Currently in UDF checkpoints PR we create UDF input table (pre-UDF table) with name that has current job_id + checkpoint hash in the name. We don't c/p this table across job re-runs since it's a constant (doesn't change) and when we search for it we need to recursivly traverse through job ancestor tree.
With introducing new run_group_id in PR we can use that one instead of current job_id when constructing table name which will simplify finding process as we won't need to do recursive traversal any more.
Blocked by: #1512
This should be implemented as follow-up of UDF checkpoint PR or directly in it if blocking PR is merged before.
Currently in UDF checkpoints PR we create UDF input table (pre-UDF table) with name that has current
job_id+ checkpointhashin the name. We don't c/p this table across job re-runs since it's a constant (doesn't change) and when we search for it we need to recursivly traverse through job ancestor tree.With introducing new
run_group_idin PR we can use that one instead of currentjob_idwhen constructing table name which will simplify finding process as we won't need to do recursive traversal any more.Blocked by: #1512
This should be implemented as follow-up of UDF checkpoint PR or directly in it if blocking PR is merged before.