I want to know how inserted and deleted temp tables in SQL server work. My question is more regarding how they work when multiple users accessing the same database. Suppose two users update the database at the same time. In that case what are the values stored in the inserted and deleted tables.
I have a trigger that records changes to the database as in an audit trail. Like any other audit trail I insert data into my audit table from the inserted and deleted temp tables in MS SQL Server. I however am not clear as to how these inserted and deleted tables store values when two users update the database at the same time. Are there separate inserted and deleted tables for each session. The users access the database thru ASP pages.
The audit trail I am trying to use is http://www.nigelrivett.net/AuditTrailTrigger.html
I actually would like to store the inserted and deleted temp tables into other temporary tables so that I can access these tables thru a stored procedure. This is when the problem of same users updating the temporary tables is more pronounced.
I am reading the WROX "Professional SQL Server 7 Programming" book. The following code appears on page 424:
CREATE TRIGGER ProductIsRationed ON Products FOR UPDATE AS IF EXISTS ( SELECT 'True' FROM Inserted i JOIN Deleted d ON i.ProductID = d.ProductID WHERE (d.UnitsInStock - i.UnitsInStock) > d.UnitsInStock / 2 AND d.UnitsInStock - i.UnitsInStock > 0 ) BEGIN RAISERROR('Cannot reduce stock by more than 50%% at once.',16,1) ROLLBACK TRAN END
The trigger fires when an UPDATE is made to Products table. The author states that the Inserted and Deleted tables only exist for the life of the trigger, not before, and not after the trigger runs. If this is true, then why would there be any rows in the Deleted table in this case? No rows were deleted within the trigger. As far as I can see, no rows have been updated either. If the condition does exist, no rows will be updated, and an error will be displayed. Otherwise, the row will be updated. Then there would be a row in the inserted table. But then the trigger is finished and the inserted table for that trigger disappears. I think my logic is flawed, which is why I am writing. I don't think I fully understand the Inserted and Deleted tables.
I just ran across an issue on a SQL 2000 sp4 db where RI was being maintained solely with triggers. I am attempting to change the primary key of a parent table and cascade the results to all its children without using the vendor-supplied trigger code (long story...) using an INSTEAD OF trigger.
My question is: does SQL Server create any kind of relationship between the inserted and deleted tables that I could exploit since the key field is unavailable?
I am trying to avoid having to add a surrogate key to each of the children just for this activity (as there are many M rows in each and no other suitable unique column combinations that span all the child tables).
I have a stored procedure that's running a little slower than I would like. I've executed the stored proc in QA and looked at the execution plan and it looks like the problem is in a trigger on one of the updated tables. The update on this table is affecting one row (I've specified the entire unique primary key, so I know this to be the case). Within my trigger there is some code to save an audit trail of the data. One of these statements does an update of the history table based on the inserted and deleted tables. For some reason this is taking 11.89% of the batch cost (MUCH more than any other statement) and within this statement 50% of the cost is for a table scan on inserted and 50% is for a table scan on deleted. These pseudo-tables should only contain one record each though.
Any ideas why this would be causing such a problem? I've included a simplified version of the update below. The "or" statements actually continue for all columns in the table. The same trigger template is used for all tables in the database and none of the others seem to exhibit this behavior as far as I can tell.
Thanks for any help! -Tom.
UPDATE H_MyTable SET HIST_END_DT = @tran_date FROM H_MyTable his INNER JOIN deleted del ON (his.PrimaryKey1 = del.PrimaryKey1) and (his.PrimaryKey2 = del.PrimaryKey2) INNER JOIN inserted ins ON (his.PrimaryKey1 = ins.PrimaryKey1) and (his.PrimaryKey2 = ins.PrimaryKey2) WHERE (his.HIST_END_DT is null) and ((IsNull(del.PrimaryKey1, -918273645) <> IsNull(ins.PrimaryKey1, -918273645)) or (IsNull(del.PrimaryKey2, -918273645) <> IsNull(ins.PrimaryKey2, -918273645)) or (IsNull(del.Column3, -918273645) <> IsNull(ins.Column3, -918273645)))
I want to compare the before and after values of an UPDATEd column using a trigger. I want to know if the value in the column has changed. Simple? No!
As you know, SqlServer puts the before image of the UPDATEd rows into the DELETED virtual table and the after image of the UPDATEd rows in the INSERTED virtual table.
So you would get the before and after data by doing a SELECT on these tables. But here is the problem - how do you join the tables? What if there are >1 rows in these 2 tables (because the UPDATE affected >1 rows) - how do i know which "old"/DELETED rows correspond to which "new"/INSERTED?" Ok - I could join the 2 tables on the primary key, but what if the primary key was updated? In that case the join would not work - the DELETED table would contain the old primary key value and the INSERTED table would contain the new (different) primary key value. In fact, ALL of the columns may have been changed by the UPDATE.
Now, there is another thing to try with triggers - the IF UPDATE ( <columname> ) test. This is designed to tell you if a specified column was UPDATEd by the last UPDATE. However, this will return TRUE for any UPDATE that mentions the column - even if the UPDATE does not change any data! So I cannot determine whether a certain column has had its value changed with this either.
So then you can try another test mentioned in the docs for CREATE TRIGGER - the IF COLUMNS_UPDATED() test. However, this will report that a column has been updated, NOT whether the data has changed as aresult of that UPDATE. So if you UPDATE the value in the column to the same value as it was beforehand (admittedly, a pointless thing to do, but it could happen in some apps), this fuction will say, yes, this column was updated.
So my question remains - how do I know if the data has changed in a column after an UPDATE, using a trigger? Any ideas?
I dont know what I am doing wrong. The trigger (see below) is doing what I want it to do when I do this:
INSERT INTO dbo.personal
(personal_id, chef_id,fornamn, efternamn)
VALUES
(40, 100, 'Malin', 'Markusson' , 'Boss')
but when I remove one value, the result in the logtable is NULL:
INSERT INTO dbo.personal (personal_id, chef_id,fornamn, efternamn)
VALUES
(40, 100, 'Malin', 'Markusson' )
How can I change the trigger so that it will give me the information of the values that have been updated, inserted or deleted when I dontchange all values (just a couple of them)?
My trigger: CREATE Trigger trigex
ON dbo.personal
FOR insert, update,delete
AS
INSERT INTO logtable (Innan_värde, Ny_värde)
SELECT
rtrim(cast(d.personal_id as varchar)+', '+cast(d.chef_id as varchar)+', '+rtrim(d.efternamn)+', '+ rtrim(d.fornamn)+ ', '+ rtrim(d.titel)),
(cast(i.personal_id as varchar)+', '+cast(i.chef_id as varchar)+', '+rtrim(i.efternamn)+', '+ rtrim(i.fornamn)+ ' '+ rtrim(i.titel))
FROM inserted i full join deleted d on i.personal_id = d.personal_id
We have an app that uses triggers for auditing. Is there a way to know the order that the records were inserted or deleted? Or maybe a clearer question is.... Can the trigger figure out if it was invoked for a transaction that "inserted and then deleted" a record versus "deleted and then inserted" a record? The order of these is important to our auding.
I have a table that sometimes has modifications to column(s) comprising the primary key [usually "end_date"]. I need to audit changes on this table, and naturally, turned to after triggers.
The problem is that for updates, when the primary key composition changes, I'm not able to relate/join using the primary key - obviously, it no longer matches across INSERTED and DELETED. Now, for a single row update, it's easy to check for updates on PK columns and then deduce what changes were made...
So the real question is: are rows in INSERTED and DELETED always in matching order (1st row in INSERTED corresponds to the 1st row in DELETED...)?
I don't want to put a surrogate key (GUID nor IDENTITY) on the base table if at all possible. INSERT... SELECT from the inserted/deleted tables into a temp table with identity column is fine, and is what I'm currently doing; I would like MVP or product engineer level confirmation that my ordering assumption is correct.
Testing using an identity surrogate key on base table, and selecting from the Ins/del tables, and the temp tables without an order by clause seems to always return in proper order (proper for my purposes). I've tested under SQL 2005 RTM, SP1, SP2, and SP2 "3152".
FYI, I've lost the debate that such auditing is better handled by the application, not the database server...
Aside: why doesn't the ROW_NUMBER() function allow an empty OVER( ORDER BY() ) clause? Will SQL ever expose an internal row_id, at least in the pseudo tables, so we can work around this situation?
I was wondering if there is a way to schedule the deletion of temporary tables that are over a certain age. Right now, temporary tables are being created, but unless I delete them manually, they end up building up and just taking up a lot of space.
I would like to wrap the following code in a function and reuse it. Â I use this code in many triggers.
DECLARE @Action as char(1); SET @Action = (CASE WHEN EXISTS(SELECT * FROM INSERTED) AND EXISTS(SELECT * FROM DELETED) THEN 'U' Â -- Set Action to Updated. WHEN EXISTS(SELECT * FROM INSERTED) THEN 'I' Â -- Set Action to Insert. WHEN EXISTS(SELECT * FROM DELETED) THEN 'D' Â -- Set Action to Deleted. ELSE NULL -- Skip. It may have been a "failed delete". Â Â END)
Is it possible to write a function and pass the INSERTED and DELETED logical tables to it?
I would like to access the data inserted or deleted within a trigger. however the built-in tables -- inserted and deleted --- are not accessible. anyone knows why? And is there any other way to do this?
