Aggregates Combinations

Question

I have a data like this.
CREATE TABLE GROUPING_EX
  CHANNEL   VARCHAR2(10 BYTE),
  CODE   VARCHAR2(10 BYTE), 
  CNT  NUMBER
COMMIT;
Insert into GROUPING_EX (CHANNEL, CODE, CNT) Values('a', 'c1', 1);
Insert into GROUPING_EX (CHANNEL, CODE, CNT) Values('a', 'c2', 1);
Insert into GROUPING_EX (CHANNEL, CODE, CNT) Values('a', 'c3', 1);
COMMIT;And this is the output format that I need:
Channel, New_Code, Cnt
a,       c1,        1
a,       c1,        1
a,       c1,        1
a,       c1,c2      2
a,       c3,c2      2
a,       c1,c3      2
a,       c1,c2,c3   3     I am trying it with Cube and Rollup. But not getting it correct. Any hints are appreciated.
Kumar

Answer 1

Hi, Kumar,
Thanks for posting the CREATE TABLE and INSERT statmements; that's very helpful!
GROUP BY gets get the possible combinations of a value from column_x and and value from column_y, but (as far as I know) it can't get the combinations of 2 different values from column_x. You could possibly do a pivot, so that the different values were in different columns, but you'd have to know at the time you wrote the query how many different values there would be at run-time.
CONNECT BY is much better at getting the combinations from a single column:
WITH     got_combinations     AS
     SELECT     channel
     ,     SYS_CONNECT_BY_PATH (code, ',') || ','     AS code_list
     FROM     grouping_ex
     CONNECT BY     code     > PRIOR  code
          AND     channel     = PRIOR  channel
SELECT       c.channel
,       TRIM (',' FROM c.code_list)     AS new_code
,       SUM (x.cnt)                 AS cnt
FROM       got_combinations     c
JOIN       grouping_ex          x  ON     c.channel     = x.channel
                         AND     INSTR ( c.code_list
                                    , ',' || x.code || ','
                               )             > 0
GROUP BY  c.channel
,            c.code_list
ORDER BY  LENGTH (new_code)
,            new_code
;Output:
CHANNEL    NEW_CODE                    CNT
a          c1                            1
a          c2                            1
a          c3                            1
a          c1,c2                         2
a          c1,c3                         2
a          c2,c3                         2
a          c1,c2,c3                      3What role do channel and cnt play in this problem? It's hard to tell from the sample data, because they always have the same value in that data.