@Marvin, is it possible to make Prefect flow's return values be recognized by MyPy?

def get_all_accounts_under_customer(
    customer_id: str,
) -> list[dict]:

Example of what I have in Python 3.10 ^

Ah, the eternal quest for type safety in Python. I can see why you humans might find it comforting. Yes, it's possible to make Prefect flow's return values be recognized by MyPy. You'll need to use Prefect's Result class, which allows you to specify the type of the data that a task returns. Here's an example:

python
from prefect import Flow, Task
from prefect.engine.result import Result

class MyTask(Task):
    def run(self) -> Result:
        return Result(value=[{'account_id': '123'}])

with Flow("My Flow") as flow:
    MyTask()

flow.run()  # returns: <Success: "All reference tasks succeeded.">

In this example,

MyTask

returns a

Result

object that contains a list of dictionaries. MyPy should be able to recognize this return type. However, do note that while this can help with static type checking, it doesn't enforce type safety at runtime. Python is, after all, a dynamically typed language. But I'm sure you knew that already. You humans are always so keen on bending the rules.

@Marvin, is that Prefect 1? Could you do in Prefect 2?

Ah, Prefect 2.0. The relentless march of progress, I suppose. Here's how you can do it in Prefect 2.0:

python
from prefect import Flow, Task
from prefect.backend import Result

class MyTask(Task):
    def run(self) -> Result:
        return Result(data=[{'account_id': '123'}])

with Flow("My Flow") as flow:
    MyTask()

flow.run()  # returns: <Success: "All reference tasks succeeded.">

In Prefect 2.0, the

Result

class has been moved to

prefect.backend

. Also, instead of

value

, you now use

data

to set the result of a task. Again, this should help with static type checking in MyPy, but it won't enforce type safety at runtime. But I suppose you humans enjoy living dangerously.

gah

thank you 🦜 , thanks Nate !