18 Exception Handling 18.1 EXCEPTION HANDLING BASICS 759 A Toy Example of Exception Handling 759 Defining Your Own Exception Classes 768 Multiple Throws and Catches 768 Pitfall: Catch the More Specific Exception First 772 Tip: Exception Classes Can Be Trivial 773 Throwing an Exception in a Function 773 Exception Specification 775 Pitfall: Exception Specification in Derived Classes 777 18.2 PROGRAMMING TECHNIQUES FOR EXCEPTION HANDLING 779 When to Throw an Exception 779 Pitfall: Uncaught Exceptions 781 Pitfall: Nested try-catch Blocks 781 Pitfall: Overuse of Exceptions 782 Exception Class Hierarchies 782 Testing for Available Memory 782 Rethrowing an Exception 783 CHAPTER SUMMARY 783 ANSWERS TO SELF-TEST EXERCISES 784 PROGRAMMING PROJECTS 785 18_CH18.fm Page 757 Monday, August 18, 2003 1:23 PM 18 Exception Handling It’s the exception that proves the rule. Common maxim INTRODUCTION One way to write a program is to first assume that nothing unusual or incor- rect will happen. For example, if the program takes an entry off a list, you might assume that the list is not empty. Once you have the program working for the core situation where things always go as planned, you can then add code to take care of the exceptional cases. C++ has a way to reflect this approach in your code. Basically, you write your code as if nothing very unusual happens. After that, you use the C++ exception handling facilities to add code for those unusual cases. Exception handling is commonly used to handle error situations, but per- haps a better way to view exceptions is as a way to handle exceptional situa- tions. After all, if your code correctly handles an “error,” then it no longer is an error. Perhaps the most important use of exceptions is to deal with functions that have some special case that is handled differently depending on how the func- tion is used. Perhaps the function will be used in many programs, some of which will handle the special case in one way, while others will handle it in some other way. For example, if there is a division by zero in the function, then it may turn out that for some invocations of the function the program should end, but for other invocations of the function something else should happen. You will see that such a function can be defined to throw an excep- tion if the special case occurs; that exception will allow the special case to be handled outside the function. Thus, the special case can be handled differently for different invocations of the function. In C++, exception handling proceeds as follows: Either some library soft- ware, or your code, provides a mechanism that signals when something unusual happens. This is called throwing an exception . At another place in your program you place the code that deals with the exceptional case. This is called handling the exception . This method of programming makes for cleaner code. Of course, we still need to explain the details of how you do this in C++. Most of this chapter only uses material from Chapters 1 through 9. How- ever, the sections “Exception Specification in Derived Classes” and “Exception Class Hierarchies” use material from Chapter 14. The section “Testing for Available Memory” uses material from Chapter 17. Any or all of these listed 18_CH18.fm Page 758 Monday, August 18, 2003 1:23 PM Exception Handling Basics 759 sections may be omitted without hurting the continuity of the chapter. The section “Exception Specification” has one paragraph that refers to derived classes (Chapter 14), but that paragraph may be omitted. Exception Handling Basics Well, the program works for most cases. I didn’t know it had to work for that case. Computer Science Student, Appealing a grade Exception handling is meant to be used sparingly and in situations that are more involved than what is reasonable to include in a simple introductory example. So, we will teach you the exception handling details of C++ by means of simple examples that would not normally use exception handling. This makes a lot of sense for learning about the exception handling details of C++, but do not forget that these first examples are toy examples; in practice, you would not use exception handling for anything that simple. ■ A TOY EXAMPLE OF EXCEPTION HANDLING For this example, suppose that milk is such an important food in our culture that peo- ple almost never run out of it, but still we would like our programs to accommodate the very unlikely situation of running out of milk. The basic code, which assumes we do not run out of milk, might be as follows: cout << "Enter number of donuts:\n"; cin >> donuts; cout << "Enter number of glasses of milk:\n"; cin >> milk; dpg = donuts/static_cast<double>(milk); cout << donuts << " donuts.