Support Passing Dataclass Values via Command Line

[Future-Outlier]

Tracking issue

flyteorg/flyte#4486

Why are the changes needed?

Currently, we don't support passing dataclass values via command line after this PR.

For example, using this command line will raise an error.
pyflyte run --remote --image localhost:30000/any:0528-1445 dataclass_task.py foo_wf --a '{ "x": 1, "y": "2", "z": { "1": "one", "2": "two" }, "w": [1, 2, 3] }'

What changes were proposed in this pull request?

If we use only dataclass decorator only for our dataclass value, convert it to dataclass_json.

How was this patch tested?

local execution, remote execution and unit test.

Setup process

from flytekit import task, workflow
from dataclasses import dataclass

@dataclass
class Datum:
    x: int
    y: str
    z: dict[int, str]
    w: list[int]

@task
def foo_dataclass(a: Datum) -> int:
    if type(a) == Datum:
        return a.x
    return 0

@workflow
def foo_wf(a: Datum) -> int:
    return foo_dataclass(a=a)

Screenshots

local execution

remote execution


unit test

Check all the applicable boxes

• I updated the documentation accordingly.
• All new and existing tests passed.
• All commits are signed-off.

[wild-endeavor]

renaming... no need for [MINOR] in the pr desc please. That typically means that the change is non-backwards compatible and will need a minor release instead of a patch release. (since we're kinda using minor as major at this point)

[Future-Outlier]

renaming... no need for [MINOR] in the pr desc please. That typically means that the change is non-backwards compatible and will need a minor release instead of a patch release. (since we're kinda using minor as major at this point)

No problem, thank you