there is a way to transform an array x1 of Union{Missing,Float64} to an Array x2 of Float64 without allocating a new array? .x1 doesn’t containg missing values. an example:
x1 = Array{Union{Missing,Float64}}(undef,3)
for i = 1:3
x1[i]=i
end
julia> x1
3-element Array{Union{Missing, Float64},1}:
1.0
2.0
3.0
you can do this by allocating a new array, but there is another way?
identity.(x) should be good. Or map(identity, x).
Or
using Missings
disallowmissing(x)
AFAIK all solutions will allocate, unless you rely on internals (but don’t do that).
There’s an open issue for this but currently no “public” API for it:
a related question, how do i detect that said vector has a missing value?, I’m using in(true,(ismissing.(x)))==true but i’m sure there is a simpler way
any(ismissing, x) will do it.
Missings.disallowmissing(x)
convert(Vector{Float64}, x)
julia> a = Union{Float64,Missing}[1.1, 2.2]
2-element Array{Union{Missing, Float64},1}:
1.1
2.2
julia> using MappedArrays
julia> b = of_eltype(Float64, a)
2-element mappedarray(x->(MappedArrays.convert)($(Expr(:static_parameter, 1)), x), y->(MappedArrays.convert)($(Expr(:static_parameter, 1)), y), ::Array{Union{Missing, Float64},1}) with eltype Float64:
1.1
2.2
julia> eltype(b)
Float64
You can easily convince yourself that there is no copy of a made:
julia> b[1] = 200
200
julia> a
2-element Array{Union{Missing, Float64},1}:
200.0
2.2
Strictly speaking, there is no transformation either but a view, and the result is not an Array{Float64} (that was asked for) but something else.
That said, I think this may be the best solution under some circumstances, as it can be very efficient. So perhaps the original question should have been not how to get an Array{Floa64}, but an array with Float64 eltype.
This function is no longer working?
It works just fine as far as I know. I just tested it and works. Did you have an issue?
OH I misunderstood. I thought the function gets rid of missings but turns out it just converts if there are no missings. Sorry my bad.