The C language provides several different kinds of constants: integer constants, such as 10
and 0x1C
, ; floating constants, such as 1.0
and 6.022e+23
, ; and character constants, such as 'a'
and '\x10'
. C also provides string literals, such as "hello, world"
and "\n"
. These may constants can all be referred to as literals.
When used in program logic, literals can reduce the readability of source code. As a result, literals , in general, and integer constants , in particular, are frequently called _magic numbers_ because their purpose is often obscured. Magic numbers may can be constant values that represent either an arbitrary value (such as a determined appropriate buffer size) or a malleable concept (such as the age at which a person is considered an adult, which can change between geopolitical boundaries). Rather than embed literals in program logic, use appropriately named symbolic constants to clarify the intent of the code. In addition, if a specific value needs to be changed, reassigning a symbolic constant once is more efficient and less error prone than replacing every instance of the value \[ [Saks 2002|AA. Bibliography#Saks 02]\]. Wiki Markup
The C programming language has several mechanisms for creating named, symbolic constants: const
-qualified objects, enumeration constants, and object-like macro definitions. Each of these mechanisms has associated advantages and disadvantages.
const
-
...
Qualified Objects
Objects that are const
-qualified have scope and can be type-checked by the compiler. Because these they are named objects (unlike macro definitions), some debugging tools can show the name of the object. The object also consumes memory.
A const
-qualified object allows you to specify the exact type of the constant. For example,
Code Block |
---|
const unsigned int buffer_size = 256;
|
...
Unfortunately, const
-qualified objects cannot be used where compile-time integer constants are required, namely to define the
- size Size of a bit-field member of a structure.
- size Size of an array (except in the case of variable length arrays).
- value Value of an enumeration constant.
- value Value of a
case
constant.
If any of these are required, then an integer constant (which would be an rvalue) must be used.
const
-qualified objects allow the programmer to take the address of the object.:
Code Block |
---|
const int max = 15; int a[max]; /* invalidInvalid declaration outside of a function */ const int *p; /* aA const-qualified object can have its address taken */ p = &max; |
...
{{const
}}\-qualified objects are likely to incur some runtime overhead \ [[Saks 2001b|AA. Bibliography#Saks 02]\]. Most C compilers, for example, allocate memory for {{const
}}\-qualified objects. {{const
}}\-qualified objects declared inside a function body may can have automatic storage duration. If so, the compiler will allocate storage for the object, and it will be on the stack. As a result, this storage will need to be allocated and initialized each time the containing function is invoked.
Enumeration Constants
Enumeration constants can be used to represent an integer constant expression that has a value representable as an int
. Unlike const
-qualified objects, enumeration constants do not consume memory. No storage is allocated for the value, so it is not possible to take the address of an enumeration constant.
Code Block |
---|
enum { max = 15 }; int a[max]; /* OK outside function */ const int *p; p = &max; /* errorError: '"&'" on enum constant */ |
Enumeration constants do not allow the type of the value to be specified. An enumeration constant whose value can be represented as an int
is always an int
.
Object-
...
like Macros
A preprocessing directive of the form:
#
define
identifier replacement-list
...
defines an _object-like_ macro that causes each subsequent instance of the macro name to be replaced by the replacement list of preprocessing tokens that constitute the remainder of the directive \[[ISO/IEC 9899:1999|AA. Bibliography#ISO/IEC 9899-1999]\].
C programmers frequently define symbolic constants as object-like macros. For example, the code
Code Block |
---|
#define buffer_size 256
|
defines buffer_size
as a macro whose value is 256. The preprocessor substitutes macros before the compiler does any other symbol processing. Later compilation phases never see macro symbols, such as buffer_size
; they see only the source text after macro substitution. As a result, many compilers do not preserve macro names among the symbols they pass on to their debuggers.
Macro names do not observe the scope rules that apply to other names. ResultantlyTherefore, macros might could substitute in unanticipated places with unexpected results.
