3: Controlling Program Flow

Like a sentient creature, a program must manipulate its world and make choices during execution.

In Java you manipulate objects and data using operators, and you make choices with execution control statements. Java was inherited from C++, so most of these statements and operators will be familiar to C and C++ programmers. Java has also added some improvements and simplifications.

If you find yourself floundering a bit in this chapter, make sure you go through the multimedia CD ROM bound into this book: Thinking in C: Foundations for Java and C++. It contains audio lectures, slides, exercises, and solutions specifically designed to bring you up to speed with the C syntax necessary to learn Java.

An operator takes one or more arguments and produces a new value. The arguments are in a different form than ordinary method calls, but the effect is the same. You should be reasonably comfortable with the general concept of operators from your previous programming experience. Addition (+), subtraction and unary minus (-), multiplication (*), division (/), and assignment (=) all work much the same in any programming language.

All operators produce a value from their operands. In addition, an operator can change the value of an operand. This is called a side effect. The most common use for operators that modify their operands is to generate the side effect, but you should keep in mind that the value produced is available for your use just as in operators without side effects.

Almost all operators work only with primitives. The exceptions are ‘=’, ‘==’ and ‘!=’, which work with all objects (and are a point of confusion for objects). In addition, the String class supports ‘+’ and ‘+=’.

Operator precedence defines how an expression evaluates when several operators are present. Java has specific rules that determine the order of evaluation. The easiest one to remember is that multiplication and division happen before addition and subtraction. Programmers often forget the other precedence rules, so you should use parentheses to make the order of evaluation explicit. For example:

has a very different meaning from the same statement with a particular grouping of parentheses:

Assignment is performed with the operator =. It means “take the value of the right-hand side (often called the rvalue) and copy it into the left-hand side (often called the lvalue). An rvalue is any constant, variable or expression that can produce a value, but an lvalue must be a distinct, named variable. (That is, there must be a physical space to store a value.) For instance, you can assign a constant value to a variable (A = 4;), but you cannot assign anything to constant value—it cannot be an lvalue. (You can’t say 4 = A;.)

Assignment of primitives is quite straightforward. Since the primitive holds the actual value and not a reference to an object, when you assign primitives you copy the contents from one place to another. For example, if you say A = B for primitives, then the contents of B are copied into A. If you then go on to modify A, B is naturally unaffected by this modification. As a programmer, this is what you’ve come to expect for most situations.

When you assign objects, however, things change. Whenever you manipulate an object, what you’re manipulating is the reference, so when you assign “from one object to another” you’re actually copying a reference from one place to another. This means that if you say C = D for objects, you end up with both C and D pointing to the object that, originally, only D pointed to. The following example will demonstrate this.

Here’s the example:

The Number class is simple, and two instances of it (n1 and n2) are created within main( ). The i value within each Number is given a different value, and then n2 is assigned to n1, and n1 is changed. In many programming languages you would expect n1 and n2 to be independent at all times, but because you’ve assigned a reference here’s the output you’ll see:

Changing the n1 object appears to change the n2 object as well! This is because both n1 and n2 contain the same reference, which is pointing to the same object. (The original reference that was in n1 that pointed to the object holding a value of 9 was overwritten during the assignment and effectively lost; its object will be cleaned up by the garbage collector.)

This phenomenon is often called aliasing and it’s a fundamental way that Java works with objects. But what if you don’t want aliasing to occur in this case? You could forego the assignment and say:

This retains the two separate objects instead of tossing one and tying n1 and n2 to the same object, but you’ll soon realize that manipulating the fields within objects is messy and goes against good object-oriented design principles. This is a nontrivial topic, so it is left for Appendix A, which is devoted to aliasing. In the meantime, you should keep in mind that assignment for objects can add surprises.

Aliasing will also occur when you pass an object into a method:

In many programming languages, the method f( ) would appear to be making a copy of its argument Letter y inside the scope of the method. But once again a reference is being passed so the line

is actually changing the object outside of f( ). The output shows this:

Aliasing and its solution is a complex issue and, although you must wait until Appendix A for all the answers, you should be aware of it at this point so you can watch for pitfalls.

The basic mathematical operators are the same as the ones available in most programming languages: addition (+), subtraction (-), division (/), multiplication (*) and modulus (%, which produces the remainder from integer division). Integer division truncates, rather than rounds, the result.

Java also uses a shorthand notation to perform an operation and an assignment at the same time. This is denoted by an operator followed by an equal sign, and is consistent with all the operators in the language (whenever it makes sense). For example, to add 4 to the variable x and assign the result to x, use: x += 4.

