Andy Ferris

As a general comment, I'm hoping that in JuliaGeometry we can define some kind of "point" supertype (or interface) that may be widely used. Points would naturally support things like...

This makes perfect sense to me, Tim :)

Hmm, I see. It probably makes sense to rename the linear approximation to something more explicit. Add a deprecation for that, and allow `AffineMap(::LinearTransformation, ::Any)` to do the "sensible thing",...

You can pass around `Val(length(v))` to get it to specialize the inner code.

Where is this one at? > Also, how does this change as a whole impact performance? Do we have an answer to this one?

Yes, this isn't unreasonable... :) There has been a long-standing plan to support static ranges, LinearIndices and CartesianIndices with StaticArrays. These things would be returned by `axes` and `indices`. I've...

OK thanks. Please bump this again soon-ish if I don't get to this.

> this is a mismatch between "structural" and "natural" gradients Yeah I see - thanks, I was wondering if it was something like that. From https://github.com/JuliaDiff/ChainRulesCore.jl/pull/446#issuecomment-1158998523 it seems resolving this...

OK thanks for the attempt @mcabbott. I kinda wish I knew where to start - is there some internals documentation for Zygote etc somewhere so I can get a rough...

> Are you running with any plugins? We are using typedoc-plugin-markdown and typedoc-plugin-merge-modules. (But the symptom likely is related to what @ebekebe wrote above - we definitely use `ReturnType` and...