Thomas Vetter

Can you comment on how/where/which context it’s safer? (when deciding whether a breaking change “is worth it”, would be good to know the perceived value of the change?)

Since I have a few nested differentiations occurring in my codes, I am supportive of a change that shortens tags and makes stacktraces more readable as a result - is...

This correctly works with the out of place version, so the issue seems to be isolated to in place calls. MWE below: ``` using DifferentialEquations function f(du,u,p,t) out = similar(u)...

I just did literally the same Pull Request on Setfield.jl (https://github.com/jw3126/Setfield.jl/pull/174) where tests pass locally (let's see what happens on CI). If it works there, maybe it helps someone with...

Just wanted to say thank you for creating this! Have a year ago I posted a hacky and ugly coded version of it on discourse (and got some help making...

I am doing something similar in a private package. I don’t think it’s official API, so I just have a few tests that will be indicative when things break. I...

It isn't really the same issue? In #316 the autocomplete doesn't work for the Ptr{T} type, but it would work for struct T directly. In this issue, there is only...

If using a simple `reshape()` instead of the `ArrayInterfaceCore.restructure()` the allocations disappear. Copy-paste version below: ```julia using ForwardDiff, PreallocationTools function my_get_tmp(dc::PreallocationTools.DiffCache, u::Vector{T}) where {T

I checked with earlier versions of PreallocationTools.jl (down to 0.2.4) where it was still ArrayInterface 5.y.z; the allocations also happen there.

Is this the right way to test it? ```julia using ArrayInterfaceCore a = zeros(10) ArrayInterfaceCore.restructure(a, reinterpret(Float32, view(a, 1:5))) ``` Then yes: ``` 10-element Vector{Float32}: 0.0 0.0 0.0 0.0 0.0 0.0...