Hello Dagster Community, I need some help with one of my ops that I have written. I am getting the 'dagster._core.errors.DagsterStepOutputNotFoundError: Core compute for op "fetch_mySQL_raw_data" did not return an output for non-optional output "result" ' error, I am trying to return a list of pandas Dataframes. I am reading files from my AWS S3 bucket into a pandas dataframe and appending them to a list which I am returning. Maybe I have some misunderstanding, but I have written a test function where I read in a single file from my S3 bucket and return that as a DF and that works perfectly fine. I am providing the code below:

@op(out=Out(dagster_type=list[pd.DataFrame]))
def fetch_mySQL_raw_data() -> list[pd.DataFrame]:
    logging.basicConfig(level=logging.DEBUG)
    boto3.set_stream_logger('botocore', level='DEBUG')

    s3_client = boto3.client('s3')
    bucket_name = 'cobbleweb-ds'

    result = []
    filenames = {}
    prefix = 'raw/nestify/mySQL/'

    # Fetch data from the S3 bucket and return the result
    response = s3_client.list_objects(Bucket=bucket_name, Prefix=prefix)

    for obj in response.get('Contents', []):
        filename = obj['Key']
        # Skipping the prefix value
        if filename == prefix:
             continue

        file = s3_client.get_object(Bucket=bucket_name, Key=filename)
        df = pd.read_csv(file['Body'])
        result.append(df)

        yield AssetMaterialization(
            asset_key=filename,
            description="Raw Nestify mySQL data from S3 bucket"
        )

    return result

# FUNCTION THAT WORKS
@op(out=Out(dagster_type=pd.DataFrame))
def load_test_dataframe() -> pd.DataFrame:
    
    s3_client = boto3.client('s3')
    bucket_name = 'cobbleweb-ds'

    # Fetch data from the S3 bucket and return the result
    response = s3_client.get_object(Bucket=bucket_name, Key='test/mySql_test.csv')
 
    df = pd.read_csv(response['Body'])
    print(df)
    return df


@job()
def fetch_data_pipeline():
    fetch_mySQL_raw_data()
    # load_test_dataframe()

I bet it’s the combination of return and yield being used here. You should be using the

context.log_event(…)

syntax to log that asset materialization instead of yielding it directly. I think the iterator coercion that we do might be getting messy in this case.

Ye I am finding the same. I think it well could be the yield that is screwing it up

yea we perform some shenanigans under the hood to coerce both regular functions and iterators to work… but when you start to mix the return syntax and yield syntax, I think that the coercion behavior gets a bit more unpredictable