I know that the SCOPE_IDENTITY() will get the last inserted row from insert statement. However, for the following case, I am not too sure is SCOPE_IDENTITY() is safe. As SELECT MAX(ID) FROM TableA will have go through scan the table to get the max id and it will have performance issue, even slightly, I believe.
Here is the case:
DECLARE #DaysInMonth INT
DECLARE #FirstID INT
DECLARE #SecondID INT
DECLARE #ThirdID INT
DECLARE #FourthID INT
SET #DaysInMonth = DAY(EOMONTH('2016-09-01'))
BEGIN TRY
BEGIN TRANSACTION
WHILE #DaysInMonth > 0
BEGIN
-- First Insert -- Begin
INSERT INTO tableA ('first insert - ' + #DaysInMonth)
-- First Insert -- End
SET #FirstID = SCOPE_IDENTITY()
-- Second Insert -- Begin
INSERT INTO tableB ('second insert - ' + #DaysInMonth)
-- Second Insert -- End
SET #SecondID = SCOPE_IDENTITY()
-- Third Insert -- Begin
INSERT INTO tableC ('third insert - ' + #DaysInMonth)
-- Third Insert -- End
SET #ThirdID = SCOPE_IDENTITY()
-- Fourth Insert -- Begin
INSERT INTO tableD ('fourth insert - ' + #DaysInMonth)
-- Fourth Insert -- End
SET #FourthID = SCOPE_IDENTITY()
SET #DaysInMonth = #DaysInMonth - 1
END
COMMIT TRANSACTION
END TRY
BEGIN CATCH
IF ##TRANCOUNT > 0
ROLLBACK TRANSACTION
THROW
END CATCH
As from the case above, I have to insert the records every loop for fourth times for how many days in the month that I have declared.
From what I know, there are 4 to get the last inserted ID:
SCOPE_IDENTITY
##IDENTITY
SELECT MAX(ID) FROM tableA
IDENT_CURRENT
From the following post:
Post
Is mentioned that SCOPE_IDENTITY() is what generally that you want to use.
What I mean with 'Safe' is, do the ID will be unique during the loop?
Thank you.
You can use OUTPUT column in the last insert statement, Ofcourse this is another option where you will get what exactly input statement executed.. Below is just an example
CREATE TABLE #tablea (
id int IDENTITY (1, 1),
val char(10)
)
DECLARE #outputtbl TABLE (
id int,
val char(10)
)
INSERT INTO #tablea (val)
OUTPUT INSERTED.* INTO #outputtbl
VALUES ('test')
SELECT id
FROM #outputtbl
Related
Not sure what I'm missing. When I debug and step through the INSERT query I've included below, I see that '%a%' is the value of #Answer, and 103 is the value for #ItemId.
IF EXISTS is always evaluating to false when I insert the values shown beneath:
CREATE TRIGGER TR_cc_Additional_Information_Answers_INS
ON cc_Additional_Information_Answers
AFTER INSERT
AS
BEGIN
CREATE TABLE temp_answers
(
TempAnswer VARCHAR(50),
TempAdditional_Information_ItemID INT
)
INSERT INTO temp_answers (TempAnswer, TempAdditional_Information_ItemID)
SELECT Description, Additional_Information_ItemID
FROM inserted
DECLARE #Answer varchar(50)
SELECT #Answer = '''%' + t.TempAnswer + '%''' FROM temp_answers t
DECLARE #ItemId int
SELECT #ItemId = t.TempAdditional_Information_ItemID FROM temp_answers t
IF EXISTS(SELECT 1
FROM cc_Additional_Information_Answers a
WHERE a.Description LIKE #Answer
AND a.Additional_Information_ItemID = #ItemId)
BEGIN
RAISERROR('Answer is too similar to pre-existing answers for this item', 16, 1)
ROLLBACK TRANSACTION
RETURN
END
DROP TABLE temp_answers
END
GO
And this is my insert query:
INSERT INTO cc_Additional_Information_Answers (Additional_Information_ItemID, Description)
VALUES (103, 'a')
And the pre-existing record:
Thanks in advance, SQL community!
EDIT: this also does not behave as expected. . .
INSERT INTO cc_Additional_Information_Answers (Additional_Information_ItemID, Description)
VALUES (103, 'a')
Given this data
Your IF EXISTS will always evaluate to true because the inserted value is already inserted (although it can be rolled back) when the trigger runs (it's an "AFTER" trigger).
