Bob Carpenter

The only reason I'm OK with this for `int` is that we don't have the vector alternative. Most of our users coming from Python and R expect arithmetic to work...

@seantalts created a design for loggers somewhere. The idea was to go even further and have all of our I/O handled by a static logger. When do you not want...

Yes, you need to treat the matrix as having 0 entries everywhere there isn't an explicit non-zero value. That means if we apply `exp()`, for example, the result is non-zero...

@nhuurre: The problem is that when you apply `exp()` to a sparse matrix, the result is dense. The question is really whether we want to auto-convert to a dense result...

> why would you want to apply exp() to a sparse matrix? I have no idea where this would come up in practice, but if it did, it would be...

I understand the reasons to return a dense matrix. Another alternative we should consider is to simply not allow functions where f(0) != 0 to apply elementwise to sparse matrices.

Good points, @nhuurre. I agree about `exp()` being an implementation detail for ``. We were just talking about this today in the Stan meeting and it's come up in other...

With no constraints, there's no problem with accidentally removing sparsity. Other constraints are consistent with sparsity, like ``. We'll eventually want to have something like a sparse Cholesky factor or...

> can it go together with the Pull Request for the new function? Yes. > Which tests are required Value tests and then derivative tests for the function. I'm not...

Thanks for the PR! I don't understand these densities well enough to review myself. I'm afraid that everyone who might review this is scrambling to fix outstanding issues for the...