January, 2016
Microsoft is making this Specification available under the Open Web Foundation Final Specification Agreement Version 1.0 (“OWF 1.0”) as of October 1, 2012. The OWF 1.0 is available at here.
TypeScript is a trademark of Microsoft Corporation.
JavaScript applications such as web e-mail, maps, document editing, and collaboration tools are becoming an increasingly important part of the everyday computing. We designed TypeScript to meet the needs of the JavaScript programming teams that build and maintain large JavaScript programs. TypeScript helps programming teams to define interfaces between software components and to gain insight into the behavior of existing JavaScript libraries. TypeScript also enables teams to reduce naming conflicts by organizing their code into dynamically-loadable modules. TypeScript’s optional type system enables JavaScript programmers to use highly-productive development tools and practices: static checking, symbol-based navigation, statement completion, and code re-factoring.
TypeScript is a syntactic sugar for JavaScript. TypeScript syntax is a superset of ECMAScript 2015 (ES2015) syntax. Every JavaScript program is also a TypeScript program. The TypeScript compiler performs only file-local transformations on TypeScript programs and does not re-order variables declared in TypeScript. This leads to JavaScript output that closely matches the TypeScript input. TypeScript does not transform variable names, making tractable the direct debugging of emitted JavaScript. TypeScript optionally provides source maps, enabling source-level debugging. TypeScript tools typically emit JavaScript upon file save, preserving the test, edit, refresh cycle commonly used in JavaScript development.
TypeScript syntax includes all features of ECMAScript 2015, including classes and modules, and provides the ability to translate these features into ECMAScript 3 or 5 compliant code.
Classes enable programmers to express common object-oriented patterns in a standard way, making features like inheritance more readable and interoperable. Modules enable programmers to organize their code into components while avoiding naming conflicts. The TypeScript compiler provides module code generation options that support either static or dynamic loading of module contents.
TypeScript also provides to JavaScript programmers a system of optional type annotations. These type annotations are like the JSDoc comments found in the Closure system, but in TypeScript they are integrated directly into the language syntax. This integration makes the code more readable and reduces the maintenance cost of synchronizing type annotations with their corresponding variables.
The TypeScript type system enables programmers to express limits on the capabilities of JavaScript objects, and to use tools that enforce these limits. To minimize the number of annotations needed for tools to become useful, the TypeScript type system makes extensive use of type inference. For example, from the following statement, TypeScript will infer that the variable ‘i’ has the type number.
var i = 0;
TypeScript will infer from the following function definition that the function f has return type string.
function f() {
return "hello";
}
To benefit from this inference, a programmer can use the TypeScript language service. For example, a code editor can incorporate the TypeScript language service and use the service to find the members of a string object as in the following screen shot.
In this example, the programmer benefits from type inference without providing type annotations. Some beneficial tools, however, do require the programmer to provide type annotations. In TypeScript, we can express a parameter requirement as in the following code fragment.
function f(s: string) {
return s;
}
f({}); // Error
f("hello"); // Ok
This optional type annotation on the parameter ’s’ lets the TypeScript type checker know that the programmer expects parameter ’s’ to be of type ‘string’. Within the body of function ‘f’, tools can assume ’s’ is of type ‘string’ and provide operator type checking and member completion consistent with this assumption. Tools can also signal an error on the first call to ‘f’, because ‘f’ expects a string, not an object, as its parameter. For the function ‘f’, the TypeScript compiler will emit the following JavaScript code:
function f(s) {
return s;
}
In the JavaScript output, all type annotations have been erased. In general, TypeScript erases all type information before emiting JavaScript.
An ambient declaration introduces a variable into a TypeScript scope, but has zero impact on the emitted JavaScript program. Programmers can use ambient declarations to tell the TypeScript compiler that some other component will supply a variable. For example, by default the TypeScript compiler will print an error for uses of undefined variables. To add some of the common variables defined by browsers, a TypeScript programmer can use ambient declarations. The following example declares the ‘document’ object supplied by browsers. Because the declaration does not specify a type, the type ‘any’ is inferred. The type ‘any’ means that a tool can assume nothing about the shape or behavior of the document object. Some of the examples below will illustrate how programmers can use types to further characterize the expected behavior of an object.
declare var document;
document.title = "Hello"; // Ok because document has been declared
In the case of ‘document’, the TypeScript compiler automatically supplies a declaration, because TypeScript by default includes a file ’lib.d.ts’ that provides interface declarations for the built-in JavaScript library as well as the Document Object Model.
The TypeScript compiler does not include by default an interface for jQuery, so to use jQuery, a programmer could supply a declaration such as:
declare var $;
Section 1.3 provides a more extensive example of how a programmer can add type information for jQuery and other libraries.
Function expressions are a powerful feature of JavaScript. They enable function definitions to create closures: functions that capture information from the lexical scope surrounding the function’s definition. Closures are currently JavaScript’s only way of enforcing data encapsulation. By capturing and using environment variables, a closure can retain information that cannot be accessed from outside the closure. JavaScript programmers often use closures to express event handlers and other asynchronous callbacks, in which another software component, such as the DOM, will call back into JavaScript through a handler function.
TypeScript function types make it possible for programmers to express the expected signature of a function. A function signature is a sequence of parameter types plus a return type. The following example uses function types to express the callback signature requirements of an asynchronous voting mechanism.
function vote(candidate: string, callback: (result: string) => any) {
// ...
}
vote("BigPig",
function(result: string) {
if (result === "BigPig") {
// ...
}
}
);
In this example, the second parameter to ‘vote’ has the function type
(result: string) => any
which means the second parameter is a function returning type ‘any’ that has a single parameter of type ‘string’ named ‘result’.
Section 3.9.2 provides additional information about function types.
TypeScript programmers use object types to declare their expectations of object behavior. The following code uses an object type literal to specify the return type of the ‘MakePoint’ function.
var MakePoint: () => {
x: number; y: number;
};
Programmers can give names to object types; we call named object types interfaces. For example, in the following code, an interface declares one required field (name) and one optional field (favoriteColor).
interface Friend {
name: string;
favoriteColor?: string;
}
function add(friend: Friend) {
var name = friend.name;
}
add({ name: "Fred" }); // Ok
add({ favoriteColor: "blue" }); // Error, name required
add({ name: "Jill", favoriteColor: "green" }); // Ok
TypeScript object types model the diversity of behaviors that a JavaScript object can exhibit. For example, the jQuery library defines an object, ‘$’, that has methods, such as ‘get’ (which sends an Ajax message), and fields, such as ‘browser’ (which gives browser vendor information). However, jQuery clients can also call ‘$’ as a function. The behavior of this function depends on the type of parameters passed to the function.
The following code fragment captures a small subset of jQuery behavior, just enough to use jQuery in a simple way.
interface JQuery {
text(content: string);
}
interface JQueryStatic {
get(url: string, callback: (data: string) => any);
(query: string): JQuery;
}
declare var $: JQueryStatic;
$.get("http://mysite.org/divContent",
function (data: string) {
$("div").text(data);
}
);
The ‘JQueryStatic’ interface references another interface: ‘JQuery’. This interface represents a collection of one or more DOM elements. The jQuery library can perform many operations on such a collection, but in this example the jQuery client only needs to know that it can set the text content of each jQuery element in a collection by passing a string to the ’text’ method. The ‘JQueryStatic’ interface also contains a method, ‘get’, that performs an Ajax get operation on the provided URL and arranges to invoke the provided callback upon receipt of a response.
Finally, the ‘JQueryStatic’ interface contains a bare function signature
(query: string): JQuery;
The bare signature indicates that instances of the interface are callable. This example illustrates that TypeScript function types are just special cases of TypeScript object types. Specifically, function types are object types that contain one or more call signatures. For this reason we can write any function type as an object type literal. The following example uses both forms to describe the same type.
var f: { (): string; };
var sameType: () => string = f; // Ok
var nope: () => number = sameType; // Error: type mismatch
We mentioned above that the ‘$’ function behaves differently depending on the type of its parameter. So far, our jQuery typing only captures one of these behaviors: return an object of type ‘JQuery’ when passed a string. To specify multiple behaviors, TypeScript supports overloading of function signatures in object types. For example, we can add an additional call signature to the ‘JQueryStatic’ interface.
(ready: () => any): any;
This signature denotes that a function may be passed as the parameter of the ‘$’ function. When a function is passed to ‘$’, the jQuery library will invoke that function when a DOM document is ready. Because TypeScript supports overloading, tools can use TypeScript to show all available function signatures with their documentation tips and to give the correct documentation once a function has been called with a particular signature.
A typical client would not need to add any additional typing but could just use a community-supplied typing to discover (through statement completion with documentation tips) and verify (through static checking) correct use of the library, as in the following screen shot.
Section 3.3 provides additional information about object types.
Object types are compared structurally. For example, in the code fragment below, class ‘CPoint’ matches interface ‘Point’ because ‘CPoint’ has all of the required members of ‘Point’. A class may optionally declare that it implements an interface, so that the compiler will check the declaration for structural compatibility. The example also illustrates that an object type can match the type inferred from an object literal, as long as the object literal supplies all of the required members.
interface Point {
x: number;
y: number;
}
function getX(p: Point) {
return p.x;
}
class CPoint {
x: number;
y: number;
constructor(x: number, y: number) {
this.x = x;
this.y = y;
}
}
getX(new CPoint(0, 0)); // Ok, fields match
getX({ x: 0, y: 0, color: "red" }); // Extra fields Ok
getX({ x: 0 }); // Error: supplied parameter does not match
See section 3.11 for more information about type comparisons.
Ordinarily, TypeScript type inference proceeds “bottom-up”: from the leaves of an expression tree to its root. In the following example, TypeScript infers ’number’ as the return type of the function ‘mul’ by flowing type information bottom up in the return expression.
function mul(a: number, b: number) {
return a * b;
}
For variables and parameters without a type annotation or a default value, TypeScript infers type ‘any’, ensuring that compilers do not need non-local information about a function’s call sites to infer the function’s return type. Generally, this bottom-up approach provides programmers with a clear intuition about the flow of type information.
However, in some limited contexts, inference proceeds “top-down” from the context of an expression. Where this happens, it is called contextual typing. Contextual typing helps tools provide excellent information when a programmer is using a type but may not know all of the details of the type. For example, in the jQuery example, above, the programmer supplies a function expression as the second parameter to the ‘get’ method. During typing of that expression, tools can assume that the type of the function expression is as given in the ‘get’ signature and can provide a template that includes parameter names and types.
