Working on a Rust based vasprun.xml parser. I would love to hear your thoughts, ideas, and concerns

[wladerer]

Background

Hi everyone. I am a huge fan of pymatgen and it's been tremendously helpful for my research and my own personal enjoyment.

I'm currently learning Rust on my free time and decided a lofty goal would be to re-create the python Vasprun object using Rust bindings. This was motivated by my own personal difficulties of having enormous vasprun.xml files (on the order of 10s of gigabytes) and a laptop with far too little memory/processing power. I know there are more reasonable ways of solving this problem, but it was a fun way to learn.

Questions

I'm particularly interested in knowing what your immediate reaction to something like this might be. Does it seem like a an unnecessary use of time? Are you tired (or suspicious) of seeing Rust put into practically anything?

Another question is, at what point would you find this tool useful? I have done some primitive benchmarking and found it reduces the run times by anywhere from 2.5-7x and has eliminated even the largest files from freezing my laptop. I was particularly keen on making sure it acted as a drop in replacement. So I've actually used some of the tests in pmg + doing some cross validation such that the python version and rust version return identical results.

Realistically, I don't imagine that parsing vasprun.xml files is a bottleneck for anyone but myself. If you do see this being an interesting project for your own purposes, I'd be more than happy to publish the code in a public repo so you may try it out yourself.

Notes

1. I have not implemented parsing VASP outputs the involve LOPTICS.
2. I am unfamiliar with best practices in Rust and therefore I would probably look for volunteers to help audit the code and the architecture itself
3. I am not requesting that my Vasrpun implementation be added to the pymatgen ecosystem - I am only looking for a realistic discussion (e.g "sounds lame" is good feedback)

[shyuep]

Thanks Will. Any enhancements to pymatgen, especially in parsing speed for Vaspruns, would be most welcome. I have no objections to using Rust if it helps, as long as the installation process is made seamless. Please submit a PR whenever you are ready in the pymatgen-core repo instead of this repo.