Nested Transaction!

The following example is given in BOL under the topic 'Nested
Transaction':
---
CREATE PROCEDURE TransProc @.PriKey INT, @.CharCol CHAR(3)
AS
BEGIN TRANSACTION InProc
INSERT INTO TestTrans VALUES (@.PriKey, @.CharCol)
INSERT INTO TestTrans VALUES (@.PriKey + 1, @.CharCol)
COMMIT TRANSACTION InProc
GO
/* Start a transaction and execute TransProc */
BEGIN TRANSACTION OutOfProc
GO
EXEC TransProc 1, 'aaa'
GO
/* Roll back the outer transaction,this will roll back TransProc's
nested transaction */
ROLLBACK TRANSACTION OutOfProc
GO
EXECUTE TransProc 3,'bbb'
GO
/* The following SELECT statement shows only rows 3 and 4 are
still in the table. This indicates that the commit of the inner
transaction from the first EXECUTE statement of TransProc was
overridden by the subsequent rollback. */
SELECT * FROM TestTrans
GO
---
The stored procedure 'TransProc' has only 1 transaction named 'InProc'.
Apart from this, there is another transaction named 'OutOfProc' (which,
if I am not mistaken, isn't a part & parcel of the stored procedure
'TransProc' since 'TransProc' ends at the first 'GO' statement) i.e.
the transactions 'InProc' & 'OutOfProc' are 2 distinct transactions. So
where or which is the nested transaction? Shouldn't a nested
transaction have 1 transaction under another transaction something like
this (similar to nested If...Else conditions):
---
CREATE PROCEDURE TransProc @.PriKey INT, @.CharCol CHAR(3)
AS
BEGIN TRANSACTION InProc
--Do Something
BEGIN TRANSACTION InProc1
--Do Something More
BEGIN TRANSACTION InProc2
--Do Something More
COMMIT TRANSACTION InProc
---
Or does the presence of more than 1 SQL statement (the 2 INSERT
queries) within the transaction 'InProc' (in the stored procedure
'TransProc') make it a nested transaction?
Thanks,
ArpanA begin tran must have either a commit/rollback tran. Thus, the TransProc's
Inproc transaction is participating in the OutProc transaction. If
commit/rollback tran OutProc is not explicitly called and the connection to
the server is dropped, the transaction will be forced to rollback.
As stated in bol, if outer most commit/rollback is what really important. It
decides the final commit all or rollback all.
-oj
"Arpan" <arpan_de@.hotmail.com> wrote in message
news:1123306747.855496.16630@.f14g2000cwb.googlegroups.com...
> The following example is given in BOL under the topic 'Nested
> Transaction':
> ---
> CREATE PROCEDURE TransProc @.PriKey INT, @.CharCol CHAR(3)
> AS
> BEGIN TRANSACTION InProc
> INSERT INTO TestTrans VALUES (@.PriKey, @.CharCol)
> INSERT INTO TestTrans VALUES (@.PriKey + 1, @.CharCol)
> COMMIT TRANSACTION InProc
> GO
> /* Start a transaction and execute TransProc */
> BEGIN TRANSACTION OutOfProc
> GO
> EXEC TransProc 1, 'aaa'
> GO
> /* Roll back the outer transaction,this will roll back TransProc's
> nested transaction */
> ROLLBACK TRANSACTION OutOfProc
> GO
> EXECUTE TransProc 3,'bbb'
> GO
> /* The following SELECT statement shows only rows 3 and 4 are
> still in the table. This indicates that the commit of the inner
> transaction from the first EXECUTE statement of TransProc was
> overridden by the subsequent rollback. */
> SELECT * FROM TestTrans
> GO
> ---
> The stored procedure 'TransProc' has only 1 transaction named 'InProc'.
> Apart from this, there is another transaction named 'OutOfProc' (which,
> if I am not mistaken, isn't a part & parcel of the stored procedure
> 'TransProc' since 'TransProc' ends at the first 'GO' statement) i.e.
> the transactions 'InProc' & 'OutOfProc' are 2 distinct transactions. So
> where or which is the nested transaction? Shouldn't a nested
> transaction have 1 transaction under another transaction something like
> this (similar to nested If...Else conditions):
> ---
> CREATE PROCEDURE TransProc @.PriKey INT, @.CharCol CHAR(3)
> AS
> BEGIN TRANSACTION InProc
> --Do Something
> BEGIN TRANSACTION InProc1
> --Do Something More
> BEGIN TRANSACTION InProc2
> --Do Something More
> COMMIT TRANSACTION InProc
> ---
> Or does the presence of more than 1 SQL statement (the 2 INSERT
> queries) within the transaction 'InProc' (in the stored procedure
> 'TransProc') make it a nested transaction?
> Thanks,
> Arpan
>|||Thanks, OJ, for your help. How stupid of me....actually I had
misinterpreted the example given in BOL :-)
Thanks once again,
Regards,
Arpan|||Transactions are not really nested. There can only be one outstanding
transaction context for a connection. The global variable @.@.TRANCOUNT
reports whether or not a transaction context is outstanding, and its value
immediately before a Transact-SQL statement begins executing determines
whether or not a new transaction context is initiated. When @.@.TRANCOUNT =
0, no transaction context is outstanding for the connection. Each time a
BEGIN TRANSACTION is executed, @.@.TRANCOUNT is incremented. Only when
@.@.TRANCOUNT = 0 immediately before a BEGIN TRANSACTION statement is a new
transaction started by that BEGIN TRANSACTION statement. Each sucessive
BEGIN TRANSACTION increments @.@.TRANCOUNT. Each COMMIT TRANSACTION
decrements @.@.TRANCOUNT. Only when @.@.TRANCOUNT = 1 immediately before a
COMMIT TRANSACTION statement are changes made within the transaction made
permanent by that COMMIT TRANSACTION statement. All Transact-SQL statements
that alter the state or schema of the database operate within the context of
a transaction. If a transaction context is not already outstanding, then a
new transaction context is initiated before executing the statement. If
IMPLICIT_TRANSACTIONS is OFF, then changes made by the statement are
committed immediately after the statement completes. This is called
"autocommit" mode. If IMPLICIT_TRANSACTIONS is ON, then an explicit COMMIT
WORK must be issued to commit the transaction. A transaction can span
multiple statements, multiple stored procedure calls, even multiple batches.
If a transaction is outstanding, dynamic SQL executed via either EXEC() or
sp_executesql executes within that transaction context. Since changes made
while a transaction context is outstanding are not made permanent until
they're committed, ROLLBACK backs all of the changes for the entire
transaction context. The only exception is when a savepoint is specified on
a ROLLBACK statement. SAVE TRANSACTION places a marker in the transaction
log that identifies a reference point which can be specified in a ROLLBACK
statement to partially backout changes made while a transaction context is
outstanding. When a ROLLBACK savepoint statement is executed, all changes
made after the save point are backed out, and the transaction context
remains outstanding.
"Arpan" <arpan_de@.hotmail.com> wrote in message
news:1123306747.855496.16630@.f14g2000cwb.googlegroups.com...
> The following example is given in BOL under the topic 'Nested
> Transaction':
> ---
> CREATE PROCEDURE TransProc @.PriKey INT, @.CharCol CHAR(3)
> AS
> BEGIN TRANSACTION InProc
> INSERT INTO TestTrans VALUES (@.PriKey, @.CharCol)
> INSERT INTO TestTrans VALUES (@.PriKey + 1, @.CharCol)
> COMMIT TRANSACTION InProc
> GO
> /* Start a transaction and execute TransProc */
> BEGIN TRANSACTION OutOfProc
> GO
> EXEC TransProc 1, 'aaa'
> GO
> /* Roll back the outer transaction,this will roll back TransProc's
> nested transaction */
> ROLLBACK TRANSACTION OutOfProc
> GO
> EXECUTE TransProc 3,'bbb'
> GO
> /* The following SELECT statement shows only rows 3 and 4 are
> still in the table. This indicates that the commit of the inner
> transaction from the first EXECUTE statement of TransProc was
> overridden by the subsequent rollback. */
> SELECT * FROM TestTrans
> GO
> ---
> The stored procedure 'TransProc' has only 1 transaction named 'InProc'.
> Apart from this, there is another transaction named 'OutOfProc' (which,
> if I am not mistaken, isn't a part & parcel of the stored procedure
> 'TransProc' since 'TransProc' ends at the first 'GO' statement) i.e.
> the transactions 'InProc' & 'OutOfProc' are 2 distinct transactions. So
> where or which is the nested transaction? Shouldn't a nested
> transaction have 1 transaction under another transaction something like
> this (similar to nested If...Else conditions):
> ---
> CREATE PROCEDURE TransProc @.PriKey INT, @.CharCol CHAR(3)
> AS
> BEGIN TRANSACTION InProc
> --Do Something
> BEGIN TRANSACTION InProc1
> --Do Something More
> BEGIN TRANSACTION InProc2
> --Do Something More
> COMMIT TRANSACTION InProc
> ---
> Or does the presence of more than 1 SQL statement (the 2 INSERT
> queries) within the transaction 'InProc' (in the stored procedure
> 'TransProc') make it a nested transaction?
> Thanks,
> Arpan
>|||Dude, paragraph breaks :)