$.get("http://mysite.org/divContent",
function (data) {
$("div").text(data); // TypeScript infers data is a string
}
);
Contextual typing is also useful for writing out object literals. As the programmer types the object literal, the contextual type provides information that enables tools to provide completion for object member names.
Section 4.23 provides additional information about contextually typed expressions.
JavaScript practice has two very common design patterns: the module pattern and the class pattern. Roughly speaking, the module pattern uses closures to hide names and to encapsulate private data, while the class pattern uses prototype chains to implement many variations on object-oriented inheritance mechanisms. Libraries such as ‘prototype.js’ are typical of this practice. TypeScript’s namespaces are a formalization of the module pattern. (The term “module pattern” is somewhat unfortunate now that ECMAScript 2015 formally supports modules in a manner different from what the module pattern prescribes. For this reason, TypeScript uses the term “namespace” for its formalization of the module pattern.)
This section and the namespace section below will show how TypeScript emits consistent, idiomatic JavaScript when emitting ECMAScript 3 or 5 compliant code for classes and namespaces. The goal of TypeScript’s translation is to emit exactly what a programmer would type when implementing a class or namespace unaided by a tool. This section will also describe how TypeScript infers a type for each class declaration. We’ll start with a simple BankAccount class.
class BankAccount {
balance = 0;
deposit(credit: number) {
this.balance += credit;
return this.balance;
}
}
This class generates the following JavaScript code.
var BankAccount = (function () {
function BankAccount() {
this.balance = 0;
}
BankAccount.prototype.deposit = function(credit) {
this.balance += credit;
return this.balance;
};
return BankAccount;
})();
This TypeScript class declaration creates a variable named ‘BankAccount’ whose value is the constructor function for ‘BankAccount’ instances. This declaration also creates an instance type of the same name. If we were to write this type as an interface it would look like the following.
interface BankAccount {
balance: number;
deposit(credit: number): number;
}
If we were to write out the function type declaration for the ‘BankAccount’ constructor variable, it would have the following form.
var BankAccount: new() => BankAccount;
The function signature is prefixed with the keyword ’new’ indicating that the ‘BankAccount’ function must be called as a constructor. It is possible for a function’s type to have both call and constructor signatures. For example, the type of the built-in JavaScript Date object includes both kinds of signatures.
If we want to start our bank account with an initial balance, we can add to the ‘BankAccount’ class a constructor declaration.
class BankAccount {
balance: number;
constructor(initially: number) {
this.balance = initially;
}
deposit(credit: number) {
this.balance += credit;
return this.balance;
}
}
This version of the ‘BankAccount’ class requires us to introduce a constructor parameter and then assign it to the ‘balance’ field. To simplify this common case, TypeScript accepts the following shorthand syntax.
class BankAccount {
constructor(public balance: number) {
}
deposit(credit: number) {
this.balance += credit;
return this.balance;
}
}
The ‘public’ keyword denotes that the constructor parameter is to be retained as a field. Public is the default accessibility for class members, but a programmer can also specify private or protected accessibility for a class member. Accessibility is a design-time construct; it is enforced during static type checking but does not imply any runtime enforcement.
TypeScript classes also support inheritance, as in the following example.* *
class CheckingAccount extends BankAccount {
constructor(balance: number) {
super(balance);
}
writeCheck(debit: number) {
this.balance -= debit;
}
}
In this example, the class ‘CheckingAccount’ derives from class ‘BankAccount’. The constructor for ‘CheckingAccount’ calls the constructor for class ‘BankAccount’ using the ‘super’ keyword. In the emitted JavaScript code, the prototype of ‘CheckingAccount’ will chain to the prototype of ‘BankAccount’.
TypeScript classes may also specify static members. Static class members become properties of the class constructor.
Section 8 provides additional information about classes.
TypeScript enables programmers to summarize a set of numeric constants as an enum type. The example below creates an enum type to represent operators in a calculator application.
const enum Operator {
ADD,
DIV,
MUL,
SUB
}
function compute(op: Operator, a: number, b: number) {
console.log("the operator is" + Operator[op]);
// ...
}
In this example, the compute function logs the operator ‘op’ using a feature of enum types: reverse mapping from the enum value (‘op’) to the string corresponding to that value. For example, the declaration of ‘Operator’ automatically assigns integers, starting from zero, to the listed enum members. Section 9 describes how programmers can also explicitly assign integers to enum members, and can use any string to name an enum member.
When enums are declared with the const
modifier, the TypeScript compiler will emit for an enum member a JavaScript constant corresponding to that member’s assigned value (annotated with a comment). This improves performance on many JavaScript engines.
For example, the ‘compute’ function could contain a switch statement like the following.
switch (op) {
case Operator.ADD:
// execute add
break;
case Operator.DIV:
// execute div
break;
// ...
}
For this switch statement, the compiler will generate the following code.
switch (op) {
case 0 /* Operator.ADD */:
// execute add
break;
case 1 /* Operator.DIV */:
// execute div
break;
// ...
}
JavaScript implementations can use these explicit constants to generate efficient code for this switch statement, for example by building a jump table indexed by case value.
An important goal of TypeScript is to provide accurate and straightforward types for existing JavaScript programming patterns. To that end, TypeScript includes generic types, discussed in the next section, and overloading on string parameters, the topic of this section.
JavaScript programming interfaces often include functions whose behavior is discriminated by a string constant passed to the function. The Document Object Model makes heavy use of this pattern. For example, the following screen shot shows that the ‘createElement’ method of the ‘document’ object has multiple signatures, some of which identify the types returned when specific strings are passed into the method.
The following code fragment uses this feature. Because the ‘span’ variable is inferred to have the type ‘HTMLSpanElement’, the code can reference without static error the ‘isMultiline’ property of ‘span’.
var span = document.createElement("span");
span.isMultiLine = false; // OK: HTMLSpanElement has isMultiline property
In the following screen shot, a programming tool combines information from overloading on string parameters with contextual typing to infer that the type of the variable ’e’ is ‘MouseEvent’ and that therefore ’e’ has a ‘clientX’ property.
Section 3.9.2.4 provides details on how to use string literals in function signatures.
Like overloading on string parameters, generic types make it easier for TypeScript to accurately capture the behavior of JavaScript libraries. Because they enable type information to flow from client code, through library code, and back into client code, generic types may do more than any other TypeScript feature to support detailed API descriptions.
To illustrate this, let’s take a look at part of the TypeScript interface for the built-in JavaScript array type. You can find this interface in the ’lib.d.ts’ file that accompanies a TypeScript distribution.
interface Array<T> {
reverse(): T[];
sort(compareFn?: (a: T, b: T) => number): T[];
// ...
}
Interface definitions, like the one above, can have one or more type parameters. In this case the ‘Array’ interface has a single parameter, ‘T’, that defines the element type for the array. The ‘reverse’ method returns an array with the same element type. The sort method takes an optional parameter, ‘compareFn’, whose type is a function that takes two parameters of type ‘T’ and returns a number. Finally, sort returns an array with element type ‘T’.
Functions can also have generic parameters. For example, the array interface contains a ‘map’ method, defined as follows:
map<U>(func: (value: T, index: number, array: T[]) => U, thisArg?: any): U[];
The map method, invoked on an array ‘a’ with element type ‘T’, will apply function ‘func’ to each element of ‘a’, returning a value of type ‘U’.
The TypeScript compiler can often infer generic method parameters, making it unnecessary for the programmer to explicitly provide them. In the following example, the compiler infers that parameter ‘U’ of the map method has type ‘string’, because the function passed to map returns a string.
function numberToString(a: number[]) {
var stringArray = a.map(v => v.toString());
return stringArray;
}
The compiler infers in this example that the ’numberToString’ function returns an array of strings.
In TypeScript, classes can also have type parameters. The following code declares a class that implements a linked list of items of type ‘T’. This code illustrates how programmers can constrain type parameters to extend a specific type. In this case, the items on the list must extend the type ‘NamedItem’. This enables the programmer to implement the ’log’ function, which logs the name of the item.
interface NamedItem {
name: string;
}
class List<T extends NamedItem> {
next: List<T> = null;
constructor(public item: T) {
}
insertAfter(item: T) {
var temp = this.next;
this.next = new List(item);
this.next.next = temp;
}
log() {
console.log(this.item.name);
}
// ...
}
Section 3.7 provides further information about generic types.
Classes and interfaces support large-scale JavaScript development by providing a mechanism for describing how to use a software component that can be separated from that component’s implementation. TypeScript enforces encapsulation of implementation in classes at design time (by restricting use of private and protected members), but cannot enforce encapsulation at runtime because all object properties are accessible at runtime. Future versions of JavaScript may provide private names which would enable runtime enforcement of private and protected members.
In JavaScript, a very common way to enforce encapsulation at runtime is to use the module pattern: encapsulate private fields and methods using closure variables. The module pattern is a natural way to provide organizational structure and dynamic loading options by drawing a boundary around a software component. The module pattern can also provide the ability to introduce namespaces, avoiding use of the global namespace for most software components.
The following example illustrates the JavaScript module pattern.
(function(exports) {
var key = generateSecretKey();
function sendMessage(message) {
sendSecureMessage(message, key);
}
exports.sendMessage = sendMessage;
})(MessageModule);
This example illustrates the two essential elements of the module pattern: a module closure and a module object. The module closure is a function that encapsulates the module’s implementation, in this case the variable ‘key’ and the function ‘sendMessage’. The module object contains the exported variables and functions of the module. Simple modules may create and return the module object. The module above takes the module object as a parameter, ’exports’, and adds the ‘sendMessage’ property to the module object. This augmentation approach simplifies dynamic loading of modules and also supports separation of module code into multiple files.
The example assumes that an outer lexical scope defines the functions ‘generateSecretKey’ and ‘sendSecureMessage’; it also assumes that the outer scope has assigned the module object to the variable ‘MessageModule’.
TypeScript namespaces provide a mechanism for succinctly expressing the module pattern. In TypeScript, programmers can combine the module pattern with the class pattern by nesting namespaces and classes within an outer namespace.
The following example shows the definition and use of a simple namespace.
namespace M {
var s = "hello";
export function f() {
return s;
}
}
M.f();
M.s; // Error, s is not exported
In this example, variable ’s’ is a private feature of the namespace, but function ‘f’ is exported from the namespace and accessible to code outside of the namespace. If we were to describe the effect of namespace ‘M’ in terms of interfaces and variables, we would write
interface M {
f(): string;
}
var M: M;
The interface ‘M’ summarizes the externally visible behavior of namespace ‘M’. In this example, we can use the same name for the interface as for the initialized variable because in TypeScript type names and variable names do not conflict: each lexical scope contains a variable declaration space and type declaration space (see section 2.3 for more details).
