MySQL - Correct approach event counting

问题

I want to list users which have a particular event count but I'm confused on which approach to take.

This is the database table:

CREATE TABLE `event` (
  `event_id` int(11) unsigned NOT NULL AUTO_INCREMENT,
  `visitor_id` int(11) DEFAULT NULL,
  `key` varchar(200) DEFAULT NULL,
  `value` text,
  `label` varchar(200) DEFAULT '',
  `datetime` datetime DEFAULT NULL,
  PRIMARY KEY (`event_id`)
) ENGINE=InnoDB AUTO_INCREMENT=6 DEFAULT CHARSET=utf8;


INSERT INTO `event` (`event_id`, `visitor_id`, `key`, `value`, `label`, `datetime`)
VALUES
    (1, 1, 'LOGIN', NULL, '', NULL),
    (2, 2, 'LOGIN', NULL, '', NULL),
    (3, 1, 'VIEW_PAGE', 'HOTEL', '', NULL),
    (4, 2, 'VIEW_PAGE', 'HOTEL', '', NULL),
    (5, 1, 'PURCHASE_HOTEL', NULL, '', NULL);

CREATE TABLE `visitor` (
  `visitor_id` int(11) unsigned NOT NULL AUTO_INCREMENT,
  `datetime` datetime DEFAULT NULL,
  PRIMARY KEY (`visitor_id`)
) ENGINE=InnoDB AUTO_INCREMENT=6 DEFAULT CHARSET=utf8;

INSERT INTO `visitor` (`visitor_id`, `datetime`)
VALUES
    (1, NULL),
    (2, NULL);

and this is my approach:

SELECT DISTINCT
    t1.`visitor_id`
FROM
    `visitor` t1

JOIN `event` t2 on t1.visitor_id = t2.visitor_id AND t2.`key` = 'LOGIN'
JOIN `event` t3 on t1.visitor_id = t3.visitor_id AND t3.`key` = 'VIEW_PAGE' AND t3.`value` = 'HOTEL'
WHERE ( SELECT COUNT(*) FROM `event` WHERE `event`.`key` = 'PURCHASE_HOTEL' ) > 0

this should only list visitor 1 but it does actually list visitor 2 too which does not have the PURCHASE_HOTEL event.

As you can imagine, there will be more "rules" like all the JOIN events for each particular case. Can we correct and improve this somehow?

BONUS: What is the name of this approach?

回答1:

I think this is a "set-within-sets" query. I like using aggregation with a having clause for this type of query. The following checks the three conditions you are looking for:

select visitor_id
from event e
group by visitor_id
having sum(e.key = 'LOGIN') > 0 and
       sum(e.key = 'VIEW_PAGE' and e.value = 'HOTEL') > 0 and
       sum(e.key = 'PURCHASE_HOTEL') > 0;

The first condition in the having clause counts the number of LOGIN records and is true when at least one is found. (If you want exactly one, change > 0 to = 0.)

The second condition checks the viewing of the hotel page.

The third counts the number of hotel purchases.

来源：https://stackoverflow.com/questions/18487327/mysql-correct-approach-event-counting

标签

mysql

join

subquery

self-join