Flyte enables production-grade orchestration for machine learning workflows and data processing created to accelerate local workflows to production.

Flyte

<https://github.com/flyteorg/flytekit/pull/1661|#1661 Pickle as a default Transformer rather than a special case of function signatures>
Pull request opened by <https://github.com/elibixby|elibixby>
*TL;DR*

Right now only "top level" missing types (those defined directly in a function) are pickled if they don't have a registered `TypeTransformer`. This PR alters that behavior to anywhere `TypeTransformer`s are used.

*Type*

☐ Bug Fix
• [X ] Feature
☐ Plugin

*Are all requirements met?*

☐ Code completed
☐ Smoke tested
☐ Unit tests added
☐ Code documentation added
☐ Any pending items have an associated Issue

*Complete description*

The special casing for pickling as a serialization/deserialization method leads to some strange behavior where, e.g

```

@task
def my_task(foo: Foo):
   ...
```

Is valid, because that type is converted, but e.g. `Foo` inside of a dataclass etc is not. Because the conversion to the placeholder `FlytePickle` type happens in a function which is only called once on top level type, and only traverses the `origin` stack, rather than being used anywhere `TypeTransformers` are used. This PR fixes that problem, while also significantly simplifying the logic to support it.

*Tracking Issue*

<https://github.com/flyteorg/flyte/issues/|https://github.com/flyteorg/flyte/issues/>

*Follow-up issue*

_NA_  
OR  
_<https://github.com/flyteorg/flyte/issues/|https://github.com/flyteorg/flyte/issues/>_
<https://github.com/flyteorg/flytekit|flyteorg/flytekit>
DCO: DCO
:white_check_mark: 1 other check has passed
1/2 successful checks