Now we will sketch the algorithm of the query rewrite system. For better illustration we show how to implement views using rules as an example.
Let the following rule be given:
create rule view_rule
as on select
to test_view
do instead
select s.sname, p.pname
from supplier s, sells se, part p
where s.sno = se.sno and
p.pno = se.pno;
The given rule will be fired whenever a select against the relation test_view is detected. Instead of selecting the tuples from test_view the select statement given in the action part of the rule is executed.
Let the following user-query against test_view be given:
select sname
from test_view
where sname <> 'Smith';
Here is a list of the steps performed by the query rewrite system whenever a user-query against test_view appears. (The following listing is a very informal description of the algorithm just intended for basic understanding. For a detailed description refer to A commentary on the POSTGRES rules system).
test_view Rewrite
Take the query given in the action part of the rule.
Adapt the targetlist to meet the number and order of attributes given in the user-query.
Add the qualification given in the where clause of the user-query to the qualification of the query given in the action part of the rule.
Given the rule definition above, the user-query will be rewritten to the following form (Note that the rewriting is done on the internal representation of the user-query handed back by the parser stage but the derived new data structure will represent the following query):
select s.sname
from supplier s, sells se, part p
where s.sno = se.sno and
p.pno = se.pno and
s.sname <> 'Smith';