Re: [cgkit-user] confusion over cg-kit's mat4 format

[Matthias B. <mat...@gm...>]

Hi Nathaniel,

it seems you got it all working now, but just to recap:
Whenever you deal with separate packages that use matrices to store
transformations, you need to check what their conventions are with
respect to how points are transformed. You can transform them by either
multiplying them to the left or the right of the matrix:

1) Maya: transformedPoint = point*Matrix
   (in this case, the translation part is the 4th *row* of the matrix)

2) cgkit: transformedPoint = Matrix*point
   (in this case, the translation part is the 4th *column* of the matrix)

It's straightforward to switch from one convention to the other, you
just have to transpose the matrix.

If those packages store matrices as flat lists instead of dedicated
matrix objects, then there's an additional stumbling block. You need to
know the order of the values, they are either in column-major or
row-major order (which has nothing to do with the above convention about
the order of the multiplication). If you get this wrong, you end up
having a transposed matrix. If you get both things wrong (the above
transformation convention and the value ordering), you are actually fine
again as one mistake cancels out the other one (which may actually
further increase the confusion).

> def cgkit_matrix( xform_list ):
>     """
>     takes a 16 element list (assumed to be the result of a xform query)
> and returns a cgkit.cgtypes.mat4 in proper form
>     """
>     matrix = cgkit.cgtypes.mat4( xform_list )
>     matrix = matrix.transpose( )
>     return matrix

That looks correct to me.

> def maya_xform_list( matrix ):
>     """
>     converts a cgkit.cgtypes.mat4 into a row list that can be used with
> mays.cmds.xform.
>     The original matrix is left unaltered.
>     """
>     #create a copy of the original matrix
>     matrix = cgkit.cgtypes.mat4( matrix )
>     matrix = matrix.transpose( )
>     return matrix.toList( rowmajor = True )

The transpose() method returns a copy anyway, so there's no need to copy
the matrix at the beginning.
Also note that this function could even be far simpler, you could just
return matrix.toList() (using the default column-major order) in which
case the transpose will be done implicitly by toList().

- Matthias -