Trouble Converting Collections from Python to Julia

[mrufsvold]

I'm sure I'm just overlooking something in the docs, so I'm sorry for the redundant question!

I have a dictionary in Python that I'd like to pass into a Julia function. But it is getting converted to type PyCall.PyDict{Pycall.Py,Pycall.Py}. I tried to explicitly convert using PyCall.convert(Dict{String, Any}, my_python_dict) but it throws:

MethodError: Cannot `convert` an object of type PythonCall.Py to an object of type String

MWE:

from juliacall import Main as jl
d = {"a" : 1}
jl.typeof(d)
# <jl PythonCall.PyDict{PythonCall.Py, PythonCall.Py}>

Just to double check, I also tried jl.typeof("a") and got String back.

Any thoughts on how to get the dictionary into a Julia Dict?

[cjdoris]

You want PythonCall.pyconvert, or its equivalent juliacall.convert from Python.

[mrufsvold]

Thank you!

[ShuhuaGao]

Could you please give a specific example? @cjdoris
For example, I have a d = {"Hello": [1, 2, 3, 4]} in Python, and I want to pass it into a Julia function

function examine(d::AbstractDict{<:AbstractString,<:AbstractVector})
    return length(first(d).second)
end

Did not figure out how to do it.

[cjdoris]

@ShuhuaGao your issue is that the type signature is too specific:

>>> from juliacall import Main as jl
>>> d = {"Hello": [1, 2, 3, 4]}
>>> jl.typeof(d)
Julia: PythonCall.PyDict{Any, Any}

As you can see, the dict is automatically converted to a PyDict{Any,Any}, which does not match the type signature of your function.

You can use juliacall.convert to instead convert to a specific type:

>>> from juliacall import convert as jlconvert
>>> jl.typeof(jlconvert(jl.PythonCall.PyDict[jl.String, jl.Vector[jl.Int]], d))
Julia: PythonCall.PyDict{String, Vector{Int64}}

(replace jl.PythonCall.PyDict with jl.Dict if you like).