Exception handling in C++ is a well-unschooled topic if you observe initial stages of the learning curve. There are numerous tutorials available online on exception handling in C++ with example. But few explains what you should not do & intricacies around it. So here, we will see some intricacies, from where & why you should not throw an exception along with some newer features introduced in on exception handling with example. I am not an expert but this is what I have gained from various sources, courses & industry experiences. Modern C++ /!\: Originally published @ . www.vishalchovatiya.com In the end, we will see the by a quick benchmark code. Finally, we will close the article with . performance cost of using an exception Best practices & some CPP Core Guidelines on exception handling Note : we will not see anything regarding a dynaic exception as it deprecated from C++11 and removed in C++17. Terminology/Jargon/Idiom you may face : may or may not throw an exception. potentially throwing : this is specifier as well as operator depending upon where & how you use it. Will see that . noexcept later : esource cquisition s nitialization is a scope-bound resource management mechanism which means resource allocation is done by the constructor & resource deallocation is done by the destructor during the defined scope of the object. I know it’s a terrible name but very powerful concept. RAII R A I I : I think this need not require any introduction. Implicitly-declared special member functions 1. Implement copy and/or move constructor while throwing user-defined type object demo() = ;
    demo(demo &&) = ;
    demo( demo &) = ;
}; { demo{}; ;
} { struct demo default delete const delete int main () throw return 0 Upon throw expression, a copy of the exception object always needs to be created as the original object goes out of the scope during the stack unwinding process. During that initialization, we may expect (see ) – omits (object constructed directly into the storage of the target object). copy elision this copy or move constructors But even though copy elision may or may not be applied you should provide proper copy constructor and/or move constructor which is what . See below compilation error for reference: C++ standard mandates(see 15.1) error: call to deleted constructor of demo{};
          ^~~~~~
note: = ;
    ^ error generated.
compiler status 'demo' throw 'demo' has been explicitly marked deleted here demo (demo &&) delete 1 exit 1 Above error stands true till C++14. Since C++17, If object thrown is a prvalue, the move/copy constructor call is optimized out i.e. . copy elision If we catch an exception by value, we may also expect copy elision(compilers are permitted to do so, but it is not mandatory). The exception object is an lvalue argument when initializing catch clause parameter. TL;DR class used for throwing the exception object needs copy and/or move constructors 2. Be cautious while throwing an exception from the constructor base(){ << ;}
    ~base(){ << ;}
}; base
{
    derive(){ << ; ;}
    ~derive(){ << ;}
}; { {
        derive{};
    } (...){} ;
} { struct base cout "base\n" cout "~base\n" : struct derive cout "derive\n" throw -1 cout "~derive\n" int main () try catch return 0 When an exception is thrown from a constructor, stack unwinding begins, destructors for the object will be called only & only if an object is created successfully. So be caution with dynamic memory allocation here. In such cases, you should use . RAII base
derive
~base As you can see in the above case, the destructor of derive is not executed, Because, it is not created successfully. base() { << ; }
    ~base() { << ; }
}; base
{
    derive() = ;
    derive( ) : derive{}
    { << ; - ;
    }
    ~derive() { << ; }
}; { {
        derive{ };
    } (...){} ;
} { struct base cout "base\n" cout "~base\n" : struct derive default int cout "derive\n" throw 1 cout "~derive\n" int main () try 0 catch return 0 In the case of constructor delegation, it is considered as the creation of object hence destructor of derive will be called. base
derive
~derive
~base When an exception is thrown from a constructor, destructors for the object will be called only & only if an object is created successfully TL;DR 3. Avoid throwing exceptions out of a destructor ~demo() { ::exception{}; }
}; { {
        demo d;
    } ( ::exception &){} ;
} { struct demo throw std int main () try catch const std return 0 Above code seems straight forward but when you run it, it will be terminated as shown below rather than catching the exception. Reason for this is destructors are by default (i.e. non-throwing) noexcept $ clang++-7 -o main main.cpp
warning: '~demo' has a non-throwing exception specification but can still
      throw [-Wexceptions]
    ~demo() { throw std::exception{}; }
              ^
note: destructor has a implicit non-throwing exception specification
