Hey all, I have an asset define like this.

@asset(
    config_schema={
        "limit": int,
        "cohort_table": str,
        "database": str
    },
    required_resource_keys={'aws'},
    # ins={
    #     "cohort": AssetIn(
    #         input_manager_key="pandas_df_manager",
    #         key=AssetKey(('qld_health', 'qld_health_simple_cohort'))
    #     )   
    # },
    non_argument_deps={
        AssetKey(('qld_health', 'qld_health_simple_cohort'))
    }

)
def cohort(context,
    # cohort
):
    pass

when setting

non_argument_deps

to the source asset everything works as expected and the UI shows the correct lineage. however when trying to supply an argument asset using the ins and the same AssetKey I get

Input asset '["qld_health", "qld_health_simple_cohort"]' for asset '["cohort"]' is not produced by any of the provided asset ops and is not one of the provided sources

any ideas what I'm missing here?

seems related to the question I had the other day. Perhaps it’ll help - https://dagster.slack.com/archives/C01U954MEER/p1659136358342089

Thanks! I don't think these are related though, my issue occurs when trying to load the repository, before I get a chance to use materialise.

hi @Oliver! Do you have a SourceAsset defined for qld_health_simple_cohort? This is required for the AssetIn case because Dagster needs to know how to load that input value (specifically, it needs an IOManager key to use). For non_argument_deps, no input value is loaded, so it's ok to not have a SourceAsset defined, which is why the behavior differs here.

ah ok, yup loads when I add that. thanks I did set the

input_manager_key

in the AssetIn definition, whats the difference between setting the io_manager in these two locations?

honestly that's a good point — it might be worth relaxing that rule when an input manager is supplied