DMatrix row_sum() cast to different type

[jon-ton]

Hello and thanks for all the good work on this crate!

Is there a preferred way to do a row_sum() cast to a different type? For example, if I have a 28x28 DMatrix of type u8, my row_sum() will overflow. I've been manually iterating over the matrix to get the row_sum() cast as a u32, but figured I'd ask if there was an easier way with the crate and some type of .into() syntax.

[sebcrozet]

Hi! You could do something like this:

self.compress_rows(|col| col.iter().map(|e| *e as u32).sum())

It’s basically the current implementation of row_sum, but with a cast inside before summing the column’s elements.

[jon-ton]

Thanks for the help.