Coding Techniques and Programming Practices
Rob Caron
Microsoft Corporation
February 2000
Summary: This article provides coding techniques and programming practices for improving the quality of source code. (12 printed pages)
Introduction
Superior coding techniques and programming practices are hallmarks of a professional programmer. The bulk of programming consists of making a large number of small choices while attempting to solve a larger set of problems. How wisely those choices are made depends largely upon the programmer's skill and expertise.
This document addresses some fundamental coding techniques and provides a collection of coding practices from which to learn. The coding techniques are primarily those that improve the readability and maintainability of code, whereas the programming practices are mostly performance enhancements.
The readability of source code has a direct impact on how well a developer comprehends a software system. Code maintainability refers to how easily that software system can be changed to add new features, modify existing features, fix bugs, or improve performance. Although readability and maintainability are the result of many factors, one particular facet of software development upon which all developers have an influence is coding technique. The easiest method to ensure that a team of developers will yield quality code is to establish a coding standard, which is then enforced at routine code reviews.
Coding Standards and Code Reviews
A comprehensive coding standard encompasses all aspects of code construction and, while developers should exercise prudence in its implementation, it should be closely followed. Completed source code should reflect a harmonized style, as if a single developer wrote the code in one session. At the inception of a software project, establish a coding standard to ensure that all developers on the project are working in concert. When the software project will incorporate existing source code, or when performing maintenance upon an existing software system, the coding standard should state how to deal with the existing code base.
Although the primary purpose for conducting code reviews throughout the development life cycle is to identify defects in the code, the reviews can also be used to enforce coding standards in a uniform manner. Adherence to a coding standard can only be feasible when followed throughout the software project from inception to completion. It is not practical, nor is it prudent, to impose a coding standard after the fact.
Coding Techniques
Coding techniques incorporate many facets of software development and, although they usually have no impact on the functionality of the application, they contribute to an improved comprehension of source code. For the purpose of this document, all forms of source code are considered, including programming, scripting, markup, and query languages.
The coding techniques defined here are not proposed to form an inflexible set of coding standards. Rather, they are meant to serve as a guide for developing a coding standard for a specific software project.
The coding techniques are divided into three sections:
Names
Perhaps one of the most influential aids to understanding the logical flow of an application is how the various elements of the application are named. A name should tell "what" rather than "how." By avoiding names that expose the underlying implementation, which can change, you preserve a layer of abstraction that simplifies the complexity. For example, you could use GetNextStudent()
instead of GetNextArrayElement()
.
A tenet of naming is that difficulty in selecting a proper name may indicate that you need to further analyze or define the purpose of an item. Make names long enough to be meaningful but short enough to avoid being wordy. Programmatically, a unique name serves only to differentiate one item from another. Expressive names function as an aid to the human reader; therefore, it makes sense to provide a name that the human reader can comprehend. However, be certain that the names chosen are in compliance with the applicable language's rules and standards.
Following are recommended naming techniques:
Routines
- Avoid elusive names that are open to subjective interpretation, such as
Analyze()
for a routine, or xxK8
for a variable. Such names contribute to ambiguity more than abstraction.
- In object-oriented languages, it is redundant to include class names in the name of class properties, such as
Book.BookTitle
. Instead, use Book.Title
.
- Use the verb-noun method for naming routines that perform some operation on a given object, such as
CalculateInvoiceTotal()
.
- In languages that permit function overloading, all overloads should perform a similar function. For those languages that do not permit function overloading, establish a naming standard that relates similar functions.
Variables
- Append computation qualifiers (Avg, Sum, Min, Max, Index) to the end of a variable name where appropriate.
- Use customary opposite pairs in variable names, such as min/max, begin/end, and open/close.
- Since most names are constructed by concatenating several words together, use mixed-case formatting to simplify reading them. In addition, to help distinguish between variables and routines, use Pascal casing (
CalculateInvoiceTotal
) for routine names where the first letter of each word is capitalized. For variable names, use camel casing (documentFormatType
) where the first letter of each word except the first is capitalized.
- Boolean variable names should contain
Is
which implies Yes/No
or True/False
values, such as fileIsFound
.
- Avoid using terms such as
Flag
when naming status variables, which differ from Boolean variables in that they may have more than two possible values. Instead of documentFlag
, use a more descriptive name such as documentFormatType
.
- Even for a short-lived variable that may appear in only a few lines of code, still use a meaningful name. Use single-letter variable names, such as
i
, or j,
for short-loop indexes only.
- If using Charles Simonyi's Hungarian Naming Convention, or some derivative thereof, develop a list of standard prefixes for the project to help developers consistently name variables. For more information, see "Hungarian Notation."
