Push!() giving Method type error

New to Julia
1

I am getting a Method error, specifically a type conversion error, specifically Array{Float64,2} to type Int64.

I don’t understand types very well yet, and definitely don’t understand why push!() would be trying to convert to an integer. Here’s my code, “principal_minor.jl” is working well in the REPL.

using Symbolics
using Latexify
using SymPy
using LinearAlgebra

include("principal_minor.jl")

function characteristic_polynomial(A)
    z = size(A)
    n = z[1]
    p = [1]
    for i = 1:n
        p_0 = (-1)^i*sum(principal_minor(A,i))
        #print(typeof(p_0))
        push!(p, p_0)
    end
    return p
end

I’ve tried casting a few of the arguments but didn’t work. Any feedback is appreciated, especially quick tips about dealing with these type errors as they have been the roadblock I come across the most.

Thanks for reading!

2

p is of type Vector{Int} and you try to push! something like a Array{Float64,2} (p_0) into this array.

julia> p = [1]
1-element Vector{Int64}:
 1

Your code is not a MWE (minimal working example) so it’s not clear what p_0 is.
What is the output of print(typeof(p_0))?
For now a naiv solution can be not to initialize p as you do it, but e.g.:

function characteristic_polynomial(A)
    z = size(A)
    n = z[1]
    for i = 1:n
        p_0 = (-1)^i*sum(principal_minor(A,i))
        if ! @isdefined p
                p=Array{typeof(p_0)}(undef,0)
        end
        push!(p, p_0)
    end
    return p
end

This assumes that the type of p_0 is always the same.

1 Like
3

thanks. I will look into the details of your response and follow up.

2 Likes
4

exactly as you said. I fixed it by changing sum(A) to sum(sum(principal_subminor(A,i))).
while this works for arrays with entries that are matrices, I know it won’t work for tensors of higher dimension, though I feel like I could write a function to do so. regardless, not relevant in this characteristic polynomial computation at the moment.

thanks for the response.

Edited that is, sum(sum(principal_subminor(A,i))

1 Like
5

small suggestion

z = size(A)
n = z[1]
for i = 1:n

can become

for i in 1:size(A, 1)

3 Likes
6

If you want there’s always the option of playing the broadcasting game:

p = (-1).^(1:size(A, 1)) .* sum.(principal_minor.(Ref(A), 1:size(A, 1)))
1 Like