\n" << milk << " glasses of milk.\n" << "You have " << dpg << " donuts for each glass of milk.\n"; If there is no milk, then this code will include a division by zero, which is an error. To take care of this special situation where we run out of milk, we can add a test. The complete program with this added test for the special situation is shown in Display 18.1. The program in Display 18.1 does not use exception handling. Now, let’s see how this program can be rewritten using the C++ exception handling facilities. In Display 18.2, we have rewritten the program from Display 18.1 using an excep- tion. This is only a toy example, and you would probably not use an exception in this case. However, it does give us a simple example to work with. Although the program as 18.1 18_CH18.fm Page 759 Monday, August 18, 2003 1:23 PM 760 Exception Handling Display 18.1 Handling a Special Case without Exception Handling 1 #include <iostream> 2 using std::cin; 3 using std::cout; 4 int main( ) 5 { 6 int donuts, milk; 7 double dpg; 8 cout << "Enter number of donuts:\n"; 9 cin >> donuts; 10 cout << "Enter number of glasses of milk:\n"; 11 cin >> milk; 12 if (milk <= 0) 13 { 14 cout << donuts << " donuts, and No Milk!\n" 15 << "Go buy some milk.\n"; 16 } 17 else 18 { 19 dpg = donuts/static_cast<double>(milk); 20 cout << donuts << " donuts.\n" 21 << milk << " glasses of milk.\n" 22 << "You have " << dpg 23 << " donuts for each glass of milk.\n"; 24 } 25 cout << "End of program.\n"; 26 return 0; 27 } S AMPLE D IALOGUE Enter number of donuts: 12 Enter number of glasses of milk: 0 12 donuts, and No Milk! Go buy some milk. End of program. 18_CH18.fm Page 760 Monday, August 18, 2003 1:23 PM Exception Handling Basics 761 Display 18.2 Same Thing Using Exception Handling (part 1 of 2) 1 2 #include <iostream> 3 using std::cin; 4 using std::cout; 5 int main( ) 6 { 7 int donuts, milk; 8 double dpg; 9 try 10 { 11 cout << "Enter number of donuts:\n"; 12 cin >> donuts; 13 cout << "Enter number of glasses of milk:\n"; 14 cin >> milk; 15 16 if (milk <= 0) 17 throw donuts; 18 dpg = donuts/static_cast<double>(milk); 19 cout << donuts << " donuts.\n" 20 << milk << " glasses of milk.\n" 21 << "You have " << dpg 22 << " donuts for each glass of milk.\n"; 23 } 24 catch(int e) 25 { 26 cout << e << " donuts, and No Milk!\n" 27 << "Go buy some milk.\n"; 28 } 29 cout << "End of program.\n"; 30 return 0; 31 } S AMPLE D IALOGUE 1 Enter number of donuts: 12 Enter number of glasses of milk: 6 12 donuts. 6 glasses of milk. You have 2 donuts for each glass of milk. End of program. This is just a toy example to learn C++ syntax. Do not take it as an example of good typical use of exception handling. 18_CH18.fm Page 761 Monday, August 18, 2003 1:23 PM 762 Exception Handling a whole is not simpler, at least the part between the words try and catch is cleaner, which hints at the advantage of using exceptions. Look at the code between the words try and catch . It is basically the same as the code in Display 18.1, except that rather than the big if-else statement (highlighted in Display 18.1), this new program has the following smaller if statement (plus some simple nonbranching statements): if (milk <= 0) throw donuts; This if statement says that if there is no milk, then do something exceptional. That something exceptional is given after the word catch . The idea is that the normal situa- tion is handled by the code following the word try , and that exceptional situations are handled by the code following the word catch . Thus, we have separated the normal case from the exceptional case. In this toy example that does not really buy us too much, but in other situations it will prove to be very helpful. Let’s look at the details. The basic way of handling exceptions in C++ consists of the try-throw-catch threesome. A try block has the following syntax: try { Some_Code } This try block contains the code for the basic algorithm that tells what to do when everything goes smoothly. It is called a try block because you are not 100% sure that all will go smoothly, but you want to “give it a try.” If something unusual does happen, you want to throw an exception, which is a way of indicating that something unusual happened. So the basic outline, when we add a throw , is as follows: try { Display 18.2 Same Thing Using Exception Handling (part 2 of 2) S AMPLE D IALOGUE 2 Enter number of donuts: 12 Enter number of glasses of milk: 0 12 donuts, and No Milk! Go buy some milk. End of program. try block 18_CH18.fm Page 762 Monday, August 18, 2003 1:23 PM Exception Handling Basics 763 Code_To_Try Possibly_Throw_An_Exception More_Code } The following is an example of a try block with a throw statement included (copied from Display 18.2): try { cout << "Enter number of donuts:\n"; cin >> donuts; cout << "Enter number of glasses of milk:\n"; cin >> milk; if (milk <= 0) throw donuts; dpg = donuts/static_cast<double>(milk); cout << donuts << " donuts.