Saving hamiltonian/geometry into an nc-file doesn't save orbital information

[jonaslb]

Describe the bug
I usually save hamiltonians in the netcdf format using sisl.
When i load them, I cannot get any information on the orbitals other than the number of them per atom.
It would make sense I think if orbital information was saved (eg. which orbital is 2sZ1 etc).

Reproducable code

print(si.get_sile("H.nc").read_geometry())

Gives orbital information like Orbital{R: -1.00000, q0: 0.0},, while

print(si.get_sile("run.fdf").read_geometry())

Gives orbital information like AtomicOrbital{5sZ1, q0: 1.0, SphericalOrbital{l: 0, R: 4.1761, q0: 1.0}}.

Version details
sisl master

[zerothi]

You are correct.
I'll see to it :)

[zerothi]

So this is not so easily done.

I need to add extra logic to enable sisl to write non-complete basis sets.
The reason is that siesta only needs to store the (equivalent) SphericalOrbital information, and from that it can generate all atomic orbitals (different m).
However, sisl allows custom choices of only some l, m variants and thus needs to store explicitly all of them. So probably the easiest is to let sisl write everything, with an optional only_l to signal whether compressed forms are stored (to be compatible with Siesta).

[jonaslb]

Is it because you are thinking of making the format identical to ion.nc files?

I think for spgeom netcdf's it would be simpler to insist on the explicit format. If Siesta needs to read it (which I don't suppose it does, but still), then it can either be taught to understand it, or the sile could have a compat_mode="siesta" option to write orbital information with the siesta assumptions of implicit m (or error out if the orbitals dont match these assumptions).

[zerothi]

I agree. I would simply make a global identifier that writes which program wrote the file. That solves the issue. ;)

I would like to focus on #221 before I proceed with this.