OBD:OBAN: Difference between revisions

:While you are technically correct about expressing the transform as a translation and a 3x3 matrix that does rotation, scaling, shearing and mirroring I have never seen a 3D graphic system (or documentation) doing it. It's much easier in practice to only deal with one 4x4 transform matrix which includes everything (the 3x3 transform part, the translation and the projection). In addition as you already noted the translation part is (m14, m24, m34) (the first 3 elements of the last column) so it is easy to spot it for someone who is only interested in object position. In a similar way a 2D graphic system uses a 3x3 matrix to express rotation, scaling, shearing and mirroring transforms for 2D vectors. This thing is sort of burned into my mind, when I saw 12 float values with 3 of them looking like a position I said to myself "alright, this must be a transform matrix without its last column" (I mostly use Direct 3D so I tend to think in row major matrix order) :).

<HR>