vararg tuple types not working when vararg not at last place

[mauro3]

This does not work:

julia> (Int,Int,Float64)<:(Int...,Float64)
ERROR: type: subtype: expected Type{T<:Top}, got (DataType,DataType)

however inside methods a tuple like that does seem to do something, at least when a Any... is involved:

julia> methods(f, (Any...,Int,IOBuffer))
1-element Array{Any,1}:
 f(a::Int64,b::Int64,c::Float64) at none:1

julia> methods(f, (Int...,Float64))
0-element Array{Any,1}

I'm not sure how it should work but probably not as above. Allowing several ... inside a type tuple could be useful. Related: #10380, #8974

[JeffBezanson]

This is absolutely not supported yet. It would be very useful, but extremely difficult. The weirdness with methods is kind of a syntax bug that will be fixed by #10380; after that it won't be possible to write that.

[timholy]

By and large there is zero support for varargs anywhere except the last slot. Many of us have had times where it could have been useful, but having dug into the code that supports varargs I agree that it would represent an enormous increase in complexity.

I suspect that a good number of things we do with varargs now will be done with tuples in the future. See, for example, the discussion in #10691. So I don't think it's worth going to the effort to implement this, certainly not until we see how the tuple revamp works out.

[mauro3]

In that case, maybe throw an error on construction?

[JeffBezanson]

Yes, that's what will happen, because (Int..., ) will mean to splice Int into a tuple.

[JeffBezanson]

fixed by #10380