The TypeScript compiler emits the following JavaScript code for the namespace:
var M;
(function(M) {
var s = "hello";
function f() {
return s;
}
M.f = f;
})(M || (M = {}));
In this case, the compiler assumes that the namespace object resides in global variable ‘M’, which may or may not have been initialized to the desired namespace object.
TypeScript also supports ECMAScript 2015 modules, which are files that contain top-level export and import directives. For this type of module the TypeScript compiler can emit both ECMAScript 2015 compliant code and down-level ECMAScript 3 or 5 compliant code for a variety of module loading systems, including CommonJS, Asynchronous Module Definition (AMD), and Universal Module Definition (UMD).
The remainder of this document is the formal specification of the TypeScript programming language and is intended to be read as an adjunct to the ECMAScript 2015 Language Specification (specifically, the ECMA-262 Standard, 6th Edition). This document describes the syntactic grammar added by TypeScript along with the compile-time processing and type checking performed by the TypeScript compiler, but it only minimally discusses the run-time behavior of programs since that is covered by the ECMAScript specification.
The syntactic grammar added by TypeScript language is specified throughout this document using the existing conventions and production names of the ECMAScript grammar. In places where TypeScript augments an existing grammar production it is so noted. For example:
Declaration: ( Modified )
…
InterfaceDeclaration
TypeAliasDeclaration
EnumDeclaration
The ‘( Modified )’ annotation indicates that an existing grammar production is being replaced, and the ‘…’ references the contents of the original grammar production.
Similar to the ECMAScript grammar, if the phrase “[no LineTerminator here]” appears in the right-hand side of a production of the syntactic grammar, it indicates that the production is not a match if a LineTerminator occurs in the input stream at the indicated position.
A core purpose of the TypeScript compiler is to track the named entities in a program and validate that they are used according to their designated meaning. Names in TypeScript can be written in several ways, depending on context. Specifically, a name can be written as
Most commonly, names are written to conform with the Identifier production, which is any IdentifierName that isn’t a reserved word.
The following keywords are reserved and cannot be used as an Identifier:
break case catch class
const continue debugger default
delete do else enum
export extends false finally
for function if import
in instanceof new null
return super switch this
throw true try typeof
var void while with
The following keywords cannot be used as identifiers in strict mode code, but are otherwise not restricted:
implements interface let package
private protected public static
yield
The following keywords cannot be used as user defined type names, but are otherwise not restricted:
any boolean number string
symbol
The following keywords have special meaning in certain contexts, but are valid identifiers:
abstract as async await
constructor declare from get
is module namespace of
require set type
The PropertyName production from the ECMAScript grammar is reproduced below:
PropertyName:
LiteralPropertyName
ComputedPropertyName
LiteralPropertyName:
IdentifierName
StringLiteral
NumericLiteral
ComputedPropertyName:
[
AssignmentExpression ]
A property name can be any identifier (including a reserved word), a string literal, a numeric literal, or a computed property name. String literals may be used to give properties names that are not valid identifiers, such as names containing blanks. Numeric literal property names are equivalent to string literal property names with the string representation of the numeric literal, as defined in the ECMAScript specification.
ECMAScript 2015 permits object literals and classes to declare members with computed property names. A computed property name specifies an expression that computes the actual property name at run-time. Because the final property name isn’t known at compile-time, TypeScript can only perform limited checks for entities declared with computed property names. However, a subset of computed property names known as well-known symbols can be used anywhere a PropertyName is expected, including property names within types. A computed property name is a well-known symbol if it is of the form
[ Symbol . xxx ]
In a well-known symbol, the identifier to the right of the dot must denote a property of the primitive type symbol
in the type of the global variable ‘Symbol’, or otherwise an error occurs.
In a PropertyName that specifies a ComputedPropertyName, the computed property name is required to denote a well-known symbol unless the property name occurs in a property assignment of an object literal (4.5) or a property member declaration in a non-ambient class (8.4).
Below is an example of an interface that declares a property with a well-known symbol name:
interface Iterable<T> {
[Symbol.iterator](): Iterator<T>;
}
TODO: Update to reflect treatment of computed property names with literal expressions.
Declarations introduce names in their associated declaration spaces. A name must be unique in its declaration space and can denote a value, a type, or a namespace, or some combination thereof. Effectively, a single name can have as many as three distinct meanings. For example:
var X: string; // Value named X
type X = number; // Type named X
namespace X { // Namespace named X
type Y = string;
}
A name that denotes a value has an associated type (section 3) and can be referenced in expressions (section 4.3). A name that denotes a type can be used by itself in a type reference or on the right hand side of a dot in a type reference (3.8.2). A name that denotes a namespace can be used one the left hand side of a dot in a type reference.
When a name with multiple meanings is referenced, the context in which the reference occurs determines the meaning. For example:
var n: X; // X references type
var s: X.Y = X; // First X references namespace, second X references value
In the first line, X references the type X because it occurs in a type position. In the second line, the first X references the namespace X because it occurs before a dot in a type name, and the second X references the variable X because it occurs in an expression.
Declarations introduce the following meanings for the name they declare:
Below are some examples of declarations that introduce multiple meanings for a name:
class C { // Value and type named C
x: string;
}
namespace N { // Value and namespace named N
export var x: string;
}
Declaration spaces exist as follows:
Top-level declarations in a source file with no top-level import or export declarations belong to the global namespace. Top-level declarations in a source file with one or more top-level import or export declarations belong to the module represented by that source file.
The container of an entity is defined as follows:
The root container of an entity is defined as follows:
Intuitively, the root container of an entity is the outermost module or namespace body from within which the entity is reachable.
Interfaces, enums, and namespaces are “open ended,” meaning that interface, enum, and namespace declarations with the same qualified name relative to a common root are automatically merged. For further details, see sections 7.2, 9.3, and 10.5.
Instance and static members in a class are in separate declaration spaces. Thus the following is permitted:
class C {
x: number; // Instance member
static x: string; // Static member
}
The scope of a name is the region of program text within which it is possible to refer to the entity declared by that name without qualification of the name. The scope of a name depends on the context in which the name is declared. The contexts are listed below in order from outermost to innermost:
Scopes may overlap, for example through nesting of namespaces and functions. When the scopes of two names overlap, the name with the innermost declaration takes precedence and access to the outer name is either not possible or only possible by qualification.
When an identifier is resolved as a PrimaryExpression (section 4.3), only names in scope with a value meaning are considered and other names are ignored.
When an identifier is resolved as a TypeName (section 3.8.2), only names in scope with a type meaning are considered and other names are ignored.
When an identifier is resolved as a NamespaceName (section 3.8.2), only names in scope with a namespace meaning are considered and other names are ignored.
TODO: Include specific rules for alias resolution.
Note that class and interface members are never directly in scope—they can only be accessed by applying the dot (’.’) operator to a class or interface instance. This even includes members of the current instance in a constructor or member function, which are accessed by applying the dot operator to this
.
As the rules above imply, locally declared entities in a namespace are closer in scope than exported entities declared in other namespace declarations for the same namespace. For example:
var x = 1;
namespace M {
export var x = 2;
console.log(x); // 2
}
namespace M {
console.log(x); // 2
}
namespace M {
var x = 3;
console.log(x); // 3
}
TypeScript adds optional static types to JavaScript. Types are used to place static constraints on program entities such as functions, variables, and properties so that compilers and development tools can offer better verification and assistance during software development. TypeScript’s static compile-time type system closely models the dynamic run-time type system of JavaScript, allowing programmers to accurately express the type relationships that are expected to exist when their programs run and have those assumptions pre-validated by the TypeScript compiler. TypeScript’s type analysis occurs entirely at compile-time and adds no run-time overhead to program execution.
All types in TypeScript are subtypes of a single top type called the Any type. The any
keyword references this type. The Any type is the one type that can represent any JavaScript value with no constraints. All other types are categorized as primitive types, object types, union types, intersection types, or type parameters. These types introduce various static constraints on their values.
The primitive types are the Number, Boolean, String, Symbol, Void, Null, and Undefined types along with user defined enum types. The number
, boolean
, string
, symbol
, and void
keywords reference the Number, Boolean, String, Symbol, and Void primitive types respectively. The Void type exists purely to indicate the absence of a value, such as in a function with no return value. It is not possible to explicitly reference the Null and Undefined types—only values of those types can be referenced, using the null
and undefined
literals.
The object types are all class, interface, array, tuple, function, and constructor types. Class and interface types are introduced through class and interface declarations and are referenced by the name given to them in their declarations. Class and interface types may be generic types which have one or more type parameters.
Union types represent values that have one of multiple types, and intersection types represent values that simultaneously have more than one type.
Declarations of classes, properties, functions, variables and other language entities associate types with those entities. The mechanism by which a type is formed and associated with a language entity depends on the particular kind of entity. For example, a namespace declaration associates the namespace with an anonymous type containing a set of properties corresponding to the exported variables and functions in the namespace, and a function declaration associates the function with an anonymous type containing a call signature corresponding to the parameters and return type of the function. Types can be associated with variables through explicit type annotations, such as
var x: number;
or through implicit type inference, as in
var x = 1;
which infers the type of ‘x’ to be the Number primitive type because that is the type of the value used to initialize ‘x’.
The Any type is used to represent any JavaScript value. A value of the Any type supports the same operations as a value in JavaScript and minimal static type checking is performed for operations on Any values. Specifically, properties of any name can be accessed through an Any value and Any values can be called as functions or constructors with any argument list.
The any
keyword references the Any type. In general, in places where a type is not explicitly provided and TypeScript cannot infer one, the Any type is assumed.
The Any type is a supertype of all types, and is assignable to and from all types.
Some examples:
var x: any; // Explicitly typed
var y; // Same as y: any
var z: { a; b; }; // Same as z: { a: any; b: any; }
function f(x) { // Same as f(x: any): void
console.log(x);
}
The primitive types are the Number, Boolean, String, Symbol, Void, Null, and Undefined types and all user defined enum types.
The Number primitive type corresponds to the similarly named JavaScript primitive type and represents double-precision 64-bit format IEEE 754 floating point values.
The number
keyword references the Number primitive type and numeric literals may be used to write values of the Number primitive type.
