I agree that the calculation of the determinant of a matrix is often not very stable in numerical terms. However, I don't think that this is the problem in this case. The division by the determinant is only involved in the calculation of the inverse of the matrix, which in term is only used for calculating the new view. So far, I have not seen any problems with the areas the views display, which is where problems with the determinant should show up. The transition itself is only a a linear combination of two matrices (i.e. Matrix1 * a + Matrix2 * (1 - a) for a = 0..1).
Should I have mistaken your point, please let me know. Also, if you have example files that show how a different calculation method improves the transition, please attach them to the bug.
I agree that the calculation of the determinant of a matrix is often not very stable in numerical terms. However, I don't think that this is the problem in this case. The division by the determinant is only involved in the calculation of the inverse of the matrix, which in term is only used for calculating the new view. So far, I have not seen any problems with the areas the views display, which is where problems with the determinant should show up. The transition itself is only a a linear combination of two matrices (i.e. Matrix1 * a + Matrix2 * (1 - a) for a = 0..1).
Should I have mistaken your point, please let me know. Also, if you have example files that show how a different calculation method improves the transition, please attach them to the bug.