We have a team of dbt developers. We use Snowflake. Each developer can build dbt models in their own isolated Snowflake database. When we want to test out model changes in Holistics, it’s a bit of a pain. The procedure is:
- Set up a new data connection that is specific to the developer and points to their developer database.
- Write AML code, but set the
data_source_nameparameter to the developer connection.
- Test the new model and go through the code review process.
- Change the
data_source_nameparameter back to the production connection.
- Merge the changes.
In my ideal world I be able to set up one developer connection that could be parameterized during AML development (e.g., change the database name or the warehouse). Then I could also choose to develop AML against my parameterized development data connection. This would reduce the number of data connections I have to set up and maintain, and wouldn’t require devs to remember to switch back-and-forth between changing the data connection during development.