Why not encode dimensions of array in the type system?

[SuperNove02]

Hi,

I am fairly new to ndarray. While using the library I was a little bit confused on why every operation between arrays panics. Why is not the size of a 2darray encoded in the type system via a const parameter?

To be honest i would have guessed that an 2 dimensional array would look a little bit like this (of course this is very simplified):

pub struct ArrayMatrix<T: num_traits::Num, const ROWS: usize, const COLS: usize> {
    pub array: [[T; COLS]; ROWS],
}

With this design the type checker can know at compile time that i.e. matrix multiplication fails due to incomplete dimensions. However ndarray does not give this guarantee:

    /// Compute the dot product of two arrays.
    ///
    /// **Panics** if the arrays' shapes are not compatible.
    fn dot(&self, rhs: &Rhs) -> Self::Output;

I do not want to criticize the project in any way i am just interested in the design decision and maybe there is a way of having this safety in ndarray, which i do not know.

Thank you.