Object-like macros do not consume memory, and ; consequently, it is not possible to create a pointer to one. Macros do not provide for type checking , as because they are textually replaced by the preprocessor.
Macros may can be passed as compile-time arguments.
Summary
The following table summarizes some of the differences between const
-qualified objects, enumeration constants, and object-like macro definitions.
Method | Evaluated |
---|
At | Consumes Memory | Viewable by Debuggers | Type Checking | Compile- |
---|
Time Constant Expression |
---|
Enumerations |
Compile time |
No |
Yes |
Yes |
Yes |
|
Runtime |
Yes |
Yes |
Yes |
No |
Macros |
Preprocessor |
No |
No |
No |
Yes |
Noncompliant Code Example
The meaning of the integer literal 18 is not clear in this example.:
Code Block | ||||
---|---|---|---|---|
| ||||
/* ... */
if (age >= 18) {
/* Take action */
}
else {
/* Take a different action */
}
/* ... */
|
...
This compliant solution replaces the integer literal 18 with the symbolic constant ADULT_AGE
to clarify the meaning of the code.:
Code Block | ||||
---|---|---|---|---|
| ||||
enum { ADULT_AGE=18 };
/* ... */
if (age >= ADULT_AGE) {
/* Take action */
}
else {
/* Take a different action */
}
/* ... */
|
...
Integer literals are frequently used when referring to array dimensions, as shown in this noncompliant code example.:
Code Block | ||||
---|---|---|---|---|
| ||||
char buffer[256];
/* ... */
fgets(buffer, 256, stdin);
|
This use of integer literals can easily result in buffer overflows if, if for example, the buffer size is reduced but the integer literal used in the call to fgets()
is not.
Compliant Solution (enum
)
In this compliant solution, the integer literal is replaced with an enumeration constant. (See guideline DCL00-C. Const-qualify immutable objects.)
Code Block | ||||
---|---|---|---|---|
| ||||
enum { BUFFER_SIZE=256 };
char buffer[BUFFER_SIZE];
/* ... */
fgets(buffer, BUFFER_SIZE, stdin);
|
...
Frequently, it is possible to obtain the desired readability by using a symbolic expression composed of existing symbols rather than by defining a new symbol. For example, a sizeof
expression can work just as well as an enumeration constant. (See guideline EXP09-C. Use sizeof to determine the size of a type or variable.)
Code Block | ||||
---|---|---|---|---|
| ||||
char buffer[256];
/* ... */
fgets(buffer, sizeof(buffer), stdin);
|
Using the {{ Wiki Markup sizeof
}} expression in this example reduces the total number of names declared in the program, which is generally a good idea \[ [Saks 2002|AA. Bibliography#Saks 02]\]. The {{sizeof
}} operator is almost always evaluated at compile time (except in the case of variable-length arrays).
When working with sizeof()
, keep in mind guideline ARR01-C. Do not apply the sizeof operator to a pointer when taking the size of an array.
...
In this noncompliant code example, the string literal "localhost"
and integer constant 1234
are embedded directly in program logic , and are consequently difficult to change.:
Code Block | ||||
---|---|---|---|---|
| ||||
LDAP *ld = ldap_init("localhost", 1234);
if (ld == NULL) {
perror("ldap_init");
return(1);
}
|
...
In this compliant solution, the host name and port number are both defined as object-like macros, so that may so they can be passed as compile-time arguments.:
Code Block | ||||
---|---|---|---|---|
| ||||
#ifndef PORTNUMBER /* mightMight be passed on compile line */ # define PORTNUMBER 1234 #endif #ifndef HOSTNAME /* mightMight be passed on compile line */ # define HOSTNAME "localhost" #endif /* ... */ LDAP *ld = ldap_init(HOSTNAME, PORTNUMBER); if (ld == NULL) { perror("ldap_init"); return(1); } |
Exceptions
DCL06-C-EX1: While Although replacing numeric constants with a symbolic constant is often a good practice, it can be taken too far. Remember that the goal is to improve readability. Exceptions can be made for constants that are themselves the abstraction you want to represent, as in this compliant solution.
