3.2. Control of flow

3.2.1. The if statement

The if statement has two forms:

if(expression) statement

if(expression) statement1
else statement2

In the first form, if (and only if) the expression is non-zero, the statement is executed. If the expression is zero, the statement is ignored. Remember that the statement can be compound; that is the way to put several statements under the control of a single if.

The second form is like the first except that if the statement shown as statement1 is selected then statement2 will not be, and vice versa.

Either form is considered to be a single statement in the syntax of C, so the following is completely legal.

if(expression)
    if(expression) statement

The first if (expression) is followed by a properly formed, complete if statement. Since that is legally a statement, the first if can be considered to read

if(expression) statement

and is therefore itself properly formed. The argument can be extended as far as you like, but it's a bad habit to get into. It is better style to make the statement compound even if it isn't necessary. That makes it a lot easier to add extra statements if they are needed and generally improves readability.

The form involving else works the same way, so we can also write this.

if(expression)
  if(expression)
    statement
  else
    statement

As Chapter 1 has said already, this is now ambiguous. It is not clear, except as indicated by the indentation, which of the ifs is responsible for the else. If we follow the rules that the previous example suggests, then the second if is followed by a statement, and is therefore itself a statement, so the else belongs to the first if.

That is not the way that C views it. The rule is that an else belongs to the first if above that hasn't already got an else. In the example we're discussing, the else goes with the second if.

To prevent any unwanted association between an else and an if just above it, the if can be hidden away by using a compound statement. To repeat the example in Chapter 1, here it is.

if(expression){
    if(expression)
            statement
}else
    statement

Putting in all the compound statement brackets, it becomes this:

if(expression){
    if(expression){
        statement
    }
}else{
    statement
}

If you happen not to like the placing of the brackets, it is up to you to put them where you think they look better; just be consistent about it. You probably need to know that this a subject on which feelings run deep.

3.2.2. The while and do statements

The while is simple:

while(expression)
    statement

The statement is only executed if the expression is non-zero. After every execution of the statement, the expression is evaluated again and the process repeats if it is non-zero. What could be plainer than that? The only point to watch out for is that the statement may never be executed, and that if nothing in the statement affects the value of the expression then the while will either do nothing or loop for ever, depending on the initial value of the expression.

It is occasionally desirable to guarantee at least one execution of the statement following the while, so an alternative form exists known as the do statement. It looks like this:

do
    statement
while(expression);

and you should pay close attention to that semicolon—it is not optional! The effect is that the statement part is executed before the controlling expression is evaluated, so this guarantees at least one trip around the loop. It was an unfortunate decision to use the keyword while for both purposes, but it doesn't seem to cause too many problems in practice.

If you feel the urge to use a do, think carefully. It is undoubtedly essential in certain cases, but experience has shown that the use of do statements is often associated with poorly constructed code. Not every time, obviously, but as a general rule you should stop and ask yourself if you have made the right choice. Their use often indicates a hangover of thinking methods learnt with other languages, or just sloppy design. When you do convince yourself that nothing else will give you just what is wanted, then go ahead - be daring—use it.

3.2.2.1. Handy hints

A very common trick in C programs is to use the result of an assignment to control while and do loops. It is so commonplace that, even if you look at it the first time and blench, you've got no alternative but to learn it. It falls into the category of ‘idiomatic’ C and eventually becomes second nature to anybody who really uses the language. Here is the most common example of all:

#include <stdio.h>
#include <stdlib.h>

main(){
    int input_c;

    /* The Classic Bit */
    while( (input_c = getchar()) != EOF){
            printf("%c was read\n", input_c);
    }
    exit(EXIT_SUCCESS);
}
Example 3.2

The clever bit is the expression assigning to input_c. It is assigned to, compared with EOF (End Of File), and used to control the loop all in one go. Embedding the assignment like that is a handy embellishment. Admittedly it only saves one line of code, but the benefit in terms of readability (once you have got used to seeing it) is quite large. Learn where the parentheses are, too. They're necessary for precedence reasons—work out why!