This example shows the use of the mathematical operators:

The first thing you will see are some shorthand methods for printing: the prt( ) method prints a String, the pInt( ) prints a String followed by an int and the pFlt( ) prints a String followed by a float. Of course, they all ultimately end up using System.out.println( ).

To generate numbers, the program first creates a Random object. Because no arguments are passed during creation, Java uses the current time as a seed for the random number generator. The program generates a number of different types of random numbers with the Random object simply by calling different methods: nextInt( ), nextLong( ), nextFloat( ) or nextDouble( ).

The modulus operator, when used with the result of the random number generator, limits the result to an upper bound of the operand minus one (99 in this case).

The unary minus (-) and unary plus (+) are the same operators as binary minus and plus. The compiler figures out which use is intended by the way you write the expression. For instance, the statement

has an obvious meaning. The compiler is able to figure out:

but the reader might get confused, so it is clearer to say:

The unary minus produces the negative of the value. Unary plus provides symmetry with unary minus, although it doesn’t have any effect.

Java, like C, is full of shortcuts. Shortcuts can make code much easier to type, and either easier or harder to read.

Two of the nicer shortcuts are the increment and decrement operators (often referred to as the auto-increment and auto-decrement operators). The decrement operator is -- and means “decrease by one unit.” The increment operator is ++ and means “increase by one unit.” If a is an int, for example, the expression ++a is equivalent to (a = a + 1). Increment and decrement operators produce the value of the variable as a result.

There are two versions of each type of operator, often called the prefix and postfix versions. Pre-increment means the ++ operator appears before the variable or expression, and post-increment means the ++ operator appears after the variable or expression. Similarly, pre-decrement means the -- operator appears before the variable or expression, and post-decrement means the -- operator appears after the variable or expression. For pre-increment and pre-decrement, (i.e., ++a or --a), the operation is performed and the value is produced. For post-increment and post-decrement (i.e. a++ or a--), the value is produced, then the operation is performed. As an example:

The output for this program is:

You can see that for the prefix form you get the value after the operation has been performed, but with the postfix form you get the value before the operation is performed. These are the only operators (other than those involving assignment) that have side effects. (That is, they change the operand rather than using just its value.)

The increment operator is one explanation for the name C++, implying “one step beyond C.” In an early Java speech, Bill Joy (one of the creators), said that “Java=C++--” (C plus plus minus minus), suggesting that Java is C++ with the unnecessary hard parts removed and therefore a much simpler language. As you progress in this book you’ll see that many parts are simpler, and yet Java isn’t that much easier than C++.

Relational operators generate a boolean result. They evaluate the relationship between the values of the operands. A relational expression produces true if the relationship is true, and false if the relationship is untrue. The relational operators are less than (<), greater than (>), less than or equal to (<=), greater than or equal to (>=), equivalent (==) and not equivalent (!=). Equivalence and nonequivalence works with all built-in data types, but the other comparisons won’t work with type boolean.

The relational operators == and != also work with all objects, but their meaning often confuses the first-time Java programmer. Here’s an example:

The expression System.out.println(n1 == n2) will print the result of the boolean comparison within it. Surely the output should be true and then false, since both Integer objects are the same. But while the contents of the objects are the same, the references are not the same and the operators == and != compare object references. So the output is actually false and then true. Naturally, this surprises people at first.

What if you want to compare the actual contents of an object for equivalence? You must use the special method equals( ) that exists for all objects (not primitives, which work fine with == and !=). Here’s how it’s used:

The result will be true, as you would expect. Ah, but it’s not as simple as that. If you create your own class, like this:

you’re back to square one: the result is false. This is because the default behavior of equals( ) is to compare references. So unless you override equals( ) in your new class you won’t get the desired behavior. Unfortunately, you won’t learn about overriding until Chapter 7, but being aware of the way equals( ) behaves might save you some grief in the meantime.

Most of the Java library classes implement equals( ) so that it compares the contents of objects instead of their references.

The logical operators AND (&&), OR (||) and NOT (!) produce a boolean value of true or false based on the logical relationship of its arguments. This example uses the relational and logical operators:

You can apply AND, OR, or NOT to boolean values only. You can’t use a non-boolean as if it were a boolean in a logical expression as you can in C and C++. You can see the failed attempts at doing this commented out with a //! comment marker. The subsequent expressions, however, produce boolean values using relational comparisons, then use logical operations on the results.

One output listing looked like this:

Note that a boolean value is automatically converted to an appropriate text form if it’s used where a String is expected.