For purposes of determining type relationships (section 3.11) and accessing properties (section 4.13), the Number primitive type behaves as an object type with the same properties as the global interface type ‘Number’.
Some examples:
var x: number; // Explicitly typed
var y = 0; // Same as y: number = 0
var z = 123.456; // Same as z: number = 123.456
var s = z.toFixed(2); // Property of Number interface
The Boolean primitive type corresponds to the similarly named JavaScript primitive type and represents logical values that are either true or false.
The boolean
keyword references the Boolean primitive type and the true
and false
literals reference the two Boolean truth values.
For purposes of determining type relationships (section 3.11) and accessing properties (section 4.13), the Boolean primitive type behaves as an object type with the same properties as the global interface type ‘Boolean’.
Some examples:
var b: boolean; // Explicitly typed
var yes = true; // Same as yes: boolean = true
var no = false; // Same as no: boolean = false
The String primitive type corresponds to the similarly named JavaScript primitive type and represents sequences of characters stored as Unicode UTF-16 code units.
The string
keyword references the String primitive type and string literals may be used to write values of the String primitive type.
For purposes of determining type relationships (section 3.11) and accessing properties (section 4.13), the String primitive type behaves as an object type with the same properties as the global interface type ‘String’.
Some examples:
var s: string; // Explicitly typed
var empty = ""; // Same as empty: string = ""
var abc = 'abc'; // Same as abc: string = "abc"
var c = abc.charAt(2); // Property of String interface
The Symbol primitive type corresponds to the similarly named JavaScript primitive type and represents unique tokens that may be used as keys for object properties.
The symbol
keyword references the Symbol primitive type. Symbol values are obtained using the global object ‘Symbol’ which has a number of methods and properties and can be invoked as a function. In particular, the global object ‘Symbol’ defines a number of well-known symbols (2.2.3) that can be used in a manner similar to identifiers. Note that the ‘Symbol’ object is available only in ECMAScript 2015 environments.
For purposes of determining type relationships (section 3.11) and accessing properties (section 4.13), the Symbol primitive type behaves as an object type with the same properties as the global interface type ‘Symbol’.
Some examples:
var secretKey = Symbol();
var obj = {};
obj[secretKey] = "secret message"; // Use symbol as property key
obj[Symbol.toStringTag] = "test"; // Use of well-known symbol
The Void type, referenced by the void
keyword, represents the absence of a value and is used as the return type of functions with no return value.
The only possible values for the Void type are null
and undefined
. The Void type is a subtype of the Any type and a supertype of the Null and Undefined types, but otherwise Void is unrelated to all other types.
NOTE: We might consider disallowing declaring variables of type Void as they serve no useful purpose. However, because Void is permitted as a type argument to a generic type or function it is not feasible to disallow Void properties or parameters.
The Null type corresponds to the similarly named JavaScript primitive type and is the type of the null
literal.
The null
literal references the one and only value of the Null type. It is not possible to directly reference the Null type itself.
The Null type is a subtype of all types, except the Undefined type. This means that null
is considered a valid value for all primitive types, object types, union types, intersection types, and type parameters, including even the Number and Boolean primitive types.
Some examples:
var n: number = null; // Primitives can be null
var x = null; // Same as x: any = null
var e: Null; // Error, can't reference Null type
The Undefined type corresponds to the similarly named JavaScript primitive type and is the type of the undefined
literal.
The undefined
literal denotes the value given to all uninitialized variables and is the one and only value of the Undefined type. It is not possible to directly reference the Undefined type itself.
The undefined type is a subtype of all types. This means that undefined
is considered a valid value for all primitive types, object types, union types, intersection types, and type parameters.
Some examples:
var n: number; // Same as n: number = undefined
var x = undefined; // Same as x: any = undefined
var e: Undefined; // Error, can't reference Undefined type
Enum types are distinct user defined subtypes of the Number primitive type. Enum types are declared using enum declarations (section 9.1) and referenced using type references (section 3.8.2).
Enum types are assignable to the Number primitive type, and vice versa, but different enum types are not assignable to each other.
Specialized signatures (section 3.9.2.4) permit string literals to be used as types in parameter type annotations. String literal types are permitted only in that context and nowhere else.
All string literal types are subtypes of the String primitive type.
TODO: Update to reflect expanded support for string literal types.
Object types are composed from properties, call signatures, construct signatures, and index signatures, collectively called members.
Class and interface type references, array types, tuple types, function types, and constructor types are all classified as object types. Multiple constructs in the TypeScript language create object types, including:
Type references (section 3.8.2) to class and interface types are classified as object types. Type references to generic class and interface types include type arguments that are substituted for the type parameters of the class or interface to produce an actual object type.
Array types represent JavaScript arrays with a common element type. Array types are named type references created from the generic interface type ‘Array’ in the global namespace with the array element type as a type argument. Array type literals (section 3.8.4) provide a shorthand notation for creating such references.
The declaration of the ‘Array’ interface includes a property ’length’ and a numeric index signature for the element type, along with other members:
interface Array<T> {
length: number;
[x: number]: T;
// Other members
}
Array literals (section 4.6) may be used to create values of array types. For example
var a: string[] = ["hello", "world"];
A type is said to be an array-like type if it is assignable (section 3.11.4) to the type any[]
.
Tuple types represent JavaScript arrays with individually tracked element types. Tuple types are written using tuple type literals (section 3.8.5). A tuple type combines a set of numerically named properties with the members of an array type. Specifically, a tuple type
[ T0, T1, ..., Tn ]
combines the set of properties
{
0: T0;
1: T1;
...
n: Tn;
}
with the members of an array type whose element type is the union type (section 3.4) of the tuple element types.
Array literals (section 4.6) may be used to create values of tuple types. For example:
var t: [number, string] = [3, "three"];
var n = t[0]; // Type of n is number
var s = t[1]; // Type of s is string
var i: number;
var x = t[i]; // Type of x is number | string
Named tuple types can be created by declaring interfaces that derive from Array<T> and introduce numerically named properties. For example:
interface KeyValuePair<K, V> extends Array<K | V> { 0: K; 1: V; }
var x: KeyValuePair<number, string> = [10, "ten"];
A type is said to be a tuple-like type if it has a property with the numeric name ‘0’.
An object type containing one or more call signatures is said to be a function type. Function types may be written using function type literals (section 3.8.8) or by including call signatures in object type literals.
An object type containing one or more construct signatures is said to be a constructor type. Constructor types may be written using constructor type literals (section 3.8.9) or by including construct signatures in object type literals.
Every object type is composed from zero or more of the following kinds of members:
new
operator to objects of the given type.Properties are either public, private, or protected and are either required or optional:
Call and construct signatures may be specialized (section 3.9.2.4) by including parameters with string literal types. Specialized signatures are used to express patterns where specific string values for some parameters cause the types of other parameters or the function result to become further specialized.
Union types represent values that may have one of several distinct representations. A value of a union type A | B is a value that is either of type A or type B. Union types are written using union type literals (section 3.8.6).
A union type encompasses an ordered set of constituent types. While it is generally true that A | B is equivalent to B | A, the order of the constituent types may matter when determining the call and construct signatures of the union type.
Union types have the following subtype relationships:
Similarly, union types have the following assignability relationships:
The || and conditional operators (section 4.19.7 and 4.20) may produce values of union types, and array literals (section 4.6) may produce array values that have union types as their element types.
Type guards (section 4.24) may be used to narrow a union type to a more specific type. In particular, type guards are useful for narrowing union type values to a non-union type values.
In the example
var x: string | number;
var test: boolean;
x = "hello"; // Ok
x = 42; // Ok
x = test; // Error, boolean not assignable
x = test ? 5 : "five"; // Ok
x = test ? 0 : false; // Error, number | boolean not assignable
it is possible to assign ‘x’ a value of type string
, number
, or the union type string | number
, but not any other type. To access a value in ‘x’, a type guard can be used to first narrow the type of ‘x’ to either string
or number
:
var n = typeof x === "string" ? x.length : x; // Type of n is number
For purposes of property access and function calls, the apparent members (section 3.11.1) of a union type are those that are present in every one of its constituent types, with types that are unions of the respective apparent members in the constituent types. The following example illustrates the merging of member types that occurs when union types are created from object types.
interface A {
a: string;
b: number;
}
interface B {
a: number;
b: number;
c: number;
}
var x: A | B;
var a = x.a; // a has type string | number
var b = x.b; // b has type number
var c = x.c; // Error, no property c in union type
Note that ‘x.a’ has a union type because the type of ‘a’ is different in ‘A’ and ‘B’, whereas ‘x.b’ simply has type number because that is the type of ‘b’ in both ‘A’ and ‘B’. Also note that there is no property ‘x.c’ because only ‘B’ has a property ‘c’.
When used as a contextual type (section 4.23), a union type has those members that are present in any of its constituent types, with types that are unions of the respective members in the constituent types. Specifically, a union type used as a contextual type has the apparent members defined in section 3.11.1, except that a particular member need only be present in one or more constituent types instead of all constituent types.
Intersection types represent values that simultaneously have multiple types. A value of an intersection type A & B is a value that is both of type A and type B. Intersection types are written using intersection type literals (section 3.8.7).
An intersection type encompasses an ordered set of constituent types. While it is generally true that A & B is equivalent to B & A, the order of the constituent types may matter when determining the call and construct signatures of the intersection type.
Intersection types have the following subtype relationships:
Similarly, intersection types have the following assignability relationships:
For purposes of property access and function calls, the apparent members (section 3.11.1) of an intersection type are those that are present in one or more of its constituent types, with types that are intersections of the respective apparent members in the constituent types. The following examples illustrate the merging of member types that occurs when intersection types are created from object types.
interface A { a: number }
interface B { b: number }
var ab: A & B = { a: 1, b: 1 };
var a: A = ab; // A & B assignable to A
var b: B = ab; // A & B assignable to B
interface X { p: A }
interface Y { p: B }
var xy: X & Y = { p: ab }; // X & Y has property p of type A & B
type F1 = (a: string, b: string) => void;
type F2 = (a: number, b: number) => void;
var f: F1 & F2 = (a: string | number, b: string | number) => { };
f("hello", "world"); // Ok
f(1, 2); // Ok
f(1, "test"); // Error
The union and intersection type operators can be applied to type parameters. This capability can for example be used to model functions that merge objects:
function extend<T, U>(first: T, second: U): T & U {
// Extend first with properties of second
}
var x = extend({ a: "hello" }, { b: 42 });
var s = x.a;
var n = x.b;
It is possible to create intersection types for which no values other than null or undefined are possible. For example, intersections of primitive types such as string & number
fall into this category.