    ~demo() { throw std::exception{}; }
    ^
1 warning generated.
$
$ ./main
terminate called after throwing an instance of 'std::exception'
  what():  std::exception
exited, aborted will solve our problem as below. noexcept(false) ~X() ( ) { ::exception{}; } 
}; { struct X noexcept false throw std But don’t do it. Destructors are by default non-throwing for a reason, and we must not throw exceptions in destructors unless we catch them inside the destructor. Why you should not throw an exception from a destructor? Because destructors are called during stack unwinding when an exception is thrown, and we are not allowed to throw another exception while the previous one is not caught – in such a case will be called. std::terminate Consider the following example for more clarity. ~base() ( ) { ; }
}; base
{
    ~derive() ( ) { ; }
}; { {
        derive d;
    } (...){ } ;
} { struct base noexcept false throw 1 : struct derive noexcept false throw 2 int main () try catch return 0 An exception will be thrown when the object d will be destroyed as a result of . But at the same time destructor of base will also be called as it is of derive which will again throw an exception. Now we have two exceptions at the same time which is invalid scenario & will be called. RAII sub-object std::terminate There are some type trait utilities like , , etc. from by which you can check whether the special member functions are exception-safe or not. std::is_nothrow_destructible std::is_nothrow_constructible #include<type_traits> { << ::boolalpha << ::is_nothrow_destructible< :: >::value << ; << ::boolalpha << ::is_nothrow_constructible< :: >::value << ; ;
} int main () cout std std std string endl cout std std std string endl return 0 1. Destructors are by default (i.e. non-throwing). 2. You should not throw exception out of destructors because destructors are called during stack unwinding when an exception is thrown, and we are not allowed to throw another exception while the previous one is not caught – in such a case will be called. TL;DR noexcept std::terminate 4. Rethrowing/Nested exception handling with std::exception_ptr( C++11) example This is more of a demonstration rather the best practice of the nested exception scenario using . Although you can simply use without complicating things much but will provide us with the leverage of handling exception out of / clause. std::exception_ptr std::exception std::exception_ptr try catch { get_nested = []( &e) -> ::exception_ptr { { < ::nested_exception &>(e).nested_ptr(); } ( ::bad_cast&) { ; }
    }; { (eptr) ::rethrow_exception(eptr);
    } ( ::exception &e){ :: << :: (level, ) << << e.what() << ;
        print_nested_exception(get_nested(e), level + ); }
} { { ::runtime_error( ); } (...) { ::throw_with_nested( ::runtime_error( )); }
} { { func2(); } (...) { ::throw_with_nested( ::runtime_error( )); }
} { { func1(); } ( ::exception&)   { print_nested_exception(); } ;
} void print_nested_exception ( ::exception_ptr &eptr= ::current_exception(), level= ) const std std size_t 0 static auto auto std try return dynamic_cast const std catch const std return nullptr try if std catch const std std cerr std string ' ' "exception: " '\n' 1 // rewind all nested exception // ----------------------------------------------------------------------------------------------- void func2 () try throw std "TESTING NESTED EXCEPTION SUCCESS" catch std std "func2() failed" void func1 () try catch std std "func1() failed" int main () try catch const std return 0 // Will only work with C++14 or above Above example looks complicated at first, but once you have implemented nested exception handler(i.e. ). Then you only need to focus on throwing the exception using function. print_nested_exception std::throw_with_nested exception: func1() failed
 exception: func2() failed
  exception: TESTING NESTED EXCEPTION SUCCESS The main thing to focus here is function in which we are rewinding nested exception using & . print_nested_exception std::rethrow_exception std::exception_ptr is a like type though dereferencing it is undefined behaviour. It can hold nullptr or point to an exception object and can be constructed as: std::exception_ptr shared pointer ::exception_ptr e1; ::exception_ptr e2 = ::current_exception(); ::exception_ptr e3 = ::make_exception_ptr( ::exception{}); std // null std std // null or a current exception std std std // std::exception Once is created, we can use it to throw or re-throw exceptions by calling as we did above, which throws the pointed exception object. std::exception_ptr std::rethrow_exception(exception_ptr) 1. extends the lifetime of a pointed exception object beyond a catch clause. 