Add an option to GlassNet to use wt% instead of mol%

[drcassar]

Suggestion of Murat Hacıibrahimoğlu.

[Lexachoc]

The wt% is already implemented (supported) in SciGlass Next. I can backport it to glasspy.

My idea is to use an additional argument in predict method, e.g.

from glasspy.predict import GlassNet

model = GlassNet()

composition = {
    "SiO2": 2,
    "Li2O": 1,
}
predictions = model.predict(composition, concentration="mol")
predictions

where concentration can be one of "mol", "wt", "at", default is "mol". (here, mol% and at% are the same, at% is just for atoms)

what do you think?

Is the argument name appropriate? I am also considering using kind="mol" (which I used internally in SciGlass Next), but that may not be very intuitive.

Other naming choice: basis="mol"?
or unit="mol", but this may cause confusion with the property unit, which may be supported in the future.

[drcassar]

Hi @Lexachoc, thanks for reaching out!

Regarding the argument name, I prefer composition_unit, even if it’s a bit verbose, do you have any thoughts on that?

Also, we already have a function in GlassPy for converting wt% to mol%: glasspy/chemistry/convert.py#L208. Do you think it would be ok to use this? Perhaps using chemistry.to_array and then chemistry.wt_to_mol?

[Lexachoc]

Regarding the argument name, I prefer composition_unit, even if it’s a bit verbose, do you have any thoughts on that?

Looks good to me! composition_unit is easy to understand.

Also, we already have a function in GlassPy for converting wt% to mol%: glasspy/chemistry/convert.py#L208. Do you think it would be ok to use this? Perhaps using chemistry.to_array and then chemistry.wt_to_mol?

Yes. That's exactly what I do.

One thing to mention: I actually did not use chemistry.to_array, and I got a typing warning:

Expected type 'list[list[float]]', got 'ndarray | ChemArray' instead for chemistry.wt_to_mol.

But it seems valid to simply pass a 2D array with list[list[float]], that's what I did actually.

Should the typing also get updated in wt_to_mol and mol_to_wt?

e.g.,

def wt_to_mol(
-    x: Union[np.ndarray, ChemArray],
+    x: Union[np.ndarray, ChemArray, list[list[float]]],
    input_cols: List[str],
    rescale_to_sum: Union[float, int, bool] = False,
- ) -> Union[np.ndarray, ChemArray]:
+ ) -> Union[np.ndarray, ChemArray, list[list[float]]]: