Lag() 函数以及 group by SQL

Question

我有一个这种格式的表格

时间戳	sys_id	产品名称	值名称	价值
2024-01-03 00:35:00+05:30	1	a	字节	10
2024-01-03 00:35:00+05:30	2	a	字节	20
2024-01-03 00:35:00+05:30	3	a	字节	30
2024-01-03 00:40:00+05:30	1	a	字节	20
2024-01-03 00:40:00+05:30	2	a	字节	20
2024-01-03 00:40:00+05:30	3	a	字节	40

对于给定的一组time_stamp、sys_id、Product_name、valuename，我想找到该组的当前条目和上一个条目之间的差异 如果差值等于 0，那么我希望该组的另一列 Break_status 被标记为 1，否则为 0 因此，当我在 where 子句中给出 sys_id 时，我使用 lag() 函数，它工作正常，下面是 sql 查询和表输出

    SELECT 
        TO_CHAR(time_stamp) as Time, 
        TO_CHAR(sys_id) as System_ID, 
        product_name, 
        valuename, 
        TO_CHAR(value), 
        (value - LAG(value, 1) OVER (ORDER BY time_stamp)) as diff,
        CASE WHEN value = 0 or diff = 0 or diff is null THEN 0 ELSE 1 END AS breach_status
    FROM table
    WHERE valuename='bytes' and system_id = '2' and 
        time_stamp BETWEEN '2024-01-03 00:35:00' AND '2024-01-03 00:40:00'
    ORDER BY time_stamp 

时间戳	sys_id	产品名称	值名称	价值	差异	违规状态
2024-01-03 00:35:00+05:30	1	a	字节	10		0
2024-01-03 00:35:00+05:30	2	a	字节	20		0
2024-01-03 00:35:00+05:30	3	a	字节	30		0
2024-01-03 00:40:00+05:30	1	a	字节	20	10	0
2024-01-03 00:40:00+05:30	2	a	字节	20	0	1
2024-01-03 00:40:00+05:30	3	a	字节	40	10	0

此处，对于时间戳 = '2024-01-03 00:40:00+05:30'、sys_id = 2、product_name = 'a' 且 valuename = 'bytes'，值差异为 0，因此，breach_status 为 1

但是这里我想使用分组依据，因为时间戳，sys_id，product_name，valuename记录有多种不同的组合。

因此我想在上面的查询中使用 group by

    SELECT 
        TO_CHAR(time_stamp) as Time, 
        TO_CHAR(system_id) as System_ID, 
        customer_name, 
        valuename, 
        TO_CHAR(max(value)) as val,
        (max(value) - LAG(max(value), 1) OVER (ORDER BY time_stamp)) as diff,
        CASE WHEN value = 0 or diff = 0 or diff is null THEN 0 ELSE 1 END AS breach_status
    FROM table
    WHERE 
        time_stamp BETWEEN '2024-01-03 00:35:00' AND '2024-01-03 00:40:00'
    group by 
        time_stamp,
        system_id,
        customer_name,
        valuename
    ORDER BY time_stamp, diff, breach_status

但这里不是使用检查其集合之前条目之间的差异，而是检查与其上面条目的差异

时间戳	sys_id	产品名称	值名称	价值	差异	违规状态
2024-01-03 00:35:00+05:30	1	a	字节	10		0
2024-01-03 00:35:00+05:30	2	a	字节	20	10	0
2024-01-03 00:35:00+05:30	3	a	字节	30	10	0
2024-01-03 00:40:00+05:30	1	a	字节	20	-10	0
2024-01-03 00:40:00+05:30	2	a	字节	20	0	1
2024-01-03 00:40:00+05:30	3	a	字节	40	-20	0

我将如何使用此查询 最后，我将在 grafana 仪表板中使用此查询来根据违规值发出警报，因此在 select 语句中我希望所有内容都在 CHAR 中，因此我使用 to_char() 将其转换 但是如果我将 diff 列转换为 char 那么我们就无法在违规条件下比较它，因为它是字符串而不是数字。

Answer 1

你可以参考这个 sql 两行之间的差异与 group by

这是您需要的查询

    SELECT 
        TO_CHAR(time_stamp) as Time, 
        TO_CHAR(system_id) as System_ID, 
        customer_name, 
        valuename, 
        TO_CHAR(max(value)) as val,
        (max(value) - LAG(max(value), 1) OVER (partition by system_id, customer_name, valuename ORDER BY time_stamp)) as diff,
        CASE WHEN value = 0 or diff = 0 or diff is null THEN 0 ELSE 1 END AS breach_status
    FROM table
    WHERE 
        time_stamp BETWEEN '2024-01-03 00:35:00' AND '2024-01-03 00:40:00'
    group by 
        time_stamp,
        system_id,
        customer_name,
        valuename
    ORDER BY time_stamp```