Reading GLTF data from model

[koiava]

https://github.com/syoyo/tinygltf/blob/dac2a89b630fcb4c0e6131a048c0e3040588bd2c/examples/raytrace/gltf-loader.cc#L47

Somewhere here should be a code which translates data from model to meshes/materials/textures

right now it returns true but actually there is nothing in the output buffers. After that it crashed inside nanort when it calculates BoundingBox with 0 elements inside mesh(This looks like a separate issue).

How

Translate Model to output buffers or log message about TODO.

[syoyo]

Ah, I remembered raytrace example is still work-in-progress, so it won't work.

I have described the current status in README.

[koiava]

I see. Still would be better to return appropriate message :) Shouldn't be too hard :)

[syoyo]

To implement raytracer for glTF model, we need scene graph data structure.

For the scene graph data structure, I have implemented nanosg: https://github.com/lighttransport/nanort/tree/master/examples/nanosg so you may use nanosg to convert glTF meshes into the scene hierarchy for the raytracer example.

[syoyo]

I see. Still would be better to return appropriate message :) Shouldn't be too hard :)

Changed to report error message :-)

[Ybalrid]

For reference, this is partially done in the PR #40 that has just been merged in. Implementation is still pretty rough at the moment. This will load all the meshes into a glTF files on the scene, but it doesn't attempt to load materials or to move theses meshes in the positions defined by the eventual "scene" inside the glTF file.

[chenliming777]

gltfBuffer and gltfImage vector? why? I feel will Influence performance. But thanks to open source.

[Ybalrid]

@chenliming777

Once memory is allocated, std::vector<T> has no performance impact over a regular array of T that was already allocated on the heap. So access to the data and traversal is as fast as it could be.

push_back(const T&) can trigger re-allocation of a larger array inside the vector. But this is only done while parsing the file and loading the data.

Once you are done with it (eg: loaded all useful data to the GPU) and you dispose of the tinygltf::Model object, all of this memory will be freed.

With this considerations in mind, I don't feel like there's a bad influence on performance from doing it this way.