You can replace the definition for int in the above program with any other primitive data type except boolean. Be aware, however, that the comparison of floating-point numbers is very strict. A number that is the tiniest fraction different from another number is still “not equal.” A number that is the tiniest bit above zero is still nonzero.

When dealing with logical operators you run into a phenomenon called “short circuiting.” This means that the expression will be evaluated only until the truth or falsehood of the entire expression can be unambiguously determined. As a result, all the parts of a logical expression might not be evaluated. Here’s an example that demonstrates short-circuiting:

Each test performs a comparison against the argument and returns true or false. It also prints information to show you that it’s being called. The tests are used in the expression:

You might naturally think that all three tests would be executed, but the output shows otherwise:

The first test produced a true result, so the expression evaluation continues. However, the second test produced a false result. Since this means that the whole expression must be false, why continue evaluating the rest of the expression? It could be expensive. The reason for short-circuiting, in fact, is precisely that; you can get a potential performance increase if all the parts of a logical expression do not need to be evaluated.

The bitwise operators allow you to manipulate individual bits in an integral primitive data type. Bitwise operators perform boolean algebra on the corresponding bits in the two arguments to produce the result.

The bitwise operators come from C’s low-level orientation; you were often manipulating hardware directly and had to set the bits in hardware registers. Java was originally designed to be embedded in TV set-top boxes, so this low-level orientation still made sense. However, you probably won’t use the bitwise operators much.

The bitwise AND operator (&) produces a one in the output bit if both input bits are one; otherwise it produces a zero. The bitwise OR operator (|) produces a one in the output bit if either input bit is a one and produces a zero only if both input bits are zero. The bitwise EXCLUSIVE OR, or XOR (^), produces a one in the output bit if one or the other input bit is a one, but not both. The bitwise NOT (~, also called the ones complement operator) is a unary operator; it takes only one argument. (All other bitwise operators are binary operators.) Bitwise NOT produces the opposite of the input bit—a one if the input bit is zero, a zero if the input bit is one.

The bitwise operators and logical operators use the same characters, so it is helpful to have a mnemonic device to help you remember the meanings: since bits are “small,” there is only one character in the bitwise operators.

Bitwise operators can be combined with the = sign to unite the operation and assignment: &=, |= and ^= are all legitimate. (Since ~ is a unary operator it cannot be combined with the = sign.)

The boolean type is treated as a one-bit value so it is somewhat different. You can perform a bitwise AND, OR and XOR, but you can’t perform a bitwise NOT (presumably to prevent confusion with the logical NOT). For booleans the bitwise operators have the same effect as the logical operators except that they do not short circuit. Also, bitwise operations on booleans include an XOR logical operator that is not included under the list of “logical” operators. You’re prevented from using booleans in shift expressions, which is described next.

The shift operators also manipulate bits. They can be used solely with primitive, integral types. The left-shift operator (<<) produces the operand to the left of the operator shifted to the left by the number of bits specified after the operator (inserting zeroes at the lower-order bits). The signed right-shift operator (>>) produces the operand to the left of the operator shifted to the right by the number of bits specified after the operator. The signed right shift >> uses sign extension: if the value is positive, zeroes are inserted at the higher-order bits; if the value is negative, ones are inserted at the higher-order bits. Java has also added the unsigned right shift >>>, which uses zero extension: regardless of the sign, zeroes are inserted at the higher-order bits. This operator does not exist in C or C++.

If you shift a char, byte, or short, it will be promoted to int before the shift takes place, and the result will be an int. Only the five low-order bits of the right-hand side will be used. This prevents you from shifting more than the number of bits in an int. If you’re operating on a long, you’ll get a long result. Only the six low-order bits of the right-hand side will be used so you can’t shift more than the number of bits in a long.

Shifts can be combined with the equal sign (<<= or >>= or >>>=). The lvalue is replaced by the lvalue shifted by the rvalue. There is a problem, however, with the unsigned right shift combined with assignment. If you use it with byte or short you don’t get the correct results. Instead, these are promoted to int and right shifted, but then truncated as they are assigned back into their variables, so you get -1 in those cases. The following example demonstrates this:

In the last line, the resulting value is not assigned back into b, but is printed directly and so the correct behavior occurs.

Here’s an example that demonstrates the use of all the operators involving bits:

The two methods at the end, pBinInt( ) and pBinLong( ) take an int or a long, respectively, and print it out in binary format along with a descriptive string. You can ignore the implementation of these for now.

You’ll note the use of System.out.print( ) instead of System.out.println( ). The print( ) method does not emit a new line, so it allows you to output a line in pieces.

