PairList constructor broken due to change in `reinterpret` behaviour

[moradza]

I have an array of 3*216 and cubic PBC. Can you help with using this package to get the neighbor list?
NeighbourLists.PairList(arrsol, 1e-9, [2 0 0 ; 0 2 0; 0 0 2], [pbc.L pbc.L pbc.L])
The comment above throws error
typeof(arrsol)
Matrix{Float64} (alias for Array{Float64, 2})

[cortner]

Hm ... it seems that reinterpret changed its behaviour and nobody ever noticed for this package since we always work with SVectors:

X = rand(3, 216);
NeighbourLists.PairList(X, 1e-2, [1 0 0 ; 0 1 0; 0 0 1], (true, true, true))
# ERROR: MethodError: no method matching reinterpret(::Type{SVector{3, Float64}}, ::Matrix{Float64}, ::Tuple{Int64})
# ... 

I suggest the following workaround:

Xs = reinterpret(SVector{3, Float64}, X)[:]
nlist = NeighbourLists.PairList(Xs, 1e-2, [1 0 0 ; 0 1 0; 0 0 1], (true, true, true))

[cortner]

I'm not too sure I'll fix this since I am considering dropping support for this package in favour of CellListMap.jl. There seems little point in maintaining two packages that do the same thing. But if you are interested in continuing to use NeighbourLists.jl for the time being then I'll welcome PRs.

[moradza]

I will take a look into CellListMap.jl. Otherwise, I will see how to work around with NeighbourLists.jl. I am new to Julia and not sure how much I can really contribute. But if you got good PRs to learn and implement hit me up, I am looking to learn and contribute!

[cortner]

Start with CellListMap.jl, I haven't had extensive time to test it but it seems quite good and fast. I'd probably be careful with highly distorted cells and test carefully that the neighbour list is correct in those cases; it took me a long time to fix all the corner cases in NeighbourLists.jl.