> Transactions are not really nested. There can only be one outstanding
> transaction context for a connection.
I don't know that this particularly means that transactions aren't nested.
By your definition IF..THEN statements are not really nested. I don't think
that nested transactions implies any technical innerworkings more than it
just simply implies that you can syntactically do:
BEGIN TRANSACTION
BEGIN TRANSACTION
BEGIN TRANSACTION
COMMIT TRANSACTION
COMMIT TRANSACTION
COMMIT TRANSACTION
It is more or less meaningless to us whether a stack is used or a counter
and whether or not ROLLBACK kills the whole stack or just goes back to the
original point. Either way I still consider them nested because of syntax.
----
Louis Davidson - http://spaces.msn.com/members/drsql/
SQL Server MVP
"Brian Selzer" <brian@.selzer-software.com> wrote in message
news:OdCa%23klmFHA.1088@.TK2MSFTNGP14.phx.gbl...
> Transactions are not really nested. There can only be one outstanding
> transaction context for a connection. The global variable @.@.TRANCOUNT
> reports whether or not a transaction context is outstanding, and its value
> immediately before a Transact-SQL statement begins executing determines
> whether or not a new transaction context is initiated. When @.@.TRANCOUNT =
> 0, no transaction context is outstanding for the connection. Each time a
> BEGIN TRANSACTION is executed, @.@.TRANCOUNT is incremented. Only when
> @.@.TRANCOUNT = 0 immediately before a BEGIN TRANSACTION statement is a new
> transaction started by that BEGIN TRANSACTION statement. Each sucessive
> BEGIN TRANSACTION increments @.@.TRANCOUNT. Each COMMIT TRANSACTION
> decrements @.@.TRANCOUNT. Only when @.@.TRANCOUNT = 1 immediately before a
> COMMIT TRANSACTION statement are changes made within the transaction made
> permanent by that COMMIT TRANSACTION statement. All Transact-SQL
> statements
> that alter the state or schema of the database operate within the context
> of
> a transaction. If a transaction context is not already outstanding, then
> a
> new transaction context is initiated before executing the statement. If
> IMPLICIT_TRANSACTIONS is OFF, then changes made by the statement are
> committed immediately after the statement completes. This is called
> "autocommit" mode. If IMPLICIT_TRANSACTIONS is ON, then an explicit
> COMMIT
> WORK must be issued to commit the transaction. A transaction can span
> multiple statements, multiple stored procedure calls, even multiple
> batches.
> If a transaction is outstanding, dynamic SQL executed via either EXEC() or
> sp_executesql executes within that transaction context. Since changes
> made
> while a transaction context is outstanding are not made permanent until
> they're committed, ROLLBACK backs all of the changes for the entire
> transaction context. The only exception is when a savepoint is specified
> on
> a ROLLBACK statement. SAVE TRANSACTION places a marker in the transaction
> log that identifies a reference point which can be specified in a ROLLBACK
> statement to partially backout changes made while a transaction context is
> outstanding. When a ROLLBACK savepoint statement is executed, all changes
> made after the save point are backed out, and the transaction context
> remains outstanding.
>
> "Arpan" <arpan_de@.hotmail.com> wrote in message
> news:1123306747.855496.16630@.f14g2000cwb.googlegroups.com...
>|||There is a difference. For example:
IF cond1
THEN IF cond2
THEN IF cond3
THEN stmt1
ELSE stmt2
Which condition does this ELSE go with? Here's another example:
for (int i = 1; i < 10; i++)
for (int j = 1; j < 10; j++)
for (int k = 1; k < 10; k++)
{
Console.WriteLine(i + j + k);
if (i + j + k == 55) break;
}
Which for does the break go with?
One of the common misunderstandings about transactions is that since there m
ust always be a matching commit transaction for every begin transaction, and
that transactions can be "nested," many newbies erroneously assume that a r
ollback only backs out the innermost block of statements--that is, to the po
int of the innermost BEGIN TRANSACTION. The assumption follows from the pat
tern etched in their brain by the second example. A break exits the innermo
st block; therefore, it is logical to assume (erroneously, of course) that a
rollback backs out the innermost transaction.
The important thing to remember is that a transaction context either exists
or it doesn't. It doesn't matter how many begin/commit pairs exist in the b
lock of code bounded by the outermost begin/commit pair, the transaction con
text is initiated by the first BEGIN TRANSACTION and is terminated either by
any ROLLBACK or by the outermost COMMIT.
Notice the pretty paragraph breaks in the above :)
"Louis Davidson" <dr_dontspamme_sql@.hotmail.com> wrote in message news:#7aRUwomFHA.2484@.TK2
MSFTNGP15.phx.gbl...
> Dude, paragraph breaks :)
>
>
> I don't know that this particularly means that transactions aren't nested.
> By your definition IF..THEN statements are not really nested. I don't thi
nk
> that nested transactions implies any technical innerworkings more than it
> just simply implies that you can syntactically do:
>
> BEGIN TRANSACTION
> BEGIN TRANSACTION
> BEGIN TRANSACTION
> COMMIT TRANSACTION
> COMMIT TRANSACTION
> COMMIT TRANSACTION
>
> It is more or less meaningless to us whether a stack is used or a counter
> and whether or not ROLLBACK kills the whole stack or just goes back to the
> original point. Either way I still consider them nested because of syntax
.
>
> --
> ----
--
> Louis Davidson - http://spaces.msn.com/members/drsql/
> SQL Server MVP
>
>
> "Brian Selzer" <brian@.selzer-software.com> wrote in message
> news:OdCa%23klmFHA.1088@.TK2MSFTNGP14.phx.gbl...
>
>|||I like to think of it as:
"Nested transactions are allowed in syntax but not semantics."
