Random via Subselect or Temporary Table

[Thomas T. Thai]

I would like to select a random record from a group of records so I'd end
up with one random record per group:

CREATE TABLE randtest (
catnum int,
title varchar(32)
);

INSERT INTO randtest VALUES (1, 'one.one');
INSERT INTO randtest VALUES (1, 'one.two');
INSERT INTO randtest VALUES (1, 'one.three');
INSERT INTO randtest VALUES (1, 'one.four');
INSERT INTO randtest VALUES (1, 'one.five');
INSERT INTO randtest VALUES (2, 'two.one');
INSERT INTO randtest VALUES (2, 'two.two');
INSERT INTO randtest VALUES (2, 'two.three');
INSERT INTO randtest VALUES (2, 'two.four');
INSERT INTO randtest VALUES (2, 'two.five');
INSERT INTO randtest VALUES (3, 'three.one');
INSERT INTO randtest VALUES (3, 'three.two');
INSERT INTO randtest VALUES (3, 'three.three');
INSERT INTO randtest VALUES (3, 'three.four');
INSERT INTO randtest VALUES (3, 'three.five');

I've got it working with using temporary tables:

DROP table t1; DROP table t2;

CREATE TEMPORARY TABLE t1 AS SELECT catnum, title, random() AS r FROM
randtest;
CREATE TEMPORARY TABLE t2 AS SELECT catnum, max(r) AS mr FROM t1 GROUP BY
catnum;
SELECT t1.catnum, t1.title, t1.r FROM t1, t2 WHERE t1.catnum = t2.catnum
AND t1.r = t2.mr;
catnum | title | r
--------+-----------+-------------------
2 | two.one | 0.5760686600469 37
3 | three.one | 0.6955522981917 26
1 | one.one | 0.9887701030768 31

catnum | title | r
--------+-------------+-------------------
1 | one.one | 0.6029698139070 39
3 | three.three | 0.8516511874513 94
2 | two.two | 0.9046093080679 93

catnum | title | r
--------+-------------+-------------------
2 | two.three | 0.6595226822698 13
3 | three.three | 0.6970271359649 61
1 | one.two | 0.8956308159491 19

The results seem pretty random per group. Is there a better way of
accomplishing this task without using temporary tables? Maybe with
subselects?

Thanks,
Thomas



---------------------------(end of broadcast)---------------------------
TIP 5: Have you checked our extensive FAQ?

http://www.postgresql.org/docs/faqs/FAQ.html

[Tom Lane]

"Thomas T. Thai" <to*@minnesota. com> writes:

CREATE TEMPORARY TABLE t1 AS SELECT catnum, title, random() AS r FROM
randtest;
CREATE TEMPORARY TABLE t2 AS SELECT catnum, max(r) AS mr FROM t1 GROUP BY
catnum;
SELECT t1.catnum, t1.title, t1.r FROM t1, t2 WHERE t1.catnum = t2.catnum
AND t1.r = t2.mr; The results seem pretty random per group. Is there a better way of
accomplishing this task without using temporary tables? Maybe with
subselects?

I'd suggest using DISTINCT ON (catnum) with ORDER BY catnum, random().
See the SELECT reference page's "weather reports" example for
background.

regards, tom lane

---------------------------(end of broadcast)---------------------------
TIP 9: the planner will ignore your desire to choose an index scan if your
joining column's datatypes do not match

[Thomas T. Thai]

<quote who="Tom Lane">

I'd suggest using DISTINCT ON (catnum) with ORDER BY catnum, random().
See the SELECT reference page's "weather reports" example for
background.

Good suggestion. Thank you Tom.

Thomas

---------------------------(end of broadcast)---------------------------
TIP 8: explain analyze is your friend