Strings comparision and get unique string in pure sql,

Question

Dear all,
Here is two strings
S1='A,B,C,D,F';
s2='C,F,H,B,A,K';
output should be like unique string values ..
S3= A,B,C,D,F,H,K;
How to get this in pure sql..
thanks in advance,
Roots

Answer 1

Hi,
In a relational database, each column of each row should store one value, not a repleating group of values, such as a delimited list. This is so basic to database design that it is called "First Normal Form". You don't have to follow rules like this, but, if you don't, your code will be complicated, inefficient, and error-prone. It sould be best to re-design your application so that each value was on a separate row.
If you can't do that, then you can start by splitting your delimitd lists into separate rows. Then you can easily fond the distinct values, and use any String Aggregation technique to combine the results into one output row.
Here's one way to do all that:
WITH     got_params     AS
     SELECT     'A,B,C,D,F' AS str, 1 AS str_id     FROM dual     UNION ALL
     SELECT     'C,F,H,B,A,K',          2            FROM dual
,     got_part_cnt     AS
     SELECT     str
     ,     1 + LENGTH (str)
            - LENGTH (REPLACE (str, ','))     AS part_cnt
     FROM    got_params
,     cntr          AS
     SELECT     LEVEL     AS n
     FROM     (
              SELECT  MAX (part_cnt)     AS max_part_cnt
              FROM    got_part_cnt
     CONNECT BY     LEVEL     <= max_part_cnt
,     got_substr     AS
     SELECT DISTINCT
            REGEXP_SUBSTR ( p.str
                           , '[^,]+'
                    , 1
                    , c.n
                    )          AS sub_str
     FROM    got_part_cnt p
     JOIN     cntr          c  ON     c.n     <= p.part_cnt
,     got_r_num     AS
     SELECT     sub_str
     ,     ROW_NUMBER () OVER (ORDER BY  sub_str)     AS r_num
     ,     ROWNUM                                        AS r
     FROM     got_substr
SELECT     MIN ( SUBSTR ( SYS_CONNECT_BY_PATH (sub_str, ',')
                      , 2
         )          AS unique_sub_strs
FROM    got_r_num
WHERE     CONNECT_BY_ISLEAF     = 1
-- START WITH     r_num     = 1
CONNECT BY     r_num     = 1 + PRIOR r_num
;This works in Oracle 10.2.0.2.0 Express Edition, which is the only database I can use right now. For the main query, you should be able to say:
SELECT     SUBSTR ( SYS_CONNECT_BY_PATH (sub_str, ',')
            , 2
            )     AS unique_sub_strs
FROM    got_r_num
WHERE     CONNECT_BY_ISLEAF     = 1
START WITH     r_num     = 1
CONNECT BY     r_num     = 1 + PRIOR r_num
;but, when I try that on my database, I only get 'A,B' as the output. CONNECT BY is very buggy in Oracle 10.2; if you have Oracle 10.1, the simpler form might work.
The query above can be shortened some, but I wrote it this way to make it easier to understand.
You can, for example, combine the sub-queries got_sub_str and got_r_num. If you do, use DENSE_RANK instead of ROW_NUMBER.
This does not assume that the sub-qtrings are all 1-character long. If they are, then the query can be simplified.
For more about string aggregation, see
http://www.oracle-base.com/articles/10g/StringAggregationTechniques.php