A type parameter represents an actual type that the parameter is bound to in a generic type reference or a generic function call. Type parameters have constraints that establish upper bounds for their actual type arguments.
Since a type parameter represents a multitude of different type arguments, type parameters have certain restrictions compared to other types. In particular, a type parameter cannot be used as a base class or interface.
Class, interface, type alias, and function declarations may optionally include lists of type parameters enclosed in < and > brackets. Type parameters are also permitted in call signatures of object, function, and constructor type literals.
TypeParameters:
<
TypeParameterList >
TypeParameterList:
TypeParameter
TypeParameterList ,
TypeParameter
TypeParameter:
BindingIdentifier Constraintopt
Constraint:
extends
Type
Type parameter names must be unique. A compile-time error occurs if two or more type parameters in the same TypeParameterList have the same name.
The scope of a type parameter extends over the entire declaration with which the type parameter list is associated, with the exception of static member declarations in classes.
A type parameter may have an associated type parameter constraint that establishes an upper bound for type arguments. Type parameters may be referenced in type parameter constraints within the same type parameter list, including even constraint declarations that occur to the left of the type parameter.
The base constraint of a type parameter T is defined as follows:
{}
.In the example
interface G<T, U extends V, V extends Function> { }
the base constraint of ‘T’ is the empty object type and the base constraint of ‘U’ and ‘V’ is ‘Function’.
For purposes of determining type relationships (section 3.11), type parameters appear to be subtypes of their base constraint. Likewise, in property accesses (section 4.13), new
operations (section 4.14), and function calls (section 4.15), type parameters appear to have the members of their base constraint, but no other members.
It is an error for a type parameter to directly or indirectly be a constraint for itself. For example, both of the following declarations are invalid:
interface A<T extends T> { }
interface B<T extends U, U extends T> { }
A type reference (section 3.8.2) to a generic type must include a list of type arguments enclosed in angle brackets and separated by commas. Similarly, a call (section 4.15) to a generic function may explicitly include a type argument list instead of relying on type inference.
TypeArguments:
<
TypeArgumentList >
TypeArgumentList:
TypeArgument
TypeArgumentList ,
TypeArgument
TypeArgument:
Type
Type arguments correspond one-to-one with type parameters of the generic type or function being referenced. A type argument list is required to specify exactly one type argument for each corresponding type parameter, and each type argument for a constrained type parameter is required to satisfy the constraint of that type parameter. A type argument satisfies a type parameter constraint if the type argument is assignable to (section 3.11.4) the constraint type once type arguments are substituted for type parameters.
Given the declaration
interface G<T, U extends Function> { }
a type reference of the form ‘G<A, B>’ places no requirements on ‘A’ but requires ‘B’ to be assignable to ‘Function’.
The process of substituting type arguments for type parameters in a generic type or generic signature is known as instantiating the generic type or signature. Instantiation of a generic type or signature can fail if the supplied type arguments do not satisfy the constraints of their corresponding type parameters.
Every class and interface has a this-type that represents the actual type of instances of the class or interface within the declaration of the class or interface. The this-type is referenced using the keyword this
in a type position. Within instance methods and constructors of a class, the type of the expression this
(section 4.2) is the this-type of the class.
Classes and interfaces support inheritance and therefore the instance represented by this
in a method isn’t necessarily an instance of the containing class—it may in fact be an instance of a derived class or interface. To model this relationship, the this-type of a class or interface is classified as a type parameter. Unlike other type parameters, it is not possible to explicitly pass a type argument for a this-type. Instead, in a type reference to a class or interface type, the type reference itself is implicitly passed as a type argument for the this-type. For example:
class A {
foo() {
return this;
}
}
class B extends A {
bar() {
return this;
}
}
let b: B;
let x = b.foo().bar(); // Fluent pattern works, type of x is B
In the declaration of b
above, the type reference B
is itself passed as a type argument for B’s this-type. Thus, the referenced type is an instantiation of class B
where all occurrences of the type this
are replaced with B
, and for that reason the foo
method of B
actually returns B
(as opposed to A
).
The this-type of a given class or interface type C implicitly has a constraint consisting of a type reference to C with C’s own type parameters passed as type arguments and with that type reference passed as the type argument for the this-type.
Classes, interfaces, enums, and type aliases are named types that are introduced through class declarations (section 8.1), interface declarations (section 7.1), enum declarations (9.1), and type alias declarations (section 3.10). Classes, interfaces, and type aliases may have type parameters and are then called generic types. Conversely, named types without type parameters are called non-generic types.
Interface declarations only introduce named types, whereas class declarations introduce named types and constructor functions that create instances of implementations of those named types. The named types introduced by class and interface declarations have only minor differences (classes can’t declare optional members and interfaces can’t declare private or protected members) and are in most contexts interchangeable. In particular, class declarations with only public members introduce named types that function exactly like those created by interface declarations.
Named types are referenced through type references (section 3.8.2) that specify a type name and, if applicable, the type arguments to be substituted for the type parameters of the named type.
Named types are technically not types—only references to named types are. This distinction is particularly evident with generic types: Generic types are “templates” from which multiple actual types can be created by writing type references that supply type arguments to substitute in place of the generic type’s type parameters. This substitution process is known as instantiating a generic type. Only once a generic type is instantiated does it denote an actual type.
TypeScript has a structural type system, and therefore an instantiation of a generic type is indistinguishable from an equivalent manually written expansion. For example, given the declaration
interface Pair<T1, T2> { first: T1; second: T2; }
the type reference
Pair<string, Entity>
is indistinguishable from the type
{ first: string; second: Entity; }
Types are specified either by referencing their keyword or name, or by writing object type literals, array type literals, tuple type literals, function type literals, constructor type literals, or type queries.
Type:
UnionOrIntersectionOrPrimaryType
FunctionType
ConstructorType
UnionOrIntersectionOrPrimaryType:
UnionType
IntersectionOrPrimaryType
IntersectionOrPrimaryType:
IntersectionType
PrimaryType
PrimaryType:
ParenthesizedType
PredefinedType
TypeReference
ObjectType
ArrayType
TupleType
TypeQuery
ThisType
ParenthesizedType:
(
Type )
Parentheses are required around union, intersection, function, or constructor types when they are used as array element types; around union, function, or constructor types in intersection types; and around function or constructor types in union types. For example:
(string | number)[]
((x: string) => string) | ((x: number) => number)
(A | B) & (C | D)
The different forms of type notations are described in the following sections.
The any
, number
, boolean
, string
, symbol
and void
keywords reference the Any type and the Number, Boolean, String, Symbol, and Void primitive types respectively.
PredefinedType:
any
number
boolean
string
symbol
void
The predefined type keywords are reserved and cannot be used as names of user defined types.
A type reference references a named type or type parameter through its name and, in the case of a generic type, supplies a type argument list.
TypeReference:
TypeName [no LineTerminator here] TypeArgumentsopt
TypeName:
IdentifierReference
NamespaceName .
IdentifierReference
NamespaceName:
IdentifierReference
NamespaceName .
IdentifierReference
A TypeReference consists of a TypeName that a references a named type or type parameter. A reference to a generic type must be followed by a list of TypeArguments (section 3.6.2).
A TypeName is either a single identifier or a sequence of identifiers separated by dots. In a type name, all identifiers but the last one refer to namespaces and the last identifier refers to a named type.
Resolution of a TypeName consisting of a single identifier is described in section 2.4.
Resolution of a TypeName of the form N.X, where N is a NamespaceName and X is an IdentifierReference, proceeds by first resolving the namespace name N. If the resolution of N is successful and the export member set (sections 10.4 and 11.3.4.4) of the resulting namespace contains a named type X, then N.X refers to that member. Otherwise, N.X is undefined.
Resolution of a NamespaceName consisting of a single identifier is described in section 2.4. Identifiers declared in namespace declarations (section 10.1) or import declarations (sections 10.3, 11.3.2, and 11.3.3) may be classified as namespaces.
Resolution of a NamespaceName of the form N.X, where N is a NamespaceName and X is an IdentifierReference, proceeds by first resolving the namespace name N. If the resolution of N is successful and the export member set (sections 10.4 and 11.3.4.4) of the resulting namespace contains an exported namespace member X, then N.X refers to that member. Otherwise, N.X is undefined.
A type reference to a generic type is required to specify exactly one type argument for each type parameter of the referenced generic type, and each type argument must be assignable to (section 3.11.4) the constraint of the corresponding type parameter or otherwise an error occurs. An example:
interface A { a: string; }
interface B extends A { b: string; }
interface C extends B { c: string; }
interface G<T, U extends B> {
x: T;
y: U;
}
var v1: G<A, C>; // Ok
var v2: G<{ a: string }, C>; // Ok, equivalent to G<A, C>
var v3: G<A, A>; // Error, A not valid argument for U
var v4: G<G<A, B>, C>; // Ok
var v5: G<any, any>; // Ok
var v6: G<any>; // Error, wrong number of arguments
var v7: G; // Error, no arguments
A type argument is simply a Type and may itself be a type reference to a generic type, as demonstrated by ‘v4’ in the example above.
As described in section 3.7, a type reference to a generic type G designates a type wherein all occurrences of G’s type parameters have been replaced with the actual type arguments supplied in the type reference. For example, the declaration of ‘v1’ above is equivalent to:
var v1: {
x: { a: string; }
y: { a: string; b: string; c: string };
};
An object type literal defines an object type by specifying the set of members that are statically considered to be present in instances of the type. Object type literals can be given names using interface declarations but are otherwise anonymous.
ObjectType:
{
TypeBodyopt }
TypeBody:
TypeMemberList ;
opt
TypeMemberList ,
opt
TypeMemberList:
TypeMember
TypeMemberList ;
TypeMember
TypeMemberList ,
TypeMember
TypeMember:
PropertySignature
CallSignature
ConstructSignature
IndexSignature
MethodSignature
The members of an object type literal are specified as a combination of property, call, construct, index, and method signatures. Object type members are described in section 3.9.
An array type literal is written as an element type followed by an open and close square bracket.
ArrayType:
PrimaryType [no LineTerminator here] [
]
An array type literal references an array type (section 3.3.2) with the given element type. An array type literal is simply shorthand notation for a reference to the generic interface type ‘Array’ in the global namespace with the element type as a type argument.
