Modern C++
Programming
6. Basic Concepts V
Functions and Preprocessing
Federico Busato
2024-03-29
Overview
A function (procedure or routine) is a piece of code that performs a specific
task
Purpose:
• Avoiding code duplication: less code for the same functionality → less
bugs
• Readability: better express what the code does
• Organization: break the code in separate modules
5/61
Function Parameter and Argument
Function Parameter [formal]
A parameter is the variable which is part of the method signature
Function Argument [actual]
An argument is the actual value (instance) of the variable that gets passed to the
function
void f(int a, char* b); // parameters: int a, char* b
// return type: void
f(3, "abc"); // arguments: 3, "abc"
6/61
Pass by-Value
Call-by-value
The object is copied and assigned to input arguments of the method f(T x)
Advantages:
• Changes made to the parameter inside the function have no effect on the argument
Disadvantages:
• Performance penalty if the copied arguments are large (e.g. a structure with several data
members)
When to use:
• Built-in data type and small objects (≤ 8 bytes)
When not to use:
• Fixed size arrays which decay into pointers
• Large objects
7/61
Pass by-Pointer
Call-by-pointer
The address of a variable is copied and assigned to input arguments of the method
f(T* x)
Advantages:
• Allows a function to change the value of the argument
• The argument is not copied (fast)
Disadvantages:
• The argument may be a null pointer
• Dereferencing a pointer is slower than accessing a value directly
When to use:
• Raw arrays (use const T* if read-only)
When not to use:
• All other cases
8/61
Pass by-Reference
Call-by-reference
The reference of a variable is copied and assigned to input arguments of the method
f(T& x)
Advantages:
• Allows a function to change the value of the argument (better readability compared with
pointers)
• The argument is not copied (fast)
• References must be initialized (no null pointer)
• Avoid implicit conversion (without const T& )
When to use:
• All cases except raw pointers
When not to use:
• Pass by-value could give performance advantages and improve the readability with built-in
data type and small objects that are trivially copyable
9/61
Examples
struct MyStruct;
void f1(int a); // pass by-value
void f2(int& a); // pass by-reference
void f3(const int& a); // pass by-const reference
void f4(MyStruct& a); // pass by-reference
void f5(int* a); // pass by-pointer
void f6(const int* a); // pass by-const pointer
void f7(MyStruct* a); // pass by-pointer
void f8(int*& a); // pass a pointer by-reference
//--------------------------------------------------------------
char c = 'a';
f1(c); // ok, pass by-value (implicit conversion)
// f2(c); // compile error different types
f3(c); // ok, pass by-value (implicit conversion)
10/61
Function Signature and Overloading 1/2
Signature
Function signature defines the input types for a (specialized) function and the
inputs + outputs types for a template function
A function signature includes the number of arguments, the types of arguments, and
the order of the arguments
• The C++ standard prohibits a function declaration that only differs in the return
type
• Function declarations with different signatures can have distinct return types
Overloading
Function overloading allows having distinct functions with the same name but with
different signatures
11/61
Function Signature and Overloading 2/2
void f(int a, char* b); // signature: (int, char*)
// char f(int a, char* b); // compile error same signature
// but different return types
void f(const int a, char* b); // same signature, ok
// const int == int
void f(int a, const char* b); // overloading with signature: (int, const char*)
int f(float); // overloading with signature: (float)
// the return type is different
12/61
Overloading Resolution Rules
• An exact match
• A promotion (e.g. char to int )
• A standard type conversion (e.g. float and int )
• A constructor or user-defined type conversion ⇝
void f(int a);
void f(float b); // overload
void f(float b, char c); // overload
//--------------------------------------------------------------
f(0); // exact match
f('a'); // promotion from char to int (promotion)
