Using InteractionType with an ExistType as interaction field leads to an Error

[ArVar]

When using an interaction field like:

(Interaction: ['street', 'house_num', 'house_num_e', 'city'])

where house_num_e is an ExistType, datamodel.DataModel.interactions() produces

(Interaction: ['((1, 1): Dummy)', 'street', 'house_num', 'city'])

based on the higher_vars. Which becomes also the name of the Interaction variable in columns as argument for interaction_indices(). There

interaction_indices = [var_names.index(f) for f in var.interaction_fields]

then runs into the error:

ValueError: '((1, 1): Dummy)' is not in list

I have no recommendation, where to change the code, but something between the naming of the interaction variable and the way the interaction indices are build might be incompatible.

[ArVar]

After some research and deep dive into the code I got a question:
Could this be related to
columns += interactions(variable_definitions, self.field_variables)
in datamodel.py? Priviously, only the so called "primary_variables" were used for interactions, exluding Custom and Interaction variables.

[ArVar]

I think, the cause is in a missing definition in the ExistType. By switching the parent class from CategoricalType to FieldType there is the standard len definition applied, which is 1.
Adding the following method to ExistType should fix this:

    def __len__(self) -> int:
        return len(self.higher_vars)

See #1196