PhasesResearchLab
Sublattice site ratios in the phase models are silently allowed to be inconsistent with JSON datasets
Currently ESPEI uses only what is in the phase models file (containing all the sublattice models), while the site ratios in the data files are not actually used other than for bookkeeping purposes.
For the sake of correctness, ESPEI's dataset checker should raise an error if there is a disagreement.
To provide some context for why this was allowed in the original design, first-principles calculations like SQS can be performed assuming different sublattice mixing (2SL, 4SL, etc) and different supercell construction methods, which may alter the site ratios. A particular Calphad modeler will be self-consistent with calculations she does herself, but if you are collecting all data available from the literature, you may end up with datasets for, e.g., BCC_B2 for different sublattice models.
The reality is that ESPEI users today probably want to curate their datasets directory to only contain physically consistent data, so an error in the dataset checker is probably the least-surprising behavior from the user perspective, but the larger issue discussed above will need to be resolved by any Calphad protodata repository such as ESPEI-datasets.
