How to loop through table and update values in pl/sql

Question

We have Oracle 12 database.

I need to loop through table and update the values based on previous row. Like below I need to get the value from row 1 and update it to rows 2 to 6. Then get the new value from row 7 and continue to the end of the table. The column1 is csv-formatted text so I'll need to extract the number-value from there...

id  column1 column2
1   xxx;yyy;zzz;123456; 
2   aaa 
3   bbb 
4   ccc 
5   ddd 
6   eee 
7   xxx;yyy;zzz;789123; 
8   aaa

After the update table should look like this:

id  column1 column2
1   xxx;yyy;zzz;123456; 
2   aaa 123456
3   bbb 123456
4   ccc 123456
5   ddd 123456
6   eee 123456
7   xxx;yyy;zzz;789123; 
8   aaa 789123

Tbh I don't have much experience with pl/sql. I tried to use pl/sql while loop but It didn't work.

If someone could kick me in right direction that would be appreciated.

Answer 1

You do not need PL/SQL for this, it can be done entirely in SQL using a combination of MERGE , the LAG( ... ) IGNORE NULLS ... analytic function and REGEXP_SUBSTR (to extract the sub-string):

MERGE INTO table_name dst
USING (
  SELECT id,
         CASE WHEN val IS NULL
              THEN LAG( val ) IGNORE NULLS OVER ( ORDER BY id )
         END AS val
  FROM   (
    SELECT id,
           REGEXP_SUBSTR( column1, ';(\d{6});$', 1, 1, NULL, 1 ) AS val
    FROM   table_name
  )
) src
ON ( dst.id = src.id )
WHEN MATCHED THEN
  UPDATE SET column2 = src.val;

Updated table:

id  column1             column2
--  ------------------- -------
1   xxx;yyy;zzz;123456; 
2   aaa                 123456
3   bbb                 123456
4   ccc                 123456
5   ddd                 123456
6   eee                 123456
7   xxx;yyy;zzz;789123; 
8   aaa                 789123

Answer 2

Something like this should do what you want. It loops through the table, passing the last bit of the first record into a variable every time it starts with 'xxx' and updates column 2 if it doesn't.

DECLARE
    L_UpdateVal VARCHAR2(10) := '';
BEGIN
    FOR REC IN (SELECT Column1, column2, ROWNUM From table) LOOP
        IF SUBSTR(REC.column1, 1, 3) = 'xxx' THEN
            L_UpdateVal := SUBSTR(column1, 13, 6);
        ELSE
            UPDATE Table SET Column2 = L_UpdateVal
                WHERE ROWNUM := REC.ROWNUM
        END IF;
    END LOOP;
END;

Answer 3

You don't need PL/SQL for this, you could do it in a single MERGE statement.

First off, work out how to get the results you're after - you can do this by using the 'LAST_VALUE()' analytic function, like so:

WITH your_table AS (SELECT 1 id, 'xxx;yyy;zzz;123456;' column1, NULL column2 FROM dual UNION ALL
                    SELECT 2 id, 'aaa' column1, NULL column2 FROM dual UNION ALL
                    SELECT 3 id, 'bbb' column1, NULL column2 FROM dual UNION ALL
                    SELECT 4 id, 'ccc' column1, NULL column2 FROM dual UNION ALL
                    SELECT 5 id, 'ddd' column1, NULL column2 FROM dual UNION ALL
                    SELECT 6 id, 'eee' column1, NULL column2 FROM dual UNION ALL
                    SELECT 7 id, 'xxx;yyy;zzz;789123;' column1, NULL column2 FROM dual UNION ALL
                    SELECT 8 id, 'aaa' column1, NULL column2 FROM dual)
select id,
       column1,
       last_value(CASE WHEN substr(column1, -1) = ';' THEN 
                            regexp_substr(column1, ';*([[:digit:]]*)(;$)', 1, 1, NULL, 1)
                  END IGNORE NULLS) OVER (ORDER BY ID) column2
from   your_table;

        ID COLUMN1             COLUMN2
---------- ------------------- -------------------
         1 xxx;yyy;zzz;123456; 123456
         2 aaa                 123456
         3 bbb                 123456
         4 ccc                 123456
         5 ddd                 123456
         6 eee                 123456
         7 xxx;yyy;zzz;789123; 789123
         8 aaa                 789123

Then you can use that in a MERGE statement to do the update, like so:

MERGE INTO your_table tgt
  USING (select id,
                column1,
                CASE WHEN substr(column1, -1) = ';' THEN 'Y' ELSE 'N' END driving_column1,
                last_value(CASE WHEN substr(column1, -1) = ';' THEN 
                                     regexp_substr(column1, ';*([[:digit:]]*)(;$)', 1, 1, NULL, 1)
                           END IGNORE NULLS) OVER (ORDER BY ID) column2 -- assuming id drives the correct order to use here
         from   your_table) src
  ON (tgt.id = src.id) -- assuming id is the primary key of your_table
WHEN MATCHED THEN
  UPDATE SET tgt.column2 = src.column2;

If you don't want to update column2 of the column1 rows where the values are the csv values (I've assumed that there will be a semicolon present if they're csv values), then you can either update the ON clause to include and driving_column1 = 'N' or you can wrap the source subquery in an outer query that filters on the driving_column1 (you can't filter on analytic functions within the same query, unfortunately) or you can add a where clause to the update section of the merge statement.