Support for FHI-AIMS (ASE)

[aronwalsh]

Dear developers. It would be very nice to make this code compatible with FHI-AIMS (or other codes through ASE). Many thanks, Aron

[ajjackson]

It's an interesting question. Obviously I'm a big fan of aims and ASE :-)

When planning Sumo we decided to avoid depending on both Pymatgen and ASE due to their large only-partially-overlapping dependency stacks and the redundancy of using them both for Vasp. @utf 's preference was for Pymatgen and given his current employment that worked out well.

However, we are also currently in the situation that I have redundantly implemented Castep band-structure support for both Sumo and ASE. It may be worth revisiting the relationship between those projects, and some kind of optional dependency setup for ASE-based codes. A lot of what Sumo does should be or is now in ASE natively, but there are some key design choices (opinionated styling and complex CLI tools) that don't fit within ASE.

[ajjackson]

An update on this after the ASE workshop: ASE now has a much lighter stack of required dependencies and there is even discussion of pulling some complex non-essential components into their own packages. So the packaging argument to avoid ASE in Sumo is weaker than it was.

There is likely to be some development and refactoring of the AIMS calculator in the near future by @flokno ; it's a good chance to make progress as long as we keep communicating.