如何在不使用 GROUP BY 或 PARTITION BY 的情况下对 Oracle SQL 中的数据进行分组

Question

My data is like this

MY_TABLE

DATE	CUSTOMER_ID	FAC_NUM	MONIES
01/Jan/2020	12345678	ABC123	125000
02/Jan/2020	12345678	ABC123	125000
03/Jan/2020	12345678	ABC123	125000
01/Feb/2020	12345678	ABC123	143000
02/Feb/2020	12345678	ABC123	143000
03/Feb/2020	12345678	ABC123	143000
04/Feb/2020	12345678	ABC123	143000
05/Feb/2020	12345678	ABC123	143000
01/Mar/2020	12345678	ABC123	125000
02/Mar/2020	12345678	ABC123	125000
03/Mar/2020	12345678	ABC123	125000
04/Mar/2020	12345678	ABC123	125000

I want the output to be like this

CUSTOMER_ID	FAC_NUM	MONIES	START_DATE	END_DATE
12345678	ABC123	125000	01/Jan/2020	03/JAN/2020
12345678	ABC123	143000	01/Feb/2020	05/Feb/2020
12345678	ABC123	125000	01/Mar/2020	04/Mar/2020

I have tried using the following

SELECT CUSTOMER_ID
      ,FAC_NUM
      ,MONIES
      ,MIN(DATE) AS START_DATE
      ,MAX(DATE) AS END_DATE
FROM MY_TABLE
GROUP BY CUSTOMER_ID
      ,FAC_NUM
      ,MONIES

However, the output I get with this approach is as follows (this is not the desired output)

CUSTOMER_ID	FAC_NUM	MONIES	START_DATE	END_DATE
12345678	ABC123	125000	1 Jan 2020	4 Mar 2020
12345678	ABC123	143000	1 Feb 2020	5 Feb 2020

Is there a way to obtain the output I am looking for without using PL/SQL (I need to run this query over a huge dataset)? If not, what would be the most efficient way to do this using PL/SQL?

I am new to stackoverflow and SQL. Your support would be highly appreciated.

Thanks in advance for your help.

Regards, Ani

Answer 1

Instead of "I want output", it would probably helped if you described what you wanted.

To me, it looks as if you'd want to group additionally by month, eg your query slightly modified:

SELECT CUSTOMER_ID
      ,FAC_NUM
      ,MONIES
      ,MIN(DATE) AS START_DATE
      ,MAX(DATE) AS END_DATE
FROM MY_TABLE
GROUP BY CUSTOMER_ID
      ,FAC_NUM
      ,MONIES
      , to_char(date, 'yyyymm')      --> this

Of course, date column name is invalid; it is reserved for date datatype so I presume it is, actually, named differently.

Answer 2

This reads like a gaps-and-islands problem. Islands are adjacent rows having the same customer, fac num and amount. Whenever the amount changes, a new group starts.

Here is one approach using the difference between row numbers to identify the islands:

select customer_id, fac_num, monies,
    min(date) as start_date, max(date) as end_date
from (
    select t.*, 
        row_number() over(partition by customer_id, fac_num order by date) as rn1,
        row_number() over(partition by customer_id, fac_num, monies order by date) as rn2
    from mytable t
) t
group by customer_id, fac_num, monies, rn1 - rn2
order by customer_id, fac_num, min(start_date)