Corner case for providing RHS node as constants with length-1 vector vs scalar issue

[perrydv]

A user reported that in model code like:

    for(i in 1:nx) {
        x[i] ~ dexp(r * lambda[i])
    }

where lambda is right-hand-side only, if nx==1 and lambda is provided by constants, we get an error about dimension mismatch. When lambda is provided as data or inits, it works fine.

I don't remember if this has come up before. I'm marking it as a bug, but it could be resolved by documentation or error-trapping if it is too much of a pain to handle.

[paciorek]

Yeah, it's because lambda specified in constants is set in the dimensionsList as a scalar. But in the code it's a vector.

If we had a way for a user to pass in a 1-element vector that was treated as a 1-d object then it would be ok.

I don't see an easy way to trap this (without awkward special-casing) beyond the fact that we already report a dimension error.