I don't know what "real" nested transactions mean, or if there is a formal d
efinition of what nested
transaction semantics means? :
Perhaps a rollback of an inner transaction would allow commit of an outer tr
ansaction? We can do
that with savepoints.
Or the other way: A commit of an inner transaction will still be committed i
f the outer transaction
does a rollback? True, we don't have this in SQL server (which, I believe, B
rian wished for in an
earlier post in some other thread). To some extent, we can work around it wi
th table variables or
opening a new connection.
Tibor Karaszi, SQL Server MVP
http://www.karaszi.com/sqlserver/default.asp
http://www.solidqualitylearning.com/
Blog: http://solidqualitylearning.com/blogs/tibor/
"Louis Davidson" <dr_dontspamme_sql@.hotmail.com> wrote in message
news:%237aRUwomFHA.2484@.TK2MSFTNGP15.phx.gbl...
> Dude, paragraph breaks :)
>
> I don't know that this particularly means that transactions aren't nested.
By your definition
> IF..THEN statements are not really nested. I don't think that nested tran
sactions implies any
> technical innerworkings more than it just simply implies that you can synt
actically do:
> BEGIN TRANSACTION
> BEGIN TRANSACTION
> BEGIN TRANSACTION
> COMMIT TRANSACTION
> COMMIT TRANSACTION
> COMMIT TRANSACTION
> It is more or less meaningless to us whether a stack is used or a counter
and whether or not
> ROLLBACK kills the whole stack or just goes back to the original point. E
ither way I still
> consider them nested because of syntax.
> --
> ----
--
> Louis Davidson - http://spaces.msn.com/members/drsql/
> SQL Server MVP
>
> "Brian Selzer" <brian@.selzer-software.com> wrote in message
> news:OdCa%23klmFHA.1088@.TK2MSFTNGP14.phx.gbl...
>|||But the one break can get us out of several scope operators. I see your
point, but as long as you have to commit them one at a time, I think the
common term nested for transactions is probably going to stick...
And you code still looks pretty good in plain text :)
----
Louis Davidson - http://spaces.msn.com/members/drsql/
SQL Server MVP
"Brian Selzer" <brian@.selzer-software.com> wrote in message
news:u9My2XpmFHA.4028@.TK2MSFTNGP10.phx.gbl...
There is a difference. For example:
IF cond1
THEN IF cond2
THEN IF cond3
THEN stmt1
ELSE stmt2
Which condition does this ELSE go with? Here's another example:
for (int i = 1; i < 10; i++)
for (int j = 1; j < 10; j++)
for (int k = 1; k < 10; k++)
{
Console.WriteLine(i + j + k);
if (i + j + k == 55) break;
}
Which for does the break go with?
One of the common misunderstandings about transactions is that since there
must always be a matching commit transaction for every begin transaction,
and that transactions can be "nested," many newbies erroneously assume that
a rollback only backs out the innermost block of statements--that is, to the
point of the innermost BEGIN TRANSACTION. The assumption follows from the
pattern etched in their brain by the second example. A break exits the
innermost block; therefore, it is logical to assume (erroneously, of course)
that a rollback backs out the innermost transaction.
The important thing to remember is that a transaction context either exists
or it doesn't. It doesn't matter how many begin/commit pairs exist in the
block of code bounded by the outermost begin/commit pair, the transaction
context is initiated by the first BEGIN TRANSACTION and is terminated either
by any ROLLBACK or by the outermost COMMIT.
Notice the pretty paragraph breaks in the above :)
"Louis Davidson" <dr_dontspamme_sql@.hotmail.com> wrote in message
news:#7aRUwomFHA.2484@.TK2MSFTNGP15.phx.gbl...
> Dude, paragraph breaks :)
>
> I don't know that this particularly means that transactions aren't nested.
> By your definition IF..THEN statements are not really nested. I don't
> think
> that nested transactions implies any technical innerworkings more than it
> just simply implies that you can syntactically do:
> BEGIN TRANSACTION
> BEGIN TRANSACTION
> BEGIN TRANSACTION
> COMMIT TRANSACTION
> COMMIT TRANSACTION
> COMMIT TRANSACTION
> It is more or less meaningless to us whether a stack is used or a counter
> and whether or not ROLLBACK kills the whole stack or just goes back to the
> original point. Either way I still consider them nested because of
> syntax.
> --
> ----
--
> Louis Davidson - http://spaces.msn.com/members/drsql/
> SQL Server MVP
>
> "Brian Selzer" <brian@.selzer-software.com> wrote in message
> news:OdCa%23klmFHA.1088@.TK2MSFTNGP14.phx.gbl...
>|||No doubt we could use a more convienient model for how transactions work,
but your statement is great.
----
Louis Davidson - http://spaces.msn.com/members/drsql/
SQL Server MVP
"Tibor Karaszi" <tibor_please.no.email_karaszi@.hotmail.nomail.com> wrote in
message news:uDRWzWqmFHA.1968@.TK2MSFTNGP14.phx.gbl...
>I like to think of it as:
> "Nested transactions are allowed in syntax but not semantics."
>
> I don't know what "real" nested transactions mean, or if there is a formal
> definition of what nested transaction semantics means? :
> Perhaps a rollback of an inner transaction would allow commit of an outer
> transaction? We can do that with savepoints.
> Or the other way: A commit of an inner transaction will still be committed
> if the outer transaction does a rollback? True, we don't have this in SQL
> server (which, I believe, Brian wished for in an earlier post in some
> other thread). To some extent, we can work around it with table variables
> or opening a new connection.
> --
> Tibor Karaszi, SQL Server MVP
> http://www.karaszi.com/sqlserver/default.asp
> http://www.solidqualitylearning.com/
> Blog: http://solidqualitylearning.com/blogs/tibor/
>
> "Louis Davidson" <dr_dontspamme_sql@.hotmail.com> wrote in message
> news:%237aRUwomFHA.2484@.TK2MSFTNGP15.phx.gbl...
>

