Vincent Barrielle

`diag_view` would need to return an entirely new type that would need to compute its values on demand for anything that it would be used for. Retrospectively I'm not sure...

It could indeed, though that would not really be a view as it would be allocating. But that could definitely be useful, especially a mutable version.

Hello, at the moment this is not supported. That would require porting a sparse QR factorization routine I guess.

Hello, As I mentionned in the users.rust-lang.org thread, I'm not sure the traits provided by alga exactly match the requirements for Krylov subspace methods, but so far I've only looked...

It looks like the vector space abstraction is quite useful to abstract over dense vectors, that's something useful to most linalg solvers, not just Krylov subspace methods, so that clearly...

Indeed, a subtrait relation looks like a good idea. Most of alga's matrix trait should be straightforward to implement, mostly methods to get a row or column may be tricky...

Indeed, a statically typed size would be far too inconvenient. Looks like the same problem should arise for dense vectors, how does nalgebra handle this?

This isomorphism to polynomials does make it sound, but I really fear that having no dimension checks would catch less errors. I'd really like to have an example where this...

Yes I think I'm ok with this more conservative solution.

Beamer supports adding videos with `\usepackage{multimedia}` then `\movie{path/to/image}{path/to/video}`. I've succsefully read that with okular, so it's not just Adobe Reader that supports it. The video is not really inside the...