malcolmgreaves / pywise / 77

95%

push

travis-ci

Enable deserialization support for dataclasses with generic types (#5)

Upgrades `deserialize` (and `serialize`) to handle @dataclass types parameterized with generics. Under the hood, the new `_align_generic_concrete`, `_fill`, and `_exec` functions in the `serialization` module handle reifying the parameterized type information at runtime. These new functions modify `_dataclass_field_types` to properly support "filling-in" the generic "holes" and output fully specified types. Extensive tests have been added to explicitly test deserialization from generic @dataclass types, including a very deeply nested example.

Note that it's not possible to support generics in `NamedTuple` deriving types. I.e. if given `class X(NamedTuple,Generic[T]):...` then `X[int]` is an invalid expression in Python. Since such expressions cannot be used, it's not possible to parametrize the generic types and thus it is not possible to gather reified type information at runtime.

This is a major API change: pre `1.x.x` rules dictate a minor version bump to `0.3.0`.

80 of 80 new or added lines in 3 files covered. (100.0%)

311 of 348 relevant lines covered (89.37%)

1.79 hits per line