Closed bendnorman closed 1 year ago
I ended up using creating a separate dagster Definition for the ferc_to_sqlite logic with an op for the dbf2sqlite and xbrl2sqlite processes. We don't know the asset names before run time for the dbf2sqlite and xbrl2sqlite processes so we can't use assets. We could maybe collect the dbf table names prior to extracting the data but I'm not sure if this is possible for the xbrl data.
This actually creates a nice delineation between sections of our full data process. The ETL and ferc extraction happens in separate Definitions. The ETL accesses the extracted ferc tables using SourceAssets. The jobs in separate definitions can then be linked together by sensors.
Convert ferc_to_sqlite and the ferc1 ETL to use dagster concepts.