DagsterTypes cannot be used to annotate a config type

[yuval-airis]

I got an unclear error running my system, and I was unsure what part of the code triggered it:
Could not load location main.py to check for sensors due to the following error: dagster._core.errors.DagsterInvalidDagsterTypeInPythonicConfigDefinitionError: Error defining Dagster config class 'MachineForwardingConfig' on field 'file_keys'. DagsterTypes cannot be used to annotate a config type. DagsterType is meant only for type checking and coercion in op and asset inputs and outputs.

I defined this:

class MachineForwardingConfig(Config):
file_keys: List[str]

I thought that perhaps I needed to use Dagster's String class instead, but I got another error when I did:

dagster._core.errors.DagsterInvalidConfigDefinitionError: Error defining config. Original value passed: <class 'inspect._empty'>. <class 'inspect._empty'> cannot be resolved.

The call itself was:

materialize(
    [machine_forwarding],
    run_config=RunConfig(
        ops={ "machine_forwarding": MachineForwardingConfig(file_keys)}
    )
)

(machine_forwarding being the name of function defined as an asset, and therefore an op)

The op itself had no other params:

`@op

def machine_forwarding(config):`

Is that an incorrect way to call that op?

[smackesey]

Hi @yuval-airis, if you import List from typing instead of dagster it should work. The name collisions of top-level exports from Dagster with typing are a relic of Python 2 support and we should probably drop them.