Assign a float value to a vector originally of integer type

General Usage
1

I created a vector:

a = [1, 2, 3 ]

Then I tried to modify its value:

a[2] = 0.1

I got an error.

This kind of performace is ok in MATLAB.

How to make it work in Julia? I have to declare that ‘a’ is a float vector at the very beginning?

2

Yes.

a = Float64[1, 2, 3 ]
1 Like
3

Alternatively you can include a float in your vector:

julia> [1, 2, 3.0]
3-element Vector{Float64}:
 1.0
 2.0
 3.0

(you could even go so far to leave out the zero - [1, 2, 3.] will work but I consider these decimal-less float numbers a bit of a foot gun)

1 Like
4

It’s OK in Matlab because [1,2,3] is already a double float array in Matlab

>> class(a)
ans =
    'double'

If you modify a you will get

>> a(2) = 0.1
a =
    1.0000    0.1000    3.0000

You have to take extra steps to make an integer array in Matlab:

>> b = int32([1,2,3])
b =
  1×3 int32 row vector
   1   2   3

Now, if you modify b, it rounds to the closest int32:

>> b(2) = 0.1
b =
  1×3 int32 row vector
   1   0   3
4 Likes
5

To be clear, the reason this works in Matlab is that Matlab by default stores all numbers as floating point under the hood.

In Julia, the numbers [1, 2, 3] are all interpreted as integers, and a vector containing only integers will be a vector of integers. You can’t insert a floating point value directly into an integer type, so you get an error.

These proposed solutions are to either specify up front that the vector should be of a floating point type, or to specify the numbers as [1.0, 2.0, 3.0] since those are interpreted as floating point values.

3 Likes