Question

Let’s assume there is a SQL Server 2008 table like below, that holds 10 million rows.
One of the fields is Id, since it’s identity it is from 1 to 10 million.

CREATE TABLE dbo.Stats
(
   id      INT IDENTITY(1,1) PRIMARY KEY,
   field1  INT,
   field2  INT,
   ...
)

Is there an efficient way by doing one select statement to get a subset of this data that satisfies the following requirements:

• contains a limited number of rows in the result set, i.e. 100, 200, etc.
• provides equal distribution of a certain column, not random, i.e. of column id

So, in our example, if we return 100 rows, the result set would look like this:

Row 1 - 100 000
Row 2 - 200 000
Row 3 - 300 000
...
Row 100 - 10 000 000

I want to avoid using cursor and storing this in a separate table.

Answer 1

Not sure how efficient it’s going to be, but thie following query will return every 100000th row (relative to ordering established by id):

SELECT *
FROM (
    SELECT *, ROW_NUMBER() OVER (ORDER BY id) RN
    FROM Stats
) T
WHERE RN % 100000 = 0
ORDER BY id

Since it does not rely on actual id values, this will work even if you have “holes” in the sequence of id values.