Check function `compute` vs. `check` variants

[sakrejda]

Putting this here to avoid thread-jacking. Re:

@Bob_Carpenter

The problem architecturally there is that there are things like checks for positivity, but when you pass a vector, you can’t just check for positivity, then throw a message, because the message should indicate the first element of the vector was found to violate the constraint.

I’ve run into this too and it makes our code messy. One wrinkle that needs investigating is whether the compute variants can always return a bool or if they sometimes need to return more information. For vectorizing the check functions it might be nice to return more (so the same thing can be called on a vector/matrix/array and still produce the correct indices at the top level).

[Bob_Carpenter]

The compute variants can return a bool, but that doesn’t include the relevant information for debugging. So I don’t see how that’d work. They might be able to return a boolean and instantiate a message, but that starts to defeat the purpose of the factorization of the underlying checker needs to format a text message.