- For variable names, it is sometimes useful to include notation that indicates the scope of the variable, such as prefixing a
g_
for global variables and m_
for module-level variables in Microsoft Visual Basic®.
- Constants should be all uppercase with underscores between words, such as
NUM_DAYS_IN_WEEK
. Also, begin groups of enumerated types with a common prefix, such as FONT_ARIAL
and FONT_ROMAN
.
Tables
- When naming tables, express the name in the singular form. For example, use
Employee
instead of Employees
.
- When naming columns of tables, do not repeat the table name; for example, avoid having a field called
EmployeeLastName
in a table called Employee
.
- Do not incorporate the data type in the name of a column. This will reduce the amount of work needed should it become necessary to change the data type later.
Microsoft SQL Server
- Do not prefix stored procedures with
sp
_, because this prefix is reserved for identifying system-stored procedures.
- In Transact-SQL, do not prefix variables with
@@
, which should be reserved for truly global variables such as @@IDENTITY
.
Miscellaneous
- Minimize the use of abbreviations. If abbreviations are used, be consistent in their use. An abbreviation should have only one meaning and likewise, each abbreviated word should have only one abbreviation. For example, if using min to abbreviate minimum, do so everywhere and do not later use it to abbreviate minute.
- When naming functions, include a description of the value being returned, such as
GetCurrentWindowName()
.
- File and folder names, like procedure names, should accurately describe what purpose they serve.
- Avoid reusing names for different elements, such as a routine called
ProcessSales()
and a variable called iProcessSales
.
- Avoid homonyms when naming elements to prevent confusion during code reviews, such as write and right.
- When naming elements, avoid using commonly misspelled words. Also, be aware of differences that exist between American and British English, such as color/colour and check/cheque.
- Avoid using typographical marks to identify data types, such as $ for strings or % for integers.
Comments
Software documentation exists in two forms, external and internal. External documentation is maintained outside of the source code, such as specifications, help files, and design documents. Internal documentation is composed of comments that developers write within the source code at development time.
One of the challenges of software documentation is ensuring that the comments are maintained and updated in parallel with the source code. Although properly commenting source code serves no purpose at run time, it is invaluable to a developer who must maintain a particularly intricate or cumbersome piece of software.
Following are recommended commenting techniques:
- When modifying code, always keep the commenting around it up to date.
- At the beginning of every routine, it is helpful to provide standard, boilerplate comments, indicating the routine's purpose, assumptions, and limitations. A boilerplate comment should be a brief introduction to understand why the routine exists and what it can do.
- Avoid adding comments at the end of a line of code; end-line comments make code more difficult to read. However, end-line comments are appropriate when annotating variable declarations. In this case, align all end-line comments at a common tab stop.
- Avoid using clutter comments, such as an entire line of asterisks. Instead, use white space to separate comments from code.
- Avoid surrounding a block comment with a typographical frame. It may look attractive, but it is difficult to maintain.
- Prior to deployment, remove all temporary or extraneous comments to avoid confusion during future maintenance work.
- If you need comments to explain a complex section of code, examine the code to determine if you should rewrite it. If at all possible, do not document bad code—rewrite it. Although performance should not typically be sacrificed to make the code simpler for human consumption, a balance must be maintained between performance and maintainability.
- Use complete sentences when writing comments. Comments should clarify the code, not add ambiguity.
- Comment as you code, because most likely there won't be time to do it later. Also, should you get a chance to revisit code you've written, that which is obvious today probably won't be obvious six weeks from now.
- Avoid the use of superfluous or inappropriate comments, such as humorous sidebar remarks.
- Use comments to explain the intent of the code. They should not serve as inline translations of the code.
- Comment anything that is not readily obvious in the code.
- To prevent recurring problems, always use comments on bug fixes and work-around code, especially in a team environment.
- Use comments on code that consists of loops and logic branches. These are key areas that will assist the reader when reading source code.
- Separate comments from comment delimiters with white space. Doing so will make comments stand out and easier to locate when viewed without color clues.
- Throughout the application, construct comments using a uniform style, with consistent punctuation and structure.
Notes Despite the availability of external documentation, source code listings should be able to stand on their own because hard-copy documentation can be misplaced.
External documentation should consist of specifications, design documents, change requests, bug history, and the coding standard that was used.
Format
Formatting makes the logical organization of the code stand out. Taking the time to ensure that the source code is formatted in a consistent, logical manner is helpful to yourself and to other developers who must decipher the source code.
Following are recommended formatting techniques:
- Establish a standard size for an indent, such as four spaces, and use it consistently. Align sections of code using the prescribed indentation.
- Use a monospace font when publishing hard-copy versions of the source code.
