Initializing an empty Julia array with unknown size

General Usage
1

I have 6 different scenarios, and under each scenario, there could be one to 100 models available for me to get the data. My goal is to store them in A as below (n is an unknown number between 1 and 100):

for i in 1:6 # Go through all of the scenarios
   # go through the 100 models
   A = []; # the step I need help.
   n = 0;
   for j = 1:100
      if a model is available 
          n .+ 1
          A[n] = Model n (35 grids, each with a size of 360 x 180);
       end
   end
   # other calculations that requires A
end

Here is my question. Under each of the scenario, I would need to reset my A to empty values. How do I do that? In Matlab, I can just do the below:
A = [];
Unfortunately, it does not work for Julia, as we know, Julia needs to know the exact size of the A, which is unknown in this example. How should I deal with this situation?

Many thanks!

2

Use push! to add elements to the array:

julia> x = Matrix{Float64}[] # an empty array of matrices (better than [])
Matrix{Float64}[]

julia> push!(x,[ 1.0  2.0 ; 3.0  4.0])
1-element Vector{Matrix{Float64}}:
 [1.0 2.0; 3.0 4.0]
2 Likes
3

This addresses my issue well :+1: :+1: Thank you very much!

4

May I ask one more question?

Actually, my model output is not a matrix with the size of 360x180. It is 35 matrices with the size of 360x180, like this: Model1[1] = grid1, Model1[2] = grid2, ..., Model1[35] = grid35.

In that case, how do I modify the below to fit my purpose?
x = Matrix{Float64}[]

Using the above directly gives me the below error:

LoadError: MethodError: no method matching Matrix{Float64}(::Vector{Any})
Closest candidates are:
  Array{T, N}(::AbstractArray{S, N}) where {T, N, S}

Many thanks!

5

One would need to see exactly what is the type of grid.

But you can use x = [] and just add anything there, it is just that this is of type Any and may have some performance effect later (although all this is highly speculative with the information we have here).

3 Likes
6

Thank you! Good to know that.

I just run my code with typeof for my 35 layer grid, and I got the below:
typeof(Model1) = Vector{Any}

7

You can use Vector{Vector{Any}} then for the container, but more for clarity than for anything, since you are already getting Any from the data you are fetching.

8

Technically, accessing the outer Vector{Vector{Any}} is not type-unstable, only accessing a position in one of the inner Vector is type-unstable. However, the type may have been deduced as Vector{Any} because some of the inner vectors are Vector{X} and other Vector{Y} (in this case the inner vectors themselves are of specific types and the outer vector must be Vector{Any}).

1 Like
9

Are you sure you typed that in correctly? That code is fine, and it shouldn’t produce an error:

julia> x = Matrix{Float64}[]
Matrix{Float64}[]

From the error you posted, it seems like you may have actually done:

julia> x = Matrix{Float64}([])
ERROR: MethodError: no method matching Matrix{Float64}(::Vector{Any})

Note the extra ().

For some more background, in Julia you can write: T[] to get a vector of elements of type T. In other words, these two things are the same:

julia> Int[]
Int64[]

julia> Vector{Int}()
Int64[]

So if you want a collection of matrices, you would substitute T = Matrix{Float64} and write:

julia> Matrix{Float64}[]
Matrix{Float64}[]

or, if you prefer:

julia> Vector{Matrix{Float64}}()
Matrix{Float64}[]