Brian Chen

No objections, we'd just have to make a breaking release with it. Anything else we'd want to get in said release?

@mrazomej when cross-posting issues, please link back to the original so that readers have some context. In this case, there's plenty of background in https://github.com/FluxML/Zygote.jl/issues/1502. We also have a Slack...

> That's because Zygote is too weak to infer type through your `[2:end]` semantics Instead of tuple slicing, you can use `Base.tail`. ~~`ntuple(i -> mytup[i + 1], length(mytup) - 1)`....

Are there benchmarks floating out there that demonstrate this? I was under the impression that rule support was not fleshed out enough to support all but the most basic of...

Thanks for the link, now I'm curious _why_ SciML hasn't picked it up for the diffeq(flux) ecosystem. I'm sure there's a perfectly legitimate reason, but I didn't see any discussions...

I thought I'd run into this and solved it with Optimisers.jl, but looks like that didn't happen: https://github.com/FluxML/Optimisers.jl/issues/25. Pinging @logankilpatrick now this has happened a couple of times and he...

If https://github.com/FluxML/Optimisers.jl/pull/142 works out, I'd recommend trying the same here.

I don't know of anything Functors can do that StructWalk couldn't on principle, but there might be small details in e.g. handling of array wrappers that would require changes.

`alignedstyle` and co. look very interesting, thanks for writing this out! I played around and managed to get `zippedpostwalk(+, a, b)` working just by changing to `f(xs...)` in zippedwalk, do...

Sorry, nothing from my end yet. I'll need to take a solid chunk of time to sit down and figure out the details of the default align and exact same...