When union, intersection, function, or constructor types are used as array element types they must be enclosed in parentheses. For example:
(string | number)[]
(() => string))[]
Alternatively, array types can be written using the ‘Array<T>’ notation. For example, the types above are equivalent to
Array<string | number>
Array<() => string>
A tuple type literal is written as a sequence of element types, separated by commas and enclosed in square brackets.
TupleType:
[
TupleElementTypes ]
TupleElementTypes:
TupleElementType
TupleElementTypes ,
TupleElementType
TupleElementType:
Type
A tuple type literal references a tuple type (section 3.3.3).
A union type literal is written as a sequence of types separated by vertical bars.
UnionType:
UnionOrIntersectionOrPrimaryType |
IntersectionOrPrimaryType
A union type literal references a union type (section 3.4).
An intersection type literal is written as a sequence of types separated by ampersands.
IntersectionType:
IntersectionOrPrimaryType &
PrimaryType
An intersection type literal references an intersection type (section 3.5).
A function type literal specifies the type parameters, regular parameters, and return type of a call signature.
FunctionType:
TypeParametersopt (
ParameterListopt )
=>
Type
A function type literal is shorthand for an object type containing a single call signature. Specifically, a function type literal of the form
< T1, T2, ... > ( p1, p2, ... ) => R
is exactly equivalent to the object type literal
{ < T1, T2, ... > ( p1, p2, ... ) : R }
Note that function types with multiple call or construct signatures cannot be written as function type literals but must instead be written as object type literals.
A constructor type literal specifies the type parameters, regular parameters, and return type of a construct signature.
ConstructorType:
new
TypeParametersopt (
ParameterListopt )
=>
Type
A constructor type literal is shorthand for an object type containing a single construct signature. Specifically, a constructor type literal of the form
new < T1, T2, ... > ( p1, p2, ... ) => R
is exactly equivalent to the object type literal
{ new < T1, T2, ... > ( p1, p2, ... ) : R }
Note that constructor types with multiple construct signatures cannot be written as constructor type literals but must instead be written as object type literals.
A type query obtains the type of an expression.
TypeQuery:
typeof
TypeQueryExpression
TypeQueryExpression:
IdentifierReference
TypeQueryExpression .
IdentifierName
A type query consists of the keyword typeof
followed by an expression. The expression is restricted to a single identifier or a sequence of identifiers separated by periods. The expression is processed as an identifier expression (section 4.3) or property access expression (section 4.13), the widened type (section 3.12) of which becomes the result. Similar to other static typing constructs, type queries are erased from the generated JavaScript code and add no run-time overhead.
Type queries are useful for capturing anonymous types that are generated by various constructs such as object literals, function declarations, and namespace declarations. For example:
var a = { x: 10, y: 20 };
var b: typeof a;
Above, ‘b’ is given the same type as ‘a’, namely { x: number; y: number; }
.
If a declaration includes a type annotation that references the entity being declared through a circular path of type queries or type references containing type queries, the resulting type is the Any type. For example, all of the following variables are given the type Any:
var c: typeof c;
var d: typeof e;
var e: typeof d;
var f: Array<typeof f>;
However, if a circular path of type queries includes at least one ObjectType, FunctionType or ConstructorType, the construct denotes a recursive type:
var g: { x: typeof g; };
var h: () => typeof h;
Here, ‘g’ and ‘g.x’ have the same recursive type, and likewise ‘h’ and ‘h()’ have the same recursive type.
The this
keyword is used to reference the this-type (section 3.6.3) of a class or interface.
ThisType:
this
The meaning of a ThisType depends on the closest enclosing FunctionDeclaration, FunctionExpression, PropertyDefinition, ClassElement, or TypeMember, known as the root declaration of the ThisType, as follows:
Note that in order to avoid ambiguities it is not possible to reference the this-type of a class or interface in a nested object type literal. In the example
interface ListItem {
getHead(): this;
getTail(): this;
getHeadAndTail(): { head: this, tail: this }; // Error
}
the this
references on the last line are in error because their root declarations are not members of a class or interface. The recommended way to reference the this-type of an outer class or interface in an object type literal is to declare an intermediate generic type and pass this
as a type argument. For example:
type HeadAndTail<T> = { head: T, tail: T };
interface ListItem {
getHead(): this;
getTail(): this;
getHeadAndTail(): HeadAndTail<this>;
}
The members of an object type literal (section 3.8.3) are specified as a combination of property, call, construct, index, and method signatures.
A property signature declares the name and type of a property member.
PropertySignature:
PropertyName ?
opt TypeAnnotationopt
TypeAnnotation:
:
Type
The PropertyName (2.2.2) of a property signature must be unique within its containing type, and must denote a well-known symbol if it is a computed property name (2.2.3). If the property name is followed by a question mark, the property is optional. Otherwise, the property is required.
If a property signature omits a TypeAnnotation, the Any type is assumed.
A call signature defines the type parameters, parameter list, and return type associated with applying a call operation (section 4.15) to an instance of the containing type. A type may overload call operations by defining multiple different call signatures.
CallSignature:
TypeParametersopt (
ParameterListopt )
TypeAnnotationopt
A call signature that includes TypeParameters (section 3.6.1) is called a generic call signature. Conversely, a call signature with no TypeParameters is called a non-generic call signature.
As well as being members of object type literals, call signatures occur in method signatures (section 3.9.5), function expressions (section 4.10), and function declarations (section 6.1).
An object type containing call signatures is said to be a function type.
Type parameters (section 3.6.1) in call signatures provide a mechanism for expressing the relationships of parameter and return types in call operations. For example, a signature might introduce a type parameter and use it as both a parameter type and a return type, in effect describing a function that returns a value of the same type as its argument.
Type parameters may be referenced in parameter types and return type annotations, but not in type parameter constraints, of the call signature in which they are introduced.
Type arguments (section 3.6.2) for call signature type parameters may be explicitly specified in a call operation or may, when possible, be inferred (section 4.15.2) from the types of the regular arguments in the call. An instantiation of a generic call signature for a particular set of type arguments is the call signature formed by replacing each type parameter with its corresponding type argument.
Some examples of call signatures with type parameters follow below.
A function taking an argument of any type, returning a value of that same type:
<T>(x: T): T
A function taking two values of the same type, returning an array of that type:
<T>(x: T, y: T): T[]
A function taking two arguments of different types, returning an object with properties ‘x’ and ‘y’ of those types:
<T, U>(x: T, y: U): { x: T; y: U; }
A function taking an array of one type and a function argument, returning an array of another type, where the function argument takes a value of the first array element type and returns a value of the second array element type:
<T, U>(a: T[], f: (x: T) => U): U[]
A signature’s parameter list consists of zero or more required parameters, followed by zero or more optional parameters, finally followed by an optional rest parameter.
ParameterList:
RequiredParameterList
OptionalParameterList
RestParameter
RequiredParameterList ,
OptionalParameterList
RequiredParameterList ,
RestParameter
OptionalParameterList ,
RestParameter
RequiredParameterList ,
OptionalParameterList ,
RestParameter
RequiredParameterList:
RequiredParameter
RequiredParameterList ,
RequiredParameter
RequiredParameter:
AccessibilityModifieropt BindingIdentifierOrPattern TypeAnnotationopt
BindingIdentifier :
StringLiteral
AccessibilityModifier:
public
private
protected
BindingIdentifierOrPattern:
BindingIdentifier
BindingPattern
OptionalParameterList:
OptionalParameter
OptionalParameterList ,
OptionalParameter
OptionalParameter:
AccessibilityModifieropt BindingIdentifierOrPattern ?
TypeAnnotationopt
AccessibilityModifieropt BindingIdentifierOrPattern TypeAnnotationopt Initializer
BindingIdentifier ?
:
StringLiteral
RestParameter:
...
BindingIdentifier TypeAnnotationopt
A parameter declaration may specify either an identifier or a binding pattern (5.2.2). The identifiers specified in parameter declarations and binding patterns in a parameter list must be unique within that parameter list.
The type of a parameter in a signature is determined as follows:
any[]
.any
.A parameter is permitted to include a public
, private
, or protected
modifier only if it occurs in the parameter list of a ConstructorImplementation (section 8.3.1) and only if it doesn’t specify a BindingPattern.
A type annotation for a rest parameter must denote an array type.
When a parameter type annotation specifies a string literal type, the containing signature is a specialized signature (section 3.9.2.4). Specialized signatures are not permitted in conjunction with a function body, i.e. the FunctionExpression, FunctionImplementation, MemberFunctionImplementation, and ConstructorImplementation grammar productions do not permit parameters with string literal types.
A parameter can be marked optional by following its name or binding pattern with a question mark (?
) or by including an initializer. Initializers (including binding property or element initializers) are permitted only when the parameter list occurs in conjunction with a function body, i.e. only in a FunctionExpression, FunctionImplementation, MemberFunctionImplementation, or ConstructorImplementation grammar production.
TODO: Update to reflect binding parameter cannot be optional in implementation signature.
TODO: Update to reflect required parameters support initializers.
If present, a call signature’s return type annotation specifies the type of the value computed and returned by a call operation. A void
return type annotation is used to indicate that a function has no return value.
When a call signature with no return type annotation occurs in a context without a function body, the return type is assumed to be the Any type.
When a call signature with no return type annotation occurs in a context that has a function body (specifically, a function implementation, a member function implementation, or a member accessor declaration), the return type is inferred from the function body as described in section 6.3.
When a parameter type annotation specifies a string literal type (section 3.2.9), the containing signature is considered a specialized signature. Specialized signatures are used to express patterns where specific string values for some parameters cause the types of other parameters or the function result to become further specialized. For example, the declaration
interface Document {
createElement(tagName: "div"): HTMLDivElement;
createElement(tagName: "span"): HTMLSpanElement;
createElement(tagName: "canvas"): HTMLCanvasElement;
createElement(tagName: string): HTMLElement;
}
states that calls to ‘createElement’ with the string literals “div”, “span”, and “canvas” return values of type ‘HTMLDivElement’, ‘HTMLSpanElement’, and ‘HTMLCanvasElement’ respectively, and that calls with all other string expressions return values of type ‘HTMLElement’.
When writing overloaded declarations such as the one above it is important to list the non-specialized signature last. This is because overload resolution (section 4.15.1) processes the candidates in declaration order and picks the first one that matches.