// f(3LL); // compile error ambiguous match
f(2.3f); // exact match
// f(2.3); // compile error ambiguous match
f(2.3, 'a'); // standard type conversion, ambiguity is not possible here
13/61
Overloading and =delete
=delete can be used to prevent calling the wrong overload
void g(int) {}
void g(double) = delete;
g(3); // ok
g(3.0); // compile error
# include <cstddef> // std::nullptr_t
void f(int*) {}
void f(std::nullptr_t) = delete;
f(nullptr); // compile error
14/61
Function Default Parameters
Default/Optional parameter
A default parameter is a function parameter that has a default value
• If the user does not supply a value for this parameter, the default value will be used
• All default parameters must be the rightmost parameters
• Default parameters must be declared only once
• Default parameters can improve compile time and avoid redundant code because they
avoid defining other overloaded functions
void f(int a, int b = 20); // declaration
//void f(int a, int b = 10) { ... } // compile error, already set in the declaration
void f(int a, int b) { ... } // definition, default value of "b" is already set
f(5); // b is 20
15/61
Function Attributes [[attribute]] 1/2
C++ allows marking functions with standard properties to better express their intent:
• C++11 [[noreturn]] indicates that a function does not return for
optimization purposes or compiler warnings
• C++14 [[deprecated]] , [[deprecated("reason")]] indicates the use of a
function is discouraged. It issues a compiler warning if used
• C++17 [[nodiscard]]
C++20 [[nodiscard("reason")]] issues a warning if the return value of a
function is discarded (not handled)
16/61
Function Attributes [[attribute]] 2/2
[[deprecated]] void my_rand() { ... }
[[deprecated("please use rnd()")]] void my_rand2() { ... }
[[nodiscard]] int f() { return 3; }
[[noreturn]] void g() { std::exit(0); }
my_rand(); // WARNING: "my_rand() is deprecated"
my_rand2(); // WARNING: "my_rand2() is deprecated, please use rnd()"
f(); // WARNING "discard return value"
int z = f(); // no warning
g(); // no code after calling this function
17/61
Function Pointer - Function as Argument 1/2
Standard C achieves generic programming capabilities and composability through the
concept of function pointer
A function can be passed as a pointer to another function and behaves as an “indirect
call”
#include <stdlib.h> // qsort
int descending(const void* a, const void* b) {
return *((const int*) a) > *((const int*) b);
}
int array[] = {7, 2, 5, 1};
qsort(array, 4, sizeof(int), descending);
// array: { 7, 5, 2, 1 }
18/61
Function Pointer - Function as Argument 2/2
int eval(int a, int b, int (*f)(int, int)) {
return f(a, b);
}
// type: int (*)(int, int)
int add(int a, int b) { return a + b; }
int sub(int a, int b) { return a - b; }
cout << eval(4, 3, add); // print 7
cout << eval(4, 3, sub); // print 1
Problems:
Safety There is no check of the argument type in the generic case (e.g. qsort )
Performance Any operation requires an indirect call to the original function. Function
inlining is not possible
19/61
Function Object (or Functor) 1/2
Function Object
A function object, or functor, is a callable object that can be treated as a
parameter
C++ provides a more efficient and convenience way to pass “procedure” to other
functions called function object
#include <algorithm> // for std::sort
struct Descending { // <-- function object
bool operator()(int a, int b) { // function call operator
return a > b;
}
};
int array[] = {7, 2, 5, 1};
std::sort(array, array + 4, Descending{});
// array: { 7, 5, 2, 1 }
20/61
Function Object (or Functor) 2/2
Advantages:
Safety Argument type checking is always possible. It could involve templates
Performance The compiler injects operator() in the code of the destination function
and then compile the routine. Operator inlining is the standard behavior
C++11 simplifies the concept by providing less verbose function objects called
lambda expressions
21/61
Lambda Expression
Lambda Expression
A C++11 lambda expression is an inline local-scope function object
auto x = [capture clause] (parameters) { body }
• The [capture clause] marks the declaration of the lambda and how the local
scope arguments are captured (by-value, by-reference, etc.)
