sql 分组查询前N名

分组查询前十名

表结构


CREATE TABLE `tb_api_app_income_rank` (
  `id` int(11) NOT NULL AUTO_INCREMENT,
  `product_id` varchar(64) DEFAULT '',
  `product_name` varchar(255) DEFAULT NULL,
  `publisher_id` varchar(64) DEFAULT NULL,
  `publisher_name` varchar(512) DEFAULT NULL,
  `company_id` varchar(64) DEFAULT NULL,
  `company_name` varchar(512) DEFAULT NULL,
  `category` varchar(64) DEFAULT NULL COMMENT '分类',
  `rank_type` varchar(32) DEFAULT NULL COMMENT '排行榜类型，0：免费，1：付费，2：畅销',
  `country` varchar(32) DEFAULT NULL,
  `rank` int(11) DEFAULT NULL COMMENT '在收入下载量排行榜的排名',
  `change_rank` int(11) DEFAULT NULL,
  `change_percent` decimal(11,6) DEFAULT NULL,
  `change_value` int(11) DEFAULT NULL,
  `price` decimal(11,2) DEFAULT NULL,
  `estimate` int(11) DEFAULT NULL,
  `downloads` int(11) DEFAULT NULL COMMENT '下载量',
  `income` int(11) DEFAULT NULL COMMENT '收入',
  `release_date` varchar(32) DEFAULT NULL,
  `is_new` int(11) DEFAULT '0' COMMENT '是否新品，0：不是，1：是。新品：距离当前时间往前推7天（昨日为第7日），release date在这7天之内，且该产品的appid是首次出现，才算做新品',
  `created_at` datetime NOT NULL DEFAULT '0000-00-00 00:00:00',
  `updated_at` datetime DEFAULT NULL,
  PRIMARY KEY (`id`,`created_at`),
  KEY `idx_apir_update_time` (`updated_at`) USING BTREE,
  KEY `idx_apir_ctime_country` (`country`,`created_at`) USING BTREE,
  KEY `idx_apir_aid` (`product_id`)
) ENGINE=InnoDB AUTO_INCREMENT=52188 DEFAULT CHARSET=utf8 ROW_FORMAT=COMPACT COMMENT='app 收入下载量的排行榜，分免费、付费、畅销排行榜'

第一种方案

SELECT
	t1.*
FROM
	tb_api_app_income_rank t1
WHERE
	created_at = '2017-11-07 00:00:00'
AND rank_type = 'grossing'
AND change_rank = 1000
AND (
	SELECT
		count(1)
	FROM
		tb_api_app_income_rank
	WHERE
		rank_type = 'grossing'
	AND change_rank = 1000
	AND created_at = '2017-11-07 00:00:00'
	AND country = t1.country
	AND category = t1.category
	AND rank < t1.rank
) < 3;

消耗时间: 7.260s

第二种方案


SELECT
	t1.*
FROM
	tb_api_app_income_rank t1
left join  tb_api_app_income_rank t2 ON t2.created_at = '2017-11-07 00:00:00'
AND t2.rank_type = 'grossing'
AND t2.change_rank = 1000
AND t1.country = t2.country
AND t1.category = t2.category
AND t1.rank > t2.rank
where t1.created_at = '2017-11-07 00:00:00'
AND t1.rank_type = 'grossing'
AND t1.change_rank = 1000
group by t1.country, t1.category, t1.product_id
having count(t1.product_id)<3
order by t1.country, t1.category,rank;

消耗时间: 8.076s

第三种

select * from 
(SELECT
	(
		@i := CASE
		WHEN @pre_country = country && @pre_category = category THEN
			@i + 1
		ELSE
			1
		END
	) rownum,
	t1.*, (@pre_country := country), (@pre_category := category)
FROM
	tb_api_app_income_rank t1,
	(
		SELECT
			@i := 0,
			@pre_country := '',
			@pre_category := ''
	) AS tmp
WHERE
	created_at = '2017-11-07 00:00:00'
AND rank_type = 'grossing'
AND change_rank = 1000
ORDER BY
	country,
	category,
	rank) t2 where t2.rownum <=3

消耗时间: 0.113s

由上得出结论，使用游标而不是连表或嵌套子查询，速度快翻几千倍

sql 分组查询前N名

分组查询前十名

参考资料