There should be a zero(::Missing) method

[yurivish]

zero(Missing) currently fails, as does one(Missing) and oneunit(Missing):

julia> zero(Missing)
ERROR: UndefVarError: T not defined

The error comes from this code in base/missing.jl:

for f in (:(Base.zero), :(Base.one), :(Base.oneunit))
    @eval function $(f)(::Type{Union{T, Missing}}) where T
        T === Any && throw(MethodError($f, (Any,)))  # To prevent StackOverflowError
        $f(T)
    end
end

Perhaps zero(::Missing) and one(::Missing) should be missing, since missing + missing == missing and the same for * and those two methods give you the additive and multiplicative identities.

@ararslan suggests adding @eval $(f)(::Type{Missing}) = missing to the loop.

[nalimilan]

Good question. It sounds a bit weird that iszero(zero(missing)) and zero(missing) == 0 would not be true. But maybe that's not an issue. Can you show an actual use case?

[yha]

I just encountered this when doing correlation plots (from StatsPlots).
It seems to arise easily by calling Statistics.cor on data with missing. Reproducing example:

julia> import Statistics: cor
julia> cor([zeros(1,1); missing])
ERROR: UndefVarError: T not defined

EDIT: for this use case, missing seems to be the right answer (at least for oneunit):

Base.oneunit(::Type{Missing}) = missing
x = rand(10)
cor([[x;missing] [0.0;x] [x;0.0] [x;0.0]])

yields a reasonable answer

4×4 Array{Union{Missing, Float64},2}:
 1.0         missing     missing     missing 
  missing   1.0        -0.0904971  -0.0904971
  missing  -0.0904971   1.0         1.0      
  missing  -0.0904971   1.0         1.0      

[nalimilan]

This has actually been fixed in March by #31303. That will be included in Julia 1.2.