• The parameters of the lambda are normal function parameters (optional)
• The body of the lambda is a normal function body
The expression to the right of = is the lambda expression, and the runtime object
x created by that expression is the closure
22/61
Lambda Expression
# include <algorithm> // for std::sort
int array[] = {7, 2, 5, 1};
auto lambda = [](int a, int b){ return a > b; }; // named lambda
std::sort(array, array + 4, lambda);
// array: { 7, 5, 2, 1 }
// in alternative, in one line of code: // unnamed lambda
std::sort(array, array + 4, [](int a, int b){ return a > b; });
// array: { 7, 5, 2, 1 }
23/61
Capture List
Lambda expressions capture external variables used in the body of the lambda in two
ways:
• Capture by-value
• Capture by-reference (can modify external variable values)
Capture list can be passed as follows
• [] no capture
• [=] captures all variables by-value
• [&] captures all variables by-reference
• [var1] captures only var1 by-value
• [&var2] captures only var2 by-reference
• [var1, &var2] captures var1 by-value and var2 by-reference
24/61
Capture List Examples
// GOAL: find the first element greater than "limit"
# include <algorithm> // for std::find_if
int limit = ...
auto lambda1 = [=](int value) { return value > limit; }; // by-value
auto lambda2 = [&](int value) { return value > limit; }; // by-reference
auto lambda3 = [limit](int value) { return value > limit; }; // "limit" by-value
auto lambda4 = [&limit](int value) { return value > limit; }; // "limit" by-reference
// auto lambda5 = [](int value) { return value > limit; }; // no capture
// compile error
int array[] = {7, 2, 5, 1};
std::find_if(array, array + 4, lambda1);
25/61
Capture List - Other Cases
• [=, &var1] captures all variables used in the body of the lambda by-value,
except var1 that is captured by-reference
• [&, var1] captures all variables used in the body of the lambda by-reference,
except var1 that is captured by-value
• A lambda expression can read a variable without capturing it if the variable is
constexpr
constexpr int limit = 5;
int var1 = 3, var2 = 4;
auto lambda1 = [](int value){ return value > limit; };
auto lambda2 = [=, &var2]() { return var1 > var2; };
26/61
Lambda Expressions - Parameters
C++14 Lambda expression parameters can be automatically deduced
auto x = [](auto value) { return value + 4; };
C++14 Lambda expression parameters can be initialized
auto x = [](int i = 6) { return i + 4; };
27/61
Lambda Expressions - Composability
Lambda expressions can be composed
auto lambda1 = [](int value){ return value + 4; };
auto lambda2 = [](int value){ return value * 2; };
auto lambda3 = [&](int value){ return lambda2(lambda1(value)); };
// returns (value + 4) * 2
A function can return a lambda (dynamic dispatch is also possible)
auto f() {
return [](int value){ return value + 4; };
}
auto lambda = f();
cout << lambda(2); // print "6"
28/61
constexpr/consteval Lambda Expression
C++17 Lambda expression supports constexpr
C++20 Lambda expression supports consteval
// constexpr lambda
auto factorial = [](int value) constexpr {
int ret = 1;
for (int i = 2; i <= value; i++)
ret *= i;
return ret;
};
auto mul = [](int v) consteval { return v * 2; };
constexpr int v1 = factorial(4) + mul(5); // '24' + '10'
29/61
template Lambda Expression ⇝
C++20 Lambda expression supports template and requires clause
auto lambda = []<typename T>(T value)
requires std::is_arithmetic_v<T> {
return value * 2;
};
auto v = lambda(3.4); // v: 6.8 (double)
struct A{} a;
// auto v = lambda(a); // compiler error
30/61
mutable Lambda Expression
Lambda capture is by-const-value
mutable specifier allows the lambda to modify the parameters captured by-value
int var = 1;
auto lambda1 = [&](){ var = 4; }; // ok
lambda1();
cout << var; // print '4'
// auto lambda2 = [=](){ var = 3; }; // compile error
// lambda operator() is const
auto lambda3 = [=]() mutable { var = 3; }; // ok
lambda3();
cout << var; // print '4', lambda3 captures by-value
31/61
[[nodiscard]] Attribute
C++23 allows adding the [[nodiscard]] attribute to lambda expressions
auto lambda = [] [[nodiscard]] (){ return 4; };
lambda(); // compiler warning
auto x = lambda(); // ok
32/61
Capture List and Classes ⇝
• [this] captures the current object (*this) by-reference (implicit in C++17)
• [x = x] captures the current object member x by-value C++14
• [&x = x] captures the current object member x by-reference C++14
• [=] default capture of this pointer by value has been deprecated C++20
class A {
int data = 1;
void f() {
int var = 2; // <-- local variable
auto lambda1 = [=]() { return var; }; // copy by-value, return 2
auto lambda2 = [=]() { int var = 3; return var; }; // return 3 (nearest scope)
auto lambda3 = [this]() { return data; }; // copy by-reference, return 1
auto lambda4 = [*this]() { return data; }; // copy by-value (C++17), return 1
// auto lambda5 = [data]() { return data; }; // compile error 'data' is not visible
auto lambda6 = [data = data]() { return data; }; // return 1
}
};
33/61
Preprocessing and Macro
A preprocessor directive is any line preceded by a hash symbol (#) which tells the
compiler how to interpret the source code before compiling it
Macro are preprocessor directives which substitute any occurrence of an identifier in
the rest of the code by replacement
Macro are evil:
Do not use macro expansion!!
