Need logic to update Varchar column in a table

Question

Hi,
Could you give me logic for below.
I have a table xx_dbc where one of the column Attribute1 is varchar2(240).
Attribute1 has values like 
==================
(AIN,PUP,GRI,NINE)
(AIN,PUP)
(AIN)
(GRI,NINE)
I have to update this column for every record with values AIN=10 , PUP=20 , GRI=30 , NINE=40
Columns should be updated like below
(10,20,30,40)
(10,20)
(10)
(30,40)
Thanks.

Answer 1

Hi,
Here's one way:
WITH  replacements  AS
    SELECT  'AIN' AS old_str, '10' AS new_str  FROM dual  UNION ALL
    SELECT  'PUP',            '20'             FROM dual  UNION ALL
    SELECT  'GRI',            '30'             FROM dual  UNION ALL
    SELECT  'NINE',           '40'             FROM dual
,    normalized_data  AS
    SELECT  p_key
    ,       LEVEL               AS sort_key
    ,       REGEXP_SUBSTR ( attribute1
                          , '[^(,)]+'
                          , 1
                          , LEVEL
                          )     AS str
    FROM    xx_dbc
--  WHERE   ...     -- If you need any filtering, put it here
    CONNECT BY   LEVEL              <= 1 + REGEXP_COUNT (attribute1, ',')
            AND  PRIOR p_key        = p_key
            AND  PRIOR SYS_GUID ()  IS NOT NULL
SELECT    '(' || LISTAGG ( NVL (r.new_str, n.str)
                         ) WITHIN GROUP (ORDER BY sort_key)
              || ')'  AS  new_attribute1
,         n.p_key                         -- If wanted
FROM             normalized_data  n
LEFT OUTER JOIN  replacements     r  ON  r.old_str  = n.str
GROUP BY  n.p_key
ORDER BY  n.p_key
Like everything else, it depends on your Oracle version.
In the query above, p_key can be any unique key from xx_dbc, including attribute1 or ROWID.
Relational databases, like Oracle, work best when each column of each row contains only 1 piece of information, not a delimited list of any number of pieces.  This is so fundamental to table design that it's called First Normal Form.  Most of the work in this problem involves converting your denormalized data into First Normal Form, and then converting it back again.  This problem, like many others, would be much simpler and more efficient if your table was in First Normal Form.
The query above assumes the replacement strings (such as 'AIN' and '10') are not already in a table.  If they are, or if they can be derived from a table, then you don't need the replacements sub-query; use your real table instead.