Question

i want to select first 10 rows then select 5 random rows of them.

Answer 1

Have to define “first 10” and it’s “something” in this snippet. This is for SQL Server 2000+

select top 5 
    * 
from 
    (
    select top 10 * from <table>
    order by something --got to have soemthing here to define "first 10"
    ) as a
order by 
    newid()

Edit:

Why you need ORDER BY in the derived table

-- large row, same result with and without index/PK
CREATE TABLE #foo (bar int /*PRIMARY KEY NONCLUSTERED (bar)*/, eejit char(8000))

--create rows with value 1-10 + some others
INSERT #foo (bar) VALUES (1)
INSERT #foo (bar)  VALUES (10)
INSERT #foo (bar)  VALUES (20)
INSERT #foo (bar)  VALUES (2)
INSERT #foo (bar)  VALUES (5)
INSERT #foo (bar)  VALUES (45)
INSERT #foo (bar)  VALUES (99)
INSERT #foo (bar)  VALUES (3)
INSERT #foo (bar)  VALUES (9)
INSERT #foo (bar)  VALUES (7)
INSERT #foo (bar)  VALUES (6)
INSERT #foo (bar)  VALUES (4)
INSERT #foo (bar)  VALUES (8)

--create logical fragmentation
DELETE #foo WHERE bar IN (1, 3, 5, 7, 9)
INSERT #foo (bar) VALUES (1)
INSERT #foo (bar) VALUES (3)
INSERT #foo (bar) VALUES (5)
INSERT #foo (bar) VALUES (7)
INSERT #foo (bar) VALUES (9)

-- run this a few times, you will see values > 10
-- "first 10" surely means values between 1 and 10?
select top 5 
    * 
from 
    (
    select top 10 * from #foo
    ) as a
order by 
    newid()

-- always <= 10 because of ORDER BY
select top 5 
    * 
from 
    (
    select top 10 * from #foo
    order by bar --got to have soemthing here to define "first 10"
    ) as a
order by 
    newid()