2. We may use to delay the handling of a current exception and transfer it to some other palaces. Though, practical usecase of is between threads. TL;DR std::exception_ptr std::exception_ptr std::exception_ptr 5. Use noexcept specifier vs operator appropriately I think this is an oblivious concept among the other concepts of the C++ exceptions. noexcept & came in C++11 to replace deprecated(removed from C++17) dynamic exception specification. specifier operator ; ; ; {} (*func_ptr)() = print; {} {} l_non_throwing = []() {}; void func () throw ( ::exception) std // dynamic excpetions, removed from C++17 void potentially_throwing () // may throw void non_throwing () noexcept // "specifier" specifying non-throwing function void print () void noexcept // Not OK from C++17, `print()` should be noexcept too, works in C++11/14 void debug_deep () noexcept ( ) false // specifier specifying throw void debug () noexcept ( (debug_deep())) noexcept // specifier & operator, will follow exception rule of `debug_deep` auto noexcept // Yeah..! lambdas are also in party I think this needs no introduction it does what its name suggests. So let’s quickly go through some pointers: noexcept specifier Can be used for normal functions, methods, & function pointer. lambda functions From C++17, function pointer with noexcept can not points to potentially throwing function. Finally, don’t use specifier for in a base class/interface because it enforces restriction for all overrides. noexcept virtual functions Don’t use noexcept unless you really need it. “Specify it when it is useful and correct” – . Google’s cppguide noexcept operator & what is it used for? Added in C++11, operator takes an expression (not necessarily constant) and performs a compile-time check determining if that expression is non-throwing ( ) or potentially throwing. noexcept noexcept The result of such compile-time check can be used, for example, to add specifier to the same category, higher-level function or in if . noexcept (noexcept(noexcept(expr))) constexpr We can use operator to check if some class has constructor, noexcept copy constructor, noexcept move constructor, and so on as follows: noexcept noexcept :
    demo() {}
    demo( demo &) {}
    demo(demo &&) {} {}
}; { << ::boolalpha << (demo()) << ; << ::boolalpha << (demo(demo())) << ; << ::boolalpha << (demo( ::declval<demo>())) << ; << ::boolalpha << ( ::declval<demo>().method()) << ; } { class demo public const void method () int main () cout std noexcept endl // C cout std noexcept endl // CC cout std noexcept std endl // MC cout std noexcept std endl // Methods // std::declval<T> returns an rvalue reference to a type You must be wondering why & how this information will be useful? This is more useful when you are using library functions inside your function to suggest compiler that your function is throwing or non-throwing depending upon library implementation. If you remove constructor, & , it will print true reason being implicitly-declared special member functions are always non-throwing. copy constructor move constructor specifier & operator are two different things. operator performs a compile-time check & doesn’t evaluate the expression. While specifier can take only constant expressions that evaluate to either true or false. TL;DR noexcept noexcept noexcept 6. Move exception-safe with std::move_if_noexcept demo() = ;
    demo( demo &) { << ; } demo(demo &&) { << ; } : :: < >    m_v;
}; {
    demo obj1; ( (demo( ::declval<demo>()))){ ; } {
        demo obj2(obj1); } ; ;
} { struct demo default const cout "Copying\n" // Exception safe move constructor noexcept cout "Moving\n" private std vector int int main () if noexcept std // if moving safe demo obj2 ( ::move(obj1)) std // then move it else // otherwise copy it demo obj3 ( ::move_if_noexcept(obj1)) std // Alternatively you can do this---------------- return 0 We can use to check if ’s move constructor exists and is in order to decide if we want to create an instance of by moving another instance of T (using ). noexcept(T(std::declval<T>())) T noexcept T std::move Alternatively, we can use , which uses operator and casts to either . Such checks are used in and other containers. std::move_if_noexcept noexcept rvalue or lvalue std::vector This will be useful while you are processing critical data which you don’t want to lose. For example, we have critical data received from the server that we do not want to lose it at any cost while processing. In such a case, we should use which will move ownership of critical data only and only if move constructor is exception-safe. std::move_if_noexcept Move critical object safely with TL;DR std::move_if_noexcept 7. Real cost of exception handling in C++ with benchmark example Despite many benefits, most people still do not prefer to use exceptions due to its overhead. So let’s clear it out of the way: { ( _ : state){ :: < > v( ); ( i = ; i < ; i++) v.at(i) = i;        
    }
}
BENCHMARK(without_exception); { ( _ : state){ :: < > v( ); ( i = ; i < ; i++){ {
                v.at(i) = i;
            } ( ::out_of_range &oor){}
        }
    }
}
