Run this script
create table #table1(steps int)
insert into #table1(steps)values(1)
insert into #table1(steps)values(3)
insert into #table1(steps)values(4)
insert into #table1(steps)values(5)

select steps from #table1
drop table #table1

You will get this
steps
1
3
4
5

I want this way.
stepFrom    stepTo
1                3
3                4
4                5
Hints: you need to join the same table multiple times.

Here you go NepaliBhai......................


create table #table1(steps int)
insert into #table1(steps)values(1)
insert into #table1(steps)values(3)
insert into #table1(steps)values(4)
insert into #table1(steps)values(5)

select distinct x.stepFrom, Y.stepTo from
(select rank()over(order by a.steps )as rank,a.steps as stepFrom from (select top 3 * from #table1 order by 1 asc) a
left outer join #table1 b
on a.steps =b.steps) x
inner join
(select rank()over(order by a.steps )as rank,a.steps as stepTo from (select top 3 * from #table1 order by 1 desc) a
left join #table1 b
on a.steps =b.steps) y
on x.rank =y.rank

drop table #table1

SQL_PRO, i don't think you should be using rank function. Can you try using only ANSI SQL :)?
Btw, if you are using sql 2012 this thing can be done with ease
Select * from
(select steps as StepsFrom,lead(steps) over (order by steps) as StepsTo from #table1 ) t1
where StepsTo is not null

But ANSI SQL would be more interesting :)

SQL_PRO Dude,
It doesn't work that way. You can't hard code like top 3. It should be dynamic.

What if the data is like this.
insert into #table1(steps)values(1)
insert into #table1(steps)values(2)
insert into #table1(steps)values(3)
insert into #table1(steps)values(4)
insert into #table1(steps)values(5)
insert into #table1(steps)values(7)
insert into #table1(steps)values(9)
insert into #table1(steps)values(10)
insert into #table1(steps)values(17)
insert into #table1(steps)values(31)

So the solution is like this

go
WITH TEMP_CTE AS (
SELECT steps, ROW_NUMBER() OVER(ORDER BY steps) AS ROW_NUM FROM #table1
)
SELECT t1.steps, t2.steps FROM TEMP_CTE t1, TEMP_CTE t2
WHERE t1.ROW_NUM < (SELECT MAX(ROW_NUM) FROM TEMP_CTE)
AND (t2.ROW_NUM - 1) = t1.ROW_NUM

Here is another puzzle.
create table #table1(steps int)
insert into #table1(steps)values(1)
insert into #table1(steps)values(3)
insert into #table1(steps)values(4)
insert into #table1(steps)values(5)
Now generate random rows for these rows.
e.g
1
3,
5,
4
or
5,
3,
1,
4

Remember, you need to produce random rows every time you run it.

I like your idea. It will be fun to work on. I will try to come up with syntax.

Here it is.

create table #table1(steps int)
insert into #table1(steps)values(1)
insert into #table1(steps)values(3)
insert into #table1(steps)values(4)
insert into #table1(steps)values(5)

SELECT cast(steps as varchar) + ',' FROM #table1 ORDER BY NEWID()

drop table #table1