The cloud-native orchestrator for the whole development lifecycle, with integrated lineage and observability.

dagster

I'm trying to run basically this example - <https://docs.dagster.io/concepts/ops-jobs-graphs/graphs#loading-an-asset-as-an-input>, loading an asset as an input to a graph. An `asset` definition returns an Output with a pandas Dataframe for its value. Then I try to pass it to an `op` that takes a dataframe and does some work to send emails based on the contents of the dataframe. But when I write a `graph` like the one in this example , I get a type error `Argument of type "SourceAsset" cannot be assigned to parameter "df" of type "DataFrame`.  And at runtime, i get an error from the `op`  that `'SourceAsset' object has no attribute 'to_string'`.

So the `op` is getting a SourceAsset, not the value of the asset. All the `op` examples just show the asset value getting passed directly as a parameter. So ... Do I need to do something in the `graph` to get the Dataframe from the `SourceAsset`  returned from my `asset`? Or does the `op`  need to take a `SourceAsset` instead of a `DataFrame` and do something to get the DataFrame out of the SourceAsset?

Like in this snippet from the example, what's the type of `emails`? is it the asset's value, like a dataframe or list, or is it a SourceAsset that the `op` needs to ... materialize or something?
```@op
def send_emails(emails) -&gt; None:
    ...```

looks like this was dealt with recently - <https://github.com/dagster-io/dagster/commit/5d1d12679c0471ac549165cb4a4f9379a9204f1e>

(<https://github.com/dagster-io/dagster/issues/10874>)

alright - i got it. One mistake was a silly one, I'd accidentally deleted or forgotten the `@op`  decorator on the email-sending op. but it also turns out the coercion of a SourceAsset to the asset value means I can't pass other non-source assets to the op function (like strings or whatever). Not sure  what that means me. Maybe i need to some sort of OpBuilder currying thing to make an op dynamically