Type declarations

This page is about type declarations, which define new type constructors. There's a separate page about the types that are produced by applying type arguments to such type constructors.


Example type declarations:

class MyClass(name) {
    shared String name;
class MyGenericClass<Arg>(Arg arg) {
interface MyInterface {
    shared formal String name;
interface MyInterface<Arg> {
    shared formal Arg arg;
object myObject {
    shared String name = "Trompon";


Types versus Type declarations

It is important the appreciate the difference between a type declaration (such as the declaration of a class) which introduces a type constructor, and an applied type (also called a produced type).

There's a separate reference page about types

Type declarations

In Ceylon, a type declaration is one of:

All these declarations can have members.

Function and value declarations to not introduce new types or type constructors and do not have members.

Declarative subtyping

The extends and satisfies clauses of a type declaration list the types that are treated as supertypes of the types produced from the type declaration.

Put another way, these clauses have the effect that every type produced from the type constructor of the class or interface being defined will be a subtype of the type in extends or satisfies clause:

class Sub() extends Generic<String>() {
class GenericSub<Parameter>() extends Generic<Parameter>() {

Thus the type Sub is a subtype of Generic<String>, and the type GenericSub<Boolean> is a subtype of Generic<Boolean>.

Declarative cases

The of clause of a type declaration enumerates the disjoint subtypes (the cases) of the types produced from the type declaration.

Different kinds of declaration

Member declarations

A declaration that occurs directly in a type declaration is called a member declaration. Member values are called attributes. Member functions are called methods. Classes and interfaces that are members do not have an alternative name they're just called member classes and interfaces.

Local declarations

A local (or nested) declaration is a declaration that is contained within another declaration or a statement.

Top-level declarations

A top level declaration is contained directly in a compilation unit and not contained within any other declaration. In other words a top level declaration is neither a member nor a local declaration.

Enumerated types

Classes can enumerate a list of their permitted subclasses.

Interfaces can enumerate a list of their permitted subtypes.

The subtypes are called the cases of the class or interface type.

Type aliases

To avoid having to repeat long type expressions, you can declare a type alias for a type using the alias keyword:

alias BasicType = String|Character|Integer|Float|Boolean;

Selected important type declarations


Anything the ultimate supertype of all types. That means that every type has Anything as a supertype, which in turn means that every reference is assignable to Anything.

You can also think of Anything as the union of all types.

Anything corresponds to the notion of universe set in mathematics.

You can't do anything with an instance of Anything, except narrow it to some more specific type.

Anything's cases are Null and Object.

Anything is the default upper bound for type parameters lacking an upper bound constraint.


Nothing is a subtype of all types. That means that every type has Nothing as a subtype (even if the type is produced from a final class declaration).

You can also thing of Nothing as the intersection of all types.

Nothing corresponds to the notion of the empty set in mathematics.

Because Nothing is the intersection of all types it is assignable to all types. Similarly because it is the intersection of all types it can have no instances.

There is a value called nothing in the language module, which has the type Nothing. At runtime trying to evaluate nothing (that is, get an instance of Nothing) will throw an exception.

Any function or value which claims to return Nothing cannot return normally, it must either:

  • throw an exception or
  • not return (for example, by looping forever, or stopping the virtual machine)


Null is the type of null.

Conceptually null is the absence of a value.

If an expression permits null then it needs Null as a supertype. This is usually expressed as using a union type such as T|Null, which can be abbreviated as T?, and we may refer to it as an optional type.

Because null represents the absence of a value (something that is not a thing), it is meaningless to ask some reference is equal to null. Thus Ceylon does not permit obj == null or null == null. In practice there are some situations where you want the answer to be true and other situations where you want the answer to be false.

Null is one of the cases of Anything, the other being Object.


Object is the class that declares equals(), hash and string.

Object is one of the cases of Anything, the other being Null.


Basic mixes Identifiable into Object.

Basic is the default superclass of classes whose declaration lacks an extends clause.


Iterable is a type that produces instances of another type when iterated.

There are two flavours of Iterable:

  • the type Iterable<T>, usually abbreviated to {T*}, may contain zero or more elements (it is possibly empty),
  • the type Iterable<T,Nothing>, usually abbreviated to {T+}, contains at least one element (it is non-empty)


Sequential is an enumerated type with subtypes Sequence and Empty. Sequential<T> is usually abbreviated to T[] or [T*].


Empty is the type of a Sequential which contains no elements.

The expression [] (or alternatively the value empty) in the language module has the type Empty.


Sequence is the type of non-empty sequences. Sequence<T> is usually abbreviated to [T+].


Tuple is a subclass of Sequence (and thus cannot be empty). It differs from Sequence in that it encodes the types of each of its elements individually.

[Integer, Boolean, String] t = [1, true, ""];
Integer first = t[0];
Boolean second = t[1];
String last = t[2];

Tuples also have a notion of variadicity:

// A tuple of at least two elements
// the first is an Integer and 
// the rest are Boolean
[Integer, Boolean+] t = [1, true, false];
// A tuple of at least element
// the first is an Integer and 
// the rest are Boolean
[Integer, Boolean*] t2 = t;

Tuple types may thus be used to represent the type of an argument or parameter list, and are therefore used to encode function types.

Unabbreviated tuple types are extremely verbose, and therefore the abbreviated form is always used.


There are two ways of modelling declarations at runtime mirroring the difference between a declaration and its type.

The package ceylon.language.meta.declaration contains interfaces which model declarations. As such, those declarations have type parameter lists. Supplying a type argument list for a declaration you can obtain an instance of a model from ceylon.language.meta.model.

See also