[Core feature] Support partial tasks

[cosmicBboy]

Motivation: Why do you think this is important?

Currently, invoking tasks in workflows or dynamic workflows requires fully-specifying the arguments to a task. This behavior is clunky in the context of map tasks, where users need to write a bunch of code to prepare the inputs to a map task (see here.

The purpose of this issue is to support partial tasks, which allow users to partially bind inputs to a task so that it can then be invoked with the remaining arguments, similar to https://docs.python.org/3/library/functools.html#functools.partial.

Goal: What should the final outcome look like, ideally?

Requirements

Partial tasks should:

1. support partial task inputs that are python values
2. support partial task inputs that are promises
3. support partial task definitions in the top-level scope of a module
4. support use of partial tasks in map tasks

Once partial tasks are implemented, it should look something like:

Support partial task inputs that are python values

from flytekit import task, workflow, partial

@task
def t1(x: int, y: float) -> float:
    return x + y

@workflow
def wf(y: float):
   partial_t1 = partial(t1, x=5)
   return partial_t1(y=y)

Support partial task inputs that are promises

from flytekit import task, workflow, partial

@task
def t1(x: int, y: float) -> float:
    return x + y

@task
def t2() -> int:
    return 5

@workflow
def wf(y: float):
   partial_t1 = partial(t1, x=t2())
   return partial_t1(y=y)

Support partials in the top-level module scope

If a user wants to specify partial tasks using python literal values in the top-level scope:

from flytekit import task, workflow, partial

@task
def t1(x: int, y: float) -> float:
    return x + y

partial_t1 = partial(t1, x=5)

@workflow
def wf(y: float):
   return partial_t1(y=y)

Support partials in map tasks

from flytekit import task, workflow, partial, map_task

@task
def t1(x: int, y: float) -> float:
    return x + y

@workflow
def wf(y: List[float]):
   partial_t1 = partial(t1, x=5)
   return map_task(partial_t1)(y=y)

Describe alternatives you've considered

Besides syntactic alternatives, not supporting partial tasks would require writing additional code to make some of Flyte's constructs work (e.g. map tasks with multiple inputs).

Propose: Link/Inline OR Additional context

No response

Are you sure this issue hasn't been raised already?

• Yes

Have you read the Code of Conduct?

• Yes