2020-09-10 MYSQL按时间段分组查询当天，每小时，15分钟数据分组

查询的需求比较麻烦，根据spreader_id和返回的状态码来判断事件的回传成功率，但是还需要新增时间段的查询，比如产品要的是每15分钟内数据的回传成功率。

这样的话就需要对时间段进行分组查询，我保存的数据是时间戳的格式。

网上找资料看了下，抄了一些方法，总结以下sql

• 按照小时分组进行查询

select FROM_UNIXTIME(timestamp, '%Y-%m-%d %H') as dt_h,	-- 这里根据分组的条件进行时间戳的格式化
spreader_id, status, count(*) as num
FROM s2s_log.s2s_callback_log
WHERE event_type_id = '8' 
AND timestamp BETWEEN '1599321600' AND '1599494400' 
group by dt_h, spreader_id, status
order by dt_h


-- 也可执行
SELECT COUNT(*),
       DATE_FORMAT(TIME,'%Y-%m-%d %H:00:00') AS TIME   
FROM tableName
GROUP BY TIME
/*
如果按小时分组 则 格式调整为 “%Y-%m-%d %H:00:00”
按分钟分组 则 格式调整为 “%Y-%m-%d %H:%i:00”
*/

但是这个满足不了产品的每15分钟去查询数据分组的数据
拆解搜索网址之后，修改如下，后续修改

• 按照15分钟分组查询

SELECT TIME,
       spreader_id,
       status,
       COUNT(*) AS num
FROM
  ( SELECT FROM_UNIXTIME(TIMESTAMP) AS data_dt,		-- 需要将时间戳进行格式的转换
           DATE_FORMAT( concat(date(FROM_UNIXTIME(TIMESTAMP)), ' ', HOUR (FROM_UNIXTIME(TIMESTAMP)), ':', floor(MINUTE (FROM_UNIXTIME(TIMESTAMP)) / 15) * 15), '%Y-%m-%d %H:%i') AS TIME,
           spreader_id,
           status
   FROM s2s_log.s2s_callback_log
   WHERE event_type_id = '8'
     AND TIMESTAMP BETWEEN '1599321600' AND '1599580800' ) a   -- 时间戳限制3日内的数据
GROUP BY DATE_FORMAT(TIME, '%Y-%m-%d %H:%i'),
         spreader_id,
         status
ORDER BY TIME

以上的sql根据日期的分钟来进行筛选（15分钟），后续分组的维度为spreader_id, status得出的数据便是不同的渠道在15分钟的时间点内数据的状态。

• mysql相关函数

DATE_FORMAT() 
'''
	DATE_FORMAT() 函数用于以不同的格式显示日期/时间数据。
	语法：  DATE_FORMAT(date,format)
	参数： https://www.w3school.com.cn/sql/func_date_format.asp

• 总结
  mysql不同时间粒度下的分组统计

    按天统计
    按小时统计
    按半小时统计
    按N分钟统计
    按分钟统计
  

按天统计

SELECT DATE(dt) AS date, COUNT(*) AS num
FROM my_table
WHERE Flag = 0 AND Duration >= 300 
GROUP BY date
ORDER BY date;

按小时统计

SELECT DATE_FORMAT(dt, '%Y-%m-%d %H:00:00') AS time, COUNT(*) AS num
-- SELECT DATE_FORMAT(dt, '%Y-%m-%d %H') AS time, COUNT(*) AS num
FROM track
WHERE Flag = 0 AND Duration >= 300
GROUP BY time
ORDER BY time;

按每分钟统计

SELECT DATE_FORMAT(dt, '%Y-%m-%d %H:%i:00') AS time, COUNT(*) AS num
-- SELECT DATE_FORMAT(dt, '%Y-%m-%d %H:%i') AS time, COUNT(*) AS num
FROM track
WHERE Flag = 0 AND Duration >= 300
GROUP BY time
ORDER BY time;

按30分钟统计

SELECT time, COUNT( * ) AS num 
FROM
	(
	SELECT Duration,
		DATE_FORMAT(
			concat( date( TimeStart ), ' ', HOUR ( TimeStart ), ':', floor( MINUTE ( TimeStart ) / 30 ) * 30 ),
			'%Y-%m-%d %H:%i' 
		) AS time 
	FROM tarck
	WHERE Flag = 0  AND Duration >= 300 
	) a 
GROUP BY DATE_FORMAT( time, '%Y-%m-%d %H:%i' ) 
ORDER BY time;

按N分钟统计

将上面的SQL语句稍微修改下，就可以实现按任意N分钟为时间片的分组统计，如按10分钟统计，先上代码：

SELECT time, COUNT( * ) AS num 
FROM
	(
	SELECT Duration,
		DATE_FORMAT(
			concat( date( TimeStart ), ' ', HOUR ( TimeStart ), ':', floor( MINUTE ( TimeStart ) / 10 ) * 10 ),
			'%Y-%m-%d %H:%i' 
		) AS time 
	FROM tarck
	WHERE Flag = 0  AND Duration >= 300 
	) a 
GROUP BY DATE_FORMAT( time, '%Y-%m-%d %H:%i' ) 
ORDER BY time;

基本思路：
将datetime类型的时间转化为相应时间片的时间，例如将‘2017-03-01 01:08:19’ 转化为‘2017-03-01 01:00:00’，然后group by即可。

• 参考博客

mysql按天，小时，半小时，N分钟，分钟进行数据分组统计
https://blog.csdn.net/qq_39268288/article/details/103735748?utm_medium=distribute.pc_relevant.none-task-blog-title-4&spm=1001.2101.3001.4242

sql中时间以5分钟半个小时任意间隔分组的实现方法
http://www.45fan.com/article.php?aid=19062448408855841712601613

https://www.cnblogs.com/fashflying/p/6023036.html