Saturday, September 04, 2004

About TSEqual function (SQL 2K)

One of the common problem the database developers face in their day to day life is record concurrency issues. Lets try to address that with the help of timestamp data type.

Lets assume that Sarah and Ram are reading a same record. First Ram updates that record with some new data. Later if Sarah also updates the record (mind you she is viewing the old content only still) then it would overwrite the changes made by Ram.

There are two ways of solving this issue they are:

1. Pessimistic Locking
2. Optimistic Locking

Pessimistic Locking: First person who reads the record would put a lock on it so that nobody else can change it until he is done with it. Only when he releases the record the other user can make use of it. This method is not recommended because it might also takes hours or days together for the first person to release his lock due to various reasons.

Optimistic Locking: The record would be locked only when a user wants to modify its content.

SQL Server has a TSEqual (I presume it means TimeStamp Equal) function which compares TimeStamp values in the table and the T-SQL statement. If the timestamp values doesn't match it would throw an error and abort the operation. Let see this in action:

Declare @tStampOriginal TimeStamp

--Here tStamp is a TimeStamp column and TimeStampExample is the name of the table
Select @tStampOriginal = tStamp from TimeStampExample

-- Compares current timestamp value with the original value before updating
Update TimeStampExample Set LastName = 'Kapil' Where LastName = 'Tendulkar' and TSEqual (tStamp, @tStampOriginal)

The above batch of code should execute fine without any issues.

Declare @tStampOriginal TimeStamp

--here tStamp is a TimeStamp column and TimeStampExample is the name of the table
Select @tStampOriginal = tStamp from TimeStampExample

--dummy update statement to change the timestamp value
Update TimeStampExample Set LastName = 'Gandhi' Where LastName = 'Gandhi'

-- Compares current timestamp value with the original value before updating
Update TimeStampExample Set LastName = 'Kapil' Where LastName = 'Tendulkar' and TSEqual (tStamp, @tStampOriginal)

This batch would fail because the @tStampOriginal contains the initial timestamp values which has changed during the dummy update statement.

Wednesday, September 01, 2004

Text functions in SQL 2k

As much as possible it is advisable to keep large text in a document on the file system and store a link to that within the database. Why do I say this?

Because for storing large chunk of data we need to depend on the TEXT, NTEXT and IMAGE data types. So What?

There are 2 disadvantages in it. They are:

1. These data types does not support commonly used string functions such as Len, Left, Right etc.,

2. It occupies more / large space in the DB which internally means there might be a performance issues if you store such data in the database.

Inspite of all these if you still want to use TEXT data type due to your business requirment then these functions might interest you!

1. PatIndex()
2. TextPtr()
3. ReadText()
4. TextValid()

PatIndex()

This function is useful with TEXT, CHAR and VARCHAR data types. This seeks for the first occurrence of a pattern within a string. If the pattern is found, it returns the character number where the first occurrence of the pattern begins.

For better understanding, the following code snippet makes use of TEXT datatype and PatIndex to search for occurrences of the "dum" pattern:

--Create a sample table with one TEXT datatype field
Create table Sample
(
LargeTextColumn TEXT
)

--Insert a sample record
Insert Sample Values ('Enter some dummy content here')

/*
Search for the text of your choice and check the output. Please note that usage of wildcards is the difference betweeen PatIndex and CharIndex. CharIndex doesn't understand wildcards and it takes those wildcards and searches for the exact symbol :)
*/


Select PatIndex('%dum%', LargeTextColumn) from Sample

TextPtr()

This function returns the binary pointer to the TEXT, NTEXT, or IMAGE column in a table. Once we get the binary pointer from TEXTPTR(), we can make use of the same in READTEXT, WRITETEXT, and UPDATETEXT statements.

The below code snippet returns the text pointer to the field LargeTextColumn.

SELECT TEXTPTR(LargeTextColumn) FROM Sample

ReadText()

As explained above, ReadText makes use of the text pointer returned by TextPtr function to retrieve Text/ntext/Image values from its corresponding datatypes. If you have a look at the below code snippet you would understand the syntax and usage of ReadText function.

Declare @getPointer varbinary(16)
Select @getPointer = TextPtr(LargeTextColumn) From Sample
ReadText Sample.LargeTextColumn @getPointer 11 5


Similarly WriteText and UpdateText are used for overwritting the existing content and updating the existing content respectively.

TextValid()

This function provides a way to see whether an existing pointer to a column with TEXT, NTEXT, or IMAGE data type is valid. It throws 1 if valid and 0 otherwise.

Declare @getPointer varbinary(16)
Select @getPointer = TextPtr(LargeTextColumn) From Sample
SELECT TEXTVALID('Sample.LargeTextColumn', @getPointer)