Re: [Assimp-discussions] light data structure

[Thomas S. <th...@dr...>]

Hi,

>>Agreed. Let's support what our target audience will be using: point,
>>directional and spot lights. So far we'd have these properties:
>>
>>position    (INF for directional lights)
>>direction   (n/a for point lights)
>>phi/theta   (2pi for point lights)
>>
>>Well .. I think it's clear that every light has a name by which it is
>>referenced by a node in the graph. So position and direction are
>>specified in the node's local transformation while all other parameters
>>are in a separate aiLight structure!?
>>Then we need light colors ... specular, diffuse and ambient I think.
>>Specular and Diffuse will be equal in most cases.

Maybe add a local transformation to each light in relation to its 
node. I guess this is neccessary because different file formats maybe 
define different default directions for light sources. In one file 
format the spotlight points to Z+, in another format it is defined as 
shining in X+ direction. A local transformation matrix for each light 
can handle this differences, while still maintaining the node 
animation possibilities for lights.

Apart from that, I agree with the proposed data structure.



>>The difficult question is the light attenuation ... linear and quadratic?

I only know the DirectX attenuation, which is defined by constant, 
linear and quadratic factors. Makes three floats for attenuation. As 
long as we don't find a better model for this, we can use this 
three-factor-model.


Bye, Thomas