SQL：按日期分组并对列中的值求和

Posted 2023-03-27

【中文标题】SQL：按日期分组并对列中的值求和

【英文标题】：SQL: Group by date and summing values in a column

【发布时间】：2015-05-07 18:46:30

【问题描述】：

我有一个 JDBC 数据库（特别是 DB2，但我正在寻找与 DB 无关的东西，至少是 DB2 和 Oracle），它有一个表，每 10 分钟获取插入记录，其中包含应用程序运行的 API 的统计信息问题。它看起来像：

StatKey, StartDate, EndDate, APIName, StatName, StatValue
201505071498224437562706    2015-05-07 14:12:44.0   2015-05-07 14:22:44.0   API5    Invocations 34
201505071498161437466684    2015-05-07 14:06:14.0   2015-05-07 14:16:14.0   API4    Invocations 79
201505071498060937466556    2015-05-07 13:56:08.0   2015-05-07 14:06:08.0   API4    Average 26,264.37
201505071497263437627286    2015-05-07 14:16:33.0   2015-05-07 14:26:34.0   API2    Invocations 24
201505071497262137620812    2015-05-07 14:16:19.0   2015-05-07 14:26:20.0   API2    Invocations 24
201505071497024537466378    2015-05-07 13:52:43.0   2015-05-07 14:02:44.0   API1    Average 6,830,050
201505071497023337466368    2015-05-07 13:52:31.0   2015-05-07 14:02:32.0   API3    Average 31,523
201505071496023337466361    2015-05-07 13:52:31.0   2015-05-07 14:02:32.0   API2    Invocations 1
201505071494263837628892    2015-05-07 14:16:36.0   2015-05-07 14:26:37.0   API5    Invocations 68
201505071493124437466656    2015-05-07 14:02:44.0   2015-05-07 14:12:44.0   API1    Invocations 2
201505071492263037625304    2015-05-07 14:16:29.0   2015-05-07 14:26:30.0   API3    Average 179,223.29

每 10 分钟，在此期间执行的任何 API 都会有一个与上述类似的条目。但是，多个 JVM 将写入同一个数据库，因此开始和结束时间不是简单地每 10 分钟一次，每小时可能有超过 6 个条目。

我要做的是创建一个 SQL，该 SQL 将每小时对所有 API 的所有调用进行分组。例如：

Date&Hour, API, Invocations
2015-05-07 12:00, API1, 100
2015-05-07 12:00, API2, 150
2015-05-07 13:00, API2, 200
etc...

我尝试在小时标记处基于主键的 SUBSTR（始终是时间戳加上一些随机数 - 但小时和分钟之间是 2 个随机数字）进行 GROUP BY，但我不是确定如何添加所有 StatName=Invocations per hour。

有人可以就我如何实现这一点提供一些想法吗？

【参考方案1】：

另一种可能的解决方案：

select to_char(StarDate,'rrrr-mm-dd HH24:')||'00'  as DateHour,
    APIName as API,
    sum(StatValue) as Invocations
from STATISTICS
where StatName = 'Invocations' 
group by to_char(StarDate,'rrrr-mm-dd HH24:')||'00', APIName

有不同的方法可以做到这一点..

祝你好运！

【讨论】：

这里给出的一切都是一个很好的解决方案，但就我的需要而言，这相对简单，似乎对我来说效果最好。谢谢。

【参考方案2】：

不太确定你要的是这个吗？

本质上，它查看 YYYYMMDDHH 10 个位置，因为它们包含要分组的值...然后仅基于调用求和

SELECT substr(statKey,1,10) as DH, APIName, Sum(Statvalue) Invocations
FROM TableName
WHERE StatName = 'Invocations'
GROUP BY substr(statKey,1,10), APIName, StatName

例子：

WITH CTE AS
  (SELECT '201505071498224437562706' AS StatKey,
    '2015-05-07 14:12:44.0'          AS StartDate,
    '2015-05-07 14:22:44.0'          AS EndDate,
    'API5'                           AS APIName,
    'Invocations'                    AS StatName,
    34                               AS statvalue
  FROM dual
  UNION ALL
  SELECT '201505071498161437466684',
    '2015-05-07 14:06:14.0',
    '2015-05-07 14:16:14.0',
    'API4',
    'Invocations',
    79
  FROM dual
  UNION ALL
  SELECT '201505071498060937466556',
    '2015-05-07 13:56:08.0',
    '2015-05-07 14:06:08.0',
    'API4',
    'Average',
    26264.37
  FROM dual
  )
SELECT substr(statKey,1,10) as DH, APIName, StatName, Sum(Statvalue) 
FROM TableName
WHERE StatName = 'Invocations'
GROUP BY substr(statKey,1,10), APIName, StatName

【参考方案3】：

至少对于 DB2，为什么不只是

select date(startdate) as start_date
       , hour(startdate) as start_hour
       , API
       , sum(statvalue) as Invocations
from mytbl
where statname = 'Invocations' 
group by date(startdate),  hour(startdate), API

如果那是你真正想要的，我将把它作为一个练习让你将日期和时间组合回时间戳......