Nested Select?

Hi all,
I have the following table
id (autonumber)
category1 (int)
category2 (int)
booking_month (int)
booking_year (int)
I have records in the table for booking_year = 2004 and booking_year = 2005,
for example
id, category1, category2, booking_month, booking_year
1, 20, 30, 4, 2004
1, 20, 31, 10, 2004
1, 20, 30, 4, 2005
I need a SQL statement there lists all those records that are in 2004 but no
in 2005 for a particular category1.
Any ideas?
Thanks,
IvanYou could use a nested sub-query, however, you can also just select from
[booking] as B04 for booking_year = 2004 and then left join [booking] as B05
on booking_year = 2005. Only include records where B05.id is NULL.
"Ivan Debono" <ivanmdeb@.hotmail.com> wrote in message
news:uW4JVI3DFHA.548@.TK2MSFTNGP14.phx.gbl...
> Hi all,
> I have the following table
> id (autonumber)
> category1 (int)
> category2 (int)
> booking_month (int)
> booking_year (int)
> I have records in the table for booking_year = 2004 and booking_year =
2005,
> for example
> id, category1, category2, booking_month, booking_year
> 1, 20, 30, 4, 2004
> 1, 20, 31, 10, 2004
> 1, 20, 30, 4, 2005
> I need a SQL statement there lists all those records that are in 2004 but
no
> in 2005 for a particular category1.
> Any ideas?
> Thanks,
> Ivan
>|||"Ivan Debono" <ivanmdeb@.hotmail.com> wrote in
news:uW4JVI3DFHA.548@.TK2MSFTNGP14.phx.gbl:

> Hi all,
> I have the following table
> id (autonumber)
> category1 (int)
> category2 (int)
> booking_month (int)
> booking_year (int)
> id, category1, category2, booking_month, booking_year
> 1, 20, 30, 4, 2004
> 1, 20, 31, 10, 2004
> 1, 20, 30, 4, 2005
> I need a SQL statement there lists all those records that are in 2004
> but no in 2005 for a particular category1.
SELECT * FROM [tablename]
WHERE (booking_year <> 2005) AND (category1 = particular_value)
HTH|||This does not work:
SELECT o1.category2 FROM offline o1
LEFT JOIN offline o2
ON o2.booking_year = 2005
WHERE (o1.category1 = 989 AND o1.booking_month = 4 AND o1.booking_year =
2004)
and o2.id_no is null
"JohnnyAppleseed" <someone@.microsoft.com> schrieb im Newsbeitrag
news:%23qJsJP3DFHA.1188@.tk2msftngp13.phx.gbl...
> You could use a nested sub-query, however, you can also just select from
> [booking] as B04 for booking_year = 2004 and then left join [booking] as
B05
> on booking_year = 2005. Only include records where B05.id is NULL.
> "Ivan Debono" <ivanmdeb@.hotmail.com> wrote in message
> news:uW4JVI3DFHA.548@.TK2MSFTNGP14.phx.gbl...
> 2005,
but
> no
>|||Or like this
SELECT * FROM [tablename]
WHERE booking_year =2004 AND category1 not in (select category1 from
[tablename] t1 where year = 2005)
Hth
"Chris Cheney" <cjc1@.nospam%ucs.cam.ac.uk%no%spam%please> wrote in message
news:Xns95F9884DB1851cjc1nospamucscamacu
k@.131.111.8.48...
> "Ivan Debono" <ivanmdeb@.hotmail.com> wrote in
> news:uW4JVI3DFHA.548@.TK2MSFTNGP14.phx.gbl:
>
> SELECT * FROM [tablename]
> WHERE (booking_year <> 2005) AND (category1 = particular_value)
> HTH|||The compare between category1 and category2 should be in the join.. on..
clause. Try this:
select
B04.booking_year,
B04.category1,
B04.category2
from
offline as B04
left join
offline as B05
-- join both aliases of offline on category1 and category2. also
filter B05 on 2005.
on B05.booking_year = 2005 and
B05.category1 = B04.category1 and
B05.category2 = B04.category 2
where
B04.booking_year = 2004 and
B05.id is NULL -- Doesn't matter which B05 column is null
"Ivan Debono" <ivanmdeb@.hotmail.com> wrote in message
news:OLdjle3DFHA.3120@.TK2MSFTNGP12.phx.gbl...
> This does not work:
> SELECT o1.category2 FROM offline o1
> LEFT JOIN offline o2
> ON o2.booking_year = 2005
> WHERE (o1.category1 = 989 AND o1.booking_month = 4 AND o1.booking_year =
> 2004)
> and o2.id_no is null
> "JohnnyAppleseed" <someone@.microsoft.com> schrieb im Newsbeitrag
> news:%23qJsJP3DFHA.1188@.tk2msftngp13.phx.gbl...
> B05
> but
>|||The original B04 records total 112.
The original B05 records total 24.
Logic and simple math tell me that I should get 112-24 = 88!!
When I run your suggested statement I get 85 records.
When I run this statement:
SELECT category2 FROM offline WHERE category1 = 989 AND booking_month = 4
AND booking_year = 2004 AND category2 NOT IN
(SELECT category2FROM offline WHERE category1 = 989 AND booking_month = 4
AND booking_year = 2005)
I get 87 records.
Strange indeed!!!
"JohnnyAppleseed" <someone@.microsoft.com> schrieb im Newsbeitrag
news:ubdN7t3DFHA.2232@.TK2MSFTNGP14.phx.gbl...
> The compare between category1 and category2 should be in the join.. on..
> clause. Try this:
> select
> B04.booking_year,
> B04.category1,
> B04.category2
> from
> offline as B04
> left join
> offline as B05
> -- join both aliases of offline on category1 and category2. also
> filter B05 on 2005.
> on B05.booking_year = 2005 and
> B05.category1 = B04.category1 and
> B05.category2 = B04.category 2
> where
> B04.booking_year = 2004 and
> B05.id is NULL -- Doesn't matter which B05 column is null
>
> "Ivan Debono" <ivanmdeb@.hotmail.com> wrote in message
> news:OLdjle3DFHA.3120@.TK2MSFTNGP12.phx.gbl...
from
as
=
2004
>|||This was my original statement that I tried but I get always 1 less than the
expected result :(
"AM" <shahdharti@.gmail.com> schrieb im Newsbeitrag
news:e8PTwj3DFHA.1496@.TK2MSFTNGP14.phx.gbl...
> Or like this
> SELECT * FROM [tablename]
> WHERE booking_year =2004 AND category1 not in (select category1 from
> [tablename] t1 where year = 2005)
> Hth
>
> "Chris Cheney" <cjc1@.nospam%ucs.cam.ac.uk%no%spam%please> wrote in message
> news:Xns95F9884DB1851cjc1nospamucscamacu
k@.131.111.8.48...
>|||Ivan,
Can you be more precise about what you want? What does
"in 2004 but no in 2005" mean? I assume it means
Find all rows where booking_year = 2004 but for which
there is not a matching row with booking_year = 2005.
But ... you haven't made it clear what "matching row" means. Does
a matching 2005 row need to have the same category1, category2,
and booking_month, or just some of those columns? And your
data here says id is "autonumber", but you show three identical id
values - does the matching 2005 row have to have the same id value?
You said you think you should get a number of rows that is
the number of 2004 rows minus the number of 2005 rows, but
how do you know every one of the 2005 rows in your table
matches exactly one 2004 row? Maybe some 2004 rows appear
twice, and maybe some 2005 rows have no corresponding 2004
row.
You may know what it means for a 2004 row to be in 2005 as
well, but unless you describe it clearly in terms of the columns of
this table, you can't expect to be able to write a query that will
give you what you want.
Steve Kass
Drew University
Ivan Debono wrote:

