Can't convert a subtype of a base Python type to a custom Julia type

[sethaxen]

When a Python package extends a basic type like a dict to add some functionality, we can't use pytype_mapping to convert it to a custom Julia type. Instead, the default conversion (in this case to Dict) takes precedent.

Example:

using PyCall
import Base: convert

py"""
class MyDict(dict):
   pass
x = MyDict({"a": "b", "c": "d"})
"""

struct MyDictWrapper
   o::PyObject
end

Base.convert(::Type{MyDictWrapper}, obj::PyObject) = MyDictWrapper(obj)

julia> pytype_mapping(py"MyDict", MyDictWrapper)
1-element Array{Tuple{PyObject,Type},1}:
 (PyObject <class '__main__.MyDict'>, MyDictWrapper)

julia> py"x" # expected: MyDictWrapper(PyObject {'a': 'b', 'c': 'd'})
Dict{Any,Any} with 2 entries:
  "c" => "d"
  "a" => "b"