LatLon(missing, missing) triggers a stack overflow error

[austinlostinboston]

This doesn't seem like an intended result for passing missing to LatLon. At the very least, this should trigger an internal exception within Geodesy. There also might be a greater question of whether a LatLon(missing, missing) should be allowed to exist.

julia> LatLon(missing, missing)
ERROR: StackOverflowError:
Stacktrace:
[1] LatLon(::Missing, ::Missing) at /Users/my_name/.julia/packages/Geodesy/qjth5/src/points.jl:37     (repeats 80000 times)

[visr]

Geodesy.jl/src/points.jl

Lines 33 to 37 in 2774631

	struct LatLon{T <: Number}
	lat::T
	lon::T
	end
	LatLon(lat, lon) = LatLon(promote(lat, lon)...)

Oh that's not so nice behavior.
Since missing is not a Number, it is not supported in LatLon currently. But when you try it, it calls the method with promote, which, since the result is still not two identical subtypes of Number, calls itself again and again, leading to the StackOverflowError. Same happens for LatLon('a', 'b') for instance.

I'm guessing that method should also be restricted to Number types:

LatLon(lat::Number, lon::Number) = LatLon(promote(lat, lon)...)

[andyferris]

I'm guessing that method should also be restricted to Number types

Yes, for the short term, let's do that!

@austinlostinboston For the moment, you might have better luck considering thet LatLon as an "entire" object that may be missing, i.e. model your data with Union{Missing, LatLon{Float64}} or whatever.

[austinlostinboston]

All good suggestions and thanks for the help @visr and @andyferris!