Databases are the heart and soul of many of the recent enterprise applications and it is very essential to pay special attention to database programming. I've seen in many occasions where database programming is overlooked, thinking that it's something easy and can be done by anyone. This is wrong. For a better performing database you need a real DBA and a specialist database programmer, let it be Microsoft SQL Server, Oracle, Sybase, DB2 or whatever! If you don't use database specialists during your development cycle, database often ends up becoming the performance bottleneck. I decided to write this article, to put together some of the database programming best practices, so that my fellow DBAs and database developers can benefit!

Here are some of the programming guidelines, best practices, keeping quality, performance and maintainability in mind. This list many not be complete at this moment, and will be constantly updated. Btw, special thanks to Tibor Karaszi (SQL Server MVP) and Linda (lindawie) for taking time to read this article, and providing suggestions.

• Decide upon a database naming convention, standardize it across your organization and be consistent in following it. It helps make your code more readable and understandable. Click here to see the database object naming convention that I follow.
• Do not depend on undocumented functionality. The reasons being:
  - You will not get support from Microsoft, when something goes wrong with your undocumented code
  - Undocumented functionality is not guaranteed to exist (or behave the same) in a future release or service pack, there by breaking your code
• Try not to use system tables directly. System table structures may change in a future release. Wherever possible, use the sp_help* stored procedures or INFORMATION_SCHEMA views. There will be situattions where you cannot avoid accessing system table though!
• Make sure you normalize your data at least till 3rd normal form. At the same time, do not compromize on query performance. A little bit of denormalization helps queries perform faster.
• Write comments in your stored procedures, triggers and SQL batches generously, whenever something is not very obvious. This helps other programmers understand your code clearly. Don't worry about the length of the comments, as it won't impact the performance, unlike interpreted languages like ASP 2.0.
• Do not use SELECT * in your queries. Always write the required column names after the SELECT statement, like SELECT CustomerID, CustomerFirstName, City. This technique results in less disk IO and less network traffic and hence better performance.
• Try to avoid server side cursors as much as possible. Always stick to 'set based approach' instead of a 'procedural approach' for accessing/manipulating data. Cursors can be easily avoided by SELECT statements in many cases. If a cursor is unavoidable, use a simpleWHILE loop instead, to loop through the table. I personally tested and concluded that a WHILE loop is faster than a cursor most of the times. But for a WHILE loop to replace a cursor you need a column (primary key or unique key) to identify each row uniquely and I personally believe every table must have a primary or unique key. Click here to see one of the many examples of using WHILE loop.
• Avoid the creation of temporary tables while processing data, as much as possible, as creating a temporary table means more disk IO. Consider advanced SQL or views or table variables of SQL Server 2000 or derived tables, instead of temporary tables. Keep in mind that, in some cases, using a temporary table performs better than a highly complicated query.
• Try to avoid wildcard characters at the beginning of a word while searching using the LIKE keyword, as that results in an index scan, which is defeating the purpose of having an index. The following statement results in an index scan, while the second statement results in an index seek:
  
  1. SELECT LocationID FROM Locations WHERE Specialities LIKE '%pples'
  2. SELECT LocationID FROM Locations WHERE Specialities LIKE 'A%s'
  
  Also avoid searching with not equals operators (<> and NOT) as they result in table and index scans. If you must do heavy text-based searches, consider using the Full-Text search feature of SQL Server for better performance.
• Use 'Derived tables' wherever possible, as they perform better. Consider the following query to find the second highest salary from Employees table:
  
  
  SELECT MIN(Salary)
  FROM Employees
  WHERE EmpID IN
  (
  SELECT TOP 2 EmpID
  FROM Employees
  ORDER BY Salary Desc
  )
  
  The same query can be re-written using a derived table as shown below, and it performs twice as fast as the above query:
  
  SELECT MIN(Salary)
  FROM
  (
  SELECT TOP 2 Salary
  FROM Employees
  ORDER BY Salary Desc
  ) AS A
  
