I have a generic ContainerTask that could receive various arguments

def dbeam_ct():
    return ContainerTask(
        "dbeam",
        image="...",
        input_data_dir="/var/inputs",
        output_data_dir="/var/outputs",
        inputs=kwtypes(
            when=datetime.datetime,
            docker_container_args=List[str],
        ),
        command=["some-command],
        arguments=["--partition={{.Inputs.when}}"] + "{{.Inputs.docker_container_args}}",
    )

is there a way i could pass the docker_container_args as a list of strings, and not as a string itself?

the error i see is

TypeError: can only concatenate list (not "str") to list

Inputs will look like this:

lp = LaunchPlan.create(
    "...",
    wf,
    fixed_inputs={
        "dockercontainer_args": [
            "--partitionColumn=whatever",
            "--output=<gs://some-bucket>",
            "--connectionUrl=jdbc:postgresql://...",
            "--table=executions",
        ]
    },
)

notice the inputs List[str]

since there can be any number of inputs to this docker container, i would prefer if i didnt have to configure default values for each, but the user could define them somehow in the Launch Plan inputs

Should i wrap them in a workflow instead?

@swift-animal-75798 at the moment, a collection is simply handled as a collection of files - https://github.com/flyteorg/flytecopilot/blob/43938f5382ae580639bb5934199c7ea7566b027d/data/download.go#L251. One per item in the collection

I think this can be improved, by handling list of primitive values differently

Should we create an issue for flytecopilot? I believe it's a solid use case for us (500+ workflows), there are a lot of teams that use docker images for certain operations already. In the non-flyte system we just schedule a docker image execution with any arguments passed from the user One good example is https://github.com/spotify/dbeam

Ideally we would like to also configure ENV VARs based on inputs. Which i think might not be possible right now

should i just try to do that by using the K8sPod plugin?

Would you be open to discussing the usecase

Copilot needs little more love

Would love to understand how it can solve your usecase and I think we would be more than happy to adapt