- Except for constants, which are best expressed in all uppercase characters with underscores, use mixed case instead of underscores to make names easier to read.
- Align open and close braces vertically where brace pairs align, such as:
for (i = 0; i < 100; i++)
{
…
}
You can also use a slanting style, where open braces appear at the end of the line and close braces appear at the beginning of the line, such as:
for (i = 0; i < 100; i++){
…
}
Whichever style is chosen, use that style throughout the source code.
- Indent code along the lines of logical construction. Without indenting, code becomes difficult to follow, such as:
If … Then
If … Then
…
Else
…
End If
Else
…
End If
Indenting the code yields easier-to-read code, such as:
If … Then
If … Then
…
Else
…
End If
Else
…
End If
- Establish a maximum line length for comments and code to avoid having to scroll the source code window and to allow for clean hard-copy presentation.
- Use spaces before and after most operators when doing so does not alter the intent of the code. For example, an exception is the pointer notation used in C++.
- Put a space after each comma in comma-delimited lists, such as array values and arguments, when doing so does not alter the intent of the code. For example, an exception is an ActiveX® Data Object (ADO) Connection argument.
- Use white space to provide organizational clues to source code. Doing so creates "paragraphs" of code, which aid the reader in comprehending the logical segmenting of the software.
- When a line is broken across several lines, make it obvious that the line is incomplete without the following line.
- Where appropriate, avoid placing more than one statement per line. An exception is a loop in C, C++, Visual J++®, or JScript®, such as
for (i = 0; i < 100; i++)
.
- When writing HTML, establish a standard format for tags and attributes, such as using all uppercase for tags and all lowercase for attributes. As an alternative, adhere to the XHTML specification to ensure all HTML documents are valid. Although there are file size trade-offs to consider when creating Web pages, use quoted attribute values and closing tags to ease maintainability.
- When writing SQL statements, use all uppercase for keywords and mixed case for database elements, such as tables, columns, and views.
- Divide source code logically between physical files.
- In ASP, use script delimiters around blocks of script rather than around each line of script or interspersing small HTML fragments with server-side scripting. Using script delimiters around each line or interspersing HTML fragments with server-side scripting increases the frequency of context switching on the server side, which hampers performance and degrades code readability.
- Put each major SQL clause on a separate line so statements are easier to read and edit, for example:
SELECT FirstName, LastName
FROM Customers
WHERE State = 'WA'
- Do not use literal numbers or literal strings, such as
For i = 1 To 7
. Instead, use named constants, such as For i = 1 To NUM_DAYS_IN_WEEK
, for ease of maintenance and understanding.
- Break large, complex sections of code into smaller, comprehensible modules.
Programming Practices
Experienced developers follow numerous programming practices or rules of thumb, which typically derived from hard-learned lessons. The practices listed below are not all-inclusive, and should not be used without due consideration. Veteran programmers deviate from these practices on occasion, but not without careful consideration of the potential repercussions. Using the best programming practice in the wrong context can cause more harm than good.
- To conserve resources, be selective in the choice of data type to ensure the size of a variable is not excessively large.
- Keep the lifetime of variables as short as possible when the variables represent a finite resource for which there may be contention, such as a database connection.
- Keep the scope of variables as small as possible to avoid confusion and to ensure maintainability. Also, when maintaining legacy source code, the potential for inadvertently breaking other parts of the code can be minimized if variable scope is limited.
- Use variables and routines for one and only one purpose. In addition, avoid creating multipurpose routines that perform a variety of unrelated functions.
- When writing classes, avoid the use of public variables. Instead, use procedures to provide a layer of encapsulation and also to allow an opportunity to validate value changes.
- When using objects pooled by MTS, acquire resources as late as possible and release them as soon as possible. As such, you should create objects as late as possible, and destroy them as early as possible to free resources.
- When using objects that are not being pooled by MTS, it is necessary to examine the expense of the object creation and the level of contention for resources to determine when resources should be acquired and released.
- Use only one transaction scheme, such as MTS or SQL Server™, and minimize the scope and duration of transactions.
- Be wary of using ASP Session variables in a Web farm environment. At a minimum, do not place objects in ASP Session variables because session state is stored on a single machine. Consider storing session state in a database instead.
- Stateless components are preferred when scalability or performance are important. Design the components to accept all the needed values as input parameters instead of relying upon object properties when calling methods. Doing so eliminates the need to preserve object state between method calls. When it is necessary to maintain state, consider using alternative methods, such as maintaining state in a database.
- Do not open data connections using a specific user's credentials. Connections that have been opened using such credentials cannot be pooled and reused, thus losing the benefits of connection pooling.
