Learning Julia

Fields with abstract types

Imagine you are creating your own type with some random field num in it. You can easily fall into doing this:

julia> struct SampleType
                        num
           end

This will allow num to take any type, and mark the beginning of some unwanted type mismatch errors too! However, from the compiler's perspective, this is bad as it will not be possible to extract more information from the object itself, which results in poor code performance:

julia> α = SampleType("stringvalue")
SampleType("stringvalue")

julia> β = SampleType(10)
SampleType(10)

julia> γ = SampleType(10.0)
SampleType(10.0)

What do you see? Well here, all the three variables are valid and make sense to the compiler, but which one would you choose (if you wrote the code targeted ...