Question

Here is my scenario:

I have a single table with 2 columns. ID and Value. ID is int and value is real.

ID    Value
1     6.7
2     8.9
3     4.5
5     3.2
8     2.5
9     2.1
10    1.0
15    2.3
18    2.4
19    4.0
20    3.2

I would like to compose a SP that receives a grouping number (Group) and an operation (Op) and returns a new table in the following manner:

Group = 2,
Op = Max

IDstart   IDend    Value
1         2        8.9
3         5        4.5
8         9        2.5
10        15       2.3
18        19       4.0 
20        20       3.2

Group = 3,
Op = Min

IDstart   IDend    Value
1         3        4.5
5         9        2.1
10        18       1.0
19        20       3.2

Group defines how many rows to combine into a single row (in the new table) and operation defines what operation to do on the group of rows, the operations I need are maximum, minimum and average.
The last group may contain less rows than all the rest of the groups. if the last group has a single value IDstart = IDEnd. ID is unique but may have ‘gaps’.

I’m looking for the fastest way to do this, any help will be appreciated.

Using SQL Server 2008 R2

Gilad.

Answer 1

Reasoning goes like this

• Using the ROW_NUMBER() function and some arithmetic allows you to create a dummy column placing each ID in a group of the size you specify.
• the result of this statement can be grouped and the operator you specify can be applied using a CASE statement. Should you need additional operators, you would only have to expand this CASE statement.

Script

DECLARE @Group INTEGER
DECLARE @Op VARCHAR(3)

SET @Group = 3
SET @Op = 'MIN'

;WITH q(ID, Value) AS (
  SELECT 1,     6.7
  UNION ALL SELECT 2,     8.9
  UNION ALL SELECT 3,     4.5
  UNION ALL SELECT 5,     3.2
  UNION ALL SELECT 8,     2.5
  UNION ALL SELECT 9,     2.1
  UNION ALL SELECT 10,    1.0
  UNION ALL SELECT 15,    2.3
  UNION ALL SELECT 18,    2.4
  UNION ALL SELECT 19,    4.0
  UNION ALL SELECT 20,    3.2
)
SELECT  [IDStart] = MIN(ID)
        , [IDEnd] = MAX(ID)
        , [Value] = CASE  WHEN @Op = 'MAX' THEN MAX(Value)
                          WHEN @Op = 'MIN' THEN MIN(Value)
                          WHEN @Op = 'AVG' THEN AVG(Value)
                    END
FROM    (
          SELECT ID
                 , Value
                 , GroupRow = (ROW_NUMBER() OVER (ORDER BY ID) - 1) / @Group
          FROM    q
        ) q          
GROUP BY
        GroupRow