@Momoko_Fan said: Also, it seems that you support providing a 4x4 matrix in all cases where a transform is expected. This makes it difficult for us since jME stores all transforms as a translation / rotation / scale tuple, and extracting that information from a 4x4 matrix is non-trivial, especially for non-uniform scales. On the other hand, deriving a 4x4 matrix from translation / rotation / scale is much easier, if the engine prefers that format. Wouldn't it make more sense to require all transforms to be stored in the separated translation / rotation / scale format then?
Self-answering this one…
When a 4x4 matrix represents rotation, scaling, and translation (even non-uniform scaling)… as long as the axes are orthogonal then the rotation, scaling, and translation can always be extracted rather trivially.
However, for non-orthogonal axes, the 4x4 matrix is the only way to represent them… and JME cannot support these kind of matrices anyway (and I’d suspect they are rare in the case this format would cover).