Обсуждение: Aggregates and Views

Is it possible to use an aggregate on a view
that contains an aggregate? I can't seem to get
this to work:

(cut & paste into psql to reproduce)

CREATE TABLE x ( n int );
INSERT INTO x VALUES(5);
INSERT INTO x VALUES(5);
INSERT INTO x VALUES(5);
INSERT INTO x VALUES(10);
INSERT INTO x VALUES(10);
INSERT INTO x VALUES(10);
INSERT INTO x VALUES(15);
INSERT INTO x VALUES(15);
INSERT INTO x VALUES(15);
INSERT INTO x VALUES(15);
INSERT INTO x VALUES(20);
INSERT INTO x VALUES(20);

SELECT count(*) from x; -- Correctly returns 12

CREATE VIEW y AS SELECT n, count(*) AS total FROM x GROUP BY n;
SELECT * FROM y;        -- Correctly returns (5,3), (10,3), (15,4),
(20,2)

-- So far so good
-- but if I do this:

SELECT count(*) FROM y;  -- Instead of getting 4, I get 3,3,4,2
SELECT sum(total) FROM y; -- Returns nonsense values 0, 9, 12, 8

DROP TABLE x;
DROP VIEW y;

(end cut & paste)

Thanks,
Mark

Mark Volpe <volpe.mark@epamail.epa.gov> writes:
> Is it possible to use an aggregate on a view
> that contains an aggregate?

Not at present.  Views are implemented by a rewriter that tries
to transform your query into another SQL query, and since aggregates
of aggregates are not possible, it doesn't work.  In current sources
I actually get an error from your second example:

regression=# SELECT sum(total) FROM y;
ERROR:  Aggregate function calls may not be nested

For the same sort of reason, GROUPed views don't play nice with an
outer query that specifies different grouping (ie, has a GROUP clause
of its own, or perhaps an aggregate).

What we need in order to fix this is subselects in FROM clauses;
if the rewriter could transform your query into

SELECT sum(total) FROM (SELECT n, count(*) AS total FROM x GROUP BY n)

then everything would just work (and the rewriter would get a lot
simpler, too ;-)).  We don't have subselects in FROM yet, but I hope
to see them in 7.1, or 7.2 at the latest.
        regards, tom lane