Concatenation according to condition

Question

Hi All,
I ahve to concatenate the rows based on other row values....
example:
Col1 Col2
A X
A Y
B Z
B Y
needs to load into target table as:
Col1 Col2
A X:Y
B Z:Y
concate col2 accoring to col1
Thanks

Answer 1

1. What are your source and target databases? Are they on same instances if they are of the same technology type?
2. I would not go to an interface if the tables are on the schema and technology would be Oracle with version later than 9i. In that case, I would use the following SQL
INSERT INTO t2 (col1, col2)
SELECT col1, REPLACE(WMSYS.WM_CONCAT(col2),',',':')
FROM t1
GROUP BY col1;
There is a way to code this in an interface but a procedure would be simplest.
3. In case, assumptions in point 2 failed, there is a long road to the general solution. The solution above will give you answers if you have maximum two rows with same col1 value. I am interested to explore what happens if there are n rows with same column value when n being a random number? In the following steps, I would try to outline how we need to proceed, but because this is complicated, you may have to work with me here (t1 is source, t2 is target and tech is Oracle):
A. Create a variable v_MaxRows. In the refresh block, write:
WITH temp AS (
SELECT col1, count(*) as cnt_rows FROM t1 GROUP BY col1)
select MAX(cnt_rows) FROM temp
If technology is not Oracle, you have to adjust SQL a bit to make it work. This variable holds maximum no of rows a particular value of col1 is repeated in source t1.
B. Create a variable v_UseRow and do not create a refresh block for it. The default value should be 1.
C. Create a temp table (or temp tables based yellow interface) called t3 with columns col1, col2 and nrows (numerical value). Create an interface I1 with t3 as target. col1 and col2 should be mapped directly. Mapping for nrows column should be -> dense_rank() OVER(partition by col1 order by col2).
D. 1. Create interface I2. t3 should be the target.Update key for table t3 should be (col1,nrows).
2. Bring t3 two times to the source side (with alias t31 and t32), join them using col1. Mark the join as Left Outer Join on staging.
3. Create a filter on t31 as t31.nrows = 1 (executable on source)
4. Create a filter on t32 as t32.nrows = #v_UseRow+1 (executable on source)
5. On target, map for col1 should be t31.col1, map for col2 should be -> t31.col2||(CASE WHEN t32.col2 IS NOT NULL THEN ':'||t32.col2 ELSE '' END)
6. Column nrows should be mapped to 1.
E. Load target t2 using interface I3 and source t3. Direct mapping should work.
F. Create a package. Refresh #v_MaxRows and then call I1. Then define a loop where v_UseRow is set to 1, call I2, increment v_UseRow by 1, see if that is less than *#v_MaxRows*, then call I2 again else call I3 and exit. That way interface I2 would be called #v_MaxRows -1 times and all repeating values would be concatenated together.
I am interested to know the result. Please let me know the result.