Page History

...

• Tdep depends on (T,T2), and (T,T2) are dependent upon by Tdep
• T and T2 have the same schema
• Tdep@ds=1..Tdep@ds=10 depends on T@ds=1..T@ds=10 respectively
• Tdep@ds=11 depends on T2@ds=11

The query on Tdep is re-written to access the underlying tables. For eg. for the query 'select count(1) from Tdep where ds = 1', once partition pruning is done (on Tdep), the operator tree (TableScan(Tdep):ds=1 -> Select) is rewritten to (TableScan(T):ds=1/hr=1 to ds=1/hr=24 -> Select). The advantage of this over external tables is that all the table properties (bucketing/sorting/list bucketing) for the underlying tables are used. This can easily extend to multiple tables. For eg. for the query 'select count(1) from Tdep where ds = 1 or ds = 11', once partition pruning is done (on Tdep), the operator tree (TableScan(Tdep):ds=1,ds=11 -> Select) is rewritten to (TableScan(T):ds=1/hr=1 to ds=1/hr=24 -> Select -> Union) and (TableScan(T2):ds=1/hr=1 to ds=1/hr=24 -> Select -> Union).

• If the schema of T changes, it can either be dis-allowed or propagated to Tdep.
• desc extended will be enhanced to show all the dependent partitions.
• The dependency of an existing partition can be changed: alter table Tdep partition (ds=10) depends on table T2