As well as demonstrating the effect of all the bitwise operators for int and long, this example also shows the minimum, maximum, +1 and -1 values for int and long so you can see what they look like. Note that the high bit represents the sign: 0 means positive and 1 means negative. The output for the int portion looks like this:

The binary representation of the numbers is referred to as signed two’s complement.

This operator is unusual because it has three operands. It is truly an operator because it produces a value, unlike the ordinary if-else statement that you’ll see in the next section of this chapter. The expression is of the form:

If boolean-exp evaluates to true, value0 is evaluated and its result becomes the value produced by the operator. If boolean-exp is false, value1 is evaluated and its result becomes the value produced by the operator.

Of course, you could use an ordinary if-else statement (described later), but the ternary operator is much terser. Although C (where this operator originated) prides itself on being a terse language, and the ternary operator might have been introduced partly for efficiency, you should be somewhat wary of using it on an everyday basis—it’s easy to produce unreadable code.

The conditional operator can be used for its side effects or for the value it produces, but in general you want the value since that’s what makes the operator distinct from the if-else. Here’s an example:

You can see that this code is more compact than what you’d need to write without the ternary operator:

The second form is easier to understand, and doesn’t require a lot more typing. So be sure to ponder your reasons when choosing the ternary operator.

The comma is used in C and C++ not only as a separator in function argument lists, but also as an operator for sequential evaluation. The sole place that the comma operator is used in Java is in for loops, which will be described later in this chapter.

There’s one special usage of an operator in Java: the + operator can be used to concatenate strings, as you’ve already seen. It seems a natural use of the + even though it doesn’t fit with the traditional way that + is used. This capability seemed like a good idea in C++, so operator overloading was added to C++ to allow the C++ programmer to add meanings to almost any operator. Unfortunately, operator overloading combined with some of the other restrictions in C++ turns out to be a fairly complicated feature for programmers to design into their classes. Although operator overloading would have been much simpler to implement in Java than it was in C++, this feature was still considered too complex, so Java programmers cannot implement their own overloaded operators as C++ programmers can.

The use of the String + has some interesting behavior. If an expression begins with a String, then all operands that follow must be Strings (remember that the compiler will turn a quoted sequence of characters into a String):

Here, the Java compiler will convert x, y, and z into their String representations instead of adding them together first. And if you say:

Java will turn x into a String.

One of the pitfalls when using operators is trying to get away without parentheses when you are even the least bit uncertain about how an expression will evaluate. This is still true in Java.

An extremely common error in C and C++ looks like this:

The programmer was trying to test for equivalence (==) rather than do an assignment. In C and C++ the result of this assignment will always be true if y is nonzero, and you’ll probably get an infinite loop. In Java, the result of this expression is not a boolean, and the compiler expects a boolean and won’t convert from an int, so it will conveniently give you a compile-time error and catch the problem before you ever try to run the program. So the pitfall never happens in Java. (The only time you won’t get a compile-time error is when x and y are boolean, in which case x = y is a legal expression, and in the above case, probably an error.)

A similar problem in C and C++ is using bitwise AND and OR instead of the logical versions. Bitwise AND and OR use one of the characters (& or |) while logical AND and OR use two (&& and ||). Just as with = and ==, it’s easy to type just one character instead of two. In Java, the compiler again prevents this because it won’t let you cavalierly use one type where it doesn’t belong.

The word cast is used in the sense of “casting into a mold.” Java will automatically change one type of data into another when appropriate. For instance, if you assign an integral value to a floating-point variable, the compiler will automatically convert the int to a float. Casting allows you to make this type conversion explicit, or to force it when it wouldn’t normally happen.

To perform a cast, put the desired data type (including all modifiers) inside parentheses to the left of any value. Here’s an example:

As you can see, it’s possible to perform a cast on a numeric value as well as on a variable. In both casts shown here, however, the cast is superfluous, since the compiler will automatically promote an int value to a long when necessary. However, you are allowed to use superfluous casts in to make a point or to make your code more clear. In other situations, a cast may be essential just to get the code to compile.

In C and C++, casting can cause some headaches. In Java, casting is safe, with the exception that when you perform a so-called narrowing conversion (that is, when you go from a data type that can hold more information to one that doesn’t hold as much) you run the risk of losing information. Here the compiler forces you to do a cast, in effect saying “this can be a dangerous thing to do—if you want me to do it anyway you must make the cast explicit.” With a widening conversion an explicit cast is not needed because the new type will more than hold the information from the old type so that no information is ever lost.

