从表 pl/sql oracle 中删除重复项

Question

I have a table that contains data and fields:

contact_id | call_siebel | start_time | operator_text | client_text | client_id | phone_num

I found duplicates:

SELECT operator_text, client_text, client_id
FROM TABLE
GROUP BY operator_text, client_text, client_id
HAVING COUNT(*) > 1;

For example, there are 4 identical data, two of which have the same client_id value (correct) and the other two have client_id = '-1'. And I need to leave only one option out of four, whose client_id is filled in correctly.

I wanted to create a test table in which I will enter data with a unique value client_id from all duplicates that exist, remove duplicates from the main table data on client_id and in the end just insert from the test table into the main one.

How would it be more correct to insert data from the main table into the test one with a unique client_id from all duplicates? In my version, I incorrectly implemented GROUP BY:

INSERT /*+ append enable_parallel_dml parallel(16)*/
INTO table_test
SELECT
 DISTINCT 
 t.contact_id, 
 t.call_siebel, 
 t.start_time, 
 t.operator_text, 
 t.client_text, 
 t.client_id, 
 t.phone_num
FROM table t
WHERE t.client_id != '-1'
GROUP BY t.operator_text, t.client_text, t.client_id
HAVING COUNT(t.client_id) > 1;

After this I could easily finish my job using:

DELETE table
WHERE client_id
IN (SELECT t.client_id
FROM table_test t);


INSERT 
INTO table
SELECT *
FROM table_test;

Answer 1

How about skipping that "temporary" table entirely?

Sample data:

SQL> select * from test;

 CLIENT_ID OPERATOR_TEXT        CLIENT_TEXT
---------- -------------------- --------------------
         1 a                    a         --> the first two rows 
         1 a                    a         --> are duplicates
        -1 a                    a         --> this is invalid row (because of a negative CLIENT_ID)
         1 b                    c         --> this is OK

Delete invalid rows (because of a negative client_id ) (line #2) and all duplicates (the rowid subquery):

SQL> delete from test a
  2  where a.client_id < 0
  3     or a.rowid > (select min(b.rowid)
  4                   from test b
  5                   where b.client_id     = a.client_id
  6                     and b.operator_text = a.operator_text
  7                     and b.client_text   = a.client_text
  8                  );

2 rows deleted.

Result:

SQL> select * from test;

 CLIENT_ID OPERATOR_TEXT        CLIENT_TEXT
---------- -------------------- --------------------
         1 a                    a
         1 b                    c

SQL>