Question

I have a case where I need to show only the top rows based on a setting in a table and the ordinal set.

Example dataset below shows two customers; each of the customers have a different product.
Since NumRowsToShow is “1” I only want to show one row (the top row based on ordinal) for EACH Customer.

| CustomerID | ProductID | Ordinal | NumRowsToShow |
+------------+-----------+---------+---------------+
| 1          |A          |1        |1              |
| 1          |B          |2        |1              |
| 1          |C          |3        |1              |
| 5          |D          |1        |1              |
| 5          |E          |2        |1              |
| 5          |F          |3        |1              |

The result set after query is run should be

| CustomerID | ProductID |
+------------+-----------+
| 1          |A          |
| 5          |D          |

In the same scenario if NumRowsToShow were 1 for customerID 1 and 2 for CustomerID 5 I would see something like.

| CustomerID | ProductID | Ordinal | NumRowsToShow |
+------------+-----------+---------+---------------+
| 1          |A          |1        |1              |
| 1          |B          |2        |1              |
| 1          |C          |3        |1              |
| 5          |D          |1        |2              |
| 5          |E          |2        |2              |
| 5          |F          |3        |2              |

The result set after query is run should be

| CustomerID | ProductID |
+------------+-----------+
| 1          |A          |
| 5          |D          |
| 5          |E          |

How can this be done?

Including a screen cap of actual result set with highlights of what I’m trying to filter down to which may be a little helpful.

_{(source: harpernet.net)}

Answer 1

It feels like “cheating in the exams”:

SELECT CustomerID, ProductID
FROM tableX
WHERE Ordinal <= NumRowsToShow

---

If, as comments suggest, the Ordinal can have 10, 20, 30 values and not only 1, ..., n values, then this will work:

SELECT t.CustomerID, t.ProductID
FROM tableX AS t
  JOIN tableX AS tt
    ON  tt.CustomerID = t.CustomerID
    AND tt.Ordinal <= t.Ordinal
GROUP BY t.CustomerID
       , t.ProductID
       , t.NumRowsToShow
HAVING COUNT(*) <= t.NumRowsToShow

or even better, the:

SELECT CustomerID, ProductID
FROM
  ( SELECT CustomerID, ProductID, NumRowsToShow
         , ROW_NUMBER() OVER( PARTITION BY CustomerID 
                              ORDER BY Ordinal
                            ) AS Rn
    FROM tableX
  ) AS tmp
WHERE Rn <= NumRowsToShow ;

Test in: SQL-Fiddle

---

Your table looks to be not normalized. The NumRowsToShow columns has duplicate infomation and that can lead to update anomalies. This:

| CustomerID | ProductID | Ordinal | NumRowsToShow |
+------------+-----------+---------+---------------+
| 1          |A          |1        |1              |
| 1          |B          |2        |1              |
| 1          |C          |3        |1              |
| 5          |D          |1        |2              |
| 5          |E          |2        |2              |
| 5          |F          |3        |2              |

could be normalized to 2 tables:

| CustomerID | ProductID | Ordinal |
+------------+-----------+---------+
| 1          |A          |1        |
| 1          |B          |2        |
| 1          |C          |3        |
| 5          |D          |1        |
| 5          |E          |2        |
| 5          |F          |3        |

and:

| CustomerID | NumRowsToShow |
+------------+---------------+
| 1          |1              |
| 5          |2              |