Implement TYPEDSIGNATURENORETURN

[ReubenJ]

Addresses #159 following the suggestion at #159 (comment).

[MichaelHatherly]

Thanks @ReubenJ, could I request though that the formatting changes aren't included to make the diff easier to review.

[ReubenJ]

Certainly! I also wanted to add another test or two before review. Thanks for the speedy reply!

[ReubenJ]

Pesky, pesky formatting. Should be good to go now @MichaelHatherly.

[fingolfin]

Perhaps first #178 should be merged and then CI re-run here?

[ReubenJ]

Also, not sure what your release workflow is—should I include a (patch) version bump in this PR?

[MichaelHatherly]

Also, not sure what your release workflow is—should I include a (patch) version bump in this PR?

Can be done separately to this one.

[ReubenJ]

Oops, something odd is happening, looks like the tests are not extensive enough yet. Give me a moment here.

[ReubenJ]

With the following docstring

"""
    $(TYPEDSIGNATURESNORETURN)

"""
function get_solver(pi::ProgramIterator)
        return pi.solver
end

I'm getting

search: get_solver

  get_solver(pi)
  

  get_solver(pi::ProgramIterator)

I wouldn't expect to see the untyped signature there. With TYPEDSIGNATURES, all I see is:

help?> get_solver
search: get_solver

  get_solver(pi::ProgramIterator) -> Any

[ReubenJ]

Stumped on this for the moment, I'll have to come back to it tomorrow.

[ReubenJ]

No luck reproducing that behavior I saw while testing the changes in another repo, so I think this is ready to go as-is.

In the process of trying to reproduce what I saw above, I switched the new tests to use ReferenceTests.jl. I found it much easier to work with for this use case, though I'm happy to revert this change if you don't want to add a test dep.

[MichaelHatherly]

though I'm happy to revert this change if you don't want to add a test dep.

Yes, please revert that.

If we're going to use ReferenceTests for testing that's fine, and I'd like to see that change, but it would need to be applied to all tests as a single change set rather than adding a few new ones that use it. Best to keep that as a completely separate PR rather than as part of this feature addition.

[ReubenJ]

Sure! No problem.

[ReubenJ]

Reference tests are removed, and the branch is up to date. Any other changes you'd like to see for this to go through?

[MichaelHatherly]

My only concern is whether this is readable enough with this naming or whether it needs any _ to break the words up. I realise that the other abbreviations don't, and you probably just went with the current naming conventions.

[ReubenJ]

I agree, it's not the most legible. Is it strictly necessary to use the const form, or is it possible to make the $(TypedMethodSignatures(false) form work? That seems much more flexible in general.