>Hi all,
>I have the following table
>id (autonumber)
>category1 (int)
>category2 (int)
>booking_month (int)
>booking_year (int)
>I have records in the table for booking_year = 2004 and booking_year = 2005
,
>for example
>id, category1, category2, booking_month, booking_year
>1, 20, 30, 4, 2004
>1, 20, 31, 10, 2004
>1, 20, 30, 4, 2005
>I need a SQL statement there lists all those records that are in 2004 but n
o
>in 2005 for a particular category1.
>Any ideas?
>Thanks,
>Ivan
>
>|||Perhaps you should be joining on booking_month too ?
"Ivan Debono" <ivanmdeb@.hotmail.com> wrote in message
news:OzdKJc4DFHA.3536@.TK2MSFTNGP15.phx.gbl...
> The original B04 records total 112.
> The original B05 records total 24.
> Logic and simple math tell me that I should get 112-24 = 88!!
> When I run your suggested statement I get 85 records.
> When I run this statement:
> SELECT category2 FROM offline WHERE category1 = 989 AND booking_month = 4
> AND booking_year = 2004 AND category2 NOT IN
> (SELECT category2FROM offline WHERE category1 = 989 AND booking_month = 4
> AND booking_year = 2005)
> I get 87 records.
> Strange indeed!!!
> "JohnnyAppleseed" <someone@.microsoft.com> schrieb im Newsbeitrag
> news:ubdN7t3DFHA.2232@.TK2MSFTNGP14.phx.gbl...
=
> from
[booking]
> as
booking_year
> =
> 2004
>

Negative unused space

I wrote simple script to check space used by tables:

CREATE TABLE #SpaceUsed(
TableName NVARCHAR(128),
NoOfRows INT,
Reserved NVARCHAR(18),
Data NVARCHAR(18),
Index_Size NVARCHAR(18),
Unused NVARCHAR(18)
)
GO
sp_msforeachtable "INSERT INTO #SpaceUsed EXEC sp_spaceused '?'"

SELECT * FROM #SpaceUsed

SELECT
CAST(Sum(CAST(Replace(Reserved,' KB','') AS INT)) AS NVARCHAR) + ' KB' AS TotalReserved,
CAST(Sum(CAST(Replace(Data,' KB','') AS INT)) AS NVARCHAR) + ' KB' AS TotalData,
CAST(Sum(CAST(Replace(Index_Size,' KB','') AS INT)) AS NVARCHAR) + ' KB' AS TotalIndex_Size,
CAST(Sum(CAST(Replace(Unused,' KB','') AS INT)) AS NVARCHAR) + ' KB' AS TotalUnused
FROM #SpaceUsed

DROP TABLE #SpaceUsed

and one of results looks strange to me:

TableName NoOfRows Reserved Data Index_Size Unused
--------------------------------------- ---- ------ ------ ------ ------
T_TableXX 50081 38024 KB 37432 KB 640 KB -48 KB

Anyone know reason of such result (negative value of unused space)?Use this :

CREATE TABLE #SpaceUsed(
TableName NVARCHAR(128),
NoOfRows INT,
Reserved NVARCHAR(18),
Data NVARCHAR(18),
Index_Size NVARCHAR(18),
Unused NVARCHAR(18)
)
GO
sp_msforeachtable "INSERT INTO #SpaceUsed EXEC sp_spaceused '?' ,@.updateusage='True'"

SELECT * FROM #SpaceUsed

SELECT
CAST(Sum(CAST(Replace(Reserved,' KB','') AS INT)) AS NVARCHAR) + ' KB' AS TotalReserved,
CAST(Sum(CAST(Replace(Data,' KB','') AS INT)) AS NVARCHAR) + ' KB' AS TotalData,
CAST(Sum(CAST(Replace(Index_Size,' KB','') AS INT)) AS NVARCHAR) + ' KB' AS TotalIndex_Size,
CAST(Sum(CAST(Replace(Unused,' KB','') AS INT)) AS NVARCHAR) + ' KB' AS TotalUnused
FROM #SpaceUsed

DROP TABLE #SpaceUsed|||Works fine. Thanx :)|||Originally posted by MST78
Anyone know reason of such result (negative value of unused space)? It happends all of the time if you don't regularly update your statistics. You can update a single table using UPDATE STATISTICS (http://msdn.microsoft.com/library/en-us/tsqlref/ts_ua-uz_1mpf.asp), or get them all at once if you have the time using DBCC UPDATEUSAGE (http://msdn.microsoft.com/library/en-us/tsqlref/ts_dbcc_24rp.asp).