So you will want to inspect only those records that existed in the table before the insertion. I always use an outer join for this. Also: I would never create a table in a trigger. The following should work as expected:
CREATE TRIGGER TR_cc_Additional_Information_Answers_INS ON cc_Additional_Information_Answers
AFTER INSERT
AS
BEGIN
IF EXISTS(
SELECT 1 FROM cc_Additional_Information_Answers a
LEFT OUTER JOIN inserted i ON a.Additional_Information_AnswerID = i.Additional_Information_AnswerID
INNER JOIN inserted temp ON a.Additional_Information_ItemID = temp.Additional_Information_ItemID
WHERE a.Description LIKE '%' + temp.Description + '%'
AND i.Additional_Information_AnswerID IS NULL
)
BEGIN
RAISERROR('Answer is too similar to pre-existing answers for this item', 16, 1)
ROLLBACK TRANSACTION
RETURN
END
END
GO
I created the following two test tables with a trigger to log all the action (Insert, Delete and Update).
Set up tables and trigger:
-- drop table test; drop table testLog
create table test (id int identity primary key, x int);
create table testLog (idx int identity primary key, Action varchar(10), id int not null,
x_deleted int, x_inserted int, uid uniqueidentifier);
go
-- Trigger to log the changes
create trigger trigger_test on test
after insert, delete, update
as
declare #id uniqueidentifier = context_info();
print #id;
insert testLog (id, Action, x_deleted, x_inserted, uid)
select isnull(d.id, i.id) ,
case when i.id is not null and d.id is not null then 'Updated'
when d.id is not null then 'Deleted'
when i.id is not null then 'Inserted'
end ,
d.x ,
i.x ,
#id
from Deleted d
full outer join inserted i on i.id = d.id;
set context_info 0;
go
Now insert some sample data
set context_info 0
insert test (x) values (10), (20), (30), (40), (50);
SELECT * FROM test;
SELECT * FROM testLog
go
The following statements work fine. The correct context_info() is saved in the log table.
begin tran
declare #newid uniqueidentifier = newid()
--
set context_info #newid
print #newid
insert test(x) values (1)
set context_info #newid
update test set x = 2 where id = 1
SELECT * FROM dbo.testLog;
rollback
go
However, only insert part of the Merge got the value in context_info()?
begin tran
declare #newid uniqueidentifier = newid()
--
set context_info #newid
print #newid;
with v as (select * from (values (1, 11), (2, 22), (6, 66)) v (id, x))
merge test as t using v on t.id = v.id
when matched then update set x = v.x
when not matched by target then insert (x) values (x);
SELECT * FROM dbo.testLog;
rollback
go
The uid of the last two updates got zeros.
Don't set context_info to zero in the trigger. Why would you do that in the first place - it is not the trigger's responsibility to "clean up". The merge statement will cause the trigger to execute for inserts separately from updates. Did you not notice the multiple "prints" in the results pane? That should have been a big clue.
I have an SSIS package which will first run my sp_doSomething. This stored procedure will select data from several different tables and join them for possible storage into dbo.someTable. But I only want that IF the select is > 1 row of selected data.
I want to then have a precedence restraint that looks at the amount of rows my stored procedure returned.
If my row count > 1, then I want to take the results of the stored procedure and insert them into one of my tables.
Otherwise, I will record an error/send an email, or whatever.
I really don't want to run this stored procedure more then once, but that is the only way I could think to do it (Run it, count the rows. Then, run it again and insert the result).
I'm a complete TSQL/SSIS newb. So I'm sorry if this question is trivial.
I can't find a good answer anywhere.
Create a variable with Package Scope of type Int32 and name rowcount.
Data Flow
Control Flow
you can try this
declare #tableVar table(col1 varchar(100))
declare #Counter int
insert into #tableVar(col1) exec CompanyNames
set #Counter = (select count(*) from #tableVar)
insert into Anytable(col) Values (#counter)
Within the Stored Proc, write the results to a #Temp. Then Select Count(*) from the #Temp, into a variable.
Select #intRows = Count(*) from myTempResults
Then evaluate the value of #intRows.
If #intRows > 1 BEGIN
Insert Into dbo.SomeTable
Select * from #Temp
End
Will a #temp table work for you?
IF OBJECT_ID('tempdb..#Holder') IS NOT NULL
begin
drop table #Holder
end
CREATE TABLE #Holder
(ID INT )
declare #MyRowCount int
declare #MyTotalCount int = 0
/* simulate your insert, you would read from your real table(s) here */
INSERT INTO #HOLDER (ID)
select 1 union all select 2 union all select 3 union all select 4
Select #MyRowCount = ##ROWCOUNT, #MyTotalCount = #MyTotalCount + #MyRowCount
Select 'TheMagicValue1' = #MyRowCount, 'TheMagicTotal' = #MyTotalCount
INSERT INTO #HOLDER (ID)
select 5 union all select 6 union all select 7 union all select 8
/* you will note that I am NOT doing a count(*) here... which is another strain on the procedure */
Select #MyRowCount = ##ROWCOUNT, #MyTotalCount = #MyTotalCount + #MyRowCount
Select 'TheMagicValue1' = #MyRowCount, 'TheMagicTotal' = #MyTotalCount
/* Optional index if needed */
CREATE INDEX IDX_TempHolder_ID ON #Holder (ID)
/* CREATE CLUSTERED INDEX IDX_TempHolder_ID ON #Holder (ID) */
if #MyTotalCount > 0
BEGIN
Select 'Put your INSERT statement here'
END
/* this will return the data to the report */
Select ID from #HOLDER
IF OBJECT_ID('tempdb..#Holder') IS NOT NULL
begin
drop table #Holder
end
I have some SQL within a stored procedure where I am updating a table based on another SELECT statement from a temp table (code below).
