Type assertion in MyPy

Question

Some functions like numpy.intersect1d return differents types (in this case an ndarray or a tuple of three ndarrays) but the compiler can only infer one of them, so if I like to make:

intersection: np.ndarray = np.intersect1d([1, 2, 3], [5, 6, 2])

It throws a type warning:

Expected type 'ndarray', got 'Tuple[ndarray, ndarray, ndarray]' instead

I could avoid this kind of problems in other languages like Typescript where I could use the as keyword to assert the type (without impact in runtime). I've read the documentation and saw the cast function, but I'd to know if there is any inline solution or something I'm missing.

mypy.readthedocs.io/en/stable/casts.html - since the docs don't mention any alternative way to do type assertions, I would assume there isn't one. — kaya3
– kaya3, Commented Sep 10, 2020 at 12:31
Thanks for your comment! If you like make an answer and i'll mark as the accepted one — Genarito
– Genarito, Commented Sep 10, 2020 at 18:45

ArtOfWarfare · Accepted Answer · 2022-06-23 19:07:04Z

19

According to the MyPy documentation, there are two ways to do type assertions:

As an inline expression, you can use the typing.cast(..., ...) function. The docs say this is "usually" done to cast from a supertype to a subtype, but doesn't say you can't use it in other cases.
As a statement, you can use assert isinstance(..., ...), but this will only work with concrete types like int or list which are represented at runtime, not more complex types like List[int] which can't be checked by isinstance.

Since the documentation doesn't mention any other ways to do type assertions, it seems like these are the only ways.

edited Jun 23, 2022 at 19:07

ArtOfWarfare

21.7k19 gold badges150 silver badges203 bronze badges

answered Sep 10, 2020 at 19:55

kaya3

51.6k7 gold badges87 silver badges118 bronze badges

Sign up to request clarification or add additional context in comments.

2 Comments

julaine Over a year ago

And there can't be a better way, because List[int] is not a class. List is a class and you can check if its elements are int, but there is nothing about a list-object that tells you it can only hold a certain type of object. In Java they call this 'type erasure', I don't know if python uses that term too, but the concept is the same. The generic parameter does not exist at runtime.

julaine Over a year ago

A runtime-type-checking-library like beartype will try to assert List[int] by checking the first element (performant default) or all elements. For lists this works, for an Iterable[int] even that approach would be impossible.

Collectives™ on Stack Overflow

Type assertion in MyPy

1 Answer 1

2 Comments

Your Answer

Linked

Hot Network Questions

Collectives™ on Stack Overflow

1 Answer 1

2 Comments

Your Answer

Sign up or log in

Post as a guest

Linked

Related