  This is just an example, the results might differ in different scenarios depending upon the database design, indexes, volume of data etc. So, test all the possible ways a query could be written and go with the efficient one. With some practice and understanding of 'how SQL Server optimizer works', you will be able to come up with the best possible queries without this trial and error method.
• While designing your database, design it keeping 'performance' in mind. You can't really tune performance later, when your database is in production, as it involves rebuilding tables/indexes, re-writing queries. Use the graphical execution plan in Query Analyzer or SHOWPLAN_TEXT or SHOWPLAN_ALL commands to analyze your queries. Make sure your queries do 'Index seeks' instead of 'Index scans' or 'Table scans'. A table scan or an index scan is a very bad thing and should be avoided where possible (sometimes when the table is too small or when the whole table needs to be processed, the optimizer will choose a table or index scan).
• Prefix the table names with owner names, as this improves readability, avoids any unnecessary confusions. Microsoft SQL Server Books Online even states that qualifying tables names, with owner names helps in execution plan reuse.
• Use SET NOCOUNT ON at the beginning of your SQL batches, stored procedures and triggers in production environments, as this suppresses messages like '(1 row(s) affected)' after executing INSERT, UPDATE, DELETE and SELECT statements. This inturn improves the performance of the stored procedures by reducing the network traffic.
• Use the more readable ANSI-Standard Join clauses instead of the old style joins. With ANSI joins the WHERE clause is used only for filtering data. Where as with older style joins, the WHERE clause handles both the join condition and filtering data. The first of the following two queries shows an old style join, while the second one shows the new ANSI join syntax:
  
  SELECT a.au_id, t.title
  FROM titles t, authors a, titleauthor ta
  WHERE
  a.au_id = ta.au_id AND
  ta.title_id = t.title_id AND
  t.title LIKE '%Computer%'
  
  
  SELECT a.au_id, t.title
  FROM authors a
  INNER JOIN
  titleauthor ta
  ON
  a.au_id = ta.au_id
  INNER JOIN
  titles t
  ON
  ta.title_id = t.title_id
  WHERE t.title LIKE '%Computer%'
  
  Be aware that the old style *= and =* left and right outer join syntax may not be supported in a future release of SQL Server, so you are better off adopting the ANSI standard outer join syntax.

• Do not prefix your stored procedure names with 'sp_'. The prefix sp_ is reserved for system stored procedure that ship with SQL Server. Whenever SQL Server encounters a procedure name starting with sp_,, it first tries to locate the procedure in the master database, then looks for any qualifiers (database, owner) provided, then using dbo as the owner. So, you can really save time in locating the stored procedure by avoiding sp_ prefix. But there is an exception! While creating general purpose stored procedures that are called from all your databases, go ahead and prefix those stored procedure names with sp_ and create them in the master database.
• Views are generally used to show specific data to specific users based on their interest. Views are also used to restrict access to the base tables by granting permission on only views. Yet another significant use of views is that, they simplify your queries. Incorporate your frequently required complicated joins and calculations into a view, so that you don't have to repeat those joins/calculations in all your queries, instead just select from the view.
• Use 'User Defined Datatypes', if a particular column repeats in a lot of your tables, so that the datatype of that column is consistent across all your tables.
• Do not let your front-end applications query/manipulate the data directly using SELECT or INSERT/UPDATE/DELETE statements. Instead, create stored procedures, and let your applications access these stored procedures. This keeps the data access clean and consistent across all the modules of your application, at the same time centralizing the business logic within the database.
• Try not to use text, ntext datatypes for storing large textual data. 'text' datatype has some inherent problems associated with it. You can not directly write, update text data using INSERT, UPDATE statements (You have to use special statements like READTEXT, WRITETEXT and UPDATETEXT). There are a lot of bugs associated with replicating tables containing text columns. So, if you don't have to store more than 8 KB of text, use char(8000) or varchar(8000)datatypes.
• If you have a choice, do not store binary files, image files (Binary large objects or BLOBs) etc. inside the database. Instead store the path to the binary/image file in the database and use that as a pointer to the actual binary file. Retrieving, manipulating these large binary files is better performed outside the database and after all, database is not meant for storing files.
• Use char data type for a column, only when the column is non-nullable. If a char column is nullable, it is treated as a fixed length column in SQL Server 7.0+. So, a char(100), when NULL, will eat up 100 bytes, resulting in space wastage. So, use varchar(100) in this situation. Of course, variable length columns do have a very little processing overhead over fixed length columns. Carefully choose between char and varchar depending up on the length of the data you are going to store.
• Avoid dynamic SQL statements as much as possible. Dynamic SQL tends to be slower than static SQL, as SQL Server must generate an execution plan every time at runtime. IF and CASE statements come in handy to avoid dynamic SQL. Another major disadvantage of using dynamic SQL is that, it requires the users to have direct access permissions on all accessed objects like tables and views. Generally, users are given access to the stored procedures which reference the tables, but not directly on the tables. In this case, dynamic SQL will not work. Consider the following scenario, where a user named 'dSQLuser' is added to the pubs database, and is granted access to a procedure named 'dSQLproc', but not on any other tables in the pubs database. The procedure dSQLproc executes a direct SELECT on titles table and that works. The second statement runs the same SELECT on titles table, using dynamic SQL and it fails with the following error:
  
  Server: Msg 229, Level 14, State 5, Line 1
  SELECT permission denied on object 'titles', database 'pubs', owner 'dbo'.
  
