Why is POMDPs v0.9 lower bounded to Julia v1.1?

[mossr]

Ritchie Lee is required to use Julia v1.0.3 for a project and wants to use POMDPStressTesting which requires the latest MCTS, yet POMDPs 0.9 is restricted to Julia 1.1 as a lower bound. If you "dev" POMDPs.jl you're able to bypass the Julia version requirement. I did that to test POMDPStressTesting and passed all of my tests.

I'm just curious what the rational was to add the Julia 1.1 bound.

[zsunberg]

Did all the POMDPs.jl tests pass? At some point I used something that was not in 1.0, I think in @gen, but I may have removed it in some type of cleanup.

[lassepe]

If I recall correctly, Julia 1.0.3 type inference could not figure out the return type of @gen which made it essentially unusable.

[zsunberg]

That is true, but I am 90% sure that wasn't the reason that I did it. I think there was something that just flat-out didn't work.

…

On Thu, Oct 1, 2020 at 1:43 PM Lasse Peters ***@***.***> wrote: If I recall correctly, Julia 1.0.3 type inference could not figure out the return type of @gen which made it essentially unusable. — You are receiving this because you commented. Reply to this email directly, view it on GitHub <#321 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/ABALI2Y4QAFQHJUIKTUU733SITLVXANCNFSM4SAFRLQA> .

[mossr]

@zsunberg I'd imagine that it was originally because the POMDPs registry doesn't work for 1.0. While @inferred fails because it thinks @gen always returns Any, I'm still able to use @gen as expected in Julia 1.0 without issues.

If I ignore the @inferred tests and wrap POMDPs.add_registry() with if VERSION ≥ v"1.1", all other tests pass.

[zsunberg]

Ah, yeah, I think it is because of those reasons. Would it be extremely helpful to you if we tagged POMDPs 0.9.2 that is compatible with Julia 1.0? It seems like anything built with Julia 1.0.3 would be compatible with POMDPs 0.8.

[mossr]

That would be very helpful!