Dinesh's Blog ::::::: Being Compiled :::::::: SQL Server 2000

Did another presentation - Optimizing TSQL

Thu, 06 Apr 2006 00:45:50 GMT

It was my turn to do a presentation at the .net forum. This time I picked a topic, that is not a new but common to all database developers. I talked about TSQL; how to write effective TSQL. Not all slides could be discussed because of the time but tried my best to deliver as much as possible.

Did you find the session important? Did you get something from my session? Like to hear something from you guys, whether the presentation is okay or not.

The first session was done by Merill – MVP about ASP.NET 2.0 tips and tricks, he showed really cool stuff.

Oldest trick but still effective :)

Tue, 24 Jan 2006 01:16:40 GMT

This is little bit an old stuff but I was questioned again regarding this. As usual, thought share it, the question of “How to handle more than one row in trigger?”. As you all know, triggers are fired once per transaction not once per row. Suppose we have implemented a logging system that logs all the update of the table “TestTable”. Fine, this can be easily implemented with a stored procedure and a trigger. We can write a trigger to handle the update and log the information by using the stored procedure that logs update. Just like below….

CREATE PROC LogTestTableTransaction (@primaryKey int)
AS
BEGIN

   -- this is not the way for loggin but for testing
   print 'The record ' + CONVERT(varchar(10), @primaryKey) + ' is changed'
END

CREATE TRIGGER TestTableUpdateTrigger ON TestTable
FOR UPDATE
AS
BEGIN
   IF UPDATE(ColumnToBeChanged)
   BEGIN
      DECLARE @primaryKey int
      SELECT @primaryKey = PrimaryKey FROM inserted
      EXEC LogTestTableTransaction @primaryKey
   END
END

UPDATE TestTable SET ColumnToBeChanged = 'NewValue' WHERE PrimaryKey = 3
UPDATE TestTable SET ColumnToBeChanged = 'NewValue' WHERE PrimaryKey IN (1, 4)

The process works fine for first update statement but second statement. Though we update two records in second statement, the trigger fires once because triggers are set-based – once per transaction. The second statement causes to log the last record (4) only. How do we fix this problem? You can find few ways but here is the one of bests.

Implement a cursor inside the trigger. Have a look on new trigger.

CREATE TRIGGER TestTableUpdateTrigger ON TestTable
FOR UPDATE
AS
BEGIN
   IF UPDATE(ColumnToBeChanged)
   BEGIN

DECLARE @primaryKey int

DECLARE PrimaryKeyCursor CURSOR

FOR

SELECT PrimaryKey FROM inserted

OPEN PrimaryKeyCursor

      FETCH NEXT FROM PrimaryKeyCursor INTO @primaryKey
      WHILE (@@FETCH_STATUS = 0)
      BEGIN
         EXEC LogTestTableTransaction @primaryKey
         FETCH NEXT FROM PrimaryKeyCursor INTO @primaryKey
      END
      CLOSE PrimaryKeyCursor
      DEALLOCATE PrimaryKeyCursor
   END
END

Now it should work fine. In most cases, we do not bother on this type of things till we get complain from the client . So, keep this procedure in mind. It might help you in your next assignment.

COMPUTE clause with Datagrid

Sat, 14 Jan 2006 11:58:45 GMT

How often you have used COMPUTE and ROLLUP when you want to show totals and subtotals of your data? I am sure; your answer will be “very rare” or “never”. There are some instances in that these are come in handy. I decided to make some series of tests that how we can use them with .net. This is my first test.

Suppose we want to populate all the orders (date, quantity and amount) with data grid. Simple way is, fill the required data with simple SELECT query to dataset and bind it to the data grid. Suppose that the grand totals of the quantity and the amount have to be shown. There are few ways to do this. You might do this by making another request to the database or looping the dataset or through ItemDataBound event. Do you ever think that this can be easily done with COMPUTE clause? Here is the way.

The below code shows how the query is written and data is bound. Make sure that the dataset can be accessed throughout the class.

