Guillaume Fraux

So if you are willing to write ugly code, you can already use manual free as well: ```julia function chemfiles_free(o) Chemfiles.lib.chfl_free(Ptr{Cvoid}(o.__handle.__ptr)) o.__handle.__ptr = C_NULL return nothing end ``` Just make...

There is also a [`read!(t, f)` method](https://chemfiles.org/Chemfiles.jl/latest/reference/trajectory/#Base.read!-Tuple{Trajectory,%20Frame}) to read the next step in place. It might also be that the rest of the memory usage is actually relevant to the...

Nice, this does look a lot better! I don't see much further improvements on the top of my head; but there is one puzzling thing here: manually calling `free` should...

The `_ndim_coords_from_arrays` function is private, so this is likely a change from one version of scipy to another.

If it is not too bad for you (if you have a good internet connection and are OK with downloading all the crates again), you can just remove the `$HOME/.cargo/registry/`...

I'm not sure I see how the link you shared would fix the issue? By allowing to define backward Python-side?

For anyone hitting the `require-hashes` pip error, the solution is to update pip to a more recent version: https://github.com/google-deepmind/alphafold3/issues/20#issuecomment-2475114061

This is a bit strange, I am working on the metatensor package, which has similar goals as the PYTORCH one, and my code uses the exact same part of the...

Hey! I never used this code in a generic context, so I'm not sure if there is a workaround for the lack of associated methods. `StructOfArray::Type` was added by @mikialex...

So it seems that the issue is actually in the `PartialOrd` bound: changing it from `T::Ref T::Ref: 'a { type Ref(&'t self, index: usize) -> Option where Self: 'a; fn...