Casting operators
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:
void casts() {
int i = 200;
long l = (long)i;
long l2 = (long)200;
}
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 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.)
Literals
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:
//: c03:Literals.java
public class Literals {
char c = 0xffff; // max char hex value
byte b = 0x7f; // max byte hex value
short s = 0x7fff; // max short hex value
int i1 = 0x2f; // Hexadecimal (lowercase)
int i2 = 0X2F; // Hexadecimal (uppercase)
int i3 = 0177; // Octal (leading zero)
// Hex and Oct also work with long.
long n1 = 200L; // long suffix
long n2 = 200l; // long suffix (but can be confusing)
long n3 = 200;
//! long l6(200); // not allowed
float f1 = 1;
float f2 = 1F; // float suffix
float f3 = 1f; // float suffix
float f4 = 1e-45f; // 10 to the power
float f5 = 1e+9f; // float suffix
double d1 = 1d; // double suffix
double d2 = 1D; // double suffix
double d3 = 47e47d; // 10 to the power
} ///:~
Hexadecimal (base 16), which works with all the integral data types, is denoted by a leading 0x or 0X followed by 0-9 or a-f either in uppercase 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 preceding 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. Uppercase or lowercase L means long, upper or lowercase F means float and uppercase 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.[17] 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
long n3 = 200;
there’s no ambiguity, so an L after the 200 would be superfluous. However, with
float f4 = 1e-47f; // 10 to the power
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.
Promotion
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.
