@Marvin Hi! I want to use from prefect.utilities.annotations.quote to speed up tasks that depend on large task outputs. How can I use this pattern together with task.submit() ? See example code below:

from prefect import task, flow
from prefect.utilities.annotations import quote

@task
def big_output():
    return [1 for _ in range(10**7)]

@task
def big_input(x):
    print(len(x))

@flow
def slow_flow():
    x = big_output.submit()
    big_input.submit(quote(x))

if __name__ == "__main__":
    slow_flow()
I'm trying to get the big_input task to start quickly after big_output task, but I get this error:

TypeError: object of type 'PrefectFuture' has no len()

It seems that while searching for relevant documents, we did not find exactly what we were looking for in relation to speeding up dependent tasks in Prefect with large outputs or using

quote

with

task.submit()

. However, searching the GitHub issues for the error you mentioned did not yield any directly related issues in the PrefectHQ repository. To address your usage question with

quote

and

task.submit()

, here's what you need to know: The

quote

function from Prefect's

utilities.annotations

module is used to avoid early evaluation of task outputs, which allows you to compose task dependencies that pass outputs between tasks without serializing and copying large data immediately. To fix the

TypeError

you're facing,

big_input.submit()

needs to properly unwrap the quoted future. When you pass

quote(x)

to

big_input

, you're passing a

PrefectFuture

object, and you cannot take the length of that future directly, as it is not the actual result of the task. One potential solution is to have

big_input

wait for the actual value of

x

inside the task. This can be done using the

get

method on the future object. Here's a simple modification to your

big_input

function:

python
@task
def big_input(x):
    real_value = x.get()  # Unwrap the actual value from the future.
    print(len(real_value))

And your

slow_flow

function should remain unchanged:

python
@flow
def slow_flow():
    x = big_output.submit()
    big_input.submit(quote(x))

This should allow your code to execute without the

TypeError

, as the

get()

method on the quoted Prefect future is used to obtain the real value once it's available. If you still encounter issues, you might need deeper insight into Prefect's execution model and the specific use of

quote

with

task.submit()

. You could consider reaching out on Prefect's community Slack or support channels for more direct assistance.