Function signatures in julia

6
votes

I am interested in understanding the actual difference, if any, between the following two function definitions

function foo(n::Integer)
    println("hello")
end

function foo{T<:Integer}(n::T)
    println("hello")
end

As far as I understand the second form triggers a new compilation every time the function is called for a new type T, but what actually happens in the first case? Is there any effect on performance associated to the first form?

Thanks

1
julia

1 Answers

6
votes

Parameters on function arguments are for method disambiguation rather than performance. From the docs http://docs.julialang.org/en/latest/manual/style-guide/#avoid-writing-overly-specific-types

The key thing to realize is that there is no performance penalty to defining only the general addone(x) = x + one(x), because Julia will automatically compile specialized versions as needed. For example, the first time you call addone(12), Julia will automatically compile a specialized addone function for x::Int arguments, with the call to one() replaced by its inlined value 1. Therefore, the first three definitions of addone above are completely redundant.

EDIT to address comment:

The difference between the two signatures is really only apparent when there are more than one argument

Consider the two functions:

julia> function foo(n::Integer, m::Integer)
           println(typeof(n), typeof(m))
       end

julia> function foo{T<:Integer}(n::T, m::T)
           println("Parameterized: ", typeof(n), typeof(m))
       end

In the first function n and m must both be integers but they do not have to be the same subtype of integer. In the second function, both m and n must be the same subtype

julia> foo(1, 2)
Parameterized: Int64Int64

julia> foo(1, Int32(1))
Int64Int32