  instead of<profile_COMMON>
> - I'm just spitting out a lot of<newparam>  for everything I'm aware of.
> - by convention, I'm splitting blender's material name into "prefix-" and
> "shader", ie "Material.001-skinning"
> - if the current blender document contains text scripts called
> "skinning.fs" and "skinning.vs", they're exported in the<code>  section of
> profile_GLES
> - otherwise an external reference to "skinning" is written.
> - additionally, I tried to hijack the current GLSL renderer to use these
> shaders to mimic my intended game effects (not working so far)
>
> So, if you like the idea I could write a clear feature design document and
> contribute some code as a starting point. Since this is a very specific
> workflow it will need some reviews - especially how to interpret all of
> this in the import case (ie load shaders as text documents and bind them to
> the GLSL renderer? convert to node graph? ...)
>
>
> Another not yet finished feature/bugfix is support for exporting
> binormals/tangents in meshes. Currently I'm doing this as a post process
> after importing a COLLADA file in my engine, checking for<bump>  and
> regenerating normals and tangents by myself. Would be great if blender
> could do this as well.
>
>
> best regards,
>    daniel
>
>
> On Wed, Mar 21, 2012 at 7:47 AM, Arystanbek Dyussenov
> <arystan.d at gmail.com>wrote:
>
>> Hi,
>>
>> This year I want to take part in improving the situation with collada
>> im/export as a part of GSoC. I plan to fix priority issues, currently faced
>> by collada users. I will post my proposal to this mailing list, after I
>> prepare it.
>>
>> I am not the only student, who would like to work on collada this year and
>> there's a possibility of other student or me adding new functionality to
>> collada support.
>>
>> Do you have ideas what collada im/export is currently missing?
>>
>> I think that carefully planned export of constrainted animations (in the
>> baked form) is something that users could benefit from.
>>
>> --
>> Best regards,
>> Arystanbek Dyussenov
>> _______________________________________________
>> Bf-committers mailing list
>> Bf-committers at blender.org
>> http://lists.blender.org/mailman/listinfo/bf-committers
>>
> _______________________________________________
> Bf-committers mailing list
> Bf-committers at blender.org
> http://lists.blender.org/mailman/listinfo/bf-committers