index scan is performed when index-only scan is possible (partial index)
Hello!
It seems to me there is a situation where index-only scan is possible
but not performed.
-----
[ACTIONS]
create table t as select a, a % 2 b, a % 2 c from generate_series(1,
1000000) a order by random();
SELECT 1000000
create index t_i1 on t (a, c) where b = 1;
CREATE INDEX
analyze t;
ANALYZE
EXPLAIN select a from t where a < 10000 and b = 1 order by a;
QUERY PLAN
---------------------------------------------------------------------
Index Scan using t_i1 on t (cost=0.42..14013.07 rows=5094 width=4)
Index Cond: (a < 10000)
(2 rows)
create index t_i2 on t (a, b) where b = 1;
CREATE INDEX
EXPLAIN select a from t where a < 10000 and b = 1 order by a;
QUERY PLAN
--------------------------------------------------------------------------
Index Only Scan using t_i2 on t (cost=0.42..14009.07 rows=5094 width=4)
Index Cond: ((a < 10000) AND (b = 1))
(2 rows)
[EXPECTED]
in both cases index-only scan is performed
[RECEIVED]
in the first case index scan is performed, though index-only scan is
possible, as no conditions rechecking is required
-----
adding (b=1) expression to index as a column does not help
Regards, Alexey
--
Sent via pgsql-bugs mailing list (pgsql-bugs@postgresql.org)
To make changes to your subscription:
http://www.postgresql.org/mailpref/pgsql-bugs
Alexey Bashtanov <bashtanov@imap.cc> writes:
create table t as select a, a % 2 b, a % 2 c from generate_series(1,
1000000) a order by random();
SELECT 1000000create index t_i1 on t (a, c) where b = 1;
CREATE INDEX
EXPLAIN select a from t where a < 10000 and b = 1 order by a;
[ doesn't do index-only scan ]
Yeah, see the comment in check_index_only():
/*
* Check that all needed attributes of the relation are available from the
* index.
*
* XXX this is overly conservative for partial indexes, since we will
* consider attributes involved in the index predicate as required even
* though the predicate won't need to be checked at runtime. (The same is
* true for attributes used only in index quals, if we are certain that
* the index is not lossy.) However, it would be quite expensive to
* determine that accurately at this point, so for now we take the easy
* way out.
*/
This code knows that b is referenced in the query, which would ordinarily
defeat using an index-only scan with this index. There's no very good way
to tell that the only such reference is in a qual that we will later
decide doesn't need to be checked at runtime.
So, yeah, it'd be nice if that worked ... but don't hold your breath.
regards, tom lane
--
Sent via pgsql-bugs mailing list (pgsql-bugs@postgresql.org)
To make changes to your subscription:
http://www.postgresql.org/mailpref/pgsql-bugs