-PatP

Needs Another help

hello,
i wanted to know to write bello with out subqueries to increase my
application performance
CREATE TABLE COURSE (
COURSESLNO INT,
COURSENAME VARCHAR)
INSERT INTO COURSE VALUES (1,'JR');
INSERT INTO COURSE VALUES (2,'SR');
INSERT INTO COURSE VALUES (3,'LT');
INSERT INTO COURSE VALUES (4,'ST');
CREATE TABLE BRANCH_COURSE(
BRANCHNAME VARCHAR,
COURSESLNO INT)
INSERT INTO BRANCH_COURSE VALUES ('BR1', 1);
INSERT INTO BRANCH_COURSE VALUES ('BR1', 2);
INSERT INTO BRANCH_COURSE VALUES ('BR1', 3);
INSERT INTO BRANCH_COURSE VALUES ('BR1', 4);
INSERT INTO BRANCH_COURSE VALUES ('BR2', 1);
INSERT INTO BRANCH_COURSE VALUES ('BR2', 3);
INSERT INTO BRANCH_COURSE VALUES ('BR3', 2);
INSERT INTO BRANCH_COURSE VALUES ('BR3', 3);
INSERT INTO BRANCH_COURSE VALUES ('BR4', 4);
and i want
BRANCHNAME COURSESLNO COURSENAME
BR1 1 JR
BR1 2 SR
BR1 3 LT
BR1 4 ST
BR2 1 JR
BR2 NULL NULL
BR2 3 LT
BR2 NULL NULL
BR3 NULL NULL
BR3 2 SR
BR3 3 LT
BR3 NULL NULL
BR4 NULL NULL
BR4 NULL NULL
BR4 NULL NULL
BR4 4 ST
thx a lot
*** Sent via Developersdex http://www.examnotes.net ***I assume that you also have a table for branches?
create table branches(branchname varchar(10) not null primary key);
insert into branches(branchname) values('br1');
insert into branches(branchname) values('br2');
insert into branches(branchname) values('br3');
insert into branches(branchname) values('br4');
If so, use:
select b.branchname, bc.courseslno,
case when bc.courseslno is null then null else c.coursename end as
coursename
from course as c
cross join branches as b
left outer join branch_course as bc
on bc.courseslno = c.courseslno
and bc.branchname = b.branchname
order by b.branchname, c.courseslno;
Otherwise simply cross with a distinct list of branches from branch_course.
BG, SQL Server MVP
www.SolidQualityLearning.com
"kamal hussain" <skkamalh@.yahoo.co.in> wrote in message
news:e2cTLu1lFHA.3144@.TK2MSFTNGP12.phx.gbl...
> hello,
> i wanted to know to write bello with out subqueries to increase my
> application performance
>
> CREATE TABLE COURSE (
> COURSESLNO INT,
> COURSENAME VARCHAR)
>
> INSERT INTO COURSE VALUES (1,'JR');
> INSERT INTO COURSE VALUES (2,'SR');
> INSERT INTO COURSE VALUES (3,'LT');
> INSERT INTO COURSE VALUES (4,'ST');
>
> CREATE TABLE BRANCH_COURSE(
> BRANCHNAME VARCHAR,
> COURSESLNO INT)
>
> INSERT INTO BRANCH_COURSE VALUES ('BR1', 1);
> INSERT INTO BRANCH_COURSE VALUES ('BR1', 2);
> INSERT INTO BRANCH_COURSE VALUES ('BR1', 3);
> INSERT INTO BRANCH_COURSE VALUES ('BR1', 4);
> INSERT INTO BRANCH_COURSE VALUES ('BR2', 1);
> INSERT INTO BRANCH_COURSE VALUES ('BR2', 3);
> INSERT INTO BRANCH_COURSE VALUES ('BR3', 2);
> INSERT INTO BRANCH_COURSE VALUES ('BR3', 3);
> INSERT INTO BRANCH_COURSE VALUES ('BR4', 4);
>
> and i want
> BRANCHNAME COURSESLNO COURSENAME
> BR1 1 JR
> BR1 2 SR
> BR1 3 LT
> BR1 4 ST
> BR2 1 JR
> BR2 NULL NULL
> BR2 3 LT
> BR2 NULL NULL
> BR3 NULL NULL
> BR3 2 SR
> BR3 3 LT
> BR3 NULL NULL
> BR4 NULL NULL
> BR4 NULL NULL
> BR4 NULL NULL
> BR4 4 ST
>
> thx a lot
>
> *** Sent via Developersdex http://www.examnotes.net ***|||try this query
SELECT T.BRANCHNAME, BC.COURSESLNO, CASE WHEN BC.COURSESLNO IS NULL THEN
NULL ELSE C.COURSENAME END FROM
(
SELECT DISTINCT BRANCHNAME
FROM BRANCH_COURSE
) T CROSS JOIN COURSE C
LEFT OUTER JOIN BRANCH_COURSE BC ON C.COURSESLNO = BC.COURSESLNO AND
T.BRANCHNAME = BC.BRANCHNAME
ORDER BY T.BRANCHNAME
ph
"kamal hussain" wrote:

> hello,
> i wanted to know to write bello with out subqueries to increase my
> application performance
>
> CREATE TABLE COURSE (
> COURSESLNO INT,
> COURSENAME VARCHAR)
>
> INSERT INTO COURSE VALUES (1,'JR');
> INSERT INTO COURSE VALUES (2,'SR');
> INSERT INTO COURSE VALUES (3,'LT');
> INSERT INTO COURSE VALUES (4,'ST');
>
> CREATE TABLE BRANCH_COURSE(
> BRANCHNAME VARCHAR,
> COURSESLNO INT)
>
> INSERT INTO BRANCH_COURSE VALUES ('BR1', 1);
> INSERT INTO BRANCH_COURSE VALUES ('BR1', 2);
> INSERT INTO BRANCH_COURSE VALUES ('BR1', 3);
> INSERT INTO BRANCH_COURSE VALUES ('BR1', 4);
> INSERT INTO BRANCH_COURSE VALUES ('BR2', 1);
> INSERT INTO BRANCH_COURSE VALUES ('BR2', 3);
> INSERT INTO BRANCH_COURSE VALUES ('BR3', 2);
> INSERT INTO BRANCH_COURSE VALUES ('BR3', 3);
> INSERT INTO BRANCH_COURSE VALUES ('BR4', 4);
>
> and i want
> BRANCHNAME COURSESLNO COURSENAME
> BR1 1 JR
> BR1 2 SR
> BR1 3 LT
> BR1 4 ST
> BR2 1 JR
> BR2 NULL NULL
> BR2 3 LT
> BR2 NULL NULL
> BR3 NULL NULL
> BR3 2 SR
> BR3 3 LT
> BR3 NULL NULL
> BR4 NULL NULL
> BR4 NULL NULL
> BR4 NULL NULL
> BR4 4 ST
>
> thx a lot
>
> *** Sent via Developersdex http://www.examnotes.net ***
>|||Hi
here is the query that u can use:
SELECT BRANCHNAME, CASE WHEN result = 0 THEN NULL ELSE COURSESLNO END,
CASE WHEN result = 0 THEN NULL ELSE COURSENAME END
FROM
(
SELECT TOP 100 PERCENT B.BranchName, C.COURSESLNO , C.COURSENAME,
SUM(CASE WHEN C.COURSESLNO=B.COURSESLNO THEN B.COURSESLNO ELSE 0 END) result
FROM BRANCH_COURSE B
CROSS JOIN COURSE C
LEFT OUTER JOIN COURSE C1 ON B.COURSESLNO = C.COURSESLNO
GROUP BY B.BranchName, C.COURSESLNO , C.COURSENAME
ORDER BY B.BranchName
)Der
Please let me know if this worked.
best Regards,
Chandra
http://chanduas.blogspot.com/
http://groups.msn.com/SQLResource/
---
"kamal hussain" wrote:

