Observables

[yomichi]

Users can define observables to be calculated by using [observer] section as the following example:

[observer]
[observer.similarity]
reference_structure = './MgAl2O4.vasp'
ignored_species = ["O"]

[[observer.solver]]
name = "magnetization"
type = 'aenet'
path= '~/opt/aenet/bin/predict.x_serial'
base_input_dir = './baseinput_mag'
perturb = 0.0
run_scheme = 'subprocess'
ignore_species = ["O"]

"Similarity" is how much the ratio of atoms in the same place as the reference configuration is. The similarity is calculated for each element, e.g., similarity_Mg.dat and similarity_Al.dat.

observer.solver takes the same input parameters as sampling.solver except for name which specifies the name of the result file.

In this PR, a new solver user is implemented. It takes a keyword function specifying a solver function from pymatgen.core.Structure to float. The function can be in some package, say, function = "mypackage.mymodule.myfunction". The module mypackage.mymodule will be automatically imported (at the time, the current directory is added to the module search path, and so ./mypackage.py is loaded).

[yomichi]

I found that pymatgen 2023.5.8 does not work for our unit test.
We will limit the supported versions to before it for a while.