NonEmpty[str] type cannot be used as sorting key

[flaeppe]

Not sure if I'm using NonEmpty incorrectly here with str. But the code below does not typecheck, and I think it should? (Since the underlying data type is a string)

from typing import NamedTuple
from phantom.sized import NonEmpty


class X(NamedTuple):
    value: NonEmpty[str]

sorted(
    [X(value=NonEmpty[str].parse("first")), X(value=NonEmpty[str].parse("second"))],
    key=lambda x: x.value
)

Results in the following error

10: error: Argument "key" to "sorted" has incompatible type "Callable[[X], NonEmpty[str]]"; expected "Callable[[X], Union[SupportsDunderLT, SupportsDunderGT]]"  [arg-type]

10: error: Incompatible return value type (got "NonEmpty[str]", expected "Union[SupportsDunderLT, SupportsDunderGT]")  [return-value]

[antonagestam]

@flaeppe NonEmpty[str] is a subtype of Sized and Iterable[str], but not a subtype of str. If you want a NonEmpty-type that is also a subtype of str you'd need to implement that like this:

# This has the added benefit that you can provide minCharacters in the schema
# which is distinct to minItems in OpenAPI! (Learned this recently).
class NonEmptyStr(str, NonEmpty[str]):
    ...

PhantomSized is generic in T, which is passed to SizedIterable which is generic in another T that is passed to Iterable. So the type argument to NonEmpty dictates the type of the value you'd get from next(NonEmpty()), it doesn't alter the super-type of the subscripted type :)

[antonagestam]

I think you've made the same assumption as the OP of this issue: #160 (comment)

I guess this is a good indication documentation could be improved.

[flaeppe]

Thanks for the clarification! That makes sense.

[antonagestam]

@flaeppe I'll close this issue for now but feel free to suggest ways to improve this 🙂