Support for dataclass fields of type Optional[List[X]]

[kylegentle]

Summary

The library currently breaks for dataclasses with fields of type Optional[List[X]], returning the following error:

  File "dcj_bug/ex.py", line 66, in <module>
    Adult.schema().loads(test3, many=True)  # Fail
  File "<home_path>/.cache/pypoetry/virtualenvs/dcj-bug-py3.7/lib/python3.7/site-packages/dataclasses_json/api.py", line 100, in schema
    infer_missing)
  File "<home_path>/.cache/pypoetry/virtualenvs/dcj-bug-py3.7/lib/python3.7/site-packages/dataclasses_json/mm.py", line 86, in _make_default_fields
    cls)
  File "<home_path>/.cache/pypoetry/virtualenvs/dcj-bug-py3.7/lib/python3.7/site-packages/dataclasses_json/mm.py", line 103, in _make_default_field
    cons = _type_to_cons[cons_type]
KeyError: typing.List[__main__.Child]

I've attached some code below that demonstrates the issue.

Please let me know if there's any information or assistance I can provide; I dug into the code a bit in mm.py where it's breaking, but wasn't able to figure out what the correct logic should be in mm._make_default_field.

Example

import json

from dataclasses import dataclass
from typing import Optional, List

from dataclasses_json import DataClassJsonMixin


@dataclass
class Child(DataClassJsonMixin):
    name: str


@dataclass
class Adult(DataClassJsonMixin):
    name: str
    children: Optional[List[Child]] = None


test1 = """
{
    "name": "Foo"

}
""".strip()

test2 = """
{
    "name": "Bar",
    "children": [
        {
            "name": "Baz"
        },
        {
            "name": "Bat"
        }
    ]
}
""".strip()

test3 = """
[
    {
        "name": "Foo"
    },
    {
        "name": "Bar",
        "children": [
            {
                "name": "Baz"
            },
            {
                "name": "Bat"
            }
        ]
    }
]
""".strip()

test4 = json.loads(test3)

Adult.from_json(test1)  # OK
Adult.from_json(test2)  # OK

# Adult.from_json(test3)  # Breaks
Adult.schema().loads(test3, many=True)  # Breaks
# Adult.schema().load(test4, many=True)  # Breaks

[kylegentle]

Here is a slightly hacky workaround to create lists of dataclass_json instances with fields of type Optional[List[X]]:

-# Adult.schema().load(test4, many=True)  # Breaks
+adults = [Adult.from_json(json.dumps(a)) for a in test4] # OK

[lidatong]

Thanks for reporting this. Fixed in #66