Perhaps....

I'd still use a table. You just need to add some type of csv list id and a sequence #.

If the function just removes an optional trailing '-' followed by other chars, you can avoid a function completely using CROSS APPLY:

UPDATE ts

SET

Col1 = CASE WHEN...

amy26 (2/11/2014)

---

Removing the transaction name does not rectify the situation.

The only thing that I could think of was we do have an if statement that calls another stored procedure...

The single index with the included columns can handle queries for both ProductKey alone and for ProductKey and one/both of the included columns. That means you only need one...

CREATE TRIGGER [dbo].[tr_a] on [dbo].[A]

AFTER UPDATE

AS

SET NOCOUNT ON

BEGIN TRY

IF UPDATE(STATUS)

BEGIN

INSERT INTO dbo.B

(

col0,

col1,

...

Look at Change Tracking; it can identity changes vastly more efficiently than what you're doing now.

If you're on Enterprise Edition, also look at Change Data Capture, which gives you "point-in-time"...

You most likely do have nested transactions, either implicitly or explicitly.

Just remove the transaction name from the ROLLBACK, which is not meaningful for SQL Server anyway:

ROLLBACK TRANSACTION /*UPDT_DATA*/

You need to run these commands (or the equivalent). Run them online if possible, and using tempdb for sort if possible:

DROP INDEX [IX_PurchaseOrderDetail_ProductKey] ON [dbo].[PurchaseOrderDetail]

CREATE NONCLUSTERED INDEX [IX_PurchaseOrderDetail_ProductKey]...

george sibbald (2/7/2014)

---

thats a bit over the top surely?

Nowhere does it say using maintenance plans per se is not...

And, if the query uses more than one table, use an alias and prefix all columns with the correct alias name. Remember, we have absolutely no idea what columns...

A CONVERT to format 101 will give us format "mm/dd/yyyy". Then, using STUFF, we can remove the "dd/", leaving you with what you want:

SELECT STUFF(CONVERT(varchar(10), GETDATE(), 101), 4, 3,...

A basic maintenance plan is just terrible for large databases, because it rebuilds every table and every index regardless. You just can't waste that much I/O in a critical...

Many of us prefer to generate such code from the table itself, something like below. Uncomment the EXEC(@sql) when ready to actually run the code:

USE css

DECLARE @nonrepeated_cols nvarchar(max)

DECLARE @repeated_cols...

You could use subqueries instead of joins, like this:

SELECT

FI.*,

(SELECT TOP (1) C.ClassName_FD

FROM DBA.Class_TB AS C...