Note that input_c is an int. This is because getchar has to be able to return not only every possible value of a char, but also an extra value, EOF. To do that, a type longer than a char is necessary.

Both the while and the do statements are themselves syntactically a single statement, just like an if statement. They occur anywhere that any other single statement is permitted. If you want them to control several statements, then you will have to use a compound statement, as the examples of if illustrated.

3.2.3. The for statement

A very common feature in programs is loops that are controlled by variables used as a counter. The counter doesn't always have to count consecutive values, but the usual arrangement is for it to be initialized outside the loop, checked every time around the loop to see when to finish and updated each time around the loop. There are three important places, then, where the loop control is concentrated: initialize, check and update. This example shows them.

#include <stdio.h>
#include <stdlib.h>
main(){
      int i;

      /* initialise */
      i = 0;
      /* check */
      while(i <= 10){
              printf("%d\n", i);
              /* update */
              i++;
      }
      exit(EXIT_SUCCESS);
}
Example 3.3

As you will have noticed, the initialization and check parts of the loop are close together and their location is obvious because of the presence of the while keyword. What is harder to spot is the place where the update occurs, especially if the value of the controlling variable is used within the loop. In that case, which is by far the most common, the update has to be at the very end of the loop: far away from the initialize and check. Readability suffers because it is hard to work out how the loop is going to perform unless you read the whole body of the loop carefully. What is needed is some way of bringing the initialize, check and update parts into one place so that they can be read quickly and conveniently. That is exactly what the for statement is designed to do. Here it is.

for (initialize; check; update) statement

The initialize part is an expression; nearly always an assignment expression which is used to initialize the control variable. After the initialization, the check expression is evaluated: if it is non-zero, the statement is executed, followed by evaluation of the update expression which generally increments the control variable, then the sequence restarts at the check. The loop terminates as soon as the check evaluates to zero.

There are two important things to realize about that last description: one, that each of the three parts of the for statement between the parentheses are just expressions; two, that the description has carefully explained what they are intended to be used for without proscribing alternative uses—that was done deliberately. You can use the expressions to do whatever you like, but at the expense of readability if they aren't used for their intended purpose.

Here is a program that does the same thing twice, the first time using a while loop, the second time with a for. The use of the increment operator is exactly the sort of use that you will see in everyday practice.

#include <stdio.h>
#include <stdlib.h>
main(){
      int i;

      i = 0;
      while(i <= 10){
              printf("%d\n", i);
              i++;
      }

      /* the same done using ``for'' */
      for(i = 0; i <= 10; i++){
              printf("%d\n", i);
      }
      exit(EXIT_SUCCESS);
}
Example 3.4

There isn't any difference betweeen the two, except that in this case the for loop is more convenient and maintainable than the while statement. You should always use the for when it's appropriate; when a loop is being controlled by some sort of counter. The while is more at home when an indeterminate number of cycles of the loop are part of the problem. As always, it needs a degree of judgement on behalf of the author of the program; an understanding of form, style, elegance and the poetry of a well written program. There is no evidence that the software business suffers from a surfeit of those qualities, so feel free to exercise them if you are able.

Any of the initialize, check and update expressions in the for statement can be omitted, although the semicolons must stay. This can happen if the counter is already initialized, or gets updated in the body of the loop. If the check expression is omitted, it is assumed to result in a ‘true’ value and the loop never terminates. A common way of writing never-ending loops is either

for(;;)

or

while(1)

and both can be seen in existing programs.

3.2.4. A brief pause

The control of flow statements that we've just seen are quite adequate to write programs of any degree of complexity. They lie at the core of C and even a quick reading of everyday C programs will illustrate their importance, both in the provision of essential functionality and in the structure that they emphasize. The remaining statements are used to give programmers finer control or to make it easier to deal with exceptional conditions. Only the switch statement is enough of a heavyweight to need no justification for its use; yes, it can be replaced with lots of ifs, but it adds a lot of readability. The others, break, continue and goto, should be treated like the spices in a delicate sauce. Used carefully they can turn something commonplace into a treat, but a heavy hand will drown the flavour of everything else.

