Thomas Vetter

I’ll have to look some more into it, but it looks like `restructure()` always allocates for non-Base.Array. Maybe it needs a general fix? Regarding `adapt()` I have to play around...

Just a note here to say that I am playing around on this. I hope to have a PR with a (at least partial) fix today or tomorrow.

https://github.com/SciML/PreallocationTools.jl/pull/32 fixes the allocations for Base Arrays now. However, I couldn't get rid of the `reshape(reinterpret(...view(...))` wrappers **and** stay allocation free. That said, I don't know whether the wrappers interfere...

I am intrigued by this comment; it might be simpler to adapt this approach here as well. I don't think it will end up in a general solution for any...

Sorry, I am not usually dealing with models where complex numbers appear (unless I am doing something wrong :)); can you provide a MWE that illustrates the issue and what...

Thanks for your advice and insights. I am aware that heterogeneous arrays are not performant and I agree that feeding in a tuple here would be preferrable here. The thing...