Julia: Efficiency and indicating argument types

user_14472815 0
Viewed 66

I have three related questions about specifying types of function arguments and outputs. I'm defining a function f that I expect to call many times, so I'm looking to improve efficiency everywhere I can. My function definition looks something like this:

function f(x, y)
    ...
    return z
end

I know that x will be an Array{Float64,1}, y will be an Array{Float64,2}, and z will be a Float64.

My questions are:

  1. Is there an efficiency benefit to specifying the types of the inputs in the function definition, i.e. function f(x::Array{Float64}, y::Array{Float64})?
  2. Is there any additional benefit to making the types more specific by specifying that x is 1-dimensional and y is 2-dimensional, i.e. function f(x::Array{Float64,1}, y::Array{Float64,2})?
  3. Is there any efficiency benefit to specifying the type of z, i.e. function f(x::Array{Float64,1}, y::Array{Float64,2})::Float64?

Thanks so much! And apologies if these questions have been addressed previously.

julia methods function types
2 Answers
user_56021490

There is a huge benefit for the human reader of the code. Consider the difference for a human reading for the first time (or perhaps after a long time away from the code) :

function f(x, y)

vs.

function f(x::Array{Float64}, y::Array{Float64})::Float64
user_2391010
Related