- Avoid the use of forced data conversion, sometimes referred to as variable coercion or casting, which may yield unanticipated results. This occurs when two or more variables of different data types are involved in the same expression. When it is necessary to perform a cast for other than a trivial reason, that reason should be provided in an accompanying comment.
- Develop and use error-handling routines. For more information on error handling in Visual Basic, see the "Error Handling and Debugging" chapter of the Microsoft Office 2000/Visual Basic Programmer's Guide, available in the MSDN Library. For more information on error handling and COM, see "Error Handling" in the Platform SDK. For more information on error handling for Web pages, see http://msdn.microsoft.com/workshop/author/script/weberrors.asp.
- Be specific when declaring objects, such as
ADODB.Recordset
instead of just Recordset
, to avoid the risk of name collisions.
- Require the use
Option Explicit
in Visual Basic and VBScript to encourage forethought in the use of variables and to minimize errors resulting from typographical errors.
- Avoid the use of variables with application scope.
- Use
RETURN
statements in stored procedures to help the calling program know whether the procedure worked properly.
- Use early binding techniques whenever possible. For more information on early versus late binding, see "A Primer on Early Binding (or How to Make OLE Automation Faster)."
- Use
Select Case
or Switch
statements in lieu of repetitive checking of a common variable using If…Then
statements.
- Explicitly release object references.
Data-Specific
- Never use
SELECT *
. Always be explicit in which columns to retrieve and retrieve only the columns that are required.
- Refer to fields implicitly; do not reference fields by their ordinal placement in a Recordset.
- Use stored procedures in lieu of SQL statements in source code to leverage the performance gains they provide.
- Use a stored procedure with output parameters instead of single-record
SELECT
statements when retrieving one row of data.
- Verify the row count when performing
DELETE
operations.
- Perform data validation at the client during data entry. Doing so avoids unnecessary round trips to the database with invalid data.
- Avoid using functions in
WHERE
clauses.
- If possible, specify the primary key in the
WHERE
clause when updating a single row.
- When using
LIKE
, do not begin the string with a wildcard character because SQL Server will not be able to use indexes to search for matching values.
- Use
WITH RECOMPILE
in CREATE PROC
when a wide variety of arguments are passed, because the plan stored for the procedure might not be optimal for a given set of parameters.
- Stored procedure execution is faster when you pass parameters by position (the order in which the parameters are declared in the stored procedure) rather than by name.
- Use triggers only for data integrity enforcement and business rule processing and not to return information.
- After each data modification statement inside a transaction, check for an error by testing the global variable
@@ERROR
.
- Use forward-only/read-only recordsets. To update data, use SQL INSERT and UPDATE statements.
- Never hold locks pending user input.
- Use uncorrelated subqueries instead of correlated subqueries. Uncorrelated subqueries are those where the inner
SELECT
statement does not rely on the outer SELECT
statement for information. In uncorrelated subqueries, the inner query is run once instead of being run for each row returned by the outer query.
ADO-Specific
- Tune the
RecordSet.CacheSize
property to what is needed. Using too small or too large a setting will adversely impact the performance of an application.
- Bind columns to field objects when looping through recordsets.
- For Command objects, describe the parameters manually instead of using
Parameters.Refresh
to obtain parameter information.
- Explicitly close ADO Recordset and Connection objects to insure that connections are promptly returned to the connection pool for use by other processes.
- Use
adExecuteNoRecords
for non-row-returning commands.
Conclusion
Using solid coding techniques and good programming practices to create high quality code plays an important role in software quality and performance. In addition, by consistently applying a well-defined coding standard and proper coding techniques, and holding routine code reviews, a team of programmers working on a software project is more likely to yield a software system that is easier to comprehend and maintain.
Suggested Reading
Books:
- McConnell, Steve. Code Complete. Redmond, WA: Microsoft Press, 1993.
- McConnell, Steve. Software Project Survival Guide. Redmond, WA: Microsoft Press, 1998.
- Maguire, Steve. Writing Solid Code. Redmond, WA: Microsoft Press, 1993.
- Ranade and Nash. The Elements of C Programming Style. New York: McGraw-Hill, 1992.
- Kernighan and Plauger. The Elements of Programming Style. New York: McGraw-Hill, 1982.
- Caner, Cem, et al. Testing Computer Software, Second Edition. New York: Wiley Computer Publishing, 1999.
- Hunt, Andrew, et al. The Pragmatic Programmer. Reading, MA: Addison-Wesley, 2000.
- Holub, Allen. Enough Rope to Shoot Yourself in the Foot: Rules for C and C++ Programming. New York: McGraw-Hill, 1995.
MSDN Library Articles:
Web Workshop Articles: