Generator method declarations are possible within classes and interfaces. A generator method declaration is indicated by an asterisk after the result type as in the following example:

char* chars(CharSequence s) {
  for (int i = 0; i < s.length(); i++) {
    yield s.charAt(i);
  }
}

Generator methods successively yield values to their invoker via the yield-statement (Section 15.5, “The yield Statement”, Section 16.2.2, “Generator Method Invocations”), in contrast to normal (non-generator) methods which return a single value via the return-statement. The asterisk can be seen as meaning "zero or more values" as in regular expressions. The body of a generator method must not contain any return-statement that returns a value.

The declaration of a generator method is not as difficult as it may seem: Basically, it is syntactic sugar. A declaration of a generator method is equivalent to the declaration of a normal method, where its header is the header of the generator method, plus an additional parameter c inserted at the beginning of the formal parameter list. The type of c is a consumer class in the package de.grogra.xl.lang: If T is the result type of the generator method, let S be the type affix (Section 6.1, “Type Affixes and Type Letters”) of T, then the consumer class is SConsumer. For the example, the equivalent method header is char chars(CharConsumer c, CharSequence s), and a complete and equivalent method declaration looks as follows:

char chars(CharConsumer c, CharSequence s) {
  for (int i = 0; i < s.length(); i++) {
    c.consume(s.charAt(i));
  }
  return 0;
}

In fact, when this specification refers to the signature of a method, the signature of this equivalent method declaration has to be used in the case of a generator method declaration.

Operator method declarations are possible within classes and interfaces. An operator method declaration is indicated by the identifier operator, followed by the token (or tokens in case of []) of one of the overloadable operators as in the following example:

Complex operator+ (Complex b) {
  return new Complex(this.real + b.real, this.imag + b.imag);
}

Note that regarding the lexical structure, operator is an identifier, not a keyword. However, it is the only identifier for a method declaration which is allowed to be followed by the token of an overloadable operator. Thus, it is a compile-time error if an operator token follows an identifier which is not equal to the character sequence operator.

The declaration of an operator method is equivalent to the declaration of a normal method, its name being derived from the operator by the following rules. At first, the arity of the method declaration is determined as follows: If the declared method is static, the arity is the number of its parameters. Otherwise, the arity is the number of its parameters plus one. An operator method declaration whose arity is one is called a unary method declaration, a declaration whose arity is two is called a binary method declaration.

For unary method declarations, the equivalent method name is derived as defined by the following table. If no identifier for the operator token is defined, a compile-time error occurs.

For binary method declarations, the equivalent method name is derived as defined by the following table. If no identifier for the operator token is defined, a compile-time error occurs.

The last parameter of binary method declarations for the operators ++ and -- must have type int, or a compile-time error occurs.

If the arity is greater than two and the operator tokens are [], the equivalent method name is operator$index. Any case which is not handled by the previous definitions leads to a compile-time error.

Instancing method declarations are possible within classes. Such a declaration provides syntactic sugar for the implementation of object instancing, which is a common technique in 3D computer graphics (but this specification makes no reference to this concrete usage). An instancing method declaration is indicated by the arrow ==>, the result type has to be void.

At class level, a declaration of an instancing method is equivalent to a declaration of a normal (non-instancing) method, where its header is the header of the instancing method, plus an additional parameter s inserted at the beginning of the formal parameter list. The type of s is the type returned by the method getInstantiationStateClass of the current compile-time model (Section 3.1, “Compile-Time Model”). The compile-time model has to return a subclass of de.grogra.xl.lang.InstantiationState.

In fact, when this specification refers to the signature of methods, the signature of this equivalent method declaration has to be used in the case of an instancing method declaration.