3.2.5. The switch statement

This is not an essential part of C. You could do without it, but the language would have become significantly less expressive and pleasant to use.

It is used to select one of a number of alternative actions depending on the value of an expression, and nearly always makes use of another of the lesser statements: the break. It looks like this.

switch (expression){
case const1:    statements
case const2:    statements
default:        statements
}

The expression is evaluated and its value is compared with all of the const1 etc. expressions, which must all evaluate to different constant values (strictly they are integral constant expressions, see Chapter 6 and below). If any of them has the same value as the expression then the statement following the case label is selected for execution. If the default is present, it will be selected when there is no matching value found. If there is no default and no matching value, the entire switch statement will do nothing and execution will continue at the following statement.

One curious feature is that the cases are not exclusive, as this example shows.

#include <stdio.h>
#include <stdlib.h>

main(){
      int i;
      for(i = 0; i <= 10; i++){
              switch(i){
                      case 1:
                      case 2:
                              printf("1 or 2\n");
                      case 7:
                              printf("7\n");
                      default:
                              printf("default\n");
              }
      }
      exit(EXIT_SUCCESS);
}
Example 3.5

The loop cycles with i having values 0–10. A value of 1 or 2 will cause the printing of the message 1 or 2 by selecting the first of the printf statements. What you might not expect is the way that the remaining messages would also appear! It's because the switch only selects one entry point to the body of the statement; after starting at a given point all of the following statements are also executed. The case and default labels simply allow you to indicate which of the statements is to be selected. When i has the value of 7, only the last two messages will be printed. Any value other than 1, 2, or 7 will find only the last message.

The labels can occur in any order, but no two values may be the same and you are allowed either one or no default (which doesn't have to be the last label). Several labels can be put in front of one statement and several statements can be put after one label.

The expression controlling the switch can be of any of the integral types. Old C used to insist on only int here, and some compilers would forcibly truncate longer types, giving rise on rare occasions to some very obscure bugs.

3.2.5.1. The major restriction

The biggest problem with the switch statement is that it doesn't allow you to select mutually exclusive courses of action; once the body of the statement has been entered any subsequent statements within the body will all be executed. What is needed is the break statement. Here is the previous example, but amended to make sure that the messages printed come out in a more sensible order. The break statements cause execution to leave the switch statement immediately and prevent any further statements in the body of the switch from being executed.

#include <stdio.h>
#include <stdlib.h>
main(){
      int i;
      for(i = 0; i <= 10; i++){
              switch(i){
                      case 1:
                      case 2:
                              printf("1 or 2\n");
                              break;
                      case 7:
                              printf("7\n");
                              break;
                      default:
                              printf("default\n");
              }
      }
      exit(EXIT_SUCCESS);
}
Example 3.6

The break has further uses. Its own section follows soon.

3.2.5.2. Integral Constant Expression

Although Chapter 6 deals with constant expressions, it is worth looking briefly at what an integral constant expression is, since that is what must follow the case labels in a switch statement. Loosely speaking, it is any expression that does not involve any value-changing operation (like increment or assignment), function calls or comma operators. The operands in the expression must all be integer constants, character constants, enumeration constants, sizeof expressions and floating-point constants that are the immediate operands of casts. Any cast operators must result in integral types.

Much what you would expect, really.

3.2.6. The break statement

This is a simple statement. It only makes sense if it occurs in the body of a switch, do, while or for statement. When it is executed the control of flow jumps to the statement immediately following the body of the statement containing the break. Its use is widespread in switch statements, where it is more or less essential to get the control that most people want.

The use of the break within loops is of dubious legitimacy. It has its moments, but is really only justifiable when exceptional circumstances have happened and the loop has to be abandoned. It would be nice if more than one loop could be abandoned with a single break but that isn't how it works. Here is an example.

