Question

所以我有一个news_news表（约16k行），每个news可以分配许多teams（约33k个关联）。

CREATE TABLE `news_news` (
  `id` int(11) NOT NULL AUTO_INCREMENT,
  `title` varchar(255) COLLATE utf8_unicode_ci NOT NULL DEFAULT '',
  `body` longtext COLLATE utf8_unicode_ci NOT NULL,
  `location` varchar(255) COLLATE utf8_unicode_ci DEFAULT NULL,
  `date` date NOT NULL DEFAULT '0000-00-00',
  `status` char(1) COLLATE utf8_unicode_ci NOT NULL DEFAULT '',
  `slug` varchar(100) COLLATE utf8_unicode_ci NOT NULL DEFAULT '',
  `announcement` tinyint(1) NOT NULL DEFAULT '0',
  `author_id` int(11) NOT NULL DEFAULT '0',
  `priority` int(10) unsigned NOT NULL,
  `embed` varchar(255) COLLATE utf8_unicode_ci DEFAULT NULL,
  `source_url` varchar(255) COLLATE utf8_unicode_ci NOT NULL,
  `source_name` varchar(255) COLLATE utf8_unicode_ci NOT NULL,
  `site_id` int(11) NOT NULL,
  PRIMARY KEY (`id`),
  KEY `news_news_slug` (`slug`),
  KEY `news_news_author_id` (`author_id`),
  KEY `news_news_site_id` (`site_id`)
) ENGINE=InnoDB AUTO_INCREMENT=16521 DEFAULT CHARSET=utf8 COLLATE=utf8_unicode_ci


CREATE TABLE `news_news_teams` (
  `id` int(11) NOT NULL AUTO_INCREMENT,
  `news_id` int(10) unsigned NOT NULL,
  `team_id` int(10) unsigned NOT NULL,
  PRIMARY KEY (`id`),
  UNIQUE KEY `news_id` (`news_id`,`team_id`),
  KEY `news_news_teams_news_id` (`news_id`),
  KEY `news_news_teams_team_id` (`team_id`)
) ENGINE=InnoDB AUTO_INCREMENT=40393 DEFAULT CHARSET=latin1;

我想获得最近5条新闻，这些新闻是从一些团队中标记出来的;

SELECT 
    `news_news`.`id`, `news_news`.`title`, `news_news`.`body`, `news_news`.`embed`,
    `news_news`.`location`, `news_news`.`date`, `news_news`.`priority`, `news_news`.`status`, 
    `news_news`.`slug`, `news_news`.`announcement`, `news_news`.`author_id`, `news_news`.`source_name`, 
    `news_news`.`source_url`, `news_news`.`site_id` 
FROM `news_news` 
INNER JOIN `news_news_teams` ON ( `news_news`.`id` = `news_news_teams`.`news_id` ) 
WHERE (
    `news_news`.`status` = 'P' 
    AND `news_news_teams`.`team_id` IN (17, 18, 21, 27, 28, 31, 32, 34, 44, 51, 53, 
        56, 60, 65, 73, 75, 77, 80, 82, 83, 87, 89, 90, 92, 95, 98, 102, 105, 106,
        110, 112, 116, 117, 120, 124, 127, 128, 130, 134, 138, 141, 146, 147, 152,
        154, 156, 158, 161, 165, 166, 169, 170, 174, 176, 179, 181, 184, 185, 188,
        194, 196, 203, 214, 220, 221, 228, 229, 230, 234, 235, 240, 245, 246, 249,
        250, 251, 252, 257, 258, 260, 264, 266, 272, 273, 275, 276, 279, 280, 281,
        283, 284, 285, 287, 294, 296, 297, 318, 319, 320, 326, 327, 330, 332, 334,
        335, 336, 337, 350, 351, 368, 369, 373, 374, 375, 376, 377, 378, 383, 390,
        393, 394, 395, 397, 398, 399, 400, 401, 405, 408, 410, 413, 416, 417, 418,
        421, 425, 426, 427, 431, 434, 439, 440, 441, 443, 448, 451, 452, 453, 454,
        457, 462, 477, 479, 482, 483, 484, 489, 491, 548, 566, 567, 10001792, 10007454))
ORDER BY `news_news`.`date` DESC 
LIMIT 5;