Java allows you to cast any primitive type to any other primitive type, except for boolean, which doesn’t allow any casting at all. Class types do not allow casting. To convert one to the other there must be special methods. (String is a special case, and you’ll find out later in this book that objects can be cast within a family of types; an Oak can be cast to a Tree and vice-versa, but not to a foreign type such as a Rock.)

Ordinarily when you insert a literal value into a program the compiler knows exactly what type to make it. Sometimes, however, the type is ambiguous. When this happens you must guide the compiler by adding some extra information in the form of characters associated with the literal value. The following code shows these characters:

Hexadecimal (base 16), which works with all the integral data types, is denoted by a leading 0x or 0X followed by 0—9 and a—f either in upper or lowercase. If you try to initialize a variable with a value bigger than it can hold (regardless of the numerical form of the value), the compiler will give you an error message. Notice in the above code the maximum possible hexadecimal values for char, byte, and short. If you exceed these, the compiler will automatically make the value an int and tell you that you need a narrowing cast for the assignment. You’ll know you’ve stepped over the line.

Octal (base 8) is denoted by a leading zero in the number and digits from 0-7. There is no literal representation for binary numbers in C, C++ or Java.

A trailing character after a literal value establishes its type. Upper or lowercase L means long, upper or lowercase F means float and upper or lowercase D means double.

Exponents use a notation that I’ve always found rather dismaying: 1.39 e-47f. In science and engineering, ‘e’ refers to the base of natural logarithms, approximately 2.718. (A more precise double value is available in Java as Math.E.) This is used in exponentiation expressions such as 1.39 x e^-47, which means 1.39 x 2.718^-47. However, when FORTRAN was invented they decided that e would naturally mean “ten to the power,” which is an odd decision because FORTRAN was designed for science and engineering and one would think its designers would be sensitive about introducing such an ambiguity.[25] At any rate, this custom was followed in C, C++ and now Java. So if you’re used to thinking in terms of e as the base of natural logarithms, you must do a mental translation when you see an expression such as 1.39 e-47f in Java; it means 1.39 x 10^-47.

Note that you don’t need to use the trailing character when the compiler can figure out the appropriate type. With

there’s no ambiguity, so an L after the 200 would be superfluous. However, with

the compiler normally takes exponential numbers as doubles, so without the trailing f it will give you an error telling you that you must use a cast to convert double to float.

You’ll discover that if you perform any mathematical or bitwise operations on primitive data types that are smaller than an int (that is, char, byte, or short), those values will be promoted to int before performing the operations, and the resulting value will be of type int. So if you want to assign back into the smaller type, you must use a cast. (And, since you’re assigning back into a smaller type, you might be losing information.) In general, the largest data type in an expression is the one that determines the size of the result of that expression; if you multiply a float and a double, the result will be double; if you add an int and a long, the result will be long.

In C and C++, the sizeof( ) operator satisfies a specific need: it tells you the number of bytes allocated for data items. The most compelling need for sizeof( ) in C and C++ is portability. Different data types might be different sizes on different machines, so the programmer must find out how big those types are when performing operations that are sensitive to size. For example, one computer might store integers in 32 bits, whereas another might store integers as 16 bits. Programs could store larger values in integers on the first machine. As you might imagine, portability is a huge headache for C and C++ programmers.

Java does not need a sizeof( ) operator for this purpose because all the data types are the same size on all machines. You do not need to think about portability on this level—it is designed into the language.

Upon hearing me complain about the complexity of remembering operator precedence during one of my seminars, a student suggested a mnemonic that is simultaneously a commentary: “Ulcer Addicts Really Like C A lot.”

Of course, with the shift and bitwise operators distributed around the table it is not a perfect mnemonic, but for non-bit operations it works.

The following example shows which primitive data types can be used with particular operators. Basically, it is the same example repeated over and over, but using different primitive data types. The file will compile without error because the lines that would cause errors are commented out with a //!.

Note that boolean is quite limited. You can assign to it the values true and false, and you can test it for truth or falsehood, but you cannot add booleans or perform any other type of operation on them.

In char, byte, and short you can see the effect of promotion with the arithmetic operators. Each arithmetic operation on any of those types results in an int result, which must be explicitly cast back to the original type (a narrowing conversion that might lose information) to assign back to that type. With int values, however, you do not need to cast, because everything is already an int. Don’t be lulled into thinking everything is safe, though. If you multiply two ints that are big enough, you’ll overflow the result. The following example demonstrates this:

The output of this is:

and you get no errors or warnings from the compiler, and no exceptions at run-time. Java is good, but it’s not that good.