SET NOCOUNT ON
DECLARE #RowCount int
UPDATE TABLEX SET
TRA = ISNULL (ir.DcTra, DCBASIC.TRA),
TRD = ISNULL(CAST(NULLIF(REPLACE(ir.DcTRD, '-', ''), '') AS datetime), DCBASIC.TRD),
LSINC = ISNULL(ir.DcLsInc, DCBASIC.LSINC),
REVSWOVR = ISNULL(ir.DcRevswovr, DCBASIC.REVSWOVR) FROM #TempData ir WHERE TABLEX.MEMBNO = ir.IntMembNo
SET #RowCount = ##ROWCOUNT
The #RowCount variable is being set to 1.
The SELECT of the #TempData table returns no rows and no rows in the TABLEX table are updated (or even exist) with the MembNo (I have added SELECT statements within the sp to debug and they confirm this)
Why is #RowCount being set to 1?
Here is an explanation:
Statements that make a simple assignment always set the ##ROWCOUNT value to 1.
More information you can find here:
##ROWCOUNT
My example:
CREATE DATABASE FirstDB
GO
USE FirstDB;
GO
CREATE TABLE Person (
personId INT IDENTITY PRIMARY KEY,
firstName varchar(20) ,
lastName varchar(20) ,
age int
)
INSERT INTO dbo.Person (firstName, lastName, age)
VALUES ('Nick', 'Smith', 30),
('Jack', 'South', 25),
('Garry', 'Perth', 20)
CREATE TABLE PersonAge (
personAgeId INT IDENTITY PRIMARY KEY ,
personId INT ,
newAge varchar(10)
)
INSERT INTO dbo.PersonAge(personId, newAge)
VALUES (1, 60),
(2, 65),
(3, 70)
ALTER TABLE dbo.PersonAge
ADD CONSTRAINT FK_PersonAgePerson FOREIGN KEY (personId)
REFERENCES dbo.Person (personId)
And then example of query:
USE FirstDB;
GO
SET NOCOUNT ON;
DECLARE #row int;
UPDATE Person
SET age = 40
FROM dbo.Person as p join dbo.PersonAge as p1
ON p.personId = p1.personId
WHERE p.age = 60
SET #row = ##ROWCOUNT
SELECT #row
I create an UPDATE query where none of rows will be affected.
At the end #row consist 0 value.
Here is another example, using INSERT and DELETE--
DECLARE #deletedRows INT = 0;
SELECT #deletedRows = ##ROWCOUNT; --no previous DML statement
SELECT #deletedRows; --##ROWCOUNT = 1 for a simple assignment
GO
DROP TABLE IF EXISTS #Test;
GO
CREATE TABLE #Test (ID INT IDENTITY, CurrentDate DATETIME DEFAULT GETDATE());
GO
INSERT #Test DEFAULT VALUES; --INSERT a single row
DECLARE #deletedRows INT = ##ROWCOUNT; --##ROWCOUNT = 1
SELECT #deletedRows;
GO
DELETE FROM #Test WHERE 1=2; --no rows deleted
DECLARE #deletedRows INT = ##ROWCOUNT; --##ROWCOUNT = 0
SELECT #deletedRows;
GO
DELETE TOP (1) t FROM #Test t WHERE 1=1; --1 row deleted
DECLARE #deletedRows INT = ##ROWCOUNT; --##ROWCOUNT = 1
SELECT #deletedRows;
GO
DELETE TOP (1) t FROM #Test t WHERE 1=1; --no rows left to delete
DECLARE #deletedRows INT = ##ROWCOUNT; --##ROWCOUNT = 0
SELECT #deletedRows;
GO
I am a little stuck with why I can not seem to get the 'new identity' of the inserted row with the statement below. SCOPE_IDENTITY() just returns null.
declare #WorkRequestQueueID int
declare #LastException nvarchar(MAX)
set #WorkRequestQueueID = 1
set #LastException = 'test'
set nocount off
DELETE dbo.WorkRequestQueue
OUTPUT
DELETED.MessageEnvelope,
DELETED.Attempts,
#LastException,
GetUtcdate(), -- WorkItemPoisened datetime
DELETED.WorkItemReceived_UTC
INTO dbo.FaildMessages
FROM dbo.WorkRequestQueue
WHERE
WorkRequestQueue.ID = #WorkRequestQueueID
IF ##ROWCOUNT = 0
RAISERROR ('Record not found', 16, 1)
SELECT Cast(SCOPE_IDENTITY() as int)
Any assistance would be most appreciated.
