Brian Goetz
Stuart Marks
2018-10-11

Why have var in Java?

Local variables are the workhorse of Java. They allow methods to compute significant results by cheaply storing intermediate values. Unlike fields, local variables are used only in the block in which they are declared, so the compiler can ensure that they're always initialized before they're used. Often, the name and initial value of a local variable are more important in understanding code than the type of the local variable. The type is important, but the way the name is used on the next line might be more important. Use of var can help make the important parts of the local variable declaration stand out.

When var is used instead of a type, the Java compiler infers the type of the variable from its initial value. This is especially worthwhile if the type has a long name, or is a complex parameterized type, or if the type is redundant with the initial value. Using var can make code more concise without sacrificing readability, and in some cases it can improve readability by removing redundancy.

Does this make Java dynamically typed? Is this like var in JavaScript?

No and no. Java is still a statically typed language, and the addition of var doesn't change this. var can be used in a local variable declaration instead of the variable's type. With var, the Java compiler infers the type of the variable at compile time, using type information obtained from the variable's initializer. The inferred type is then used as the static type of the variable. Typically, this is the same as the type you would have written explicitly, so a variable declared with var behaves exactly as if you had written the type explicitly.

Java compilers have performed type inference for many years. For example, in Java 8, the parameters of a lambda expression do not need explicit types because the compiler infers their types from how the lambda expression is used:

List<Person> list = ...
list.stream().filter(p -> p.getAge() > 18) ...

In the code snippet above, the lambda parameter p is inferred to have the static type Person. If the Person class is changed so that it no longer has a getAge method, or if the list is changed to be a list of type other than Person, type inference will fail with a compile-time error.

Is a var variable final?

No. Local variables declared with var are non-final by default. However, the final modifier can be added to var declarations:

final var person = new Person();

There is no shorthand for final var in Java. Languages such as Scala use val to declare immutable (final) variables. This works well in Scala because all variables - locals and fields alike - are declared using a syntax of the form

val name : type

or

var name : type

You can include or omit the ": type" part of the declaration depending on whether or not you want type inference. In Scala, the choice between mutability and immutability is orthogonal to type inference.

In Java, var can be used only where type inference is desired; it cannot be used where a type is declared explicitly. If val were added, it too could be used only where type inference is used. The use of var or val in Java could not be used to control immutability if the type were declared explicitly.

In addition, Java allows the use of var only for local variables, not for fields. Immutability is much more significant for fields, whereas immutable local variables are comparatively rarely used.

Using var/val keywords to control immutability is a feature that seems like it ought to carry over cleanly from Scala to Java. In Java, however, it would be much less useful than it is in Scala.

Won't bad developers misuse this feature to write terrible code?

Yes, bad developers will write terrible code no matter what we do. Withholding a feature won't prevent them from doing so. But, when used properly, using type inference allows developers to also write better code.

One way that var may encourage developers to write better code is that it lowers the overhead of declaring a new variable. If the overhead of declaring a variable is high, developers will often avoid doing so, and create complex nested or chained expressions that impair readability solely in order to avoid declaring more variables. With var, the overhead of pulling a subexpression into a named variable is lower, so developers are more likely to do so, resulting in more cleanly factored code.

When a feature is introduced, it is common that at first, programmers will use, overuse, and maybe even abuse that feature, and it takes some time for the community to converge on a reasonable set of guidelines for what uses are reasonable and what uses are not. It's probably reasonable to use var fairly frequently though not for the majority of local variable declarations.

Starting with Local Variable Type Inference (LVTI), we're publishing material about its intent and recommended usage (such as this FAQ, and the LVTI Style Guidelines) around the same time the feature is delivered. We hope that this will accelerate the community's convergence on what constitutes reasonable usage, and that it will help avoid most cases of abuse.

Where can var be used?

var can be used for declaring local variables, including index variables of for-loops and resource variables of the try-with-resources statement.

var cannot be used for fields, method parameters, and method return types. The reason is that types in these locations appear explicitly in class files and in javadoc specifications. With type inference, it's quite easy for a change to an initializer to cause the variable's inferred type to change. For local variables, this is not a problem, because local variables are limited in scope, and their types are not recorded directly into class files. However, type inference could easily cause a problem if types for fields, method parameters, and method return types were inferred.

For example, suppose that the return type of a method were inferred from the expression in the method's return statement. A change to the method's implementation might end up changing the type of the expression in the return statement. This in turn might change the method's return type. This could result in a source or binary incompatibility. Such incompatible changes should not arise from harmless-looking changes to the implementation.

Suppose a field's type were inferred. A change to the field's initializer could change the field's type, which might unexpectedly break reflective code.

Type inference is ok within the implementation, but not in APIs. API contracts should be declared explicitly.

What about private fields and methods, which are not part of APIs? In theory, we could have chosen to support var for private fields and for the return type of private methods, without worry that this would cause incompatibilities due to separate compilation and dynamic linkage. We chose to limit the scope of type inference in this way for simplicity. Trying to push the boundary to include some fields and some method returns makes the feature considerably more complex and harder to reason about, but only marginally more useful.

Why is an initializer required on the right-hand side of var?

The type of the variable is inferred from the type of the initializer. This means, of course, that var can only be used when there is an initializer. We could have chosen to infer the type from the assignments to the variable, but that would have made the feature considerably more complex, and it could potentially lead to misleading or hard-to-diagnose errors. In order to keep things simple, we've defined var so that only local information is used for type inference.

Suppose that we allowed type inference based on assignment in multiple locations, separate from the variable declaration. Consider this example:

var x;
...
x = "foo";
...
x = 17;

If a type were chosen based on (say) the first assignment, it might cause an error at another statement that's quite distant from the cause of the error. (This is sometimes referred to as the "action-at-a-distance" problem.)

Alternatively, a type could be chosen that's compatible with all assignments. In this case one might expect that the inferred type would be Object, because that's the common superclass of String and Integer. Unfortunately, the situation is more complicated than that. Since both String and Integer are Serializable and Comparable, the common supertype would be an odd intersection type that's something like

Serializable & Comparable<? extends Serializable & Comparable<...>>

(Note that it isn't possible to declare a variable of this type explicitly.) Also note that this results in a boxing conversion when 17 is assigned to x, which might be unexpected and undesirable.

To avoid these problems, it seems preferable to require that the type be inferred using an explicit initializer.

Why can't you use var with null?

Consider this declaration (which is illegal):

var x = null; // ERROR

The null literal denotes a value of a special null type (JLS 4.1) that is the subtype of all reference types in Java. The only value of the null type is null itself, therefore, the only value that could ever be assigned to a variable of the null type is null. This isn't very useful.

A special rule could be made so that a var declaration initialized to null is inferred to have type Object. This could be done, but it raises the question of what the programmer intended. Presumably the variable is initialized to null so that it can be assigned to some other value later. In that case it seems unlikely that inferring the variable's type as Object is the correct choice.

Instead of creating some special rules to handle this case, we've disallowed it. If you want a variable of type Object, declare it explicitly.

Can you use var with a diamond on the right-hand side?

Yes, it works, but it's probably not what you want. Consider:

var list = new ArrayList<>();

This will infer the type of list to be ArrayList<Object>. In general, it's preferable use an explicit type on the left with diamond on the right, or use var on the left with an explicit type on the right. See the LVTI Style Guidelines, guideline G6, for further information.