Compound assignments do not require casts for char, byte, or short, even though they are performing promotions that have the same results as the direct arithmetic operations. On the other hand, the lack of the cast certainly simplifies the code.

You can see that, with the exception of boolean, any primitive type can be cast to any other primitive type. Again, you must be aware of the effect of a narrowing conversion when casting to a smaller type, otherwise you might unknowingly lose information during the cast.

Java uses all of C’s execution control statements, so if you’ve programmed with C or C++ then most of what you see will be familiar. Most procedural programming languages have some kind of control statements, and there is often overlap among languages. In Java, the keywords include if-else, while, do-while, for, and a selection statement called switch. Java does not, however, support the much-maligned goto (which can still be the most expedient way to solve certain types of problems). You can still do a goto-like jump, but it is much more constrained than a typical goto.

All conditional statements use the truth or falsehood of a conditional expression to determine the execution path. An example of a conditional expression is A == B. This uses the conditional operator == to see if the value of A is equivalent to the value of B. The expression returns true or false. Any of the relational operators you’ve seen earlier in this chapter can be used to produce a conditional statement. Note that Java doesn’t allow you to use a number as a boolean, even though it’s allowed in C and C++ (where truth is nonzero and falsehood is zero). If you want to use a non-boolean in a boolean test, such as if(a), you must first convert it to a boolean value using a conditional expression, such as if(a != 0).

The if-else statement is probably the most basic way to control program flow. The else is optional, so you can use if in two forms:

or

The conditional must produce a boolean result. The statement means either a simple statement terminated by a semicolon or a compound statement, which is a group of simple statements enclosed in braces. Any time the word “statement” is used, it always implies that the statement can be simple or compound.

As an example of if-else, here is a test( ) method that will tell you whether a guess is above, below, or equivalent to a target number:

The return keyword has two purposes: it specifies what value a method will return (if it doesn’t have a void return value) and it causes that value to be returned immediately. The test( ) method above can be rewritten to take advantage of this:

There’s no need for else because the method will not continue after executing a return.

while, do-while and for control looping and are sometimes classified as iteration statements. A statement repeats until the controlling Boolean-expression evaluates to false. The form for a while loop is

The Boolean-expression is evaluated once at the beginning of the loop and again before each further iteration of the statement.

Here’s a simple example that generates random numbers until a particular condition is met:

This uses the static method random( ) in the Math library, which generates a double value between 0 and 1. (It includes 0, but not 1.) The conditional expression for the while says “keep doing this loop until the number is 0.99 or greater.” Each time you run this program you’ll get a different-sized list of numbers.

The form for do-while is

The sole difference between while and do-while is that the statement of the do-while always executes at least once, even if the expression evaluates to false the first time. In a while, if the conditional is false the first time the statement never executes. In practice, do-while is less common than while.

A for loop performs initialization before the first iteration. Then it performs conditional testing and, at the end of each iteration, some form of “stepping.” The form of the for loop is:

Any of the expressions initialization, Boolean-expression or step can be empty. The expression is tested before each iteration, and as soon as it evaluates to false execution will continue at the line following the for statement. At the end of each loop, the step executes.

for loops are usually used for “counting” tasks:

Note that the variable c is defined at the point where it is used, inside the control expression of the for loop, rather than at the beginning of the block denoted by the open curly brace. The scope of c is the expression controlled by the for.

Traditional procedural languages like C require that all variables be defined at the beginning of a block so when the compiler creates a block it can allocate space for those variables. In Java and C++ you can spread your variable declarations throughout the block, defining them at the point that you need them. This allows a more natural coding style and makes code easier to understand.

You can define multiple variables within a for statement, but they must be of the same type:

The int definition in the for statement covers both i and j. The ability to define variables in the control expression is limited to the for loop. You cannot use this approach with any of the other selection or iteration statements.

Earlier in this chapter I stated that the comma operator (not the comma separator, which is used to separate definitions and function arguments) has only one use in Java: in the control expression of a for loop. In both the initialization and step portions of the control expression you can have a number of statements separated by commas, and those statements will be evaluated sequentially. The previous bit of code uses this ability. Here’s another example:

Here’s the output:

You can see that in both the initialization and step portions the statements are evaluated in sequential order. Also, the initialization portion can have any number of definitions of one type.

Inside the body of any of the iteration statements you can also control the flow of the loop by using break and continue. break quits the loop without executing the rest of the statements in the loop. continue stops the execution of the current iteration and goes back to the beginning of the loop to begin the next iteration.

