Describe your issue.

Maybe it's undefined behavior, but calling bool on an instance of scipy.spatial.transform.Rotation yields an unclear error.

Perhaps any identity rotations could be Falsy, and anything else be truthy? (That's also why I tried bool on it)

Reproducing Code Example

from scipy.spatial.transform import Rotation
bool(Rotation.from_euler("x", 0))

See:
gonum/gonum#845 (comment)

What is the desired addition or change?

provide a cmake c++17 switch

What is the motivation for this feature?

to not depend on the std_backport

If applicable, describe how this feature would be implemented.

change those lines from 11 to 17 conditional to a global define

https://github.com/mlpack/mlpack/blob/de94eee6881e33f1145d9d4a8b5380a5be8af36a/CMakeLists.txt#L

Current implementation of join can be improved by performing the operation in a single call to the backend kernel instead of multiple calls.

This is a fairly easy kernel and may be a good issue for someone getting to know CUDA/ArrayFire internals. Ping me if you want additional info.

Currently spack does not support the following packages, all of these packages are installed outside of Spack at Cori, we would like to get support for these packages if possible.

• cpmd https://www.lcrc.anl.gov/for-users/software/available-software/cpmd/
• ComputeCPP https://github.com/codeplaysoftware/computecpp-sdk need to confirm link
• ds9 https://sites.google.com/cfa.harvard.

Add #[track_caller] to some panicking methods. Slicing methods seem the most prudent (Index/Mut does not apply, sorry).

Needs benchmarks to investigate impact, preferably holistic benchmarks.

Now that we've made BLAS support optional on several linfa crates, we should compare the performance of those crates with and without BLAS. Doing this requires those crates to have a complete set of benchmarks that represent realistic workloads. If BLAS turns out to have no performance improvements, we can even remove BLAS support, improving code quality.

Benchmark status for each crate that

CDash has builds with C++17 enabled, but they are not triggered when github builds opened PR's.

It would be great to have at least one build (gcc or clang) in github wih C++17.

https://llvm.org/docs/NewPassManager.html

The tricky part is to keep our custom command line options working.