> hello,
> i wanted to know to write bello with out subqueries to increase my
> application performance
>
> CREATE TABLE COURSE (
> COURSESLNO INT,
> COURSENAME VARCHAR)
>
> INSERT INTO COURSE VALUES (1,'JR');
> INSERT INTO COURSE VALUES (2,'SR');
> INSERT INTO COURSE VALUES (3,'LT');
> INSERT INTO COURSE VALUES (4,'ST');
>
> CREATE TABLE BRANCH_COURSE(
> BRANCHNAME VARCHAR,
> COURSESLNO INT)
>
> INSERT INTO BRANCH_COURSE VALUES ('BR1', 1);
> INSERT INTO BRANCH_COURSE VALUES ('BR1', 2);
> INSERT INTO BRANCH_COURSE VALUES ('BR1', 3);
> INSERT INTO BRANCH_COURSE VALUES ('BR1', 4);
> INSERT INTO BRANCH_COURSE VALUES ('BR2', 1);
> INSERT INTO BRANCH_COURSE VALUES ('BR2', 3);
> INSERT INTO BRANCH_COURSE VALUES ('BR3', 2);
> INSERT INTO BRANCH_COURSE VALUES ('BR3', 3);
> INSERT INTO BRANCH_COURSE VALUES ('BR4', 4);
>
> and i want
> BRANCHNAME COURSESLNO COURSENAME
> BR1 1 JR
> BR1 2 SR
> BR1 3 LT
> BR1 4 ST
> BR2 1 JR
> BR2 NULL NULL
> BR2 3 LT
> BR2 NULL NULL
> BR3 NULL NULL
> BR3 2 SR
> BR3 3 LT
> BR3 NULL NULL
> BR4 NULL NULL
> BR4 NULL NULL
> BR4 NULL NULL
> BR4 4 ST
>
> thx a lot
>
> *** Sent via Developersdex http://www.examnotes.net ***
>

need urgent help in T-SQL

-- TOP MEDIAN
BEGIN
DECLARE @.medvarcnt int
DECLARE @.medianValue float
DECLARE @.medianfield varchar(255)
DECLARE @.SQLSTR Nvarchar(800)
SET @.medianfield = 'Cluster_Top'

CREATE TABLE #medianlist (rid int IDENTITY(1,1), medianval int)

SET @.SQLSTR = ('INSERT #medianlist SELECT ' + @.medianfield + ' AS medianval FROM ' + @.result_table_name + ' ORDER BY

' + @.medianfield + ' DESC')
SET @.SQLSTR = CAST (@.SQLSTR AS NVARCHAR(800))
EXECUTE sp_executesql @.SQLSTR
SET @.medvarcnt = (SELECT COUNT(*) FROM #medianlist)

IF @.medvarcnt % 2 = 0
BEGIN
--even
line17 set @.medianValue = (SELECT SUM(medianval)/2 FROM #medianlist WHERE rid >=(@.medvarcnt/2) and rid <=

(@.medvarcnt/2)+1)
line19 set @.sql = 'Update ' + @.result_table_statistic + ' set Top_Median = ' + CAST(@.medianValue AS NVARCHAR(20))
set @.sql = @.sql + ' Where Testcell = ''' + @.testcell + ''' '
print(@.medianValue) --exec(@.sql)
END
ELSE
BEGIN
--odd
set @.medianValue = (SELECT medianval FROM #medianlist where rid =(@.medvarcnt/2)+1)
set @.sql = 'Update ' + @.result_table_statistic + ' set Top_Median = ' + CAST(@.medianValue AS NVARCHAR(20))
set @.sql = @.sql + ' Where Testcell = ''' + @.testcell + ''' '
print(@.medianValue) --exec(@.sql)
END
DROP TABLE #medianlist

END

[addedon]March 17, 2007, 7:31 pm[/addedon]i'm trying to create a median solution with stored procedure...and from the coding i post above, i encounter the error stating failure to change from varchar to float everytime i execute it...

i suspect its happen on line 17 and 18 ...T-SQL got it as varchar...and when i use it back as variable in line 19 (whereas it suppose to take it as a float value....) thus error occur....

make it simple...@.medianvalue should be an int (let's say 18) in line 17 and 18...but at line 19 ,system still take it as whole sentence in varchar...hope everyone can understand what i try to tell...

anyone expert can provide me with solution? thanx

Try this as a replacement for Line 19:

set @.sql = 'Update ' + @.result_table_statistic + ' set Top_Median = ' + CAST(@.medianValue AS NVARCHAR(20))

..and here's an explanation as to why you should do this:

http://msdn2.microsoft.com/en-us/library/ms190309.aspx

You need to review the code sample that you provided for further occurrences of attempting to append an INT to an NVARCHAR - with a quick glance I spotted one more.

Chris

|||

yes, thanks for your solution, chris...it works perfectly right now.

i should be more observant next time.

|||

there's a problem again after undergone some testing...

i expect @.medianvalue to be equals to 5.50 (should be accurate until 2 decimal) but the result coming out is shown as 5.

is there anything wrong with my coding? i have try CAST the SUM value into FLOAT but it doesnt work...

For more information, data type of the field Top_Median that i'm going to update is FLOAT.

any help is very much appreciated.

|||

Could it be the datatype specified in this line that's causing the problem?
CREATE TABLE #medianlist (rid int IDENTITY(1,1), medianval int)
Chris

|||i've changed it to type float. This solve the problem , thanks again Chris.