Plans leading to v0.17.0

[mikeingold]

Breaking changes under consideration:

• Remove deprecated integral(f, geometry, ::GaussKronrod()) for geometries with more than one parametric dimension.
• Lift minimum compat for Meshes.jl to v0.52.12
  • Remove conditional version number check in @testitems for Frustum, Pyramid, and Wedge
  • Lift minimum compat for CoordRefSystems.jl to v0.16 for consistency with this min version of Meshes.
• Use of public keyword/macro for exported references?

Can probably still be implemented under v0.16.x: [updated]

• Add @test_throws on unsupported Enzyme differentials to monitor for resolution of this upstream issue.
• Implement Base.show methods

[JoshuaLampert]

Sounds good to me.

[JoshuaLampert]

Do you have an estimate when to release v0.16.4, @mikeingold? I think we could finish the @test_throws for Enzyme.jl and then release the new version. I see you already have a branch prepared, which looks like it is close to be ready.

[mikeingold]

Merging that branch is my last planned step before release. It’s nearly complete, but I had something I wanted to think over before opening a PR and haven’t gotten back to it yet. If I have time I may be able to knock this out later today.

[JoshuaLampert]

Alright, no worries.

[mikeingold]

@JoshuaLampert do you have any opinions on how we could implement public declarations while maintaining compatibility with Julia 1.10? I've done some light reading and it seems like opinions have been mixed on whether to use Compat.jl, a macro, a conditional include, or any number of other options.

[JoshuaLampert]

I have not really looked into that yet mostly because I didn't want to think about the hassle to find a good and easy to maintain solution, which works on Julia v1.10 and Julia v1.11. Personally, I would be totally fine with implementing public declarations once we switch to a minimum Julia version of v1.11.

[mikeingold]

Yeah, I'm a bit disappointed by the state of this feature because it sounds nice in theory but won't be supported by an LTS for a while. I've been trying to figure out an elegant way to implement this anyway, but I'm not really a fan of the solutions I've seen so far. Without any better ideas, I don't see a significant enough benefit to bother with this for now.

[JoshuaLampert]

I agree.