【参考方案4】：

SQL Fiddle

Oracle 11g R2 架构设置：

CREATE TABLE Data AS
          SELECT '201505071498224437562706' AS StatKey, TO_DATE( '2015-05-07 14:12:44', 'YYYY-MM-DD HH24:MI:SS' ) AS StartDate, TO_DATE( '2015-05-07 14:22:44', 'YYYY-MM-DD HH24:MI:SS' ) AS EndDate, 'API5' AS APIName, 'Invocations' AS StatName, 34 AS StatValue FROM DUAL
UNION ALL SELECT '201505071498161437466684' AS StatKey, TO_DATE( '2015-05-07 14:06:14', 'YYYY-MM-DD HH24:MI:SS' ) AS StartDate, TO_DATE( '2015-05-07 14:16:14', 'YYYY-MM-DD HH24:MI:SS' ) AS EndDate, 'API4' AS APIName, 'Invocations' AS StatName, 79 AS StatValue FROM DUAL
UNION ALL SELECT '201505071498060937466556' AS StatKey, TO_DATE( '2015-05-07 13:56:08', 'YYYY-MM-DD HH24:MI:SS' ) AS StartDate, TO_DATE( '2015-05-07 14:06:08', 'YYYY-MM-DD HH24:MI:SS' ) AS EndDate, 'API4' AS APIName, 'Average' AS StatName, 26264.37 AS StatValue FROM DUAL
UNION ALL SELECT '201505071497263437627286' AS StatKey, TO_DATE( '2015-05-07 14:16:33', 'YYYY-MM-DD HH24:MI:SS' ) AS StartDate, TO_DATE( '2015-05-07 14:26:34', 'YYYY-MM-DD HH24:MI:SS' ) AS EndDate, 'API2' AS APIName, 'Invocations' AS StatName, 24 AS StatValue FROM DUAL
UNION ALL SELECT '201505071497262137620812' AS StatKey, TO_DATE( '2015-05-07 14:16:19', 'YYYY-MM-DD HH24:MI:SS' ) AS StartDate, TO_DATE( '2015-05-07 14:26:20', 'YYYY-MM-DD HH24:MI:SS' ) AS EndDate, 'API2' AS APIName, 'Invocations' AS StatName, 24 AS StatValue FROM DUAL
UNION ALL SELECT '201505071497024537466378' AS StatKey, TO_DATE( '2015-05-07 13:52:43', 'YYYY-MM-DD HH24:MI:SS' ) AS StartDate, TO_DATE( '2015-05-07 14:02:44', 'YYYY-MM-DD HH24:MI:SS' ) AS EndDate, 'API1' AS APIName, 'Average' AS StatName, 6830050 AS StatValue FROM DUAL
UNION ALL SELECT '201505071497023337466368' AS StatKey, TO_DATE( '2015-05-07 13:52:31', 'YYYY-MM-DD HH24:MI:SS' ) AS StartDate, TO_DATE( '2015-05-07 14:02:32', 'YYYY-MM-DD HH24:MI:SS' ) AS EndDate, 'API3' AS APIName, 'Average' AS StatName, 31523 AS StatValue FROM DUAL
UNION ALL SELECT '201505071496023337466361' AS StatKey, TO_DATE( '2015-05-07 13:52:31', 'YYYY-MM-DD HH24:MI:SS' ) AS StartDate, TO_DATE( '2015-05-07 14:02:32', 'YYYY-MM-DD HH24:MI:SS' ) AS EndDate, 'API2' AS APIName, 'Invocations' AS StatName, 1 AS StatValue FROM DUAL
UNION ALL SELECT '201505071494263837628892' AS StatKey, TO_DATE( '2015-05-07 14:16:36', 'YYYY-MM-DD HH24:MI:SS' ) AS StartDate, TO_DATE( '2015-05-07 14:26:37', 'YYYY-MM-DD HH24:MI:SS' ) AS EndDate, 'API5' AS APIName, 'Invocations' AS StatName, 68 AS StatValue FROM DUAL
UNION ALL SELECT '201505071493124437466656' AS StatKey, TO_DATE( '2015-05-07 14:02:44', 'YYYY-MM-DD HH24:MI:SS' ) AS StartDate, TO_DATE( '2015-05-07 14:12:44', 'YYYY-MM-DD HH24:MI:SS' ) AS EndDate, 'API1' AS APIName, 'Invocations' AS StatName, 2 AS StatValue FROM DUAL
UNION ALL SELECT '201505071492263037625304' AS StatKey, TO_DATE( '2015-05-07 14:16:29', 'YYYY-MM-DD HH24:MI:SS' ) AS StartDate, TO_DATE( '2015-05-07 14:26:30', 'YYYY-MM-DD HH24:MI:SS' ) AS EndDate, 'API3' AS APIName, 'Average' AS StatName, 179223.29 AS StatValue FROM DUAL;

查询 1：

SELECT   TRUNC( EndDate, 'HH' ) AS "Date&Hour",
         APIName,
         SUM( StatValue ) AS Invocations
FROM     Data
WHERE    StatName = 'Invocations'
GROUP BY TRUNC( EndDate, 'HH' ),
         APIName

Results：

|             Date&Hour | APINAME | INVOCATIONS |
|-----------------------|---------|-------------|
| May, 07 2015 14:00:00 |    API2 |          49 |
| May, 07 2015 14:00:00 |    API5 |         102 |
| May, 07 2015 14:00:00 |    API1 |           2 |
| May, 07 2015 14:00:00 |    API4 |          79 |

【讨论】：

DB2 似乎也有TRUNC function，所以我认为这应该适用于两者（但无法测试）。

【参考方案5】：

日期函数似乎很难以与数据库无关的方式实现。

对于与 DB 无关的解决方案，我建议在数据库中创建视图，以隐藏特定于 DB 的代码的实现，因此您可以使用直接选择而不会出现任何语法问题。