Pietro Vertechi

Concerning lazy versus eager: I agree that it only makes sense if the timestamps are sorted (or if one stores the `sortperm` in the lazy object). I'm actually not 100%...

I think it's just a problem with displaying things. The actual `fftshift` method shouldn't call `getindex`. Might be worth to inquire over at CUDA.jl as to whether there's a way...

See #37 (esp. https://github.com/JuliaArrays/ShiftedArrays.jl/pull/37#issuecomment-1083089037) for the rationale behind this decision.

I see how this would be useful (esp. for the high-level data manipulation syntax of DataFrames & co.). One issue is that for the `shifts` argument ShiftedArrays accepts also `Tuples`...

Sorry I did not reply to your comment on #175, I'll write my thoughts here. The problem with `replace_storage(f, s)`, if `f` mutates the eltype, is that it is not...

Yes, a PR is definitely welcome. As long as we mark this new interface as experimental (IMO the design is not completely clear) I think there's no harm in adding...

Ah, `LazyRow` is actually a bit too low-level, because it assumes that the passed index type is the optimal one for the array you are using. ```julia lh = LazyRows(t)[i,...

Adding the 1.0 milestone here because this would be a breaking change. There is a keyword constructor `StructArray{T}(; kwargs...)` to pass the named tuple of columns via `kwargs` (otherwise the...

CC: @simonbyrne, who I think wrote the implementation of the `dims` method.

Do other array libraries do this? It seems to me that `rand` always returns a plain array, regardless of the type of the container with the options: ```julia julia> using...