minimum, maximum and extrema with specifing dims for an array that have all missing value dim doesn't work

[Paalon]

minimum, maximum and extrema with specifing dims for an array that have all missing value dim doesn't work.

julia> minimum([1 missing; 2 missing], dims=1)
ERROR: TypeError: non-boolean (Missing) used in boolean context
Stacktrace:
 [1] reducedim_init(::Function, ::typeof(min), ::Array{Union{Missing, Int64},2}, ::Int64) at ./reducedim.jl:148
 [2] _mapreduce_dim(::Function, ::Function, ::NamedTuple{(),Tuple{}}, ::Array{Union{Missing, Int64},2}, ::Int64) at ./reducedim.jl:317
 [3] #mapreduce#580 at ./reducedim.jl:307 [inlined]
 [4] _minimum at ./reducedim.jl:679 [inlined]
 [5] _minimum at ./reducedim.jl:678 [inlined]
 [6] #minimum#589 at ./reducedim.jl:652 [inlined]
 [7] top-level scope at REPL[6]:1

[pdeffebach]

Note that

julia> minimum([1, 2, missing])
missing

julia> min(1, missing)
missing

So this is just about the keyword argument for dims. This is not a case of adding missing propagation to a function that otherwise errors.

[nalimilan]

Good catch. I have an inefficient fix at #35323. An efficient one would take much more work.

[stev47]

seems to be fixed by 76952a8

[nalimilan]

As noted at #35323, this isn't completely fixed unfortunately. See the note there:

                # TODO: Some types, like BigInt, don't support typemin/typemax.
                # So a Matrix{Union{BigInt, Missing}} can still error here.