查询连续记录并对这些连续数据统计取出指定连续次数的记录,这类操做并很少,但出现时会比较棘手。mysql
查询思想是:sql
顺序行号 - 减首差值 = 连续差块 微信
顺序行号 如同 oracle 中的 rownum 但mysql目前尚未这个功能,因此只能经过局部变量来实现,oracle
减首差值 就是每条记录与最开始记录的差(须要保证这个差值与顺序行号递增值相同,固然若是原本就是自增值则不须要单独计算)ide
只要 顺序行号与减首差值保持相同递增值则 连续差块 值相同,就能够统计出连续长度函数
示例表:(以简单的签到表为例)fetch
create table user_sign( id int unsigned primary key auto_increment, user_id int unsigned not null comment '用户ID', date date not null comment '签到日期', created_time int unsigned not null comment '建立时间', updated_time int unsigned not null comment '修改时间' )engine=innodb default charset=utf8 comment '用户签到';
随机生成数据(建立函数随机生成签到数据)
spa
create function insert_sign_data(num int) returns int begin declare _num int default 0; declare _date date; declare _tmpdate date; declare _user_id int; declare line int default 0; declare _get_last cursor for select date from user_sign where user_id=_user_id order by date desc limit 1; declare continue handler for SQLSTATE '02000' set line = 1; while _num < num do set _user_id = CEIL( RAND( ) * 500 ); open _get_last; fetch _get_last into _tmpdate; IF line THEN set _date = FROM_UNIXTIME( unix_timestamp( ) - 86400 * round( RAND( ) * 200 ), '%Y-%m-%d' ); set line = 0; ELSE set _date = FROM_UNIXTIME( unix_timestamp( _tmpdate ) + 86400 * round( RAND( ) * 2 + 1), '%Y-%m-%d' ); END IF; INSERT INTO user_sign ( user_id, date, created_time, updated_time ) VALUES (_user_id, _date, unix_timestamp( ), unix_timestamp( )); set _num = _num + 1; close _get_last; end while; return _num; end
生成数据(因为生成时有判断最近打卡日期生成有会点慢)unix
select insert_sign_data(20000);
提取出连续打卡超过6天的用户orm
SELECT user_id, val - ( @rownum := @rownum + 1 ) AS type, group_concat( date ) AS date_join, count( 1 ) num FROM ( SELECT us1.date, us1.user_id, ( unix_timestamp( us1.date ) - min_timestamp ) / 86400 + 1 AS val FROM user_sign AS us1 LEFT JOIN ( SELECT UNIX_TIMESTAMP( min( date ) ) AS min_timestamp, user_id, min( date ) AS min_date FROM user_sign GROUP BY user_id ) AS us2 ON us1.user_id = us2.user_id ORDER BY us1.user_id ASC, us1.date ASC ) AS t1, ( SELECT @rownum := 0 ) AS t2 GROUP BY user_id, type HAVING num > 6
这里查询的是全表里连续超过3次打卡的,并把日期展现出来。
查询的思路是:
提取出全表用户每次打卡记录与第一次打卡记录的差值但按用户与日期正排序
增长一个局部变量rownum与上面查询数据进行连查
在结果字段集里使用日期差值减去自增顺序行号值获得连续差块
经过分组用户与连续差块获取连续签到次数
经过having来提取超过6次签到的用户