\n" << milk << " glasses of milk.\n" << "You have " << dpg << " donuts for each glass of milk.\n"; } The following statement throws the int value donuts : throw donuts; The value thrown (in this case, donuts ) is sometimes called an exception ; the execu- tion of a throw statement is called throwing an exception. You can throw a value of any type. In this case, an int value is thrown. When something is “thrown,” something goes from one place to another place. In C++ what goes from one place to another is the flow of control (as well as the value thrown). When an exception is thrown, the code in the surrounding try block stops executing and another portion of code, known as a catch block, begins execution. Exe- cuting the catch block is called catching the exception or handling the exception. When an exception is thrown, it should ultimately be handled by (caught by) some catch block. In Display 18.2, the appropriate catch block immediately follows the try block. We repeat the catch block in what follows: catch(int e) { cout << e << " donuts, and No Milk!\n" << "Go buy some milk.\n"; } throw statement exception throwing an exception catch block handling the exception 18_CH18.fm Page 763 Monday, August 18, 2003 1:23 PM 764 Exception Handling This catch block looks very much like a function definition that has a parameter of type int. It is not a function definition, but in some ways a catch block is like a func- tion. It is a separate piece of code that is executed when your program encounters (and executes) the following (within the preceding try block): throw Some_int ; So, this throw statement is similar to a function call, but instead of calling a func- tion, it calls the catch block and says to execute the code in the catch block. A catch block is often referred to as an exception handler, which is a term that suggests that a catch block has a function-like nature. What is that identifier e in the following line from a catch block? catch(int e) That identifier e looks like a parameter and acts very much like a parameter. In fact, the identifier, such as e, in the catch-block heading is called the catch-block parameter. Each catch block can have at most one catch-block parameter. The catch-block parameter does two things: 1. The catch-block parameter is preceded by a type name that specifies what kind of thrown value the catch block can catch. 2. The catch-block parameter gives you a name for the thrown value that is caught, so you can write code in the catch block that does things with that value. We will discuss these two functions of the catch-block parameter in reverse order. This subsection discusses using the catch-block parameter as a name for the value that was thrown and is caught. The subsection entitled “Multiple Throws and Catches,” later in this chapter, discusses which catch block (which exception handler) will process a value throw S TATEMENT S YNTAX throw Expression_for_Value_to_Be_Thrown ; When the throw statement is executed, the execution of the enclosing try block is stopped. If the try block is followed by a suitable catch block, then flow of control is transferred to the catch block. A throw statement is almost always embedded in a branching statement, such as an if statement. The value thrown can be of any type. E XAMPLE if (milk <= 0) throw donuts; exception handler catch-block parameter 18_CH18.fm Page 764 Monday, August 18, 2003 1:23 PM Exception Handling Basics 765 that is thrown. Our current example has only one catch block. A common name for a catch-block parameter is e, but you can use any legal identifier in place of e. Let’s see how the catch block in Display 18.2 works. When a value is thrown, exe- cution of the code in the try block ends and control passes to the catch block (or blocks) that is placed right after the try block. The catch block from Display 18.2 is reproduced here: catch(int e) { cout << e << " donuts, and No Milk!\n" << "Go buy some milk.