Likelihood names are ordered, resulting in errors when data is added on a different order

[sebvetter]

In a model with a config like

likelihood_config: likelihood_terms: - name: a [...] - name: b [...]

when we try to add data (e.g. to fit something)

data = {'b': data_b, 'a': data_a}

model.data = data


will throw an error ValueError: Likelihood names do not match, however,

data = {'a': data_a, 'b': data_b}

model.data = data


will work. As the data is given as a dict, this is unexpected.

Expected behaviour would be to cross check the likelihood names in a manner where the order does not matter, or raise a warning that the order matters.

[hammannr]

This is indeed unexpected from the user perspective. We need to enforce the correct order to match the order of the likelihood_list. Otherwise, we might set datasets to the wrong likelihood terms.
I believe the best solution would be to make the likelihood_list a dict, then we can ignore the order of both. A few things would need to be adjusted though to ensure nothing breaks in blueice. Definitely doable though I think 👍