3
CREATE table comp_tb 
(
a tinyint
)


insert into comp_tb values('4')
insert into comp_tb values('1')
insert into comp_tb values('5')
insert into comp_tb values('6')
insert into comp_tb values('10')
insert into comp_tb values('3')
insert into comp_tb values('2')
insert into comp_tb values('8')


SELECT * from comp_tb ct --as order of insert 

create NONCLUSTERED INDEX NCI_a ON comp_tb(a)

SELECT * from comp_tb ct --sorts in acending order

drop INDEX NCI_a ON comp_tb

create CLUSTERED INDEX CI_a ON comp_tb(a)

SELECT * from comp_tb ct -- sorts in acending order

drop INDEX CI_a ON comp_tb

So both clustered index as well as non-clustered index sorts data physically as soon as it created?

As per my reading only clustered index does physically sorting?

Registered User
  • 1,554
  • 3
  • 22
  • 37

2 Answers2

10

The non clustered index does not sort the data on disk like a clustered index.

But the query optimiser generates a query plan that scans this non-clustered index, which gives the data in the order of the index. This happens because the index exactly matches the table: one column in both table and index. SO it uses the index and you apparent ordering of data.

If you add more columns so that the index is useless for a full table scan, a query plan is generated that scans the actual heap data. The index is not used.

CREATE table #comp_tb 
(
a tinyint,
payload1 char(3000) NOT NULL,
payload2 varchar(100) NOT NULL,
)

insert into #comp_tb values('4', '4' /*padded*/, '44444444444')
insert into #comp_tb values('1', '1' /*padded*/, '1111111111111111111111111111')
insert into #comp_tb values('5', '5' /*padded*/, '55')
insert into #comp_tb values('6', '6' /*padded*/, '666666666666666666666666666666666666')
insert into #comp_tb values('10', '10' /*padded*/, 'a')
insert into #comp_tb values('3', '3' /*padded*/, '3333333')
insert into #comp_tb values('2', '2' /*padded*/, REPLICATE('2', 50))
insert into #comp_tb values('8', '8' /*padded*/, '8888888888888888888888')

SELECT * from #comp_tb ct --as order of insert 

create NONCLUSTERED INDEX NCI_a ON #comp_tb(a DESC)

SELECT * from #comp_tb ct --as order of insert 

drop INDEX NCI_a ON #comp_tb
gbn
  • 422,506
  • 82
  • 585
  • 676
2

Physically, table data is arranged only by clustered index. Different indexes are used depending on query plan. In case non-clustered index is used, data is retrieved not from a table and, as a result, you see that the data is ordered.

Query Plan

Devart
  • 119,203
  • 23
  • 166
  • 186