COUNT() 返回分组表中的总行数

Posted 2023-03-23

【中文标题】COUNT() 返回分组表中的总行数

【英文标题】：COUNT() returns the total number of rows in the grouped table

【发布时间】：2021-12-31 09:35:27

【问题描述】：

我有两张桌子：

工作表：

[![在此处输入图片描述][1]][1]

FailedReason 表被 Job 表引用：

[![在此处输入图片描述][2]][2]

我的目标是根据故障原因计算故障率。

我的期望是得到一个结果，第一列包含failure reason name，第二列包含total number of all jobs ('successful' + 'failed')，第三列包含total number of failed jobs ，第四列包含failure ratio，使用以下公式计算：失败计数（3 列）/总计数（2 列）* 100。

我的sql查询：

SELECT
  FailedReason.main_reason as "Failure reason",
  COUNT(job.name) AS "Total jobs",
  SUM(CASE WHEN job.status='failed' THEN 1 ELSE 0 END) AS "Total failed jobs",
  SUM(CASE WHEN job.status='failed' THEN 1 ELSE 0 END) / COUNT(job.name) * 100 AS "Failure ratio"
FROM job
LEFT JOIN FailedReason
ON job.id=FailedReason.job_id
GROUP BY 1
ORDER BY 3 DESC

我得到的结果是汇总表中的作业总数被计算在内。结果，失败率是百分之一百。

[![在此处输入图片描述][3]][3]

我应该修改什么以获得正确的作业数量（“成功”+“失败”）并计算正确的比率值

样本数据：

CREATE TABLE failedreason (id INT PRIMARY KEY AUTO_INCREMENT,
                  job_id INT REFERENCES job(id),
                  main_reason varchar(255)
                 );



INSERT INTO failedreason (job_id, main_reason) VALUES (13095427, 'test case failure'),
                                    (13095407, 'test case failure'),
                                    (13095533, 'connection error'),
                                    (13095546, 'connection error'),
                                    (13098367, 'runner connection error'),
                                    (13101522, 'script error');

CREATE TABLE job (id INT PRIMARY KEY,
                  created_at date,
                  finished_at date,
                  status varchar(255)
                 );
INSERT INTO job (id,
                  created_at ,
                  finished_at ,
                  status
                 )
VALUES (13095427,  '2021-05-03 02:50:41', '2021-05-03 03:47:27', 'failed'),
       (13095407,  '2021-05-03 02:50:39', '2021-05-03 03:46:41', 'failed'),
       (13095533,  '2021-05-03 02:50:41', '2021-05-03 03:47:27', 'failed'),
       (13095546,  '2021-05-03 02:50:41', '2021-05-03 03:47:27', 'failed'),
       (13098367,  '2021-05-03 02:50:41', '2021-05-03 03:47:27', 'failed'),
       (13101522,  '2021-05-03 02:50:41', '2021-05-03 03:47:27', 'failed');
       (13101444,  '2021-05-03 02:50:41', '2021-05-03 03:47:27', 'success');
       (13101445,  '2021-05-03 02:50:41', '2021-05-03 03:47:27', 'success');
       (13101446,  '2021-05-03 02:50:41', '2021-05-03 03:47:27', 'success');



  [1]: https://i.stack.imgur.com/CYnPg.png
  [2]: https://i.stack.imgur.com/t3baS.png
  [3]: https://i.stack.imgur.com/LC7Hp.png

【参考方案1】：

您的查询未给出正确的总数，因此 % 不正确。

请试试这个

select reason, alltotal, failed,
    failedtotal,
    cast( (100.00 * failed /failedtotal) as numeric(10,2)) failedRatio , 
    cast( (100.00 * failed /alltotal) as numeric(10,2)) failedPercent from 
    (SELECT
      FailedReason.[main_reason] reason,
      MAX(tots.total_jobs)  alltotal,
      MAX(ftots.total_failed)  failedtotal,
      COUNT(DISTINCT CASE WHEN job.status='failed' THEN job.id END) failed
    FROM job, FailedReason,
    (SELECT COUNT(*) AS total_failed FROM FailedReason) ftots , 
    (SELECT COUNT(*) AS total_jobs FROM job) tots
     where job.id = FailedReason.job_id 
    GROUP BY FailedReason.[main_reason]) finaltable

共享样本查询数据的结果

【讨论】：

此查询返回错误 '查询中的错误 (1064): Syntax error near 'numeric(10,2))/total) * 100 as numeric(10,2)) failedPercent from (SELECT fa '在第 2 行'

你可以删除强制转换并尝试 failed/total * 100

不进行强制转换，查询返回整数值，因此结果为 0，最好强制转换。或者你可以使用 (100.00 * failed/total) failedPercent

非常感谢！有没有办法计算比率 connection_error_jobs / total_failed_jobs * 100 ？因此，在每个组原因下，工作失败的比例。例如connection error jobs计数是2。Total failed jobs这里是5。所以ratio 应该是40%

更新了查询以获取比率，即 2/6 和百分比，即 2/9

【参考方案2】：

使用SUM 而不是COUNT

SELECT f.main_reason AS "main reason",
       COUNT(*) AS "Total jobs",
       SUM(CASE WHEN j.status='failed' THEN 1 ELSE 0 END) AS "Total failed jobs",
       (SUM(CASE WHEN j.status='failed' THEN 1 ELSE 0 END) / COUNT(*)) * 100 AS "Failure ratio"
FROM job j
LEFT JOIN failedreason f ON j.id = f.job_id
GROUP BY f.main_reason

db<>fiddle中的演示

【讨论】：

这个查询给了我同样的结果

请输入样本数据约15条记录

请查看样本数据

@shapale 我根据样本数据编辑了答案。我认为它现在有效。使用 SUM 而不是 COUNT

它返回了相同的结果

【参考方案3】：

计算唯一的作业，并汇总以获得总数。

SELECT
  CASE WHEN GROUPING(job.status) = 1 THEN 'TOTAL' ELSE job.status END AS `Job Status`
, CASE WHEN GROUPING(fail.main_reason) = 1 THEN 'TOTAL' ELSE fail.main_reason END AS `Failure Reason`
, COUNT(DISTINCT job.id) AS `Total Jobs`
, COUNT(DISTINCT job.id) / MAX(tots.total_jobs) * 100 AS `Ratio`
FROM job
CROSS JOIN (SELECT COUNT(*) AS total_jobs FROM job) tots
LEFT JOIN failedreason fail
ON job.id = fail.job_id
GROUP BY job.status, fail.main_reason WITH ROLLUP
ORDER BY GROUPING(job.status), GROUPING(fail.main_reason), `Ratio` DESC;

工作状态 |失败原因 |总职位 |比率 :--------- | :------------------------ | ---------: | --------: 成功| 空 | 3 | 33.3333 失败 |连接错误 | 2 | 22.2222 失败 |测试用例失败 | 2 | 22.2222 失败 |跑步者连接错误| 1 | 11.1111 失败 |脚本错误 | 1 | 11.1111 失败 |总计 | 6 | 66.6667 成功|总计 | 3 | 33.3333 总计 |总计 | 9 | 100.0000

dbfiddle here

上的演示

【讨论】：

同样的结果。计数应用于分组表。

好吧，好吧......如果他们有一个 main_reason，那么他们就是失败的。因此，按 main_reason 分组将有总作业数 = 总故障数。

查看更新、交叉连接并按比例使用

我检查了更新的查询。现在总作业数和总失败作业数仍然相等，但比率是奇数。

但是“Total Jobs”应该是MAX(tots.total_jobs)吗？你期待什么结果？