Selecting single dataset from multiple datasets based on prioritized column

Question

I apologize if a similar question has been answered before. I don't seem to find to the correct search terms to find the solution.
On to the question - I have the following data:
+---ItemID---+---ItemName----------------------+---CompanyID---+
|    1001    | Item  1001                      |       1       |
|    1001    | Tuote 1001                      |       2       |
|    1001    | Artikel 1001                    |       3       |
|    1002    | Item  1002                      |       1       |
|    1002    | Artikel 1002                    |       3       |
|    1150    | Tuote 1150                      |       2       |
|    2000    | Item  2000                      |       1       |
|    2000    | Tuote 2000                      |       2       |
|    2222    | Tuote 2222                      |       2       |
|    2222    | Artikel 2222                    |       3       |
+------------+---------------------------------+---------------+
with t as (
        select 1001 ItemID, 'Item  1001' ItemName, 1 CompanyID from dual union all
        select 1001,    'Tuote 1001',   2 from dual union all
        select 1001,    'Artikel 1001', 3 from dual union all
        select 1002,    'Item  1002',   1 from dual union all
        select 1002,    'Artikel 1002', 3 from dual union all
        select 1150,    'Tuote 1150',   2 from dual union all
        select 2000,    'Item  2000',   1 from dual union all
        select 2000,    'Tuote 2000',   2 from dual union all
        select 2222,    'Tuote 2222',   2 from dual union all
        select 2222,    'Artikel 2222', 3 from dual
select
from
    tAs you can see, not all companies contain the same items. I would need to get a list which contains each itemid only once, and item name (and other fields not visible in the example) from the first company in which the item is found. The priority would be CompanyID 1, 3 and lastly 2. So the above example would result in the following list:
+---ItemID---+---ItemName----------------------+---CompanyID---+
|    1001    | Item  1001                      |       1       |
|    1002    | Item  1002                      |       1       |
|    1150    | Tuote 1150                      |       2       |
|    2000    | Item  2000                      |       1       |
|    2222    | Artikel 2222                    |       3       |
+------------+---------------------------------+---------------+

Answer 1

Hi,
That's called a Top-N Query , and here's one way to do it:
WITH     got_r_num   AS
     SELECT     ItemID, ItemName, CompanyID
     ,     ROW_NUMBER () OVER ( PARTITION BY  ItemID
                               ORDER BY          CASE  CompanyID
                                               WHEN  1  THEN  1
                                    WHEN  3  THEN  2
                                    WHEN  2  THEN  3
                                         END
                             ) AS r_num
     FROM    t
--     WHERE     ...     -- Any filtering goes here
SELECT     ItemID, ItemName, CompanyID
FROM     got_r_num
WHERE     r_num     = 1
;The tricky part here is getting an expression that will sort the CompanIDs in the order in which you want them (3 comes before 2). If that order applies to other queries besides this one, then you should really not hard-code it into any query. Instead, add a column to the company table (or create a company table, if you don't already have one), that indicates the sort order, and join to that table.
The aggregate KEEP (DENSE_RANK ...) feature could also be used to get these results, but you mentioned that there were "other fields not visible in the example", which means you would probably have to repeat a fairly complicated expression for each one of them. The query above is easily extended; add whatever columns you want to both SELECT clauses.
Edited by: Frank Kulash on Jun 21, 2010 4:41 PM