BENCHMARK(with_exception); { ( _ : state){ :: < > v( ); ( i = ; i < ; i++){ {
                v.at(i) = i;
            } ( ::out_of_range &oor){}
        }
    }
}
BENCHMARK(throwing_exception); static void without_exception (benchmark::State &state) for auto std vector uint32_t 10000 for uint32_t 0 10000 //---------------------------------------- static void with_exception (benchmark::State &state) for auto std vector uint32_t 10000 for uint32_t 0 10000 try catch const std //-------------------------------------------- static void throwing_exception (benchmark::State &state) for auto std vector uint32_t 10000 for uint32_t 1 10001 try catch const std //----------------------------------------- As you can see above, & has only a single difference i.e. exception syntax. But none of them throws any exceptions. with_exception without_exception While does the same task except it throws an exception of type in the last iteration. throwing_exception std::out_of_range As you can see in below bar graph, the last bar is slightly high as compared to the previous two which shows the cost of throwing an exception. But the here, as the previous two bars are identical. cost of using exception is zero I am not considering the optimization here which is the separate case as it trims some of the assembly instructions completely. Also, implementation of compiler & ABI plays a crucial role. But still, it is far better than losing time by setting up a guard( strategy) and explicitly checking for the presence of error everywhere. if(error) While in case of exception, the compiler generates a side table that maps any point that may throw an exception (program counter) to the list of handlers. When an exception is thrown, this list is consulted to pick the right handler (if any) and the stack is unwound. See for in-depth knowledge. this By the way, I am using a & which internally uses , if you want to explore more. quick benchmark Google Benchmark First and foremost, remember that using try and catch doesn’t actually decrease performance unless an exception is thrown.It’s “zero cost” exception handling – no instruction related to exception handling is executed until one is thrown. But, at the same time, it contributes to the size of executable due to unwinding routines, which may be important to consider for embedded systems. No instruction related to exception handling is executed until one is thrown so using / doesn’t actually decrease performance. TL;DR try catch Best practices & some CPP Core Guidelines on exception handling Best practices for C++ exception handling Ideally, you should not throw an exception from the destructor, move constructor or swap like functions. – data in an invalid state, i.e. data that cannot be further read & used; – leaked resources such as memory, files, ids, or anything else that needs to be allocated and released; – corrupted memory; – broken invariants, e.g. size function returns more elements than actually held in a container. Prefer RAII idiom for the exception safety because in case of exception you might be left with Avoid using raw new & delete . Use solutions from the standard library, e.g. std::unique_pointer , std::make_unique , std::fstream , std::lock_guard , etc. Moreover, it is useful to split your code into modifying and non-modifying parts, where only the non-modifying part can throw exceptions. Never throw exceptions while owing some resource. Some CPP Core Guidelines E.1: Develop an error-handling strategy early in a design E.3: Use exceptions for error handling only E.6: Use RAII to prevent leaks E.13: Never throw while being the direct owner of an object E.16: Destructors, deallocation, and swap must never fail E.17: Don’t try to catch every exception in every function E.18: Minimize the use of explicit try / catch 26: If you can’t throw exceptions, consider failing fast E.31: Properly order your catch -clauses

Assembly

Google

Rewind

Target

How C program stored in RAM memory

All About Lambda Functions in C++: from C++11 to C++17

www.vishalchovatiya.com

Read My Stories

Too Long; Didn't Read

7 best practices for exception handling in C++

7 best practices for exception handling in C++

About Author

Comments

TOPICS

THIS ARTICLE WAS FEATURED IN

Related Stories

7 Advanced C++ Concepts You Should Know

The Noonification: Will ChatGPT Change Coding Interviews Forever? (12/24/2022)

46 Stories To Learn About C

6 Best C# Programming Books Ranked by Review Score

6 Best C Programming Books Ranked by Review Scores

7 Reasons to Use Buck Build

7 Advanced C++ Concepts You Should Know

The Noonification: Will ChatGPT Change Coding Interviews Forever? (12/24/2022)

46 Stories To Learn About C

6 Best C# Programming Books Ranked by Review Score

6 Best C Programming Books Ranked by Review Scores

7 Reasons to Use Buck Build

Light-Mode

Classic

Newspaper

Minty

Dark-Mode

Neon Noir

Minty

HN StartUps