...or use as little as possible
• Macro cannot be directly debugged
• Macro expansions can have unexpected side effects
• Macro have no namespace or scope
34/61
Preprocessors
All statements starting with #
• #include "my file.h"
Inject the code in the current file
• #define MACRO <expression>
Define a new macro
• #undef MACRO
Undefine a macro
(a macro should be undefined as early as possible for safety reasons)
Multi-line Preprocessing: \ at the end of the line
Indent: # define
35/61
Conditional Compiling
• #if <condition>
code
#elif <condition>
code
#else
code
#endif
• #if defined(MACRO) equal to #ifdef MACRO
#elif defined(MACRO) equal to #elifdef MACRO C++23
Check if a macro is defined
• #if !defined(MACRO) equal to #ifndef MACRO
#elif !defined(MACRO) equal to #elifndef MACRO C++23
Check if a macro is not defined
36/61
Common Error 1
A Define macros in header files and before includes!!
# include <iostream>
# define value // very dangerous!!
# include "big_lib.hpp"
int main() {
std::cout << f(4); // should print 7, but it always prints 3
}
big lib.hpp:
int f(int value) { // 'value' disappears
return value + 3;
}
It is very hard to see this problem when the macro is in a header
37/61
Common Error 2
#if defined can introduce bugs related to macro visibility
// #include "macro_definition.hpp" // forget to add the header that defines ENABLE_DEBUG
# if defined(ENABLE_DEBUG)
void f(int v) { cout << v << endl; return v * 3; }
# else
void f(int v) { return v * 3; }
# endif
# if ENABLE_DEBUG // evaluated to 0 or 1
void f(int v) { cout << v << endl; return v * 3; }
# else
void f(int v) { return v * 3; }
# endif
38/61
Common Error 3
Forget to use parenthesis in macro definitions!!
# include <iostream>
# define SUB1(a, b) a - b // WRONG
# define SUB2(a, b) (a - b) // WRONG
# define SUB3(a, b) ((a) - (b)) // correct
int main() {
std::cout << (5 * SUB1(2, 1)); // print 9 not 5!!
std::cout << SUB2(3 + 3, 2 + 2); // print 6 not 2!!
std::cout << SUB3(3 + 3, 2 + 2); // print 2
}
39/61
Common Error 4
Macros make hard to find compile errors!!
1: # include <iostream>
2:
3: # define F(a) { \
4: ... \
5: ... \
6: return v;
7:
8: int main() {
9: F(3); // compile error at line 9!!
10: }
• In which line is the error??!*
*modern compilers are able to roll out the macro
40/61
Common Error 5
Macro can introduce bugs related to the evaluation of their expressions!!
# if defined(DEBUG)
# define CHECK(EXPR) // do something with EXPR
void check(bool b) { /* do something with b */ }
# else
# define CHECK(EXPR) // do nothing
void check(bool) {} // do nothing
# endif
bool f() { /* return a boolean value */ }
check( f() )
CHECK( f() ) // <-- problem here
• What happens when DEBUG is not defined?
f() is not evaluated by using the macro
41/61
Common Error 6
Forget curly brackets in multi-lines macros!!
# include <iostream>
# include <nuclear_explosion.hpp>
# define NUCLEAR_EXPLOSION \ // {
std::cout << "start nuclear explosion"; \
nuclear_explosion();
// }
int main() {
bool never_happen = false;
if (never_happen)
NUCLEAR_EXPLOSION
} // BOOM!!
A
The second line is executed!!
42/61
Common Error 7
Macros do not have scope!!
