The call stack is an area of a program's memory used to manage the currently executing function and all pending function calls. Figure 3 lends some insight into what is meant by a "pending function call." Looking back at figure 3, we can see that even though the instance of function factorial with the parameter 5 was the first to begin execution, it is the last to finish execution. The program pauses the execution of factorial(5) "pending" the completion of the function call to factorial(4). Likewise, the program pauses the execution of factorial(4) pending the completion of the function call to factorial(3). This series of nested function calls is managed using the call stack. Consider the following example C++ program.
| 1: 2: 3: 4: 5: 6: 7: 8: 9: 10: 11: 12: 13: 14: 15: 16: 17: 18: 19: 20: 21: | #include <iostream> #include <cstdlib> using namespace std; void method3(void) { cout << "Method 3" << endl; } void method2(void) { method3(); } void method1(void) { method2(); } int main(int argc, char* argv[]) { method1(); return EXIT_SUCCESS; } |
| Listing 3A program with nested function calls |
Stepping through the function calls in Listing 3 demonstrates how a program uses the call stack to manage the function calls in a program. As in all C++ programs, the first function that executes in the program in Listing 3 is function main. Figure 1(a) represents the state of the call stack after function main begins execution, but before it calls method1. Since function main is the routine currently in execution, the information needed to run this function sits on the top of the call stack. This information, which includes among other things the local variables of the function, is known as an activation record.
Figure 1 The call stack during various points of Listing 3
When main calls function method1, the run-time system pushes an activation record for method1 onto the top of the call stack. The run-time system then halts execution of function main, pending the completion of method1. At this point in the program, Figure 1(b) represents the state of the call stack. Function method1 then immediately calls method2. The run-time system pauses execution of method1 and pushes an activation record for method2 onto the top of the stack. This state of the call stack at this point in the program corresponds to Figure 1(c). After function method2 calls method3, the call stack resembles Figure 1(d).
At this point during the execution of the program, function calls are nested four levels deep. The program currently is executing function method3, with functions method2, method1, and main all suspended. After the program finishes execution of function method3, the run-time system pops off the activation record for method3 from the top of the stack. Execution of method2 then resumes. The call stack would then again resemble Figure 1(c). When execution of method2 completes, the run-time system pops it off the stack also, putting the call stack back to the state represented in Figure 1(b). As the nested functions in the program end, the call stack grows shorter and shorter. Eventually, function main finishes and the run-time system pops its activation record off the stack, ending the execution of the program.
The call stack operates in the same manner when dealing with recursive functions as it does with regular functions. Revisiting the factorial example, let's trace the call stack as it manages the recursive calls to function factorial. The program execution begins with function main. Function main then calls function factorial with the parameter 5. The run-time system pushes the information needed to execute factorial with the parameter 5 on top of the stack. The call stack, at this point resembles Figure 2.
Figure 2 The call stack while factorial(5) executes
The function call factorial(5), as we know from earlier examination, makes a call to factorial(4), which makes a call to factorial(3), and so on until factorial(0) begins execution. At this point, the call stack resembles Figure 3.
Figure 3 The call stack while factorial(0) executes
Function factorial(0) is the base case of the recursion. When this instance of the function finishes, it returns the value 1 to the function whose activation record sits below it. Then the run-time system pops the activation record for factorial(0) off the stack and resumes execution with function factorial(1). The nested functions continue to "unwind" in this manner, passing values back to their calling functions, until function main completes and the program terminates.
Debuggers often provide a feature that allows programmers to inspect the call stack of a program. This feature can prove invaluable when trying to debug a recursive function. Figure 4 shows an example call-stack window from Microsoft Visual C++.
Figure 4 Microsoft Visual C++ call stack window