Dominik Gresch

Fun! For the release candidate, at the very least I think we should nail down the interface. How about we have a ``get_constants(qe_version: Optional[str]) -> Dict[str, float]`` function for the...

So.. we just pack the constants into a `SimpleNamespace` to make sure we can make multiple instances if needed? What about naming? Should the default be at top-level `qe_tools.CONSTANTS`? Or...

> > > > SimpleNamespace > > I don't know what this is, I trust your decision here :-) It's basically the same as the AiiDA `AttributeDict`, except without recursively...

I'm using QE version 6.4.1, in case the change of output format has something to do with this.

Note also that I didn't check for ``nspin=2``, which (from looking at the code), might be an exception.

I think the purpose of the calculation should definitely be included. About the subprocess, I'm not sure. Here's a hypothetical to consider: What happens if we decide to implement a...

Yeah, I've definitely had cases where the error was "random", and simply restarting did the trick. I think it might only make sense for the exit codes where `invalidates_cache` is...

I have a high-level question that might be slightly off-topic, so apologies in advance: The underlying issue here is that QE requires each magnetic moment to be specified as a...

> input but doesn't set nspin to 2 just a quick side note: non-collinear magnetism (`noncolin=.TRUE.`, which automatically sets `nspin=4`) can also have initial magnetization.

And yet another note: QE has a hard-coded limit on the number of kinds that is allowed (`ntypx`, defined in `parameters.F90`). The default is 10, to use more kinds QE...