How can I eliminate duplicate subquery/expression in DATEDIFF? - mysql

Question

My database is a sample of login dates. I want to calculate a datediff between consecutive login dates on different rows. For example:
user_id login_date
1 2012-05-22
1 2012-05-25
/* difference is 3 days */
I was able to figure out two queries to do this calculation, but in both queries I needed to duplicate a subquery/expression to get my desired results.
I tried to use 'nextdate' in the datediff, but get an error:
#1054 - Unknown column 'nextdate' in 'field list'
Is there a way to eliminate the duplication? A completely new query is acceptable if it produces the desired results.
Sample database
CREATE TABLE IF NOT EXISTS `tbl` (
`user_id` int(11) DEFAULT NULL,
`login_date` date DEFAULT NULL
) ENGINE=MyISAM DEFAULT CHARSET=utf8 COLLATE=utf8_unicode_ci;
INSERT INTO `tbl` (`user_id`, `login_date`) VALUES
(1, '2012-04-01'),
(1, '2012-04-25'),
(1, '2012-05-03'),
(1, '2012-05-04'),
(1, '2012-05-05'),
(1, '2012-05-06'),
(1, '2012-05-07'),
(1, '2012-05-09'),
(1, '2012-05-10'),
(1, '2012-05-11'),
(1, '2012-05-12'),
(1, '2012-05-16'),
(1, '2012-05-19'),
(1, '2012-05-20'),
(1, '2012-05-21'),
(1, '2012-05-22'),
(1, '2012-05-25'),
(2, '2012-04-02'),
(2, '2012-04-03'),
(2, '2012-04-04'),
(2, '2012-05-04'),
(2, '2012-05-06'),
(2, '2012-05-08'),
(2, '2012-05-09'),
(2, '2012-05-11'),
(2, '2012-05-17'),
(2, '2012-05-18'),
(2, '2012-05-19'),
(2, '2012-05-20'),
(2, '2012-05-21'),
(2, '2012-05-22'),
(2, '2012-05-25'),
(2, '2012-05-26'),
(2, '2012-05-27'),
(2, '2012-05-28'),
(2, '2012-05-29'),
(2, '2012-05-30'),
(2, '2012-05-31'),
(2, '2012-06-01'),
(2, '2012-06-02');
working query #1
SELECT
a.`user_id`,
a.`login_date`,
(SELECT
MIN(b.`login_date`)
FROM `tbl` b
WHERE a.`login_date` < b.`login_date`
AND a.`user_id` = b.`user_id`
) AS `nextdate`,
DATEDIFF((SELECT
MIN(b.`login_date`)
FROM `tbl` b
WHERE a.`login_date` < b.`login_date`
AND a.`user_id` = b.`user_id`
), a.`login_date`) AS `timespan`
FROM `tbl` a
WHERE 1=1
AND (a.`login_date` >= '2012-05-10' AND a.`login_date` <= '2012-05-25')
HAVING `nextdate` IS NOT NULL
ORDER BY a.`user_id` ASC, a.`login_date` ASC
working query #2
SELECT
a.`user_id`,
a.`login_date`,
MIN(b.`login_date`) AS `nextdate`,
DATEDIFF(MIN(b.`login_date`), a.`login_date`) AS `timespan`
FROM
(
SELECT
`user_id`,
`login_date`
FROM `tbl`
) a
JOIN
(
SELECT
`user_id`,
`login_date`
FROM `tbl`
) b
ON a.`user_id` = b.`user_id`
AND a.`login_date` < b.`login_date`
WHERE 1=1
AND (a.`login_date` >= '2012-05-10' AND a.`login_date` <= '2012-05-25')
GROUP BY a.`user_id`,a.`login_date`
desired results
user_id login_date nextdate timespan
1 2012-05-10 2012-05-11 1
1 2012-05-11 2012-05-12 1
1 2012-05-12 2012-05-16 4
1 2012-05-16 2012-05-19 3
1 2012-05-19 2012-05-20 1
1 2012-05-20 2012-05-21 1
1 2012-05-21 2012-05-22 1
1 2012-05-22 2012-05-25 3
2 2012-05-11 2012-05-17 6
2 2012-05-17 2012-05-18 1
2 2012-05-18 2012-05-19 1
2 2012-05-19 2012-05-20 1
2 2012-05-20 2012-05-21 1
2 2012-05-21 2012-05-22 1
2 2012-05-22 2012-05-25 3
2 2012-05-25 2012-05-26 1

Answer 1

This query is basically the same as your query #2, but just using a simple self join. The self join with group by and min(login_date) is about the simplest you can reduce this query down to, range scan on tbl a and then a key lookup on tbl b.
select a.user_id, a.login_date, min(b.login_date), datediff(min(b.login_date), a.login_date)
from tbl a
join tbl b on a.user_id = b.user_id and a.login_date < b.login_date
where (a.login_date >= '2012-05-10' AND a.login_date <= '2012-05-25')
group by a.user_id, a.login_date
order by a.user_id, a.login_date, b.login_date
;