This program shows examples of break and continue within for and while loops:

In the for loop the value of i never gets to 100 because the break statement breaks out of the loop when i is 74. Normally, you’d use a break like this only if you didn’t know when the terminating condition was going to occur. The continue statement causes execution to go back to the top of the iteration loop (thus incrementing i) whenever i is not evenly divisible by 9. When it is, the value is printed.

The second portion shows an “infinite loop” that would, in theory, continue forever. However, inside the loop there is a break statement that will break out of the loop. In addition, you’ll see that the continue moves back to the top of the loop without completing the remainder. (Thus printing happens in the second loop only when the value of i is divisible by 10.) The output is:

The value 0 is printed because 0 % 9 produces 0.

A second form of the infinite loop is for(;;). The compiler treats both while(true) and for(;;) in the same way so whichever one you use is a matter of programming taste.

The goto keyword has been present in programming languages from the beginning. Indeed, goto was the genesis of program control in assembly language: “if condition A, then jump here, otherwise jump there.” If you read the assembly code that is ultimately generated by virtually any compiler, you’ll see that program control contains many jumps. However, a goto is a jump at the source-code level, and that’s what brought it into disrepute. If a program will always jump from one point to another, isn’t there some way to reorganize the code so the flow of control is not so jumpy? goto fell into true disfavor with the publication of the famous “Goto considered harmful” paper by Edsger Dijkstra, and since then goto-bashing has been a popular sport, with advocates of the cast-out keyword scurrying for cover.

As is typical in situations like this, the middle ground is the most fruitful. The problem is not the use of goto, but the overuse of goto—in rare situations goto is actually the best way to structure control flow.

Although goto is a reserved word in Java, it is not used in the language; Java has no goto. However, it does have something that looks a bit like a jump tied in with the break and continue keywords. It’s not a jump but rather a way to break out of an iteration statement. The reason it’s often thrown in with discussions of goto is because it uses the same mechanism: a label.

A label is an identifier followed by a colon, like this:

The only place a label is useful in Java is right before an iteration statement. And that means right before—it does no good to put any other statement between the label and the iteration. And the sole reason to put a label before an iteration is if you’re going to nest another iteration or a switch inside it. That’s because the break and continue keywords will normally interrupt only the current loop, but when used with a label they’ll interrupt the loops up to where the label exists:

In case 1, the break breaks out of the inner iteration and you end up in the outer iteration. In case 2, the continue moves back to the beginning of the inner iteration. But in case 3, the continue label1 breaks out of the inner iteration and the outer iteration, all the way back to label1. Then it does in fact continue the iteration, but starting at the outer iteration. In case 4, the break label1 also breaks all the way out to label1, but it does not re-enter the iteration. It actually does break out of both iterations.

Here is an example using for loops:

This uses the prt( ) method that has been defined in the other examples.

Note that break breaks out of the for loop, and that the increment-expression doesn’t occur until the end of the pass through the for loop. Since break skips the increment expression, the increment is performed directly in the case of i == 3. The continue outer statement in the case of i == 7 also goes to the top of the loop and also skips the increment, so it too is incremented directly.

Here is the output:

If not for the break outer statement, there would be no way to get out of the outer loop from within an inner loop, since break by itself can break out of only the innermost loop. (The same is true for continue.)

Of course, in the cases where breaking out of a loop will also exit the method, you can simply use a return.

Here is a demonstration of labeled break and continue statements with while loops:

The same rules hold true for while:

The output of this method makes it clear:

It’s important to remember that the only reason to use labels in Java is when you have nested loops and you want to break or continue through more than one nested level.

In Dijkstra’s “goto considered harmful” paper, what he specifically objected to was the labels, not the goto. He observed that the number of bugs seems to increase with the number of labels in a program. Labels and gotos make programs difficult to analyze statically, since it introduces cycles in the program execution graph. Note that Java labels don’t suffer from this problem, since they are constrained in their placement and can’t be used to transfer control in an ad hoc manner. It’s also interesting to note that this is a case where a language feature is made more useful by restricting the power of the statement.

The switch is sometimes classified as a selection statement. The switch statement selects from among pieces of code based on the value of an integral expression. Its form is:

Integral-selector is an expression that produces an integral value. The switch compares the result of integral-selector to each integral-value. If it finds a match, the corresponding statement (simple or compound) executes. If no match occurs, the default statement executes.