Every specialized call or construct signature in an object type must be assignable to at least one non-specialized call or construct signature in the same object type (where a call signature A is considered assignable to another call signature B if an object type containing only A would be assignable to an object type containing only B). For example, the ‘createElement’ property in the example above is of a type that contains three specialized signatures, all of which are assignable to the non-specialized signature in the type.
A construct signature defines the parameter list and return type associated with applying the new
operator (section 4.14) to an instance of the containing type. A type may overload new
operations by defining multiple construct signatures with different parameter lists.
ConstructSignature:
new
TypeParametersopt (
ParameterListopt )
TypeAnnotationopt
The type parameters, parameter list, and return type of a construct signature are subject to the same rules as a call signature.
A type containing construct signatures is said to be a constructor type.
An index signature defines a type constraint for properties in the containing type.
IndexSignature:
[
BindingIdentifier :
string
]
TypeAnnotation
[
BindingIdentifier :
number
]
TypeAnnotation
There are two kinds of index signatures:
string
, define type constraints for all properties and numeric index signatures in the containing type. Specifically, in a type with a string index signature of type T, all properties and numeric index signatures must have types that are assignable to T.number
, define type constraints for all numerically named properties in the containing type. Specifically, in a type with a numeric index signature of type T, all numerically named properties must have types that are assignable to T.A numerically named property is a property whose name is a valid numeric literal. Specifically, a property with a name N for which ToString(ToNumber(N)) is identical to N, where ToString and ToNumber are the abstract operations defined in ECMAScript specification.
An object type can contain at most one string index signature and one numeric index signature.
Index signatures affect the determination of the type that results from applying a bracket notation property access to an instance of the containing type, as described in section 4.13.
A method signature is shorthand for declaring a property of a function type.
MethodSignature:
PropertyName ?
opt CallSignature
If the PropertyName is a computed property name (2.2.3), it must specify a well-known symbol. If the PropertyName is followed by a question mark, the property is optional. Otherwise, the property is required. Only object type literals and interfaces can declare optional properties.
A method signature of the form
f < T1, T2, ... > ( p1, p2, ... ) : R
is equivalent to the property declaration
f : { < T1, T2, ... > ( p1, p2, ... ) : R }
A literal type may overload a method by declaring multiple method signatures with the same name but differing parameter lists. Overloads must either all be required (question mark omitted) or all be optional (question mark included). A set of overloaded method signatures correspond to a declaration of a single property with a type composed from an equivalent set of call signatures. Specifically
f < T1, T2, ... > ( p1, p2, ... ) : R ;
f < U1, U2, ... > ( q1, q2, ... ) : S ;
...
is equivalent to
f : {
< T1, T2, ... > ( p1, p2, ... ) : R ;
< U1, U2, ... > ( q1, q2, ... ) : S ;
...
} ;
In the following example of an object type
{
func1(x: number): number; // Method signature
func2: (x: number) => number; // Function type literal
func3: { (x: number): number }; // Object type literal
}
the properties ‘func1’, ‘func2’, and ‘func3’ are all of the same type, namely an object type with a single call signature taking a number and returning a number. Likewise, in the object type
{
func4(x: number): number;
func4(s: string): string;
func5: {
(x: number): number;
(s: string): string;
};
}
the properties ‘func4’ and ‘func5’ are of the same type, namely an object type with two call signatures taking and returning number and string respectively.
A type alias declaration introduces a type alias in the containing declaration space.
TypeAliasDeclaration:
type
BindingIdentifier TypeParametersopt =
Type ;
A type alias serves as an alias for the type specified in the type alias declaration. Unlike an interface declaration, which always introduces a named object type, a type alias declaration can introduce a name for any kind of type, including primitive, union, and intersection types.
A type alias may optionally have type parameters (section 3.6.1) that serve as placeholders for actual types to be provided when the type alias is referenced in type references. A type alias with type parameters is called a generic type alias. The type parameters of a generic type alias declaration are in scope and may be referenced in the aliased Type.
Type aliases are referenced using type references (3.8.2). Type references to generic type aliases produce instantiations of the aliased type with the given type arguments. Writing a reference to a non-generic type alias has exactly the same effect as writing the aliased type itself, and writing a reference to a generic type alias has exactly the same effect as writing the resulting instantiation of the aliased type.
The BindingIdentifier of a type alias declaration may not be one of the predefined type names (section 3.8.1).
It is an error for the type specified in a type alias to depend on that type alias. Types have the following dependencies:
Given this definition, the complete set of types upon which a type depends is the transitive closure of the directly depends on relationship. Note that object type literals, function type literals, and constructor type literals do not depend on types referenced within them and are therefore permitted to circularly reference themselves through type aliases.
Some examples of type alias declarations:
type StringOrNumber = string | number;
type Text = string | { text: string };
type NameLookup = Dictionary<string, Person>;
type ObjectStatics = typeof Object;
type Callback<T> = (data: T) => void;
type Pair<T> = [T, T];
type Coordinates = Pair<number>;
type Tree<T> = T | { left: Tree<T>, right: Tree<T> };
Interface types have many similarities to type aliases for object type literals, but since interface types offer more capabilities they are generally preferred to type aliases. For example, the interface type
interface Point {
x: number;
y: number;
}
could be written as the type alias
type Point = {
x: number;
y: number;
};
However, doing so means the following capabilities are lost:
Types in TypeScript have identity, subtype, supertype, and assignment compatibility relationships as defined in the following sections.
The apparent members of a type are the members observed in subtype, supertype, and assignment compatibility relationships, as well as in the type checking of property accesses (section 4.13), new
operations (section 4.14), and function calls (section 4.15). The apparent members of a type are determined as follows:
If a type is not one of the above, it is considered to have no apparent members.
In effect, a type’s apparent members make it a subtype of the ‘Object’ or ‘Function’ interface unless the type defines members that are incompatible with those of the ‘Object’ or ‘Function’ interface—which, for example, occurs if the type defines a property with the same name as a property in the ‘Object’ or ‘Function’ interface but with a type that isn’t a subtype of that in the ‘Object’ or ‘Function’ interface.
Some examples:
var o: Object = { x: 10, y: 20 }; // Ok
var f: Function = (x: number) => x * x; // Ok
var err: Object = { toString: 0 }; // Error
The last assignment is an error because the object literal has a ’toString’ method that isn’t compatible with that of ‘Object’.
Two types are considered identical when
Two members are considered identical when
Two call or construct signatures are considered identical when they have the same number of type parameters with identical type parameter constraints and, after substituting type Any for the type parameters introduced by the signatures, identical number of parameters with identical kind (required, optional or rest) and types, and identical return types.
Note that, except for primitive types and classes with private or protected members, it is structure, not naming, of types that determines identity. Also, note that parameter names are not significant when determining identity of signatures.
Private and protected properties match only if they originate in the same declaration and have identical types. Two distinct types might contain properties that originate in the same declaration if the types are separate parameterized references to the same generic class. In the example
class C<T> { private x: T; }
interface X { f(): string; }
interface Y { f(): string; }
var a: C<X>;
var b: C<Y>;
the variables ‘a’ and ‘b’ are of identical types because the two type references to ‘C’ create types with a private member ‘x’ that originates in the same declaration, and because the two private ‘x’ members have types with identical sets of members once the type arguments ‘X’ and ‘Y’ are substituted.
S is a subtype of a type T, and T is a supertype of S, if S has no excess properties with respect to T (3.11.5) and one of the following is true:
When comparing call or construct signatures, parameter names are ignored and rest parameters correspond to an unbounded expansion of optional parameters of the rest parameter element type.
Note that specialized call and construct signatures (section 3.9.2.4) are not significant when determining subtype and supertype relationships.
Also note that type parameters are not considered object types. Thus, the only subtypes of a type parameter T are T itself and other type parameters that are directly or indirectly constrained to T.
Types are required to be assignment compatible in certain circumstances, such as expression and variable types in assignment statements and argument and parameter types in function calls.
S is assignable to a type T, and T is assignable from S, if S has no excess properties with respect to T (3.11.5) and one of the following is true:
When comparing call or construct signatures, parameter names are ignored and rest parameters correspond to an unbounded expansion of optional parameters of the rest parameter element type.
Note that specialized call and construct signatures (section 3.9.2.4) are not significant when determining assignment compatibility.
The assignment compatibility and subtyping rules differ only in that
The assignment compatibility rules imply that, when assigning values or passing parameters, optional properties must either be present and of a compatible type, or not be present at all. For example:
function foo(x: { id: number; name?: string; }) { }
foo({ id: 1234 }); // Ok
foo({ id: 1234, name: "hello" }); // Ok
foo({ id: 1234, name: false }); // Error, name of wrong type
foo({ name: "hello" }); // Error, id required but missing
The subtype and assignment compatibility relationships require that source types have no excess properties with respect to their target types. The purpose of this check is to detect excess or misspelled properties in object literals.
A source type S is considered to have excess properties with respect to a target type T if
A property P is said to be expected in a type T if one of the following is true:
The type inferred for an object literal (as described in section 4.5) is considered a fresh object literal type. The freshness disappears when an object literal type is widened (3.12) or is the type of the expression in a type assertion (4.16).
Consider the following example:
interface CompilerOptions {
strict?: boolean;
sourcePath?: string;
targetPath?: string;
}
var options: CompilerOptions = {
strict: true,
sourcepath: "./src", // Error, excess or misspelled property
targetpath: "./bin" // Error, excess or misspelled property
};
The ‘CompilerOptions’ type contains only optional properties, so without the excess property check, any object literal would be assignable to the ‘options’ variable (because a misspelled property would just be considered an excess property of a different name).
In cases where excess properties are expected, an index signature can be added to the target type as an indicator of intent:
interface InputElement {
name: string;
visible?: boolean;
[x: string]: any; // Allow additional properties of any type
}
var address: InputElement = {
name: "Address",
visible: true,
help: "Enter address here", // Allowed because of index signature
shortcut: "Alt-A" // Allowed because of index signature
};
During type argument inference in a function call (section 4.15.2) it is in certain circumstances necessary to instantiate a generic call signature of an argument expression in the context of a non-generic call signature of a parameter such that further inferences can be made. A generic call signature A is instantiated in the context of non-generic call signature B as follows:
In certain contexts, inferences for a given set of type parameters are made from a type S, in which those type parameters do not occur, to another type T, in which those type parameters do occur. Inferences consist of a set of candidate type arguments collected for each of the type parameters. The inference process recursively relates S and T to gather as many inferences as possible:
When comparing call or construct signatures, signatures in S correspond to signatures of the same kind in T pairwise in declaration order. If S and T have different numbers of a given kind of signature, the excess first signatures in declaration order of the longer list are ignored.
