How to convert this query from MySQL to SQL Server? - mysql

Question

Here is the source table:
+----+-------+
| Id | Score |
+----+-------+
| 1 | 3.50 |
| 2 | 3.65 |
| 3 | 4.00 |
| 4 | 3.85 |
| 5 | 4.00 |
| 6 | 3.65 |
+----+-------+
Here is the result table:
+-------+------+
| Score | Rank |
+-------+------+
| 4.00 | 1 |
| 4.00 | 1 |
| 3.85 | 2 |
| 3.65 | 3 |
| 3.65 | 3 |
| 3.50 | 4 |
+-------+------+
I have the MySQL version query, how to convert it to SQL server version? I tried to do the declare but I have no idea how to update the value of the variables.
SELECT Score, ranking AS Rank
FROM
(
SELECT
Score,
CASE
WHEN #dummy = Score
THEN #ranking := #ranking
ELSE #ranking := #ranking + 1
END as ranking,
#dummy := Score
FROM Scores, (SELECT #ranking := 0, #dummy := -1) init
ORDER BY Score DESC
)AS Result

Answer 1

Your code is a MySQL work-around for the ANSI standard DENSE_RANK() function (as explained by Sean Lange in a comment). The code simply looks like:
SELECT s.*, DENSE_RANK() OVER (ORDER BY Score DESC) as rank
FROM Scores s
ORDER BY Score DESC;
Incidentally, the MySQL code itself is not really accurate. The following is much safer:
SELECT Score, ranking AS Rank
FROM (SELECT Score,
(#rn := if(#dummy = Score, #rn,
if(#dummy := Score, #rn + 1, #rn + 1)
)
) as ranking
FROM Scores CROSS JOIN
(SELECT #rn := 0, #dummy := -1) init
ORDER BY Score DESC
) s;
The key difference is that all the assignments to variables occur in a single expression. MySQL does not guarantee the order of evaluations of expressions in a SELECT, so you should not use #dummy in one expression and then assign it in another.

Answer 2

SQL Server doesn't support variables used in that manner, so the syntax doesn't translate. Scalar variables are set once by a query, not set once for each row of the result set. MySQL's syntax here is a hack to get analytic-function-like behavior without analytic function support. You should just use:
SELECT Score,
DENSE_RANK() OVER(ORDER BY Score DESC) AS Rank
FROM Scores
ORDER BY Score DESC;
If you insist on not using DENSE_RANK(), you can use the SQL Server syntax from SQL Server 2000:
SELECT s1.Score,
(SELECT COUNT(DISTINCT s2.Score) FROM Scores s2 WHERE s1.Score <= s2.Score) AS Ranking
FROM Scores s1
ORDER BY s1.Score DESC;