Java Basics: Types

A computer only knows about ones and zeros (bits). Still, it's able to tell you 1+1 equals 2. How is that possible?

The answer is that it can put many bits together and interpret them as something else. 1100001 for example can be interpreted as a number in base 2: 97. The curious part is that the same bits can just as well be interpreted as an ASCII character, in which case they represent the letter a!

So, if I have a variable containing 1100001, what does it represent? The letter a or 97?!

Here's where types come in. Each variable has a type associated with it, which specifies how its data should be interpreted.

int  i = 0b1100001; // interpret as an integer
char c = 0b1100001; // interpret as a character

System.out.println(i); // prints 97
System.out.println(c); // prints 'a'

Type Checking

Since the computer knows how to interpret the data, it can check whether or not the program treats the data sensibly. For example, multiplying an integer with a boolean value (true/false) doesn't make sense. If you try to do so you get a type error.

int i = 5;
boolean b = true;
System.out.println(i * b);

// Error: Operator '*' cannot be
// applied to 'int', 'boolean'

Parts of the Java type system is quite complicated. Some type errors are in fact so complicated that even experienced programmers have trouble deciphering the meaning.

Dynamic vs Static Typing

Some programming languages (Python, JavaScript, PHP, …) lets you run code containing type errors, and doesn't complain unless the faulty instruction is actually executed. This is called dynamic typing. Others, like Java (and C, C++, Scala, …), refuses to even start executing code that contain a type error. This is called static typing.

print("Begin")
if False:
    print("2" / 2)
print("End")

System.out.println("Begin");
if (false)
    System.out.println("2" / 2);
System.out.println("End");

The static approach trades some flexibility for robustness: It's more restrictive in terms of what's acceptable to write, but a program that passes compilation is guaranteed to be free of a whole class of common errors.

Weak vs Strong Typing

In some programming languages (like JavaScript or Perl) the type of a value is determined by how it's used. If you for instance do if (x) …, then x is treated as a boolean, but if you do y = x * 5, then x is treated as a number. This is called weak typing.

In other languages, like Java, it's the other way around: The type determines what you can do with a value. If, for example, the type of x is a number, then if (x) … will result in a compilation error. This is called strong typing.

Primitive vs Reference Types

Java provides 8 built-in primitive types and lets you define your own custom reference types.

Primitive Types

The primitive types are:

• Integers
  • byte (−128–127)
  • short (±32 thousand)
  • int (±2 billion)
  • long (±9 quintillion)
• Reals
  • float (~6 digits precision)
  • double (~15 digits precision)
• char (unicode characters: a 5, ξ, ♪, …)
• boolean (true/false)

Java is able to convert between some of these types automatically, but apart from that, they have nothing in common.

Further details: Java: Primitive Types

Reference Types

Sometimes simple numbers and characters aren't enough. You may want your data to represent products, people, buttons, etc. For these situations we have classes. Classes can be thought of as blueprints for objects. Classes and objects are the foundation of what is called Object Oriented Programming and is a whole topic of it's own. In this article we'll just cover the basics and refer to other articles for details.

class Person {
    String name;
    int age;
}

Person me = new Person();
me.name = "Andreas";
me.age = 34;

What's important in regards to Javas type system is the fact that all classes are arranged in a class hierarchy. Classes like Banana or Apple could for example be subclasses of Fruit. The Fruit class could in turn be a subclass of something like Edible.

At the top of this hierarchy we have a class called Object which has special status in the language. It is for example what you implicitly subclass if you don't explicitly specify anything else. Unlike C++, the hierarchy forms a proper tree; Each class has precisely one "parent" class.

If you have a variable of type Fruit, you could assign a Banana, or an Apple reference to it.

Fruit toEat = new Banana();

// Changed my mind
toEat = new Apple();

It would however be an error to try to assign a Car reference to the toEat variable, since a Car is not a Fruit.

Fruit toEat = new Car();
// error: incompatible types: Car cannot be converted to Fruit

Finally, a slighly more intricate example:

Edible e = new Banana();
Fruit toEat = e;
// error: incompatible types: Edible cannot be converted to Fruit

Looking at the snippet as a whole, it's clear that the second assignment is safe. The type checker however has a quite narrow field of view, and looks at each assignment in isolation. In the last assignment it refuses to let us assign an arbitrary Edible to a Fruit. It basically want's to play it safe and make sure we don't assign something like a Cake reference to a Fruit variable.

What's different here compared to the example where we tried to assign a Car to a Fruit is that an Edible could conceivably contain a valid Fruit reference. While the type checker won't allow it by default, you can force it through with a cast as explained in the next section.

Further Reading: Java: Primitives vs Objects and References.

Conversions

Java allows you to convert between types that are similar. There are different ways such conversions can be expressed.

Explicit Casts

You can tell the compiler to reinterpret the data by doing an explicit cast. The syntax is (TargetType) expression.

int i = 97;
char c = (char) i; // cast from int to char
// c == 'a'

Perhaps more common is to use casts with reference types. As mentioned earlier, it might for example make sense to force an Edible into a Fruit.

Edible e = …
if (e instanceof Fruit) {
    Fruit fruit = (Fruit) e;
    System.out.println(fruit.getNumSeeds());
}

If the cast is illegal, i.e. if you try to cast a Cake into a Fruit you would get a ClassCastException.

Edible e = new Cake();
Fruit f = (Fruit) e;

Exception in thread "main" java.lang.ClassCastException: Cake cannot be cast to Fruit

If the e variable was of type Integer the compiler would even reject the program, since it's guaranteed that it would give an error when executed.

Integer e = 5;
Fruit f = (Fruit) e;
// incompatible types: Integer cannot be converted to Fruit

Implicit Widening

A long can store all int values and then some. It has a "wider" range. Therefore it's always safe to assign an int value to a long variable, thus the compiler will do some conversion implicitly for you.

int i = 1147;
long l = i; // no explicit cast needed

Similarly, a byte can be widened to a short or an int, a float to a double and so on.

Narrowing

In cases where it's clearly safe, the compiler will implicitly perform the opposite conversion as well. A numeric literal such as 7 is for example strictly speaking an int. Since it fits in a byte the compiler converts it for you.

byte b = 7; // no explicit cast needed

Boxing

Each primitive type (byte, int, double, …) has a corresponding wrapper class called the boxed type (Byte, Integer, Double, …). Java can automatically convert between the two forms. This is called autoboxing and unboxing

int i = 7;

// Automatically box an int into an Integer
Integer j = i;

// Unbox j, perform addition, box result
Integer k = j + 5;

Continue reading here:

Java: Autoboxing / unboxing

Promotion

Promotion is a collective name for widening and unboxing conversions. For example, + is defined for int, float and double operands. If you add two byte variables together, they are first promoted to int values.

Byte b1 = 7;
byte b2 = 8;
int i = b1 + b2; // i == 15

short s1 = 5;
short s2 = 7;
short s3 = s1 + s2; // error: incompatible types: possible
                    // lossy conversion from int to short

Summary

• Types give meaning to data
• All variables (and expressions) have a type
• The entire program must be free of type errors for it to be compiled and executed
• There are 8 different primitive types representing numbers, characters and booleans
• Classes are arranged in a hierrachy
• Sometimes you're allowed to convert the type of the data from one type to another