For now I use a workaround this like so.
declare #WorkRequestQueueID int
declare #LastException nvarchar(MAX)
set #WorkRequestQueueID = 7
set #LastException = 'test'
set nocount on
set xact_abort on
DECLARE #Failed TABLE
(
MessageEnvelope xml,
Attempts smallint,
LastException nvarchar(max),
WorkItemPoisened_UTC datetime,
WorkItemReceived_UTC datetime
)
BEGIN TRAN
DELETE dbo.WorkRequestQueue
OUTPUT
DELETED.MessageEnvelope,
DELETED.Attempts,
#LastException,
GetUtcdate(), -- WorkItemPoisened datetime
DELETED.WorkItemReceived_UTC
INTO
#Failed
FROM
dbo.WorkRequestQueue
WHERE
WorkRequestQueue.ID = #WorkRequestQueueID
IF ##ROWCOUNT = 0 BEGIN
RAISERROR ('Record not found', 16, 1)
Rollback
END ELSE BEGIN
insert into dbo.FaildMessages select * from #Failed
COMMIT TRAN
SELECT Cast(SCOPE_IDENTITY() as int)
END
EDITED FEB'2013
#MartinSmith alerts us that this bug don't want be fixed by Microsoft.
"Posted by Microsoft on 2/27/2013 at 2:18 PM Hello Martin, We
investigated the issue and found that changing the behavior is not an
easy thing to do. It would basically require redefining some of the
behavior when both INSERT & OUTPUT INTO target has identity columns.
Given the nature of the problem & the uncommon scenario, we have
decided not to fix the issue. -- Umachandar, SQL Programmability
Team"
EDITED OCT'2012
This is caused by a bug:
Testing bug:
Quoting OUTPUT Clause doc:
##IDENTITY, SCOPE_IDENTITY, and IDENT_CURRENT return identity values
generated only by the nested DML statement, and not those generated by
the outer INSERT statement.
After test it It seems that scope_identity() only works if outer operation is an insert in a table with identity columns:
Test 1: Delete
create table #t ( a char(1) );
create table #d ( a char(1), i int identity );
insert into #t
values ('a'),('b'),('c');
delete #t
output deleted.a into #d;
select SCOPE_IDENTITY(), * from #d;
a i
---- - -
null a 1
null b 2
null c 3
Test 2: Inserting in outer table with identity
create table #t ( a char(1), i int identity );
create table #d ( a char(1), i int identity );
insert into #t
values ('x'),('x'),('x');
insert into #t
output inserted.a into #d
values ('a'),('b');
select scope_identity(), * from #d;
a i
- - -
2 a 1
2 b 2
Test 3: Inserting in outer table without identity
create table #t ( a char(1) );
create table #d ( a char(1), i int identity );
insert into #t
values ('x'),('x'),('x');
insert into #t
output inserted.a into #d
values ('a'),('b');
select scope_identity(), * from #d;
a i
---- - -
null a 1
null b 2
You might try to use a table variable for your output clause, thus allowing you to explicitly insert into FaildMessages:
declare #WorkRequestQueueID int
declare #LastException nvarchar(MAX)
set #WorkRequestQueueID = 1
set #LastException = 'test'
set nocount off
-- Declare a table variable to capture output
DECLARE #output TABLE (
MessageEnvelope VARCHAR(50), -- Guessing at datatypes
Attempts INT, -- Guessing at datatypes
WorkItemReceived_UTC DATETIME -- Guessing at datatypes
)
-- Run the deletion with output
DELETE dbo.WorkRequestQueue
OUTPUT
DELETED.MessageEnvelope,
DELETED.Attempts,
DELETED.WorkItemReceived_UTC
-- Use the table var
INTO #output
FROM dbo.WorkRequestQueue
WHERE
WorkRequestQueue.ID = #WorkRequestQueueID
-- Explicitly insert
INSERT
INTO dbo.FaildMessages
SELECT
MessageEnvelope,
Attempts,
#LastException,
GetUtcdate(), -- WorkItemPoisened datetime
WorkItemReceived_UTC
FROM #output
IF ##ROWCOUNT = 0
RAISERROR ('Record not found', 16, 1)
SELECT Cast(SCOPE_IDENTITY() as int)