  To reproduce the above problem, use the following commands:
  
  sp_addlogin 'dSQLuser'
  GO
  sp_defaultdb 'dSQLuser', 'pubs'
  USE pubs
  GO
  sp_adduser 'dSQLUser', 'dSQLUser'
  GO
  CREATE PROC dSQLProc
  AS
  BEGIN
  SELECT * FROM titles WHERE title_id = 'BU1032' --This works
  DECLARE @str CHAR(100)
  SET @str = 'SELECT * FROM titles WHERE title_id = ''BU1032'''
  EXEC (@str) --This fails
  END
  GO
  GRANT EXEC ON dSQLProc TO dSQLuser
  GO
  
  Now login to the pubs database using the login dSQLuser and execute the procedure dSQLproc to see the problem.
• Consider the following drawbacks before using IDENTITY property for generating primary keys. IDENTITY is very much SQL Server specific, and you will have problems if you want to support different database backends for your application.IDENTITY columns have other inherent problems. IDENTITY columns run out of numbers one day or the other. Numbers can't be reused automatically, after deleting rows. Replication and IDENTITY columns don't always get along well. So, come up with an algorithm to generate a primary key, in the front-end or from within the inserting stored procedure. There could be issues with generating your own primary keys too, like concurrency while generating the key, running out of values. So, consider both the options and go with the one that suits you well.
• Minimize the usage of NULLs, as they often confuse the front-end applications, unless the applications are coded intelligently to eliminate NULLs or convert the NULLs into some other form. Any expression that deals with NULL results in a NULL output. ISNULL and COALESCE functions are helpful in dealing with NULL values. Here's an example that explains the problem:
  
  Consider the following table, Customers which stores the names of the customers and the middle name can be NULL.
  
  CREATE TABLE Customers
  (
  FirstName varchar(20),
  MiddleName varchar(20),
  LastName varchar(20)
  )
  
  Now insert a customer into the table whose name is Tony Blair, without a middle name:
  
  INSERT INTO Customers
  (FirstName, MiddleName, LastName)
  VALUES ('Tony',NULL,'Blair')
  
  The following SELECT statement returns NULL, instead of the customer name:
  
  SELECT FirstName + ' ' + MiddleName + ' ' + LastName FROM Customers
  
  To avoid this problem, use ISNULL as shown below:
  
  SELECT FirstName + ' ' + ISNULL(MiddleName + ' ','') + LastName FROM Customers
• Use Unicode datatypes like nchar, nvarchar, ntext, if your database is going to store not just plain English characters, but a variety of characters used all over the world. Use these datatypes, only when they are absolutely needed as they need twice as much space as non-unicode datatypes.
• Always use a column list in your INSERT statements. This helps in avoiding problems when the table structure changes (like adding a column). Here's an example which shows the problem.
  
  Consider the following table:
  
  CREATE TABLE EuropeanCountries
  (
  CountryID int PRIMARY KEY,
  CountryName varchar(25)
  )
  
  Here's an INSERT statement without a column list , that works perfectly:
  
  INSERT INTO EuropeanCountries
  VALUES (1, 'Ireland')
  
  Now, let's add a new column to this table:
  
  
  ALTER TABLE EuropeanCountries
  ADD EuroSupport bit
  
  Now run the above INSERT statement. You get the following error from SQL Server:
  
  Server: Msg 213, Level 16, State 4, Line 1
  Insert Error: Column name or number of supplied values does not match table definition.
  
  This problem can be avoided by writing an INSERT statement with a column list as shown below:
  
