Mutating function and array assignment in Julia

Shuvomoy Das Gupta

January 18, 2021

A mutating function in Julia changes one of its inputs. A common convention is to use ! after the name of the function if it is a mutating function.

If we are mutating an array type object, for proper mutation we must change what the array contains, not what the array points to.

function frst_inpt_dbl_of_2nd_inpt!(A::AbstractMatrix{<: Real}, B::AbstractMatrix{<: Real})
    A .= 2*B # this .= notation ensures that we are changing the content of A
    # 💀 if we do A = B, then it will only change the reference to the array A
end

Let us test the function above.

B = ones(5,5)
A = similar(A)
frst_inpt_dbl_of_2nd_inpt!(A, B)
A

which indeed gives us the output:

A =
 2.0  2.0  2.0  2.0  2.0
 2.0  2.0  2.0  2.0  2.0
 2.0  2.0  2.0  2.0  2.0
 2.0  2.0  2.0  2.0  2.0
 2.0  2.0  2.0  2.0  2.0

So the contents of A has indeed changed.

Just to experiment, let us experiment with a wrong implementation below.

function wrong_frst_inpt_dbl_of_2nd_inpt_💀!(A::AbstractMatrix{<: Real}, B::AbstractMatrix{<: Real})
    A = 2*B # this will change the references to A, not the values
end

B = ones(5,5)
A = similar(A)
wrong_frst_inpt_dbl_of_2nd_inpt_💀!(A, B)
println(A)

which gives us the disastrous output:

A =
3.5e-323  8.4e-323   1.43e-322  2.17e-322  2.8e-322
4.0e-323  9.4e-323   1.73e-322  2.4e-322   2.87e-322
4.4e-323  1.1e-322   1.9e-322   2.5e-322   3.1e-322
7.0e-323  1.14e-322  2.03e-322  2.57e-322  3.16e-322
8.0e-323  1.33e-322  2.08e-322  2.67e-322  3.2e-322

For a more detailed explanation of why this happens, see the blog by John White here.