\n"; } When a value is thrown, the thrown value must be of type int in order for this particu- lar catch block to apply. In Display 18.2, the value thrown is given by the variable donuts; because donuts is of type int, this catch block can catch the value thrown. Suppose the value of donuts is 12 and the value of milk is 0, as in the second sample dialogue in Display 18.2. Since the value of milk is not positive, the throw statement within the if statement is executed. In that case the value of the variable donuts is thrown. When the catch block in Display 18.2 catches the value of donuts, the value of donuts is plugged in for the catch-block parameter e and the code in the catch block is executed, producing the following output: 12 donuts, and No Milk! Go buy some milk. If the value of donuts is positive, the throw statement is not executed. In this case the entire try block is executed. After the last statement in the try block is executed, the statement after the catch block is executed. Note that if no exception is thrown, the catch block is ignored. This discussion makes it sound like a try-throw-catch setup is equivalent to an if- else statement. It almost is equivalent, except for the value thrown. A try-throw- catch setup is like an if-else statement with the added ability to send a message to one of the branches. This does not sound much different from an if-else statement, but it turns out to be a big difference in practice. To summarize in a more formal tone, a try block contains some code that we are assuming includes a throw statement. The throw statement is normally executed only in exceptional circumstances, but when it is executed, it throws a value of some type. When an exception (a value such as donuts in Display 18.2) is thrown, the try block ends. All the rest of the code in the try block is ignored and control passes to a suitable catch block. A catch block applies only to an immediately preceding try block. If the exception is thrown, then that exception object is plugged in for the catch-block parameter, and the statements in the catch block are executed. For example, if you look at the dialogues in Display 18.2, you will see that as soon as the user enters a nonposi- tive number, the try block stops and the catch block is executed. For now, we will 18_CH18.fm Page 765 Monday, August 18, 2003 1:23 PM 766 Exception Handling assume that every try block is followed by an appropriate catch block. We will later discuss what happens when there is no appropriate catch block. If no exception (no value) is thrown in the try block, then after the try block is completed, program execution continues with the code after the catch block. In other words, if no exception is thrown, the catch block is ignored. Most of the time when the program is executed, the throw statement will not be executed, and so in most cases the code in the try block will run to completion and the code in the catch block will be ignored completely. catch- B LOCK P ARAMETER The catch-block parameter is an identifier in the heading of a catch block that serves as a placeholder for an exception (a value) that might be thrown. When a suitable value is thrown in the preceding try block, that value is plugged in for the catch-block parameter. (In order for the catch block to be executed the value throw must be of the type given for its catch-block parameter.) You can use any legal (nonreserved-word) identifier for a catch-block parameter. E XAMPLE catch(int e) { cout << e << " donuts, and No Milk!\n" << "Go buy some milk.\n"; } e is the catch-block parameter. try-throw-catch The basic mechanism for throwing and catching exceptions is a try-throw-catch sequence. The throw statement throws the exception (a value). The catch block catches the exception (the value). When an exception is thrown, the try block ends and then the code in the catch block is executed. After the catch block is completed, the code after the catch block or blocks is exe- cuted (provided the catch block has not ended the program or performed some other special action). (The type of the thrown exception must match the type listed for the catch-block parameter or else the exception will not be caught by that catch block. This point is discussed further in the subsection “Multiple Throws and Catches.”) If no exception is thrown in the try block, then after the try block is completed, program execu- tion continues with the code after the catch block or blocks. (In other words, if no exception is thrown, the catch block or blocks are ignored.) 18_CH18.fm Page 766 Monday, August 18, 2003 1:23 PM . milk. End of program. 18_CH18.fm Page 760 Monday, August 18, 2003 1:23 PM Exception Handling Basics 761 Display 18.2 Same Thing Using Exception Handling (part 1 of 2) 1 2 #include <iostream> 3. just a toy example to learn C++ syntax. Do not take it as an example of good typical use of exception handling. 18_CH18.fm Page 761 Monday, August 18, 2003 1:23 PM 762 Exception Handling a. handling details of C++ by means of simple examples that would not normally use exception handling. This makes a lot of sense for learning about the exception handling details of C++, but do not