`only` or `what`?

[tfrederiksen]

539 made me notice that the keyword in Geometry.swapaxes is what, while in Geometry.rotate it is only. Would it make sense to unify? I like what better.

[zerothi]

I agree. Unification! Grep for what, only, which, how, others we could have used, to ensure we get all.

[tfrederiksen]

Good. By the way, I think what='xyz' could be the better default in rotate. At least I have never used rotation of axes and thus need to set this keyword every time.

[zerothi]

sisl % grep -E "[^A-Za-z0-9_](only|which|what|how)[^A-Za-z0-9_]" **/*.py | grep " def "                                                                                                          main
geometry.py:    def rotate(self, angle, v, origin=None, atoms: Optional[AtomsArg]=None, only='abc+xyz', rad=False) -> Geometry:
geometry.py:    def center(self, atoms: Optional[AtomsArg]=None, what="xyz") -> ndarray:
geometry.py:    def scale(self, scale, what="abc", scale_atoms=True) -> Geometry:
io/tbtrans/tbt.py:    def orbital_transmission(self, E, elec=0, kavg=True, isc=None, only="all", orbitals=None) -> csr_matrix:
io/tbtrans/tbt.py:    def orbital_current(self, elec=0, elec_other=1, kavg=True, isc=None, only="all", orbitals=None) -> csr_matrix:
io/tbtrans/tbt.py:    def bond_transmission(self, E, elec=0, kavg=True, isc=None, only="all", orbitals=None, uc=False) -> csr_matrix:
io/tbtrans/tbt.py:    def bond_current(self, elec=0, elec_other=1, kavg=True, isc=None, only="all", orbitals=None, uc=False) -> csr_matrix:
io/tbtrans/tbt.py:    def vector_transmission(self, E, elec=0, kavg=True, isc=None, only="all", orbitals=None) -> ndarray:
io/tbtrans/tbt.py:    def vector_current(self, elec=0, elec_other=1, kavg=True, isc=None, only="all", orbitals=None) -> ndarray:
sparse_geometry.py:    def rij(self, what='orbital', dtype=np.float64):
sparse_geometry.py:    def Rij(self, what='orbital', dtype=np.float64):
supercell.py:    def rotate(self, angle, v, only='abc', rad=False):
supercell.py:    def scale(self, scale, what="abc"):
viz/input_fields/orbital.py:    def _split_query(self, query, on, only=None, exclude=None, query_gen=None, ignore_constraints=False, **kwargs):
viz/input_fields/orbital.py:    def _generate_queries(self, on, only=None, exclude=None, query_gen=None, **kwargs):
viz/plots/fatbands.py:    def split_groups(self, on="species", only=None, exclude=None, clean=True, colors=(), **kwargs):
viz/plots/pdos.py:    def split_requests(self, *i_or_names, on="species", only=None, exclude=None, remove=True, clean=False, ignore_constraints=False, **kwargs):
viz/plots/pdos.py:    def split_DOS(self, on="species", only=None, exclude=None, clean=True, **kwargs):

[zerothi]

@tfrederiksen do you have a suggestion for tbtrans, should we change those?

[tfrederiksen]

I think what would work in all cases above.

[tfrederiksen]

Well, maybe the meaning is somewhat different in viz with only and exclude?

[zerothi]

Well, maybe the meaning is somewhat different in viz with only and exclude?

true, lets keep them.

[zerothi]

Good. By the way, I think what='xyz' could be the better default in rotate. At least I have never used rotation of axes and thus need to set this keyword every time.

actually, changing the default to "xyz" breaks the nanoribbon and bilayer creation routines since they rely on also rotating the lattice vectors.
I can see that there may be multiple use-cases.

1. When rotating geometries for attaching them, "xyz+abc" is preferred,
2. When rotating sub-segments "xyz" is preferred.

So which has precedence? I think you are correct that 2. is mostly used, so lets stick with that, this is just to at least be aware of the implications.

[tfrederiksen]

OK I see. Would it make sense that what="xyx" is the default if atoms is specified, otherwise "xyz+abc"?

[zerothi]

I think that complicates things... I have this somewhere, but when you don't use the routines often, then it makes more sense to have a stable default... well, I don't know.

Is my above understanding even correct, are there other cases I haven't thought of?

[zerothi]

OK I see. Would it make sense that what="xyx" is the default if atoms is specified, otherwise "xyz+abc"?

Implementing this and fixing stuff all-around made me convinced that it is ok to have two different defaults, I have done as you suggested ;)