How to deserialize multidict to data model?

multidict: https://github.com/aio-libs/multidict

import dataclasses
import typing
import cattrs
import multidict


@dataclasses.dataclass
class Model:
    a: typing.List[str]
    b: str


d = multidict.MultiDict([('a', '111'), ('b', '2'), ('a', '333')])
obj = cattrs.structure(d, Model)
print(obj)

d = multidict.MultiDict([('a', '111'), ('b', '2')])
obj = cattrs.structure(d, Model)
print(obj)

want:

Model(a=['111', '333'], b='222')
Model(a=['111'], b='222')

actual:

Model(a=['1', '1', '1'], b='222')
Model(a=['1', '1', '1'], b='222')

Start with the Python reproduction in issue #407 and inspect cattrs's structuring path for multidict mappings and list-typed dataclass fields. Done means repeated keys produce all values in Model.a, a single key produces a one-item list, and the scalar b value remains correct; add or run a regression test if the repository provides a relevant test location.