automatic sequence number by id

Hi,
Is it possible to have SQL server automatically generate a sequence number
based on another column, both forming the table's primary key. So another
kind of auto-increment field.
What is mean is something like this:
Code Seq Name ...
A100 1 a
A100 2 b
A100 3 c
G432 1 x
G432 2 y
H008 1 p
H008 2 q
H008 3 r
...
Thanks a lot for your help.
EdgarHi
In SQL 2005 when retrieving data you can use the ROWNUMBER function see
http://msdn2.microsoft.com/en-us/library/ms189798.aspx, but you could not
store them. You could use a subquery when inserting the records
e.g.
CREATE TABLE mytable ( [Code] CHAR(4) NOT NULL, [Seq] INT NOT NULL,
[Name]
CHAR(1) )
INSERT INTO MyTable ( [Code], [Seq], [Name] )
SELECT 'A100', 1, 'a'
UNION ALL SELECT 'A100', 2, 'b'
UNION ALL SELECT 'A100', 3, 'c'
UNION ALL SELECT 'G432', 1, 'x'
UNION ALL SELECT 'G432', 2, 'y'
UNION ALL SELECT 'H008', 1, 'p'
UNION ALL SELECT 'H008', 2, 'q'
SELECT * FROM MyTable
INSERT INTO MyTable ( [Code], [Seq], [Name] )
SELECT 'H008', ISNULL( ( SELECT COUNT(*)+1 FROM mytable WHERE [Code]
=
'H008'),0), 'r'
SELECT * FROM MyTable
INSERT INTO MyTable ( [Code], [Seq], [Name] )
SELECT 'H010', ISNULL( ( SELECT COUNT(*)+1 FROM mytable WHERE [Code]
=
'H010'),0), 'g'
SELECT * FROM MyTable
This could be incorporated into an INSTEAD OF TRIGGER
John
"Edgar" wrote:

> Hi,
> Is it possible to have SQL server automatically generate a sequence number
> based on another column, both forming the table's primary key. So another
> kind of auto-increment field.
> What is mean is something like this:
> Code Seq Name ...
> A100 1 a
> A100 2 b
> A100 3 c
> G432 1 x
> G432 2 y
> H008 1 p
> H008 2 q
> H008 3 r
> ...
>
> Thanks a lot for your help.
> Edgar|||Why does this data need to be stored, when you could always retrieve Seq at
query time?
The problem with storing it in the table is that now it has to be
maintained. DELETE table WHERE Code = 'A100' AND Name = 'a' and now you are
mising Seq=1 for that combination. If A100 has 80,000 rows and you need to
decrease all of their Seq values by 1, that becomes a very, very, very
expensive delete operation.
A
"Edgar" <Edgar@.discussions.microsoft.com> wrote in message
news:8178F22E-B631-413B-8BF1-32CB0E6B06B3@.microsoft.com...
> Hi,
> Is it possible to have SQL server automatically generate a sequence number
> based on another column, both forming the table's primary key. So another
> kind of auto-increment field.
> What is mean is something like this:
> Code Seq Name ...
> A100 1 a
> A100 2 b
> A100 3 c
> G432 1 x
> G432 2 y
> H008 1 p
> H008 2 q
> H008 3 r
> ...
>
> Thanks a lot for your help.
> Edgar|||I need the sequence numbers, because they indicate the order of the records
related to their parent record.
Thanks,
Edgar
"Aaron Bertrand [SQL Server MVP]" wrote:

> Why does this data need to be stored, when you could always retrieve Seq a
t
> query time?
> The problem with storing it in the table is that now it has to be
> maintained. DELETE table WHERE Code = 'A100' AND Name = 'a' and now you a
re
> mising Seq=1 for that combination. If A100 has 80,000 rows and you need t
o
> decrease all of their Seq values by 1, that becomes a very, very, very
> expensive delete operation.
> A
>
> "Edgar" <Edgar@.discussions.microsoft.com> wrote in message
> news:8178F22E-B631-413B-8BF1-32CB0E6B06B3@.microsoft.com...
>
>|||"Edgar" <Edgar@.discussions.microsoft.com> wrote in message
news:ADBF8CBC-78F0-4E89-AF0A-318BC806A56B@.microsoft.com...
>I need the sequence numbers, because they indicate the order of the records
> related to their parent record.
>
> --
Ok, that's legit. You can just use an IDENTITY column for the sequence
numbers. They won't be sequential, and they won't start over for each
parent, but they will give you the relative ordering
EG
Code Seq Name ...
A100 1132 a
A100 1314 b
A100 5991 c
G432 7202 x
G432 82929 y
H008 1002 p
H008 89231 q
H008 999231 r
David|||David,
Thanks for your help. Good suggestion.
I will create the normal 1, 2, 3 when i retrieve the data.
(But it would be a nice addition to the product :-))
Thanks,
Edgar
"David Browne" wrote:

>
> "Edgar" <Edgar@.discussions.microsoft.com> wrote in message
> news:ADBF8CBC-78F0-4E89-AF0A-318BC806A56B@.microsoft.com...
>
> Ok, that's legit. You can just use an IDENTITY column for the sequence
> numbers. They won't be sequential, and they won't start over for each
> parent, but they will give you the relative ordering
> EG
> Code Seq Name ...
> A100 1132 a
> A100 1314 b
> A100 5991 c
> G432 7202 x
> G432 82929 y
> H008 1002 p
> H008 89231 q
> H008 999231 r
> David
>|||Hi Edgar
If you do that then the sequencing may not reflect the true order in which
they were inserted, for example if an entry is deleted subsequent entries
will be moved up. If you are ok with this then using the identity is ok, you
may also want to only allocate the sequence number on the client which would
save you doing the subquery.
John
"Edgar" wrote:
[vbcol=seagreen]
> David,
> Thanks for your help. Good suggestion.
> I will create the normal 1, 2, 3 when i retrieve the data.
> (But it would be a nice addition to the product :-))
> Thanks,
> Edgar
>
> "David Browne" wrote:
>