从具有零值的单个表返回多列结果答案

【问题标题】：Return multi columns result from single table with zero values从具有零值的单个表返回多列结果
【发布时间】：2016-05-28 07:39:44
【问题描述】：

我有一个像这样的表：

mysql> select RefID,State,StartTime,EndTime from execReports limit 5; +----------------------------+---------- -+----------------------+----------+ |参考号 |状态 |开始时间 |结束时间 | +----------------------------+---------- -+----------------------+----------+ | 00019a52-8480-4431-9ad2-3767c3933627 |已完成 | 2016-04-18 13:45:00 | 2016-04-18 13:45:01 | | 00038a8a-995e-4cb2-a335-cb05d5b3e92d |中止 | 2016-05-03 04:00:00 | 2016-05-03 04:00:02 | | 001013f8-0b86-456f-bd59-a7ef066e565f |已完成 | 2016-04-14 03:30:00 | 2016-04-14 03:30:11 | | 001f8d23-3022-4271-bba0-200494de678a |失败 | 2016-04-30 05:00:00 | 2016-04-30 05:00:02 | | 0027ba42-1c37-4e50-a7d6-a4e24056e080 |已完成 | 2016-04-18 03:45:00 | 2016-04-18 03:45:02 | +----------------------------+---------- -+----------------------+----------+

我可以提取每个状态的 exec 计数：

mysql> select distinct State,count(StartTime) as nbExec from execReports group by State; +-----------+--------+ |状态 | nbExec | +-----------+--------+ |中止 | 3 | |已完成 | 14148 | |失败 | 49 | +-----------+--------+ 一组 4 行（0.02 秒）

我可以使用以下方法提取每周的 exec 计数：

mysql> select distinct extract(week from StartTime) as Week, count(StartTime) as nbExec from execReports group by Week; +--------+--------+ |周 | nbExec | +--------+--------+ | 14 |第1317章 | 15 | 3051 | | 16 | 3066 | | 17 | 3059 | | 18 | 3059 | | 19 |第652章 +--------+--------+ 6 行（0.01 秒）

但我想提取一个交叉表，例如：

+--------+---------+------------+--------+---------+- --------+ |周 | nbExec |已完成 |失败 |跑步 |中止 | +--------+---------+------------+--------+---------+- --------+ | 14 |第1317章1312 | 3 | 1 | 1 | | 15 | 3051 | 3050 | 1 | 0 | 0 | | 16 | 3066 | 3060 | 3 | 2 | 1 | | 17 | 3059 | 3058 | 0 | 1 | 0 | | 18 | 3059 | 3057 | 1 | 0 | 1 | | 19 |第652章第652章0 | 0 | 0 | +--------+---------+------------+--------+---------+- --------+

我被困了几天。任何帮助表示赞赏。

最好的问候

【问题讨论】：

MySQL pivot table的可能重复
请注意 distinct 在这里没有做任何事情

标签： mysql left-join zero cross-join

【解决方案1】：

select extract(week from StartTime) as Week, count(StartTime) as nbExec, 
       sum(if(state="Completed",1,0)) Completed,
       sum(if(state="Failed",1,0)) Failed,
       sum(if(state="Aborted",1,0)) Aborted
    from execReports group by Week;

demo

【讨论】：

商品解决方案。在 MySQL 中：true= 1，false = 0。因此您可以简单地写成sum(state='Completed') as Completed。
@ThorstenKettner 我从那个开始，但可能在 smthg 上错了，然后又回到了更经典的方式:)

【解决方案2】：

您可以为此加入多个表。如果您想要动态的行到列，请检查：MySQL pivot row into dynamic number of columns

SELECT 
   a.week, 
   count(a.StartTime) as nbExec,
   count(b1.StartTime) as Completed,
   count(b2.StartTime) as Failed,
   count(b3.StartTime) as Running,
   count(b4.StartTime) as Aborted,

FROM execReports a
LEFT JOIN execReports b1 ON a.refID = b1.refID and b1.state ='Completed'
LEFT JOIN execReports b2 ON a.refID = b2.refID and b2.state ='Failed'
LEFT JOIN execReports b3 ON a.refID = b3.refID and b3.state ='Running'
LEFT JOIN execReports b4 ON a.refID = b4.refID and b4.state ='Aborted'
GROUP BY 1

【讨论】：