  INSERT INTO EuropeanCountries
  (CountryID, CountryName)
  VALUES (1, 'England')
• Perform all your referential integrity checks, data validations using constraints (foreign key and check constraints). These constraints are faster than triggers. So, use triggers only for auditing, custom tasks and validations that can not be performed using these constraints. These constraints save you time as well, as you don't have to write code for these validations and the RDBMS will do all the work for you.
• Always access tables in the same order in all your stored procedures/triggers consistently. This helps in avoiding deadlocks. Other things to keep in mind to avoid deadlocks are: Keep your transactions as short as possible. Touch as less data as possible during a transaction. Never, ever wait for user input in the middle of a transaction. Do not use higher level locking hints or restrictive isolation levels unless they are absolutely needed. Make your front-end applications deadlock-intelligent, that is, these applications should be able to resubmit the transaction incase the previous transaction fails with error 1205. In your applications, process all the results returned by SQL Server immediately, so that the locks on the processed rows are released, hence no blocking.
• Offload tasks like string manipulations, concatenations, row numbering, case conversions, type conversions etc. to the front-end applications, if these operations are going to consume more CPU cycles on the database server (It's okay to do simple string manipulations on the database end though). Also try to do basic validations in the front-end itself during data entry. This saves unnecessary network roundtrips.
• If back-end portability is your concern, stay away from bit manipulations with T-SQL, as this is very much RDBMS specific. Further, using bitmaps to represent different states of a particular entity conflicts with the normalization rules.
• Consider adding a @Debug parameter to your stored procedures. This can be of bit data type. When a 1 is passed for this parameter, print all the intermediate results, variable contents using SELECT or PRINT statements and when 0 is passed do not print debug information. This helps in quick debugging of stored procedures, as you don't have to add and remove these PRINT/SELECT statements before and after troubleshooting problems.
• Do not call functions repeatedly within your stored procedures, triggers, functions and batches. For example, you might need the length of a string variable in many places of your procedure, but don't call the LEN function whenever it's needed, instead, call the LEN function once, and store the result in a variable, for later use.
• Make sure your stored procedures always return a value indicating the status. Standardize on the return values of stored procedures for success and failures. The RETURN statement is meant for returning the execution status only, but not data. If you need to return data, use OUTPUT parameters.
• If your stored procedure always returns a single row resultset, consider returning the resultset using OUTPUT parameters instead of a SELECT statement, as ADO handles output parameters faster than resultsets returned by SELECT statements.
• Always check the global variable @@ERROR immediately after executing a data manipulation statement (like INSERT/UPDATE/DELETE), so that you can rollback the transaction in case of an error (@@ERROR will be greater than 0 in case of an error). This is important, because, by default, SQL Server will not rollback all the previous changes within a transaction if a particular statement fails. This behavior can be changed by executing SET XACT_ABORT ON. The @@ROWCOUNT variable also plays an important role in determining how many rows were affected by a previous data manipulation (also, retrieval) statement, and based on that you could choose to commit or rollback a particular transaction.
• To make SQL Statements more readable, start each clause on a new line and indent when needed. Following is an example:
  
  SELECT title_id, title
  FROM titles
  WHERE title LIKE 'Computing%' AND
  title LIKE 'Gardening%'
• Though we survived the Y2K, always store 4 digit years in dates (especially, when using char or int datatype columns), instead of 2 digit years to avoid any confusion and problems. This is not a problem with datetime columns, as the century is stored even if you specify a 2 digit year. But it's always a good practice to specify 4 digit years even with datetime datatype columns.
• In your queries and other SQL statements, always represent date in yyyy/mm/dd format. This format will always be interpreted correctly, no matter what the default date format on the SQL Server is. This also prevents the following error, while working with dates:
  
  Server: Msg 242, Level 16, State 3, Line 2
  The conversion of a char data type to a datetime data type resulted in an out-of-range datetime value.
• As is true with any other programming language, do not use GOTO or use it sparingly. Excessive usage of GOTO can lead to hard-to-read-and-understand code.

• Do not forget to enforce unique constraints on your alternate keys.

• Always be consistent with the usage of case in your code. On a case insensitive server, your code might work fine, but it will fail on a case sensitive SQL Server if your code is not consistent in case. For example, if you create a table in SQL Server or database that has a case-sensitive or binary sort order, all references to the table must use the same case that was specified in the CREATE TABLE statement. If you name the table as 'MyTable' in the CREATE TABLE statement and use 'mytable' in the SELECT statement, you get an 'object not found' or 'invalid object name' error.
• Though T-SQL has no concept of constants (like the ones in C language), variables will serve the same purpose. Using variables instead of constant values within your SQL statements, improves readability and maintainability of your code. Consider the following example:
  
  UPDATE dbo.Orders
  SET OrderStatus = 5
  WHERE OrdDate < '2001/10/25'
  
  The same update statement can be re-written in a more readable form as shown below:
  
  DECLARE @ORDER_PENDING int
  SET @ORDER_PENDING = 5
  
  UPDATE dbo.Orders
  SET OrderStatus = @ORDER_PENDING
  WHERE OrdDate < '2001/10/25'
  
  
• Do not use the column numbers in the ORDER BY clause as it impairs the readability of the SQL statement. Further, changing the order of columns in the SELECT list has no impact on the ORDER BY when the columns are referred by names instead of numbers. Consider the following example, in which the second query is more readable than the first one:
  
  SELECT OrderID, OrderDate
  FROM Orders
  ORDER BY 2
  
  SELECT OrderID, OrderDate
  FROM Orders
  ORDER BY OrderDate
  
  

Well, this is all for now folks. I'll keep updating this page as and when I have something new to add. I welcome your feedback on this, so feel free to email me. Happy database programming!