如何更新具有最常值的表

Question

I have an update query that should update a field with the most frequent values from another table this is how i do it in postgreSQL

UPDATE TABLE1 T1 
SET COLUMN_B = (SELECT COLUMN_B
                         FROM 
                         (SELECT COLUMN_A,COLUMN_B, COUNT(1) AS FREQUENCY
                                 FROM TABLE2 T2
                                GROUP BY COLUMN_A,COLUMN_B
                                 ORDER BY COLUMN_A,FREQUENCY DESC) QUERY1
                          WHERE QUERY1.COLUMN_A= T1.COLUMN_A

                    GROUP BY COLUMN_A,COLUMN_B,FREQUENCY
                    ORDER BY FREQUENCY DESC LIMIT 1
                   )

this works fine in postgresql , i want to do the same query in Oracle using rownum =1 but I have several issues :

1. I can't put an ORDER BY inside the select of an update
2. If i decide to put the order by in the nested select (QUERY1), the nested select doesn't understand the reference to table T1 ( T2.COLUMN_A = T1.COLUMN_A gives an error that T1.COLUMN_A invalid identifier)

How can i do this in oracle? what i want is that T1.COLUMN_B be filled with the most frequent common value of COLUMN_B FROM T2 for each COLUMN_A.

Thank you for any help or suggestion

Answer 1

I'm pretty sure there is some much less complicated way to do this but this should be good starting point and should work both in Oracle and in Postgres

UPDATE TABLE1 T1
SET COLUMN_B = 
select COLUMN_B from (
(SELECT DISTINCT COLUMN_A, COLUMN_B
    FROM TABLE2 T2
    WHERE T2.COLUMN_A= T1.COLUMN_A
    GROUP BY COLUMN_A, COLUMN_B
    HAVING COUNT(1) = (
        SELECT MAX(CNT) FROM (
            SELECT COLUMN_B, COUNT(1) AS CNT
            FROM TABLE2 T3
            WHERE T3.COLUMN_A= T2.COLUMN_A)
        )
    )
)

Answer 2

This can be done in Oracle uses analytic functions; don't know whether the same code would work in PostgreSQL.

Creating a target table:

SQL> create table t23
  2    as select deptno, cast (null as number(7,2)) max_sal from dept
  3  /

Table created.
SQL> 

Now, the update:

SQL> update t23
  2  set max_sal = ( select sal from
  3                   ( select deptno, sal
  4                     , row_number() over (partition by deptno order by sal desc) rn
  5                      from emp )
  6                 where rn = 1
  7                 and deptno = t23.deptno )
  8  /

6 rows updated.

SQL>

Here is the outcome:

SQL> select * from t23
  2  /

    DEPTNO    MAX_SAL
---------- ----------
        10       5000
        20       3000
        30       3750
        40
        50       4500
        60

6 rows selected.

SQL> 

Just to confirm the results...

SQL> select deptno, max(sal)
  2  from emp
  3  where sal is not null
  4  group by deptno
  5  order by deptno
  6  /

    DEPTNO   MAX(SAL)
---------- ----------
        10       5000
        20       3000
        30       3750
        50       4500

SQL>