# include <iostream>
void f() {
# define value 4
std::cout << value;
}
int main() {
f(); // 4
std::cout << value; // 4
# define value 3
f(); // 4
std::cout << value; // 3
}
43/61
Common Error 8
Macros can have side effect!!
# define MIN(a, b) ((a) < (b) ? (a) : (b))
int main() {
int array1[] = { 1, 5, 2 };
int array2[] = { 6, 3, 4 };
int i = 0;
int j = 0;
int v1 = MIN(array1[i++], array2[j++]); // v1 = 5!!
int v2 = MIN(array1[i++], array2[j++]); // undefined behavior/
} // segmentation fault
A
arne-mertz.de/2019/03/macro-evil
44/61
Common Error 9
Macros can have undefined behavior themselves!!
# define MY_MACRO defined(EXTERNAL_MACRO)
# if MY_MACRO
# define MY_VALUE 1
# else
# define MY_VALUE 0
# endif
int f() { return MY_VALUE; } // undefined behavior
45/61
When Preprocessors are Necessary
• Conditional compiling: different architectures, compiler features, etc.
• Mixing different languages: code generation (example: asm assembly)
• Complex name replacing: see template programming
Otherwise, prefer const and constexpr for constant values and functions
# define SIZE 3 // replaced with
const int SIZE = 3; // only C++11 at global scope
# define SUB(a, b) ((a) - (b)) // replaced with
constexpr int sub(int a, int b) {
return a - b;
}
Are We Macro free Yet, CppCon2019
46/61
Source Location Macros 1/3
LINE Integer value representing the current line in the source code file
being compiled
FILE A string literal containing the name of the source file being
compiled
FUNCTION (non-standard, gcc, clang) A string literal containing the name of
the function in the ‘macro scope’
PRETTY FUNCTION (non-standard, gcc, clang) A string literal containing the full
signature of the function in the ‘macro scope’
func (C++11 keyword) A string containing the name of the function in
the ‘macro scope’
47/61
Source Location Macros 2/3
source.cpp:
# include <iostream>
void f(int p) {
std::cout << __FILE__ << ":" << __LINE__; // print 'source.cpp:4'
std::cout << __FUNCTION__; // print 'f'
std::cout << __func__; // print 'f'
}
// see template lectures
template<typename T>
float g(T p) {
std::cout << __PRETTY_FUNCTION__; // print 'float g(T) [T = int]'
return 0.0f;
}
void g1() { g(3); }
48/61
Source Location Macros 3/3
C++20 provides source location utilities for replacing macro-based approach
#include <source location>
current() get source location info (static member)
line() source code line
column() line column
file name() current file name
function name() current function name
# include <source_location>
void f(std::source_location s = std::source_location::current()) {
cout << "function: " << s.function_name() << ", line " << s.line();
}
f(); // print: "function: f, line 6"
49/61
Condition Compiling Macros 1/2
Select code depending on the C/C++ version
• #if defined( cplusplus) C++ code
• #if cplusplus == 199711L ISO C++ 1998/2003
• #if cplusplus == 201103L ISO C++ 2011*
• #if cplusplus == 201402L ISO C++ 2014*
• #if cplusplus == 201703L ISO C++ 2017
Select code depending on the compiler
• #if defined( GNUG ) The compiler is gcc/g++
†
• #if defined( clang ) The compiler is clang/clang++
• #if defined( MSC VER) The compiler is Microsoft Visual C++
* MSVC defines cplusplus == 199711L even for C++11/14
† GNUC is defined by many compilers, e.g clang
50/61
Condition Compiling Macros 2/2
Select code depending on the operating system or environment
• #if defined( WIN64) OS is Windows 64-bit
• #if defined( linux ) OS is Linux
• #if defined( APPLE ) OS is Mac OS
• #if defined( MINGW32 ) OS is MinGW 32-bit
• ...and many others
51/61
Other Macros
DATE A string literal in the form ”MMM DD YYYY” containing the date in which
the compilation process began
TIME A string literal in the form ”hh:mm:ss” containing the time at which the
compilation process began
Very comprehensive macro list:
• sourceforge.net/p/predef/wiki/Home/
• Compiler predefined macros
• Abseil platform macros
52/61
Feature Testing Macro
C++17 introduces has include macro which returns 1 if header or source file
with the specified name exists
# if __has_include(<iostream>)
# include <iostream>
# endif
C++20 introduces a set of macros to evaluate if a given feature is supported by the
compiler
# if __cpp_constexpr
constexpr int square(int x) { return x * x; }
# endif
Feature Testing Macros
53/61
Common Error 10 ⇝
Macros depend on compilers and environment!!
