How to optimze this Mysql simple query

问题

This query is taking 450ms

SELECT `u`.`user_id`, `c`.`company`
FROM `users` AS `u`
LEFT JOIN `companies` AS `c` ON `c`.`user_id` = `u`.`user_id`
WHERE `u`.`user_id` = 'search_term' 
  OR `u`.`lname` LIKE 'search_term%'    
  OR `u`.`email` LIKE 'search_term%'        
  OR `c`.`company` LIKE 'search_termeo%'

tables:

users (260250 rows)
companies (570 rows)

structures:

- users:

   CREATE TABLE `users` (
    `user_id` int(10) unsigned NOT NULL AUTO_INCREMENT,
    `region_id` int(10) unsigned NOT NULL,
    `fname` varchar(30) COLLATE utf8mb4_unicode_ci DEFAULT NULL,
    `lname` varchar(30) COLLATE utf8mb4_unicode_ci DEFAULT NULL,
    `email` varchar(50) COLLATE utf8mb4_unicode_ci NOT NULL,
    `password` varchar(100) COLLATE utf8mb4_unicode_ci DEFAULT NULL,
    `phone` varchar(20) COLLATE utf8mb4_unicode_ci DEFAULT NULL,
    `active` tinyint(1) NOT NULL DEFAULT '0',
     PRIMARY KEY (`user_id`),
     KEY `idx_lname` (`lname`),
     KEY `idx_email` (`email`),
     UNIQUE KEY `unq_region_id_email` (`region_id`, `email`)
   ) ENGINE=InnoDB DEFAULT CHARSET=utf8mb4 COLLATE=utf8mb4_unicode_ci;

- companies:

 CREATE TABLE `companies` (
  `user_id` int(10) unsigned NOT NULL,
  `company` varchar(35) COLLATE utf8mb4_unicode_ci NOT NULL,
  `vat_num` varchar(20) COLLATE utf8mb4_unicode_ci DEFAULT NULL,
   PRIMARY KEY (`user_id`),
   KEY `idx_company` (`company`) USING BTREE,
   CONSTRAINT `users_companies_ibfk_1` FOREIGN KEY (`user_id`) REFERENCES `users` (`user_id`) ON DELETE CASCADE ON UPDATE CASCADE,
   ) ENGINE=InnoDB DEFAULT CHARSET=utf8mb4 COLLATE=utf8mb4_unicode_ci;

The result of explain query

I think 450ms is too much for such query and such little amount of data and I want to know if there is somthing to optimize

query run in querious v3 under iMac 2017, 3,4 GHz, 16Go

Mysql: 5.7.26 on MAMP pro v5.7

回答1:

OR conditions when not on the same field or range based (such as <, >, LIKE) really decrease MySQL's ability to take advantage of indexes; you can restructure queries by breaking them down into separate simpler ones that you can then UNION. Separating it out like this allows MySQL to take advantage of a different index of each query within the UNIONs

SELECT `u`.`user_id`, `c`.`company`
FROM `users` AS `u` LEFT JOIN `companies` AS `c` ON `c`.`user_id` = `u`.`user_id`
WHERE `u`.`user_id` = 'search_term' 
UNION DISTINCT 
SELECT `u`.`user_id`, `c`.`company`
FROM `users` AS `u` LEFT JOIN `companies` AS `c` ON `c`.`user_id` = `u`.`user_id`
WHERE `u`.`lname` LIKE 'search_term%'    
UNION DISTINCT 
SELECT `u`.`user_id`, `c`.`company`
FROM `users` AS `u` LEFT JOIN `companies` AS `c` ON `c`.`user_id` = `u`.`user_id`
WHERE `u`.`email` LIKE 'search_term%'
UNION DISTINCT 
SELECT `u`.`user_id`, `c`.`company`
FROM `users` AS `u` INNER JOIN `companies` AS `c` ON `c`.`user_id` = `u`.`user_id`
WHERE `c`.`company` LIKE 'search_termeo%'
;

Also, note that I changed the last one's JOIN to an INNER since any condition on the right-hand table of a LEFT JOIN (that isn't "without a match from that table") is basically an INNER JOIN anyway.

UNION DISTINCT is used to prevent records that satisfied multiple conditions from being repeated, however... if companies.company is not unique (i.e. company id 1 called "Blah" and company id 12 also called "Blah") then those will also be merged where they would not be in your original query; if it is a potential issue, that can be remedied by also including company_id in each SELECT.

来源：https://stackoverflow.com/questions/60674471/how-to-optimze-this-mysql-simple-query

标签

mysql

query-optimization