Code Block | ||||
---|---|---|---|---|
| ||||
x = (-b + sqrt(b*b - 4*a*c)) / (2*a);
|
Replacing numeric constants with symbolic constants in this example does nothing to improve the readability of the code and may can actually make the code more difficult to read.
Code Block |
---|
enum { TWO = 2 }; /* aA scalar */ enum { FOUR = 4 }; /* aA scalar */ enum { SQUARE = 2 }; /* anAn exponent */ x = (-b + sqrt(pow(b, SQUARE) - FOUR*a*c))/ (TWO * a); |
...
Note that this example does not check for invalid operations (taking the sqrt()
of a negative number). See guideline FLP32-C. Prevent or detect domain and range errors in math functions for more information on detecting domain and range errors in math functions.
...
Using numeric literals makes code more difficult to read and understand. Buffer overruns are frequently a consequence of a magic number being changed in one place (like such as in an array declaration) but not elsewhere (like such as in a loop through an array).
Recommendation | Severity | Likelihood | Remediation Cost | Priority | Level |
---|---|---|---|---|---|
DCL06-C |
Low |
Unlikely |
Medium | P2 | L3 |
Automated Detection
Tool | Version | Checker | Description |
---|
Axivion Bauhaus Suite |
|
|
|
CertC-DCL06 | |
Compass/ROSE |
Could detect violations of this recommendation |
merely by searching for the use of |
"magic numbers |
" and magic strings in the code itself. That is, any number ( |
except a few canonical numbers: |
−1, 0, 1, 2) that appears in the code |
anywhere besides where assigned to a variable is a magic number |
and should instead be assigned to a |
besides where assigned to a | |||||||||
| CC2.DCL06 | Fully implemented | |||||||
Helix QAC |
| C3120, C3121, C3122, C3123, C3131, C3132 | |||||||
LDRA tool suite |
| 201 S | Fully implemented | ||||||
Parasoft C/C++test |
| CERT_C-DCL06-a | Use meaningful symbolic constants to represent literal values | ||||||
Polyspace Bug Finder |
| Checks for:
Rec. fully covered. |
Related Vulnerabilities
Search for vulnerabilities resulting from the violation of this rule on the CERT website.
Related Guidelines
...
...
Bibliography
Wiki Markup |
---|
\[[Henricson 1992|AA. Bibliography#Henricson 92]\] Chapter 10, "[Constants|http://www.doc.ic.ac.uk/lab/cplus/c++.rules/chap10.html]"
\[[ISO/IEC 9899:1999|AA. Bibliography#ISO/IEC 9899-1999]\] Section 6.3.2.1, "Lvalues, arrays, and function designators," Section 6.7, "Declarations," Section 6.7.2.2, "Enumeration specifiers," and Section 6.10.3, "Macro replacement"
\[[ISO/IEC PDTR 24772|AA. Bibliography#ISO/IEC PDTR 24772]\] "BRS Leveraging human experience"
\[[MITRE 2007|AA. Bibliography#MITRE 07]\] [CWE ID 547|http://cwe.mitre.org/data/definitions/547.html], "Use of Hard-coded, Security-relevant Constants"
\[[Saks 2001a|AA. Bibliography#Saks 01]\]
\[[Saks 2001b|AA. Bibliography#Saks 02]\]
\[[Saks 2002|AA. Bibliography#Saks 02]\]
\[[Summit 2005|AA. Bibliography#Summit 05]\] [Question 10.5b|http://c-faq.com/cpp/constvsdefine.html] |
Bibliography
[Henricson 1992] | Chapter 10, "Constants" |
[Saks 2001a] | |
[Saks 2001b] | |
[Saks 2002] | |
[Summit 2005] | Question 10.5b |
...
DCL05-C. Use typedefs to improve code readability 02. Declarations and Initialization (DCL)