Is there an alternative to using FETCH to loop through the Inserted/Delete Tables within a trigger? Does this work?
SELECT * FROM Inserted BEGIN
if INSERTED.IsActive then ...
END
Would this only see the first record?
Currently I'm doing the following;
AS DECLARE @JobID INTEGER; DECLARE @IsActive BIT; DECLARE Temp CURSOR FOR SELECT JobID, IsActive FROM Inserted; BEGIN OPEN Temp; FETCH NEXT FROM Temp INTO @JobID, @IsActive; WHILE (@@FETCH_STATUS = 0) BEGIN
if @IsActive then ...
FETCH NEXT FROM Temp INTO @JobID, @IsActive; END; CLOSE Temp; DEALLOCATE Temp;
Is this the best method for looping through the Deleted/Inserted or any other table within a trigger?
I need to add the row number or record number to the 'inserted' and 'deleted' tables in a trigger.
Among other things, I have tried-
SELECT 1 as rowId, i.* INTO #ins FROM inserted i if @@ROWCOUNT > 1 --if multiple rows, set row number begin
SELECT @pkWhere = coalesce(@pkWhere + ' and ', ' where ') + PKF.colName + ' <= i.' + PKF.colName FROM #primaryKeyFields PKF set @strSQL = 'update #ins set rowId = (Select Count(*) From #ins i' + @pkWhere + ')'
exec (@strSql)
end
-the above sets rowId for every record to the total instead of sequential. Keep in mind that this code is used to create a trigger for any table, hence I cannot specify any column names literally.
I created manage update trigger to react on one column changes. There is an application which is working with DB, so I don't have access to SQL query which changes this column. In most cases trigger works fine, but in one case when this column changes, trigger is fired and IsUpdatedColumn is true for this column, but both inserted and deleted table are empty, so I can't get new value for this column. Any idea why is it happened? Is any way around?
This column type is uniqueidentifier. Inserted and deleted tables are empty when application is changing value from NULL to not null value, but if I change it myself from Management Studio inserted table contains right values. Most like problem is in query which is changing that value.
HI, I am wondering if it is possible to retreive this information without using row count transform. Can I get the # of rows inserted/updated or deleted by destination from the log?
Hi all!In a insert-trigger I have two joins on the table named inserted.Obviously this construction gives a name collition beetween the twojoins (since both joins starts from the same table)Ofcourse I thougt the usingbla JOIN bla ON bla bla bla AS a_different_name would work, but itdoes not. Is there a nice solution to this problem?Any help appriciated
Is there a configuration or a trick to improve the speed of the access to inserted and deleted tables whithin a trigger? Whenever a trigger is called, the access to inserted or deleted constitute approximatly 95% of the execution time.
Is there a way to have access to inserted and to deleted improved other than copying the data to another table?
When i debug a trigger is it possible to add a WATCHon the INSERTED or DELETED?I think not, at least I couldn't figure out a way to do so.Does someone have a suggestion on how I can see the values?I did try to do something likeINSERT INTO TABLE1(NAME)SELECT NAME FROM INSERTEDbut this didn't work. When the trigger completed and Iwent to see the TABLE1, there were no records in it.Are there any documents, web links that describe waysof debugging the trigger's INSERTED and DELETED?Thank you
I have a publisher database set up for a merge replication. This is using parameterized filter with join filters.
I also have a stored procedure that does deletes & inserts on the table where the parameterized filter is applied to aid in changing a subscriber's eligibility to receive so and so data. I have observed that running the stored procedure takes extraordinarily long and as a result, the log file grows to a size 1.5 - 2.5 times the database size.
At first I reasoned that this might because I had it set up to use precomputed partitions and changing it requires recalculating the partitions. As a test, I turned off the precomputed partitions. Didn't work. I turned on "optimize synchronization" AKA "keep_partition_changes", which normally is not available when you have precomputed partition on, and that didn't work, either.
At this point, I think I can rule out precomputed partitions being a problem here but I'm stumped now what else I should do to reduce the amount of log writes being required. We do need the parameterized filters & join tables, so that can't go.
For some reason in Enterprise Manager for SQL Server 2000, I cannotput the following line into a trigger:select * into #deleted from deletedWhen I hit the Apply button I get the following error:Cannot use text, ntext, or image columns in the 'inserted' or'deleted' tablesThis seems like a weird error, since I am not actually doing anythingto the inserted or deleted tables, I am just trying to make a tempcopy.I have another workaround but I am just curious why this happens.Thanks,Rebecca
Looking at BOL for temp tables help, I discover that a local temp table (I want to only have life within my stored proc) SHOULD be visible to all (child) stored procs called by the papa stored proc.
