I believe we could include transform information (e.g. this, but also something like "this is a simplex"). Note that the actual bound may not be a simple number, so in general the output may not be that useful. E.g,

parameters {
  real a;
  real <lower=exp(a) - 3> b;
}

The output would now have to contain something like "lower": "exp(a) - 3", so you can't assume you just get a number there

That’s very true. I suppose it’s also true of the dimension.

I would find this very useful - I want to be able to infer programmatically (statically, not at runtime) whether we’re operating on a constrained space or not.

I suppose it’s also true of the dimension.

For dimension, it has to be static data. So we can resolve the sizes as soon as the data is fixed. The bounds can't be resolved until we have parameter draws.

#904 asked about the dimension information. It was never implemented.

It does seem like it is probably reasonable to know that a variable has a lower bound, even if the exact lower bound is some expression you can't resolve on the outside.