Andreas Borgen Longva

I think this raises the question of how far `nalgebra` should go to implement conversion traits. Every impl increases maintenance burdens for `nalgebra`. Arrays are *the* natural representation for vector...

Unfortunately the user guide has not yet been updated, but the macros `vector!` and `matrix!` should do what you want: ```rust const ZERO: Vector3 = vector![0.0, 0.0, 0.0]; ``` Hope...

This is something that's been bothering me too. The question is if introducing these constants can break any code - I don't think so, because the type aliases and the...

> Doesn't this solve the problem in concern https://github.com/dimforge/nalgebra/blob/dev/src/base/dimension.rs#L281 ? The problem is that you can't write e.g ```rust matrix.reshape_generic((0, 0), (U3, U1)); ``` because `U3` and `U1` are type...

The example matrix you use is clearly singular (it has a zero column). From a *numerical* perspective, there is no meaningful way of distinguishing between a singular matrix and a...

In fact, maybe we should deprecate `solve` and provide `solve_invertible` and `solve_least_squares` or something like that, to make the intention clearer.

SVD least squares produces the minimum norm solution. So obtaining the identity here is not the expected solution, the solution that is returned (with a zero middle column) *is* what...

> Sadly, this is the kind of values I do expect in realistic scenarios of systems I'm solving, and they're well off from `f64::EPSILON`, so I expected I'd be fine......

We could perhaps make a new issue with a more actionable set of suggestions, but yeah, let's keep this issue open for the time being, as you've raised some valid...

I personally think not filtering out zero entries in `push_matrix` is the preferred behavior. If you have many zero entries in the dense matrix you're pushing then you should probably...