Would it be possible to get an imperative function like

defs.load_asset_value

that can be accessed inside an op/job? I'm aware of the very recent implementation that lets you load an asset into an op using

.to_source_asset()

, with the big caveat that it does not allow for partitioned assets. Secondly, it's a real bear to have to send in to the job all the

resource_defs

, the

config

that maps all of those resources, then in the

ops

containing the assets you have to include

required_resource_keys

and a noop

ins

. Hopefully I'm wrong about all that.

I'm aware of the very recent implementation that lets you load an asset into an op using

.to_source_asset()

, with the big caveat that it does not allow for partitioned assets

this is fixed in our release that's landing today

I was going to look into doing just that but had a feeling it wouldn't "just work". I'll give it a try, thank you

Try this:

@op
def my_op(context):
    from .somewhere import defs
    asset_value = defs.load_asset_value(..., instance=context.instance)

isn't this likely to cause a circular dependency? If you want to use the above op in a job, you have

op->job->def

as an import chain, but then you also have

def->op

now, creating a loop. Maybe you make it work if you only use the op in another code location, but that seems restrictive.

that's why you need to do the import inside the body of the op instead of at module-level

ah, fair, but I think non-top imports are not exactly best practice and I'd personally be uncomfortable relying on a solution like this one, but maybe I'm being irrational 🙂