SqlConnection sqlConnection = new SqlConnection("Server=(local);Database=Northwind;Uid=sa;Pwd=Dinesh");
string sql = @" SELECT o.OrderDate, Quantity, UnitPrice*Quantity AS Amount
                     FROM Orders o
                     INNER JOIN [Order Details] od
                        ON o.OrderId = od.OrderId
                     COMPUTE SUM(Quantity), SUM(UnitPrice*Quantity )
";

SqlCommand sqlCommand = new SqlCommand(sql, sqlConnection);
SqlDataAdapter sqlDataAdapter = new SqlDataAdapter(sqlCommand);
try
{
   sqlConnection.Open();
   sqlDataAdapter.Fill(dataSet);
   DataGrid1.DataSource = dataSet.Tables[0];
   DataGrid1.DataBind();
}
catch (Exception ex)
{
   Response.Write(ex.Message);
}
finally
{
   sqlConnection.Close();
}

This code shows the way of setting the grand totals of the quantity and amount in the footer of the grid. Note that this has to be written inside the ItemDataBound event.

if (e.Item.ItemType == ListItemType.Footer)
{
   Label labelTotalQuantity = (Label) e.Item.FindControl("LabelTotalQuantity");
   labelTotalQuantity.Text = dataSet.Tables[1].Rows[0][0].ToString();
   Label labelTotalAmount = (Label) e.Item.FindControl("LabelTotalAmount");
   labelTotalAmount.Text = string.Format("{0:c}", dataSet.Tables[1].Rows[0][1]);
}

Do you think that this is easier than the usual way we used to? The COMPUTE BY is being tested with the grid. Let me published it once I finished with it.

Default size of char and varchar

Mon, 24 Oct 2005 18:07:04 GMT

When a question “What is the default size of char or varchar if the size is not specified?” was raised in the middle of SQL session, as I expected, almost all answers were 1. Of course they are not experienced developers with SQL Server (some are new-bies). Do you know the answer? In a way, it is correct. But it changes its behavior little bit with convert operation. Have a look on the code below.

DECLARE @char char, @varchar varchar

SET @char = 'abcd'
SET @varchar = 'abcd'

SELECT @char , @varchar
-- result
-- a a

SELECT LEN(CONVERT(char, REPLICATE('a', 100))), LEN(CONVERT(varchar, REPLICATE('a', 100)))
-- result
-- 30 30

The last statements return 30 for both char and varchar types. This because of, SQL Server set the size (if not specified) of char and varchar as 30 when they are used in CONVERT operation. Now, was your answer correct? Let me know your answer.

Placing conditions: in the WHERE or ON clause?

Sun, 16 Oct 2005 03:54:24 GMT

Yesterday, one of my friends called me and put a question regarding JOINs. He had joined two tables with OUTER JOIN to get his desired result but the output is not as what he expected. When I saw the query, I immediately found the problem; one of the key conditions is placed in the WHERE clause instead ON clause. Does it really affect for the output. If it is OUTER JOIN, yes, it might affect to the output. Have look on below queries.

--create Students table and insert some data
CREATE TABLE #Students
(Id int PRIMARY KEY,
FirstName varchar(50) NOT NULL)

INSERT INTO #Students VALUES (1, 'Dinesh')
INSERT INTO #Students VALUES (2, 'Asela')
INSERT INTO #Students VALUES (3, 'Sanjaya')

--create StudentCources table and insert some data
CREATE TABLE #StudentCources
(StudentId int NOT NULL,
CourseName varchar(50) NOT NULL)

INSERT INTO #StudentCources VALUES (1, 'A')
INSERT INTO #StudentCources VALUES (1, 'B')
INSERT INTO #StudentCources VALUES (1, 'C')

Assume we need to list out all the students and see whether they are enrolled for course ‘A’. There are few ways to the list out but we will use OUTER JOIN to see the different.

--query1
SELECT s.FirstName, ISNULL(sc.CourseName, '-') AS CourseName
FROM #Students s
LEFT OUTER JOIN #StudentCources sc
ON s.Id = sc.StudentId and sc.CourseName = 'A'

--output
FirstName CourseName
--------- ----------
Dinesh    A
Asela     -
Sanjaya   -

Fine. We get what we wanted. See the place of the condition of sc.CourseName = 'A'. It is placed in the ON clause. What will happen if it is placed in the WHERE clause.

--query2
SELECT s.FirstName, ISNULL(sc.CourseName, '-') AS CourseName
FROM #Students s
LEFT OUTER JOIN #StudentCources sc
ON s.Id = sc.StudentId
WHERE sc.CourseName = 'A'

--output
FirstName CourseName
--------- ---------
Dinesh A

Now the output is completely different. Do you know the reason?

When the condition is placed in the ON clause, it is applied before joining tables. That’s why we get all the students for query 1. But if the condition is placed in the WHERE clause, it is applied for the final result set. Keep this in mind because this is a common mistake we do.

SET or SELECT

Fri, 30 Sep 2005 06:39:40 GMT

I was questioned by one of my colleagues regarding SET statement but, my answer for him was incomplete and made me to brush up on this. If I sum up the background of the question tersely, here it is: Should "SET" be always used when need to assign a value into local variable instead of "SELECT"? (This is not the exact question he made.) I went through couple of points to distinguish between "SET" and "SELECT" as below, might help you too.