以下是该查询的解释。

+----+-------------+-----------------+--------+---------------------------------------------------------+-------------------------+---------+----------------------------+------+--------------------------------------------------------+
| id | select_type | table           | type   | possible_keys                                           | key                     | key_len | ref                        | rows | Extra                                                  |
+----+-------------+-----------------+--------+---------------------------------------------------------+-------------------------+---------+----------------------------+------+--------------------------------------------------------+
|  1 | SIMPLE      | news_news_teams | range  | news_id,news_news_teams_news_id,news_news_teams_team_id | news_news_teams_team_id | 4       | NULL                       | 4075 | Using index condition; Using temporary; Using filesort |
|  1 | SIMPLE      | news_news       | eq_ref | PRIMARY                                                 | PRIMARY                 | 4       | cs.news_news_teams.news_id |    1 | Using where                                            |
+----+-------------+-----------------+--------+---------------------------------------------------------+-------------------------+---------+----------------------------+------+--------------------------------------------------------+

执行需要2秒多时间，有没有办法对其进行优化？

Answer 1

在@ news_news.date上添加索引作为@Octopi建议是一个好主意（因为我们按该字段排序），但在这种情况下仅仅说服mysql使用{{1}是不够的索引。对我有用的解决方案是在date中再添加一个过滤器，告诉WHERE仅在上个月添加的mysql中进行搜索。这使查询速度提高了5-7倍，现在执行约250毫秒。

news

PS。添加了SELECT DISTINCT `news_news`.`id`, `news_news`.`title`, `news_news`.`body`, `news_news`.`embed`, `news_news`.`location`, `news_news`.`date`, `news_news`.`priority`, `news_news`.`status`, `news_news`.`slug`, `news_news`.`announcement`, `news_news`.`author_id`, `news_news`.`source_name`, `news_news`.`source_url`, `news_news`.`site_id` FROM `news_news` INNER JOIN `news_news_teams` ON ( `news_news`.`id` = `news_news_teams`.`news_id` ) WHERE ( `news_news`.`status` = 'P' AND `news_news`.`date` >= DATE_SUB(CURRENT_DATE, INTERVAL 1 MONTH) AND `news_news_teams`.`team_id` IN (17, 18, 21, 27, 28, 31, 32, 34, 44, 51, 53, 56, 60, 65, 73, 75, 77, 80, 82, 83, 87, 89, 90, 92, 95, 98, 102, 105, 106, 110, 112, 116, 117, 120, 124, 127, 128, 130, 134, 138, 141, 146, 147, 152, 154, 156, 158, 161, 165, 166, 169, 170, 174, 176, 179, 181, 184, 185, 188, 194, 196, 203, 214, 220, 221, 228, 229, 230, 234, 235, 240, 245, 246, 249, 250, 251, 252, 257, 258, 260, 264, 266, 272, 273, 275, 276, 279, 280, 281, 283, 284, 285, 287, 294, 296, 297, 318, 319, 320, 326, 327, 330, 332, 334, 335, 336, 337, 350, 351, 368, 369, 373, 374, 375, 376, 377, 378, 383, 390, 393, 394, 395, 397, 398, 399, 400, 401, 405, 408, 410, 413, 416, 417, 418, 421, 425, 426, 427, 431, 434, 439, 440, 441, 443, 448, 451, 452, 453, 454, 457, 462, 477, 479, 482, 483, 484, 489, 491, 548, 566, 567, 10001792, 10007454) ) ORDER BY `news_news`.`date` DESC LIMIT 5; //and the explain +----+-------------+-----------------+-------+---------------------------------------------------------+----------------+---------+-----------------+------+-----------------------------------------------------+ | id | select_type | table | type | possible_keys | key | key_len | ref | rows | Extra | +----+-------------+-----------------+-------+---------------------------------------------------------+----------------+---------+-----------------+------+-----------------------------------------------------+ | 1 | SIMPLE | news_news | range | PRIMARY,news_news_date | news_news_date | 3 | NULL | 190 | Using index condition; Using where; Using temporary | | 1 | SIMPLE | news_news_teams | ref | news_id,news_news_teams_news_id,news_news_teams_team_id | news_id | 4 | cs.news_news.id | 1 | Using where; Using index; Distinct | +----+-------------+-----------------+-------+---------------------------------------------------------+----------------+---------+-----------------+------+-----------------------------------------------------+以排除重复的DISTINCT。

Answer 2

添加这两个;目前还不清楚哪个更好：

INDEX(status, news_id)
INDEX(status, date)

关于其他说明......

UNIQUE KEY这是多余的，可以删除：

KEY `news_news_teams_news_id` (`news_id`),

这占用3个字节，虽然我怀疑你从不使用任何东西，但ascii标志（更改为CHARACTER SET ascii）：

`status` char(1) COLLATE utf8_unicode_ci NOT NULL DEFAULT '',

使用相关表

2 个答案: