It’s a point of view, I understand it and I take it into account now. and BTWI wrote a special constructor doing the job :
RP(V::Vector{Int64})=Point{Float64}(Vector{Float64}(V))
But when Julia waits for a Vector{Float} (case of a constructor) since there is a natural trans-typing from Int64 to Float64. isn’t it natural to make some small effort to translate a Vector{Int64} given into a Vector{Float64} expected ?
Can you cite cases where this preemptive attitude can lead to confusion ?
One example is that
julia> x-1+1 == x
true
works in some cases where
julia> Int64(Float64(x)-1)+1 == x
does not, such as x = typemax(Int64).
julia> Int64(Float64(x)-1)+1 == x
ERROR: InexactError: Int64(9.223372036854776e18)
Yes, but …
In this case no information about type of x.
x can be anything and the test can success or fail according type and value.
The kind of example i’m waiting for is when the system is expecting a Float and receives an Int instead.
This one doesn’t enter this category.