You will notice in the above definition that each case ends with a break, which causes execution to jump to the end of the switch body. This is the conventional way to build a switch statement, but the break is optional. If it is missing, the code for the following case statements execute until a break is encountered. Although you don’t usually want this kind of behavior, it can be useful to an experienced programmer. Note the last statement, following the default, doesn’t have a break because the execution just falls through to where the break would have taken it anyway. You could put a break at the end of the default statement with no harm if you considered it important for style’s sake.

The switch statement is a clean way to implement multi-way selection (i.e., selecting from among a number of different execution paths), but it requires a selector that evaluates to an integral value such as int or char. If you want to use, for example, a string or a floating-point number as a selector, it won’t work in a switch statement. For non-integral types, you must use a series of if statements.

Here’s an example that creates letters randomly and determines whether they’re vowels or consonants:

Since Math.random( ) generates a value between 0 and 1, you need only multiply it by the upper bound of the range of numbers you want to produce (26 for the letters in the alphabet) and add an offset to establish the lower bound.

Although it appears you’re switching on a character here, the switch statement is actually using the integral value of the character. The singly-quoted characters in the case statements also produce integral values that are used for comparison.

Notice how the cases can be “stacked” on top of each other to provide multiple matches for a particular piece of code. You should also be aware that it’s essential to put the break statement at the end of a particular case, otherwise control will simply drop through and continue processing on the next case.

deserves a closer look. Math.random( ) produces a double, so the value 26 is converted to a double to perform the multiplication, which also produces a double. This means that ‘a’ must be converted to a double to perform the addition. The double result is turned back into a char with a cast.

What does the cast to char do? That is, if you have the value 29.7 and you cast it to a char, is the resulting value 30 or 29? The answer to this can be seen in this example:

The output is:

So the answer is that casting from a float or double to an integral value always truncates.

A second question concerns Math.random( ). Does it produce a value from zero to one, inclusive or exclusive of the value ‘1’? In math lingo, is it (0,1), or [0,1], or (0,1] or [0,1)? (The square bracket means “includes” whereas the parenthesis means “doesn’t include.”) Again, a test program might provide the answer:

To run the program, you type a command line of either:

or

In both cases you are forced to break out of the program manually, so it would appear that Math.random( ) never produces either 0.0 or 1.0. But this is where such an experiment can be deceiving. If you consider[26] that there are about 2⁶² different double fractions between 0 and 1, the likelihood of reaching any one value experimentally might exceed the lifetime of one computer, or even one experimenter. It turns out that 0.0 is included in the output of Math.random( ). Or, in math lingo, it is [0,1).

This chapter concludes the study of fundamental features that appear in most programming languages: calculation, operator precedence, type casting, and selection and iteration. Now you’re ready to begin taking steps that move you closer to the world of object-oriented programming. The next chapter will cover the important issues of initialization and cleanup of objects, followed in the subsequent chapter by the essential concept of implementation hiding.

Solutions to selected exercises can be found in the electronic document The Thinking in Java Annotated Solution Guide, available for a small fee from www.BruceEckel.com.

[25] John Kirkham writes, “I started computing in 1962 using FORTRAN II on an IBM 1620. At that time, and throughout the 1960s and into the 1970s, FORTRAN was an all uppercase language. This probably started because many of the early input devices were old teletype units that used 5 bit Baudot code, which had no lowercase capability. The ‘E’ in the exponential notation was also always upper case and was never confused with the natural logarithm base ‘e’, which is always lowercase. The ‘E’ simply stood for exponential, which was for the base of the number system used—usually 10. At the time octal was also widely used by programmers. Although I never saw it used, if I had seen an octal number in exponential notation I would have considered it to be base 8. The first time I remember seeing an exponential using a lowercase ‘e’ was in the late 1970s and I also found it confusing. The problem arose as lowercase crept into FORTRAN, not at its beginning. We actually had functions to use if you really wanted to use the natural logarithm base, but they were all uppercase.”

[26] Chuck Allison writes: The total number of numbers in a floating-point number system is
2(M-m+1)b^(p-1) + 1
where b is the base (usually 2), p is the precision (digits in the mantissa), M is the largest exponent, and m is the smallest exponent. IEEE 754 uses:
M = 1023, m = -1022, p = 53, b = 2
so the total number of numbers is
2(1023+1022+1)2^52
= 2((2^10-1) + (2^10-1))2^52
= (2^10-1)2^54
= 2^64 - 2^54
Half of these numbers (corresponding to exponents in the range [-1022, 0]) are less than 1 in magnitude (both positive and negative), so 1/4 of that expression, or 2^62 - 2^52 + 1 (approximately 2^62) is in the range [0,1). See my paper at http://www.freshsources.com/1995006a.htm (last of text).