When these data sources were migrated to the Plugin Framework, three collection fields that were sets under SDKv2 (slice_set) silently became lists, because the generated PF models default every slice to a list. Since the underlying APIs don't guarantee ordering, a list can produce spurious diffs.
Affected:
databricks_shares — sharesdatabricks_volumes — idsdatabricks_share — object
These should go back to sets. It's a schema change on released data sources, so it's 2.0 material rather than a patch.
For reference, the new databricks_recipients data source (#5868) uses a set for exactly this reason.
When these data sources were migrated to the Plugin Framework, three collection fields that were sets under SDKv2 (
slice_set) silently became lists, because the generated PF models default every slice to a list. Since the underlying APIs don't guarantee ordering, a list can produce spurious diffs.Affected:
databricks_shares—sharesdatabricks_volumes—idsdatabricks_share—objectThese should go back to sets. It's a schema change on released data sources, so it's 2.0 material rather than a patch.
For reference, the new
databricks_recipientsdata source (#5868) uses a set for exactly this reason.