使用 sum 时避免使用 group by

Question

I'm new to SQL/ORACLE . I am trying to wrote a query to get the desired result. I am using a sum function also in the select statement bu then I also have to write all columns from select in the group by clause. Is there any way to avoid using this approach. Below is my query and it is giving me the desired result, I just want to get rid of group by clause.

SELECT
    T1.ORDER_CODE "ORDER NUMBER",
    T1.ORDER_ITEM "ORDER ITEM",
    T1.ROLL_SHEET,
    T1.ORDER_STATUS "ORDER ITEM STATUS",
    T1.ORDER_TYPE "ORDER_TYPE",
    T2.ORDER_STATUS_SALES "SALES ORDER STATUS",
    T1.CLASSIFICATION "GRADE",
    T1.PRODUCT_CODE "PRODUCT", 
    T1.ARTICLE_CODE "ARTICLE",
    T1.WIDTH || 'x' || T1.LENGTH || 'x' || NVL((NULLIF(T1.HEIGHT_PALLET,0)),T1.DIAMETER) "SIZE", 
    T2.DATE_CUST_ISSUED "ORDERED DATE", 
    T2.DATE_CONFIRMED "CONFIRMATION DATE",
    T1.DATE_EXMILL "EXMILL DATE",
    T1.MASS_ORDERED "ORDERED",
    T3.MASS_TRIMMED "TRIMMED", 
    T3.MASS_WOUND "PRODUCED",
    T3.MASS_SCALED "WRAPPED",
    SUM(T4.MASS) "IN WH"
FROM ORDERITEM T1 
INNER JOIN ORDERHEADER T2 ON T1.ORDER_CODE = T2.ORDER_CODE
LEFT OUTER JOIN MACHINERUNORDERITEM T3 ON T1.ORDER_CODE = T3.ORDER_CODE AND T1.ORDER_ITEM = T3.ORDER_ITEM
LEFT OUTER JOIN ORDERITEMINVENTORY T4 ON T1.ORDER_CODE = T4.ORDER_CODE AND T1.ORDER_ITEM = T4.ORDER_ITEM
WHERE T1.ORDER_CODE = '1000000294'
GROUP BY T1.ORDER_CODE, T1.ORDER_ITEM, T1.ROLL_SHEET, T1.ORDER_STATUS,
         T1.ORDER_TYPE, T2.ORDER_STATUS_SALES, T1.CLASSIFICATION,
         T1.PRODUCT_CODE, T1.ARTICLE_CODE,
         T1.WIDTH || 'x' || T1.LENGTH || 'x' || NVL((NULLIF(T1.HEIGHT_PALLET,0)),T1.DIAMETER),
         T2.DATE_CUST_ISSUED, T2.DATE_CONFIRMED, T1.DATE_EXMILL,
         T1.MASS_ORDERED, T3.MASS_TRIMMED, T3.MASS_WOUND, T3.MASS_SCALED

Answer 1

Well, you can't. It's part of the SYNTAX.

If you use SUM, SQL, which thinks in sets (not line by line), must have a GROUP BY clause. When you ask for a column in the SELECT statement, SQL (Oracle, in this case), brings you the columns. If you want to SUM, it need to know what to GROUP BY, otherwise, it would just SUM everything. That's all. I suggest you'd read more about the GROUP BY clause and the SUM function in the W3 schools in order to understand the issue better.

Answer 2

A typicall way of constructing a non-trivial GROUP BY statement is to write the first part and copy the grouping columns in the GROUP BY clause

select 
COL1 ||'-' || COL2 as x,
case when COL1 = 'A1' then COL2 ||'-' || COL1 else COL1 || '-' || COL2 end as y,
sum(COL3)
from tab
group by COL1 ||'-' || COL2 as x,
case when COL1 = 'A1' then COL2 ||'-' || COL1 else COL1 || '-' || COL2 end as y;

This leads to an error:

ORA-00933: SQL command not properly ended

and is known pain point (interestingly not for all) - see the discussion in Oracle Ideas Page

So the next step is to remove the aliases from the GROUP BY clause and the statement works ...

select 
COL1 ||'-' || COL2 as x,
case when COL1 = 'A1' then COL2 ||'-' || COL1 else COL1 || '-' || COL2 end as y,
sum(COL3)
from tab
group by COL1 ||'-' || COL2,
case when COL1 = 'A1' then COL2 ||'-' || COL1 else COL1 || '-' || COL2 end;

... but is a classical example of an stinking code where every change must be made in identical way in two places.

Less know, but it works, is that you do not need to copy the exact expressions of the GROUP BY columns , list of all used original columns is enough .

Here an example - COL1, COL2 in GROUP BY covers all the expression used in the SELECT

select 
COL1 ||'-' || COL2 as x,
case when COL1 = 'A1' then COL2 ||'-' || COL1 else COL1 || '-' || COL2 end as y,
sum(COL3)
from tab
group by COL1, COL2

X     Y      SUM(COL3)
----- ----- ----------
A1-B1 B1-A1          1 
A3-B3 A3-B3          3 
A2-B2 A2-B2          2 

My test data

create table tab as
select 'A1' col1, 'B1' col2, 1 col3 from dual union all
select 'A2' col1, 'B2' col2, 2 col3 from dual union all
select 'A3' col1, 'B3' col2, 3 col3 from dual;

Answer 3

Assuming that T4 is what is causing "duplication", you can solve this problem simply by using a correlated subquery:

select . . .,
       (select sum(oit.mass)
        from ORDERITEMINVENTORY oii 
        where T1.ORDER_CODE = oii.ORDER_CODE AND T1.ORDER_ITEM = oii.ORDER_ITEM
       )
from . . .  -- leave `T4` out of this clause

No GROUP BY is then necessary. This will probably also have better performance.

Note that this "seems" to work only for one aggregation column -- which solves your immediate problem. Oracle 12C supports lateral joins, which would allow multiple columns in a "correlated" subquery.

Also, I strongly recommend that you use meaningful table aliases. In this code snippet, I replaced the meaningless T4 with oii , which are the first letters of each word in the table name.