Why must local variables, including primitives, always be initialized in Java?
In Java, class and instance variables assume a default value (null, 0, false) if they are not initialized manually. However, local variables don't have a default value. Unless a local variable has been assigned a value, the compiler will refuse to compile the code that reads it. IMHO, this leads to the conclusion, that initializing a local variable with some default value (like null, which might lead to a NullPointerException later) when it is declared is actually a bad thing. Consider the following example:
Object o;
if (<some boolean condition>)
o = <some value>;
else
o = <some other value>;
System.out.println(o);
An initialization of o
with null is completely unnecessary, since the Java compiler checks at compile time, that any code-path initializes o
(with either null or some non-null value) before the variable is read. That means, that the compiler will refuse to compile the line System.out.println(o);
if you would comment out any of the two initializations of the variable o
in the code snippet above.
This holds for Java, and maybe for Java only. I don't know about language like C#. In good old C (and maybe C++) however, it is still recommended to always initialize variables when declaring them, AFAIK. Such "old-school" programming languages might be the reason, that the recommendation to always initialize variables pops up in books and discussions about modern languages like Java, where the compiler keeps track of whether a variable has been initialized or not.
Not totally true. Local variables only need to be initialized if being reference. A local variable can be left uninitialized if never referenced. For example:
int x; // Valid
int y;
println("y=" + y); // Not valid since y's value has never been assigned
Basically, requiring a variable to be assigned a value before you read it is a Good Thing. It means you won't accidentally read something you didn't intend to. Yes, variables could have default values - but isn't it better for the compiler to be able to catch your bug instead, if it can prove that you're trying to read something which might not have been assigned yet? If you want to give a local variable a default value, you can always assign that explicitly.
Now that's fine for local variables - but for instance and static variables, the compiler has no way of knowing the order in which methods will be called. Will a property "setter" be called before the "getter"? It has no way of knowing, so it has no way of alerting you to the danger. That's why default values are used for instance/static variables - at least then you'll get a known value (0, false, null etc) instead of just "whatever happened to be in memory at the time." (It also removes the potential security issue of reading sensitive data which hadn't been explicitly wiped.)
There was a question about this very recently for C#... - read the answers there as well, as it's basically the same thing. You might also find Eric Lippert's recent blog post interesting; it's at least around the same area, even though it has a somewhat different thrust.