[Feature] I want to add support for AMBER coordinates and topologies

[helmutcarter]

A few things:

I've been using rust for less than a year, but I successfully used it for a project involving processing MD trajectories. However, because no rust library supports AMBER trajectories, I ended up doing all the trajectory setup in python, which was a big bottleneck. I would like to add support for it to molio, and because I am relatively new to rust (and completely new to contributing to rust libraries) I would like some direction.

Also, unlike all the filetypes supported by molio so far, AMBER stores atom coordinates/velocities in a separate file from the atom identities, bond connections, etc. What do you think is the best way to handle this in the library? I.e. should I just force the user to always load the parameter file at the same time as the coordinate file?

The most widely-used AMBER trajectory format is a NetCDF file, are you okay with including netcdf as a dependency?

Thank you!

[chem-william]

Hi, super cool that you want to contribute! Happy to assist wherever I can.

It would be super cool to have support for AMBER. Currently, the molio is more geared towards text files and I think it would require some refactoring to make it work nicely with binary files. Skimming the code (as it's been a while since I worked on it), it might be doable with the current interface, but take it as a heads-up

Also, unlike all the filetypes supported by molio so far, AMBER stores atom coordinates/velocities in a separate file from the atom identities, bond connections, etc. What do you think is the best way to handle this in the library? I.e. should I just force the user to always load the parameter file at the same time as the coordinate file?

I'd probably take the same approach as https://github.com/chemfiles/chemfiles/ although I don't quite know what they do. Most of molio is essentially a rewrite of chemfiles into Rust. Are there situations where it's sensible to not provide the parameter file with coordinate file? Or maybe switch it halfway through the program? Since they constructed it in such a way, I guess there could be, but I don't know. If not, I think it would make sense to force the user to always load the parameter file at the same time. And perhaps it's even a good start as we can probably lift the restriction later on (possibly through a breaking change)

The most widely-used AMBER trajectory format is a NetCDF file, are you okay with including netcdf as a dependency?
Yup! I'm by no means trying to make this a light-weight lib :) The potential challenge with that lib is that it is a wrapper on a netcdf-c. I've never tried bundling that with a Rust library as dependency. I don't know how that works for users or CI. Do we have to ship a build.rs file that installs netcdf-c? And if so, we should probably gate the AMBER feature behind a flag

I'm sorry that I don't have anything more concrete for you, but I hope it helps! And if you decide this is too much work/too complicated/too vague, that's perfectly!