The recommended way to assign local variable is "SET" rather than "SELECT" because "SET" is the ANSI standard way of assigning values into variables.
SET cannot be used with statements that return multi values.
//this will end up with an error if table has more than one row
SET @Id = (SELECT Id FROM Table1)
//this will execute without any errors but the result might not be desired one.
SELECT @Id = Id FROM table1
SELECT will save the previous assigned value if new assign statement will not set new value.
//set a value to local variable
SET @Id = 0
//new assignment statement with SELECT but will not return any rows
SELECT @Id = Id FROM Table1 WHERE Name = 'Dinesh'
//Will return 0 because the previous value is saved.
SELECT @Id
//new assignment with SET. This statement will return no rows.
SET @Id = (SELECT @Id FROM Table1 WHERE Name = 'Dinesh')
//Will return NULL
SELECT @Id
SELECT allows assigning more than one local variables with one statement.
SELECT @Id = Id, @Name = Name FROM Table1

In addition to that, I do not think that there is a performance different between these two but it might boost the performance for last point with "SELECT".

I highly appreciate if you have more points other than the mentioned ones on this and share with me.

Delete a record with all of its dependancies

Mon, 08 Aug 2005 00:18:58 GMT

Have you ever wanted to delete some records manually in your database while testing? If yes, how long did you take if the record had much more dependencies? I know, if the record has dependencies, and all the dependencies have dependencies, and so on… it makes you mad unless “Cascade delete” is on.

To avoid the time it takes, I just wrote a small code that supports the deletion of all dependencies. Here is the code I did. More thing to be enhanced and much to be tested but it works fine. Have a copy if you need but remember that it still supports table that has single integer type primary key. Improvement for handling string type and composite primary keys are being done.

IF OBJECT_ID('DeleteRecord') IS NOT NULL
DROP PROC DeleteRecord
GO
CREATE PROCEDURE DeleteRecord @ownerName varchar(100), @tableName varchar(200), @columnName varchar(200), @PrimaryKeyValue varchar(1000), @IsPrimaryKeyValueTypesString bit
AS
BEGIN

SET NOCOUNT ON
SET ANSI_WARNINGS OFF

   --*************************************************--
   -- Region: Find Primary key values
   --*************************************************--

DECLARE @tempPKeys_ColumnName sysname, @primaryKeyValueInsertStatement varchar(5000), @primaryKeyValues varchar(1000), @Value varchar(10)

IF EXISTS (SELECT name FROM tempdb..sysobjects WHERE name LIKE '#tempPKeys%')
DROP TABLE #tempPKeys

   CREATE TABLE #tempPKeys
   (tempPKeys_TableQualifier sysname,
   tempPKeys_TableOwner sysname,
   tempPKeys_TableName sysname,
   tempPKeys_ColumnName sysname,
   tempPKeys_KeySequence smallint,
   tempPKeys_PKName sysname)

INSERT INTO #tempPKeys
EXEC sp_pkeys @tableName

SET @tempPKeys_ColumnName = ''
SELECT TOP 1 @tempPKeys_ColumnName = tempPKeys_ColumnName FROM #tempPKeys

SET @primaryKeyValues = ''

DROP TABLE #tempPKeys

   IF (@columnName = '') -- firsttime sp execution.
   BEGIN
      SET @primaryKeyValues = @PrimaryKeyValue
   END
   ELSE
   BEGIN
      IF (@tempPKeys_ColumnName != '') -- no primary key defined for the tabele
      BEGIN
         IF EXISTS (SELECT name FROM tempdb..sysobjects WHERE name LIKE '#tempPKeyValues%')
         DROP TABLE #tempPKeyValues

CREATE TABLE #tempPKeyValues
(tempPKeyValues_Value varchar(1000))

         SET @primaryKeyValueInsertStatement = ''
         SET @primaryKeyValueInsertStatement = 'INSERT INTO #tempPKeyValues SELECT ' + @tempPKeys_ColumnName
         SET @primaryKeyValueInsertStatement = @primaryKeyValueInsertStatement + ' FROM ' + @tableName + ' '
         SET @primaryKeyValueInsertStatement = @primaryKeyValueInsertStatement + 'WHERE ' + @columnName + ' IN ('
         SET @primaryKeyValueInsertStatement = @primaryKeyValueInsertStatement + @PrimaryKeyValue + ')'
         EXEC (@primaryKeyValueInsertStatement)

