In practice do we actually sequentially move up the stack of matrices multiplying by each one for each object or are the matrices memoized while we go, storing both the transformation from the last layer and the overall transformation.
raymondx
Looking at this example, would A1 be a shear with respect to A0 and A2 would be another shear with respect to A1?
Isaaz
@aylu I think how we store the transforms should be related to what we are doing. If we are developing a skeletal animation editor, then store transforms in the parent node's space is better because it's easier to modify the property of a single node. But when we finally export the animation to be used in a read-only manner, I think it is reasonable to bake the transform chain into one transform for avoiding runtime calculation.
In practice do we actually sequentially move up the stack of matrices multiplying by each one for each object or are the matrices memoized while we go, storing both the transformation from the last layer and the overall transformation.
Looking at this example, would A1 be a shear with respect to A0 and A2 would be another shear with respect to A1?
@aylu I think how we store the transforms should be related to what we are doing. If we are developing a skeletal animation editor, then store transforms in the parent node's space is better because it's easier to modify the property of a single node. But when we finally export the animation to be used in a read-only manner, I think it is reasonable to bake the transform chain into one transform for avoiding runtime calculation.