Proposal
Hive will consider a view updatable if:
- The view refers to exactly one base table or updatable view in the FROM clause without a WHERE clause.
- Each column in the view is a column in the underlying table/updatable view with no underlying columns duplicated.
- Views must have the same partition columns as the underlying table/updatable view.
When inserting into a view:
- If a view does not specify all underlying columns, NULL will be inserted for each column not specified.
Given
create table t1 (id int, key string, value string) partitioned by (ds string, hr string); create view v partitioned on (ds, hr) as select id, value, ds, hr from t1;
we can insert into v where NULL will be used for key.
Notes:
- Should we try to support views like (create view v partitioned on (ds, hr) as select id, value, ds where ds='2011-01-01' and hr='12') for updating, where we infer values for ds and hr?
- With non-dynamic partitioning, do we require the partition be on each view in the updatable view chain? This seems burdensome if you don't have write access to all the views?
- When we specify dynamic partitions for a view, do we create partitions on each view in a chain of updatable views? If we don't, there may be strange behavior where SHOW PARTITIONS may not show anything on a view, but we can insert into such partitions of a view. If we do, drop partition on the view actually does nothing to the data.
See Hive Views for general information about views.