Inconsistent predict object output from data.frame RHS

[finnlindgren]

When the formula RHS is a data.frame, the result is (usually) a prediction&list of data.frame:s, but should likely be a list of prediction objects. Also unclear what the intended behaviour is for different situations, involving integration etc.

[finnlindgren]

I think this has been sorted, but need to check that the documentation is clear about the intended behaviour.