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

dagster

I have a partitioned upstream asset, and I'm trying to decide whether its downstream asset should also be partitioned. The downstream asset is the resulting table from a dbt merge, so it's not really partitioned, but it does need to use the partition key from the upstream asset in order to run the merge. Are there any advantages to partitioning the downstream asset in this case (or vice-versa)?

more direct question: if a non-partitioned asset is downstream from a partitioned asset, does `context.asset_partition_keys` return every possible all-time partition or a subset?

right now I'm seeing the former, but is that because I haven't done any materializations of the downstream asset yet?