#include <stdio.h>
#include <stdlib.h>
main(){
      int i;

      for(i = 0; i < 10000; i++){
              if(getchar() == 's')
                      break;
              printf("%d\n", i);
      }
      exit(EXIT_SUCCESS);
}
Example 3.7

It reads a single character from the program's input before printing the next in a sequence of numbers. If an ‘s’ is typed, the break causes an exit from the loop.

If you want to exit from more than one level of loop, the break is the wrong thing to use. The goto is the only easy way, but since it can't be mentioned in polite company, we'll leave it till last.

3.2.7. The continue statement

This statement has only a limited number of uses. The rules for its use are the same as for break, with the exception that it doesn't apply to switch statements. Executing a continue starts the next iteration of the smallest enclosing do, while or for statement immediately. The use of continue is largely restricted to the top of loops, where a decision has to be made whether or not to execute the rest of the body of the loop. In this example it ensures that division by zero (which gives undefined behaviour) doesn't happen.

#include <stdio.h>
#include <stdlib.h>
main(){
      int i;

      for(i = -10; i < 10; i++){
              if(i == 0)
                      continue;
              printf("%f\n", 15.0/i);
              /*
               * Lots of other statements .....
               */
      }
      exit(EXIT_SUCCESS);
}
Example 3.7

You could take a puritanical stance and argue that, instead of a conditional continue,, the body of the loop should be made conditional instead—but you wouldn't have many supporters. Most C programmers would rather have the continue than the extra level of indentation, particularly if the body of the loop is large.

Of course the continue can be used in other parts of a loop, too, where it may occasionally help to simplify the logic of the code and improve readability. It deserves to be used sparingly.

Do remember that continue has no special meaning to a switch statement, where break does have. Inside a switch, continue is only valid if there is a loop that encloses the switch, in which case the next iteration of the loop will be started.

There is an important difference between loops written with while and for. In a while, a continue will go immediately to the test of the controlling expression. The same thing in a for will do two things: first the update expression is evaluated, then the controlling expresion is evaluated.

3.2.8. goto and labels

Everybody knows that the goto statement is a ‘bad thing’. Used without care it is a great way of making programs hard to follow and of obscuring any structure in their flow. Dijkstra wrote a famous paper in 1968 called ‘Goto Statement Considered Harmful’, which everybody refers to and almost nobody has read.

What's especially annoying is that there are times when it is the most appropriate thing to use in the circumstances! In C, it is used to escape from multiple nested loops, or to go to an error handling exit at the end of a function. You will need a label when you use a goto; this example shows both.

goto L1;
/* whatever you like here */
L1: /* anything else */

A label is an identifier followed by a colon. Labels have their own ‘name space’ so they can't clash with the names of variables or functions. The name space only exists for the function containing the label, so label names can be re-used in different functions. The label can be used before it is declared, too, simply by mentioning it in a goto statement.

Labels must be part of a full statement, even if it's an empty one. This usually only matters when you're trying to put a label at the end of a compound statement—like this.

label_at_end: ; /* empty statement */
}

The goto works in an obvious way, jumping to the labelled statements. Because the name of the label is only visible inside its own function, you can't jump from one function to another one.

It's hard to give rigid rules about the use of gotos but, as with the do, continue and the break (except in switch statements), over-use should be avoided. Think carefully every time you feel like using one, and convince yourself that the structure of the program demands it. More than one goto every 3–5 functions is a symptom that should be viewed with deep suspicion.

Summary

Now we've seen all of the control of flow statements and examples of their use. Some should be used whenever possible, some are not for use line by line but for special purposes where their particular job is called for. It is possible to write elegant and beautiful programs in C if you are prepared to take the extra bit of care necessary; the specialized control of flow statements give you the chance to add the extra polish that some other languages lack.

All that remains to be done to complete the picture of flow of control in C is to finish off the logical operators.