Index element in column of arrays

I have a dataframe, where one of the columns contains 3d arrays. I want to get a vector containing a single value from each array. For example:

# create some data
ar = cat([1 2; 3 4], [5 6; 7 8], dims=3)
df = DataFrame(var1 = Int64[1,2,3], var2 = Int64[2,3,4], var3 = Array{Int64, 3}[ar,ar,ar])

Now I want to get the the second element from each array. I can do this manually e.g.:

df[:,3][1][2,1,1]
df[:,3][2][2,1,1]
df[:,3][3][2,1,1]

But I want to do it in a single go, I thought the following would work:

df[:,3][1:3][2,1,1]

But that returns an array. The desired output would be a vector containing 3 3 3. Any advice would be much appreciated.

you can use the getindex function along with a comprehension:

vector = [getindex(df[:,3][i], 2, 1, 1) for i in 1:size(df, 1)]
print(vector)

This code iterates over each row of the specified column (df[:,3]), retrieves the desired element using getindex, and stores it in the vector. The result will contain the first element from each array in the column.

So, the output vector should be [3, 3, 3].

Hope it helps.

You can do:

julia> [a[2, 1, 1] for a in df[!, :var3]]
3-element Vector{Int64}:
 3
 3
 3

though my first instinct would be:

julia> map(a -> a[2, 1, 1], df.var3)
3-element Vector{Int64}:
 3
 3
 3

Note that it's usually preferrable to use df[!, :column] (or df.column, which is equivalent) rather than df[:, :column] or df[:, columnnumber]. This is because df[:, .. makes a copy of that column first, which is unnecessary here and can affect the performance of your code. df[!, .. instead refers to the original column in the DataFrame itself, with no extra unnecessary copies.