Help with calling a julia function from python when the input is an array of arrays

[mestinso]

If I have a Julia function that looks like the following:
myfunc(x::AbstractVector{<:AbstractVector{<:Real}}), how do I call that from python using juliacall in the simplest way possible?

I've tried jl.myfunc(np.array([np.array([1,2,3])])) but that was unsuccessful.

Got an error like the following: "...expected AbstractVector{<:AbstractVector{<:Real}}, got a value of type PyArray{Int64, 2, true, false, Int64}..."

@cjdoris I will note I found this thread and I did get your option 4 working. But I was hoping I could get something along the lines of your option 1 working. Note that for situations where my input is a vector I've been successful, where I've failed is a vector of vectors.

[cjdoris]

AFAIR np.array([np.array([1,2,3])]) is actually a matrix, not a vector of vectors. There are no conversions in PythonCall from a matrix to a vector of vectors, so you'll need to make it a vector of vectors yourself.

You could just have a list [np.array([1,2,3])] or do np.array([np.array([1,2,3])], dtype=object). However neither of these will convert to a AbstractVector{<:AbstractVector{<:Real}} because now the type of the elements of the given list or array cannot be inferred (the element type is just object).

I think you'll need to do some manual conversion using juliacall.convert or do the conversion Julia-side.

[mestinso]

Thanks. For anybody who might be reading this, the following is basically what I did:
jl.myfunc(jl.convert(jl.Vector[jl.Vector[jl.Float64]],[[1,2,3],[20,50,90]]))