However, the following code works just peachy when I use a GLOBAL temp table (i.e., ##MyTempTbl) but fails when I use a local temp table (i.e., #MyTempTable). Through trial and error, and careful weeding efforts, I know that the error I get on the local version is coming from the xp_sendmail call. The error I get is: ODBC error 208 (42S02) Invalid object name '#MyTempTbl'.
Here is the code that works:SET NOCOUNT ON
CREATE TABLE ##MyTempTbl (SeqNo int identity, MyWords varchar(1000)) INSERT ##MyTempTbl values ('Put your long message here.') INSERT ##MyTempTbl values ('Put your second long message here.') INSERT ##MyTempTbl values ('put your really, really LONG message (yeah, every guy says his message is the longest...whatever!') DECLARE @cmd varchar(256) DECLARE @LargestEventSize int DECLARE @Width int, @Msg varchar(128) SELECT @LargestEventSize = Max(Len(MyWords)) FROM ##MyTempTbl
SET @cmd = 'SELECT Cast(MyWords AS varchar(' + CONVERT(varchar(5), @LargestEventSize) + ')) FROM ##MyTempTbl order by SeqNo' SET @Width = @LargestEventSize + 1 SET @Msg = 'Here is the junk you asked about' + CHAR(13) + '----------------------------' EXECUTE Master.dbo.xp_sendmail 'YoMama@WhoKnows.com', @query = @cmd, @no_header= 'TRUE', @width = @Width, @dbuse = 'MyDB', @subject='none of your darn business', @message= @Msg DROP TABLE ##MyTempTbl
The only thing I change to make it fail is the table name, change it from ##MyTempTbl to #MyTempTbl, and it dashes the email hopes of the stored procedure upon the jagged rocks of electronic despair.
Any insight anyone? Or is BOL just full of...well..."stuff"?
Hi SQL fans,I realized that I often encounter the same situation in a relationdatabase context, where I really don't know what to do. Here is anexample, where I have 2 tables as follow:__________________________________________ | PortfolioTitle|| Portfolio |+----------------------------------------++-----------------------------+ | tfolio_id (int)|| folio_id (int) |<<-PK----FK--| tfolio_idfolio (int)|| folio_name (varchar) | | tfolio_idtitle (int)|--FK----PK->>[ Titles]+-----------------------------+ | tfolio_weight(decimal(6,5)) |+-----------------------------------------+Note that I also have a "Titles" tables (hence the tfolio_idtitlelink).My problem is : When I update a portfolio, I must update all theassociated titles in it. That means that titles can be either removedfrom the portfolio (a folio does not support the title anymore), addedto it (a new title is supported by the folio) or simply updated (atitle stays in the portfolio, but has its weight changed)For example, if the portfolio #2 would contain :[ PortfolioTitle ]id | idFolio | idTitre | poids1 2 1 102 2 2 203 2 3 30and I must update the PortfolioTitle based on these values :idFolio | idTitre | poids2 2 202 3 352 4 40then I should1 ) remove the title #1 from the folio by deleting its entry in thePortfolioTitle table2 ) update the title #2 (weight from 30 to 35)3 ) add the title #4 to the folioFor now, the only way I've found to do this is delete all the entriesof the related folio (e.g.: DELETE TitrePortefeuille WHERE idFolio =2), and then insert new values for each entry based on the new givenvalues.Is there a way to better manage this by detecting which value has to beinserted/updated/deleted?And this applies to many situation :(If you need other examples, I can give you.thanks a lot!ibiza
I have an application that I am working on that uses some small temptables. I am considering moving them to Table Variables - Would thisbe a performance enhancement?Some background information: The system I am working on has numeroustables but for this exercise there are only three that really matter.Claim, Transaction and Parties.A Claim can have 0 or more transactions.A Claim can have 1 or more parties.A Transaction can have 1 or more parties.A party can have 1 or more claim.A party can have 1 or more transactions. Parties are really many tomany back to Claim and transaction tables.I have three stored procsinsertClaiminsertTransactioninsertPartiesFrom an xml point of view the data looks like this<claim><parties><info />insertClaim takes 3 sets of paramters - All the claim levelinformation (as individual parameters), All the parties on a claim (asone xml parameter), All the transactions on a claim(As one xmlparameter with Parties as part of the xml)insertClaim calls insertParties and passes in the parties xml -insertParties returns a recordset of the newly inserted records.insertClaim then uses that table to join the claim to the parties. Itthen calls insertTransaction and passes the transaction xml into thatsproc.insertTransaciton then inserts the transactions in the xml, and alsocalls insertParties, passing in the XML snippet
I want to pass the 'inserted' table from a trigger into an SP, I think I need to do this by dumping inserted table into a temporary table and passing the temp table. However, I need to do this for many tables, and don't want to list all the column names for each table/trigger (maintenance nightmare).
Can I dump the 'inserted' table to a temp table WITHOUT specifying the column names?
