The current trait bounds restrict us taking the metric of two slices. It looks like this:
impl<'a, 'b, U, T, M1, M2> MatrixMetric<'a, 'b, T, M1, M2> for U
where U: MatrixNorm<T, M1>,
M1: 'a + BaseMatrix<T>,
M2: 'b + BaseMatrix<T>,
&'a M1: Sub<&'b M2, Output=M1> {
fn metric(&self, m1: &'a M1, m2: &'b M2) -> T {
self.norm(&(m1 - m2))
}
}
The issue is the last line of the where clause. We require that the output be of type M1 but this is not the case for slices. I think we can remedy this by requiring that the Output is a Matrix<T>.
The current trait bounds restrict us taking the metric of two slices. It looks like this:
The issue is the last line of the
whereclause. We require that the output be of typeM1but this is not the case for slices. I think we can remedy this by requiring that the Output is aMatrix<T>.