In SQL server the errors are always written to sysmessages table, so no exception can be handled here. The type of access to the SQL server from the front end determines the state of the error and how the error information can be retrieved.

Eg.

a).  For ODBC type of access the error messages are stored in the  SQLError object of the SQL server, the front end should query that object to get the error message.

b).  For OLE-DB type of access the error messages are stored in the ISQLServerErrorInfo object. So the OLE-DB provider library

RAISERROR('Test Only',  1,  2) WITH SETERROR
SELECT @@ERROR
RAISERROR (101,  1,  2) WITH SETERROR
SELECT @@ERROR
Here is the result set:
Msg 50000,  Level 1,  State 50000
Test Only
-----------
50000
(1 row(s) affected)
Msg 101,  Level 1,  State 101
Line 0: SQL syntax error.
-----------
101
(1 row(s) affected)

The first RAISERROR returns an @@ERROR value of 50000. The second returns the syntax error message used by Microsoft® SQL Server™ with an @@ERROR value of 101.

Formatting T-SQL Blocks:

T-SQL blocks have the following sections:

1. Header
2. Declaration section
3. Executable section.
4. Exception section.

PROCEDURE (Parameters)
RETURN DATATYPE
AS
Variables DATATYPE;

BEGIN
Executable_statements;
If err then goto Err_block
RETURN Value

Err_block:
Process error and return the error string
END
GO

Recommendations:

1. All the executable statements after the BEGIN and error handling block are indented in from the BEGIN
2. Include a blank line after each section.
3. IS statement to be on a new line where everused.

Formatting SQL  statement:

1. The column name should be of the form XX<Abbr/Name><suffix> (XXX represents the abbreviation of the Table Name.It should be a meaningful abbreviation of the table name to which the column belongs.<suffix> is a meaningful suffix).
2. Column name should  not be  longer than 30  Alphanumeric characters.
3. Column name (<Abbr/Name> only) should be same throughout an  application even though the field may be present in more than one table.

The following suffixes can be used for better understanding.

1. The table name should be the same as the logical entity name or should bear some significance to the data stored therein.
2. The table must have a primary key that uniquely identifies each row.
3. The column names will be same as logical attribute name.
4. Columns must be defined as NOT NULL wherever possible.
5. Table names should be no more than 30 alphanumeric characters.

For Examples creating the tables : AccountMaster , CashBalance , TransactionDetail

Guidelines to create the tables:

1. Keep table names to around 20 characters. This will prevent the truncation of table name in the naming of other objects that contain the table name as part of the object name. e.g. Indexes
2. Columns may only be denormalized to improve performance, but this should be very carefully balanced against the additional update that may be required.