struct A {
int x; // enable C++11 code
# if __cplusplus >= 201103
A() = default;
# else
A() {}
# endif
};
// should return ≈ 10.0f
float safe_function() {
A a{}; // zero-initialization
for (int i = 0; i < 10; i++)
a.x += 1.0f;
return a.x;
}
// what is the behavior ???
The code works fine on Linux, but not under Windows MSVC. MSVC sets cplusplus to
199711 even if C++11/14/17 flag is set!! in this case the code can return NaN
see Lecture “Object-Oriented Programming II - Zero Initialization” and MSVC now correctly
reports cplusplus
54/61
Stringizing Operator (#)
The stringizing macro operator ( # ) causes the corresponding actual argument to be
enclosed in double quotation marks "
# define STRING_MACRO(string) #string
cout << STRING_MACRO(hello); // equivalent to "hello"
# define INFO_MACRO(my_func) \
{ \
my_func \
cout << "call " << #my_func << " at " \
<< __FILE__ << ":" __LINE__; \
}
void g(int) {}
INFO_MACRO( g(3) ) // print: "call g(3) at my_file.cpp:7"
55/61
Common Error 11
Code injection
# include <cstdio>
# define CHECK_ERROR(condition) \
{ \
if (condition) { \
std::printf("expr: " #condition " failed at line %d\n",\
__LINE__); \
} \
}
int t = 6, s = 3;
CHECK_ERROR(t > s) // print "expr: t > s failed at line 13"
CHECK_ERROR(t % s == 0) // segmentation fault!!!
A
// printf interprets "% s" as a format specifier
56/61
#error and #warning
• #error "text" The directive emits a user-specified error message at compile
time when the compiler parse it and stop the compilation process
• C++23 #warning "text" The directive emits a user-specified warning message
at compile time when the compiler parse it without stopping the compilation
process
57/61
#pragma
The #pragma directive controls implementation-specific behavior of the compiler. In
general, it is not portable
• #pragma message "text" Display informational messages at compile time
(every time this instruction is parsed)
• #pragma GCC diagnostic warning "-Wformat"
Disable a GCC warning
• Pragma(<command>) (C++11)
It is a keyword and can be embedded in a #define
#define MY_MESSAGE \
_Pragma("message(\"hello\")")
58/61
Token-Pasting Operator (##)
⋆
The token-concatenation (or pasting) macro operator ( ## ) allows combining two
tokens (without leaving no blank spaces)
# define FUNC_GEN_A(tokenA, tokenB) \
void tokenA##tokenB() {}
# define FUNC_GEN_B(tokenA, tokenB) \
void tokenA##_##tokenB() {}
FUNC_GEN_A(my, function)
FUNC_GEN_B(my, function)
myfunction(); // ok, from FUNC_GEN_A
my_function(); // ok, from FUNC_GEN_B
59/61
Variadic Macro
⋆
A variadic macro C++11 is a special macro accepting a variable number of arguments
(separated by comma)
Each occurrence of the special identifier VA ARGS in the macro replacement list is
replaced by the passed arguments
Example:
void f(int a) { printf("%d", a); }
void f(int a, int b) { printf("%d %d", a, b); }
void f(int a, int b, int c) { printf("%d %d %d", a, b, c); }
# define PRINT(...) \
f( VA ARGS );
PRINT(1, 2)
PRINT(1, 2, 3)
60/61
Macro Trick
⋆
Convert a number literal to a string literal
# define TO_LITERAL_AUX(x) #x
# define TO_LITERAL(x) TO_LITERAL_AUX(x)
Motivation: avoid integer to string conversion (performance)
int main() {
int x1 = 3 * 10;
int y1 = __LINE__ + 4;
char x2[] = TO_LITERAL(3);
char y2[] = TO_LITERAL(__LINE__);
}
61/61