         DECLARE curPrimaryKeyValues CURSOR FORWARD_ONLY
         FOR
         SELECT * FROM #tempPKeyValues

SET @primaryKeyValues = ''

         OPEN curPrimaryKeyValues
         FETCH NEXT FROM curPrimaryKeyValues INTO @Value
         WHILE (@@FETCH_STATUS = 0)
         BEGIN

            SET @primaryKeyValues = @primaryKeyValues + ',' + @Value
            FETCH NEXT FROM curPrimaryKeyValues INTO @Value
         END
      CLOSE curPrimaryKeyValues
      DEALLOCATE curPrimaryKeyValues

DROP TABLE #tempPKeyValues

IF (@primaryKeyValues != '')
SET @primaryKeyValues = SUBSTRING(@primaryKeyValues, 2, LEN(@primaryKeyValues)-1)

      END
   END
   --*************************************************--
   -- EndRegion: Find Primary key values
   --*************************************************--

   --*************************************************--
   -- Region: Check dependancies
   --*************************************************--

IF (@primaryKeyValues != '')
BEGIN

IF EXISTS (SELECT name FROM tempdb..sysobjects WHERE name LIKE '#tempFKeys%')
DROP TABLE #tempFKeys

      CREATE TABLE #tempFKeys
      (PKTableQualifier sysname,
      PKTable_Owner sysname,
      PKTableName sysname,
      PKColumnName sysname,
      FKTableQualifier sysname,
      FKTableOwner sysname,
      FKTableName sysname,
      FKColumnName varchar(100),
      KeySequence smallint,
      UpdateRule smallint,
      DeleteRule smallint,
      FKName sysname,
      PKName sysname,
      Deferrability sql_variant)

DECLARE @primaryKeyTableQualifier sysname, @primaryKeyTableOwner sysname, @primaryKeyTableName sysname, @primaryKeyColumnName sysname, @foreignKeyTableQualifier sysname, @foreignKeyTableOwner sysname,
@foreignKeyTableName sysname, @foreignKeyColumnName varchar(200), @keySequence smallint, @updateRule smallint, @deleteRule smallint, @foreignKeyName sysname,
@primaryKeyName sysname, @deferrability sql_variant

INSERT INTO #tempFKeys
EXEC sp_fkeys @tableName, @ownerName

      DECLARE curForeignKeys CURSOR LOCAL STATIC
      FOR
      SELECT * FROM #tempFKeys

OPEN curForeignKeys
FETCH NEXT FROM curForeignKeys INTO @primaryKeyTableQualifier, @primaryKeyTableOwner, @primaryKeyTableName, @primaryKeyColumnName, @foreignKeyTableQualifier, @foreignKeyTableOwner, @foreignKeyTableName, @foreignKeyColumnName, @keySequence, @updateRule, @deleteRule,
@foreignKeyName, @primaryKeyName, @deferrability

      WHILE (@@FETCH_STATUS = 0)
      BEGIN
         -- Call the sp again for dependancies
         EXEC DeleteRecord @foreignKeyTableOwner, @foreignKeyTableName, @foreignKeyColumnName, @primaryKeyValues, @IsPrimaryKeyValueTypesString

FETCH NEXT FROM curForeignKeys INTO @primaryKeyTableQualifier, @primaryKeyTableOwner,
@primaryKeyTableName, @primaryKeyColumnName, @foreignKeyTableQualifier,
@foreignKeyTableOwner, @foreignKeyTableName, @foreignKeyColumnName,
@keySequence, @updateRule, @deleteRule,
@foreignKeyName, @primaryKeyName, @deferrability

      END
      CLOSE curForeignKeys
      DEALLOCATE curForeignKeys

END

   --*************************************************--
   -- End region: Check dependancies
   --*************************************************--

   --*************************************************--
   -- Region: Delete records
   --*************************************************--
   DECLARE @deleteStatement varchar(5000)

SET @deleteStatement = 'DELETE ' + @ownerName + '.' + @tableName + ' WHERE '

   IF (@primaryKeyValues != '')
      SET @deleteStatement = @deleteStatement + @tempPKeys_ColumnName + ' IN (' + @primaryKeyValues + ')'
   ELSE -- no primary keys. Add conditions with foreign keys
      SET @deleteStatement = @deleteStatement + @columnName + ' IN (' + @primaryKeyValue + ')'

EXEC (@deleteStatement)
--*************************************************--
-- End region: Delete records
--*************************************************--

END

To delete a record by using the SP;

EXEC DeleteRecord 'dbo', 'MyTable', '', '7', 0

If you come across any errors with this SP, please let me. Again, this has to be enhanced.