TODO: Update to reflect improved union and intersection type inference.
Classes and interfaces can reference themselves in their internal structure, in effect creating recursive types with infinite nesting. For example, the type
interface A { next: A; }
contains an infinitely nested sequence of ’next’ properties. Types such as this are perfectly valid but require special treatment when determining type relationships. Specifically, when comparing types S and T for a given relationship (identity, subtype, or assignability), the relationship in question is assumed to be true for every directly or indirectly nested occurrence of the same S and the same T (where same means originating in the same declaration and, if applicable, having identical type arguments). For example, consider the identity relationship between ‘A’ above and ‘B’ below:
interface B { next: C; }
interface C { next: D; }
interface D { next: B; }
To determine whether ‘A’ and ‘B’ are identical, first the ’next’ properties of type ‘A’ and ‘C’ are compared. That leads to comparing the ’next’ properties of type ‘A’ and ‘D’, which leads to comparing the ’next’ properties of type ‘A’ and ‘B’. Since ‘A’ and ‘B’ are already being compared this relationship is by definition true. That in turn causes the other comparisons to be true, and therefore the final result is true.
When this same technique is used to compare generic type references, two type references are considered the same when they originate in the same declaration and have identical type arguments.
In certain circumstances, generic types that directly or indirectly reference themselves in a recursive fashion can lead to infinite series of distinct instantiations. For example, in the type
interface List<T> {
data: T;
next: List<T>;
owner: List<List<T>>;
}
‘List<T>’ has a member ‘owner’ of type ‘List<List<T»’, which has a member ‘owner’ of type ‘List<List<List<T»>’, which has a member ‘owner’ of type ‘List<List<List<List<T»»’ and so on, ad infinitum. Since type relationships are determined structurally, possibly exploring the constituent types to their full depth, in order to determine type relationships involving infinitely expanding generic types it may be necessary for the compiler to terminate the recursion at some point with the assumption that no further exploration will change the outcome.
In several situations TypeScript infers types from context, alleviating the need for the programmer to explicitly specify types that appear obvious. For example
var name = "Steve";
infers the type of ’name’ to be the String primitive type since that is the type of the value used to initialize it. When inferring the type of a variable, property or function result from an expression, the widened form of the source type is used as the inferred type of the target. The widened form of a type is the type in which all occurrences of the Null and Undefined types have been replaced with the type any
.
The following example shows the results of widening types to produce inferred variable types.
var a = null; // var a: any
var b = undefined; // var b: any
var c = { x: 0, y: null }; // var c: { x: number, y: any }
var d = [ null, undefined ]; // var d: any[]
This chapter describes the manner in which TypeScript provides type inference and type checking for JavaScript expressions. TypeScript’s type analysis occurs entirely at compile-time and adds no run-time overhead to expression evaluation.
TypeScript’s typing rules define a type for every expression construct. For example, the type of the literal 123 is the Number primitive type, and the type of the object literal { a: 10, b: “hello” } is { a: number; b: string; }. The sections in this chapter describe these rules in detail.
In addition to type inference and type checking, TypeScript augments JavaScript expressions with the following constructs:
Unless otherwise noted in the sections that follow, TypeScript expressions and the JavaScript expressions generated from them are identical.
Expressions are classified as values or references. References are the subset of expressions that are permitted as the target of an assignment. Specifically, references are combinations of identifiers (section 4.3), parentheses (section 4.8), and property accesses (section 4.13). All other expression constructs described in this chapter are classified as values.
The type of this
in an expression depends on the location in which the reference takes place:
this
is of the this-type (section 3.6.3) of the containing class.this
is the constructor function type of the containing class.this
is of type Any.this
is of type Any.In all other contexts it is a compile-time error to reference this
.
Note that an arrow function (section 4.11) has no this
parameter but rather preserves the this
of its enclosing context.
When an expression is an IdentifierReference, the expression refers to the most nested namespace, class, enum, function, variable, or parameter with that name whose scope (section 2.4) includes the location of the reference. The type of such an expression is the type associated with the referenced entity:
An identifier expression that references a variable or parameter is classified as a reference. An identifier expression that references any other kind of entity is classified as a value (and therefore cannot be the target of an assignment).
Literals are typed as follows:
null
literal is the Null primitive type.true
and false
is the Boolean primitive type.Object literals are extended to support type annotations in methods and get and set accessors.
PropertyDefinition: ( Modified )
IdentifierReference
CoverInitializedName
PropertyName :
AssignmentExpression
PropertyName CallSignature {
FunctionBody }
GetAccessor
SetAccessor
GetAccessor:
get
PropertyName (
)
TypeAnnotationopt {
FunctionBody }
SetAccessor:
set
PropertyName (
BindingIdentifierOrPattern TypeAnnotationopt )
{
FunctionBody }
The type of an object literal is an object type with the set of properties specified by the property assignments in the object literal. A get and set accessor may specify the same property name, but otherwise it is an error to specify multiple property assignments for the same property.
A shorthand property assignment of the form
prop
is equivalent to
prop : prop
Likewise, a property assignment of the form
f ( ... ) { ... }
is equivalent to
f : function ( ... ) { ... }
Each property assignment in an object literal is processed as follows:
The type of a property introduced by a property assignment of the form Name :
Expr is the type of Expr.
A get accessor declaration is processed in the same manner as an ordinary function declaration (section 6.1) with no parameters. A set accessor declaration is processed in the same manner as an ordinary function declaration with a single parameter and a Void return type. When both a get and set accessor is declared for a property:
If a get accessor is declared for a property, the return type of the get accessor becomes the type of the property. If only a set accessor is declared for a property, the parameter type (which may be type Any if no type annotation is present) of the set accessor becomes the type of the property.
When an object literal is contextually typed by a type that includes a string index signature, the resulting type of the object literal includes a string index signature with the union type of the types of the properties declared in the object literal, or the Undefined type if the object literal is empty. Likewise, when an object literal is contextually typed by a type that includes a numeric index signature, the resulting type of the object literal includes a numeric index signature with the union type of the types of the numerically named properties (section 3.9.4) declared in the object literal, or the Undefined type if the object literal declares no numerically named properties.
If the PropertyName of a property assignment is a computed property name that doesn’t denote a well-known symbol (2.2.3), the construct is considered a dynamic property assignment. The following rules apply to dynamic property assignments:
An array literal
[ expr1, expr2, ..., exprN ]
denotes a value of an array type (section 3.3.2) or a tuple type (section 3.3.3) depending on context.
Each element expression in a non-empty array literal is processed as follows:
The resulting type an array literal expression is determined as follows:
A spread element must specify an expression of an array-like type (section 3.3.2), or otherwise an error occurs.
TODO: The compiler currently doesn’t support applying the spread operator to a string (to spread the individual characters of a string into a string array). This will eventually be allowed, but only when the code generation target is ECMAScript 2015 or later.
TODO: Document spreading an iterator into an array literal.
The rules above mean that an array literal is always of an array type, unless it is contextually typed by a tuple-like type. For example
var a = [1, 2]; // number[]
var b = ["hello", true]; // (string | boolean)[]
var c: [number, string] = [3, "three"]; // [number, string]
When the output target is ECMAScript 3 or 5, array literals containing spread elements are rewritten to invocations of the concat
method. For example, the assignments
var a = [2, 3, 4];
var b = [0, 1, ...a, 5, 6];
are rewritten to
var a = [2, 3, 4];
var b = [0, 1].concat(a, [5, 6]);
TODO: Template literals.
A parenthesized expression
( expr )
has the same type and classification as the contained expression itself. Specifically, if the contained expression is classified as a reference, so is the parenthesized expression.
The super
keyword can be used in expressions to reference base class properties and the base class constructor.
Super calls consist of the keyword super
followed by an argument list enclosed in parentheses. Super calls are only permitted in constructors of derived classes, as described in section 8.3.2.
A super call invokes the constructor of the base class on the instance referenced by this
. A super call is processed as a function call (section 4.15) using the construct signatures of the base class constructor function type as the initial set of candidate signatures for overload resolution. Type arguments cannot be explicitly specified in a super call. If the base class is a generic class, the type arguments used to process a super call are always those specified in the extends
clause that references the base class.
The type of a super call expression is Void.
The JavaScript code generated for a super call is specified in section 8.7.2.
A super property access consists of the keyword super
followed by a dot and an identifier. Super property accesses are used to access base class member functions from derived classes and are permitted in contexts where this
(section 4.2) references a derived class instance or a derived class constructor function. Specifically:
this
references a derived class instance, a super property access is permitted and must specify a public instance member function of the base class.this
references the constructor function object of a derived class, a super property access is permitted and must specify a public static member function of the base class.Super property accesses are not permitted in other contexts, and it is not possible to access other kinds of base class members in a super property access. Note that super property accesses are not permitted inside function expressions nested in the above constructs because this
is of type Any in such function expressions.
Super property accesses are typically used to access overridden base class member functions from derived class member functions. For an example of this, see section 8.4.2.
The JavaScript code generated for a super property access is specified in section 8.7.2.
TODO: Update section to include bracket notation in super property access.
Function expressions are extended from JavaScript to optionally include parameter and return type annotations.
FunctionExpression: ( Modified )
function
BindingIdentifieropt CallSignature {
FunctionBody }
The descriptions of function declarations provided in chapter 6 apply to function expressions as well, except that function expressions do not support overloading.
The type of a function expression is an object type containing a single call signature with parameter and return types inferred from the function expression’s signature and body.
When a function expression with no type parameters and no parameter type annotations is contextually typed (section 4.23) by a type T and a contextual signature S can be extracted from T, the function expression is processed as if it had explicitly specified parameter type annotations as they exist in S. Parameters are matched by position and need not have matching names. If the function expression has fewer parameters than S, the additional parameters in S are ignored. If the function expression has more parameters than S, the additional parameters are all considered to have type Any.
Likewise, when a function expression with no return type annotation is contextually typed (section 4.23) by a function type T and a contextual signature S can be extracted from T, expressions in contained return statements (section 5.10) are contextually typed by the return type of S.
A contextual signature S is extracted from a function type T as follows:
In the example
var f: (s: string) => string = function (s) {
return s.toLowerCase();
};
the function expression is contextually typed by the type of ‘f’, and since the function expression has no type parameters or type annotations its parameter type information is extracted from the contextual type, thus inferring the type of ’s’ to be the String primitive type.
Arrow functions are extended from JavaScript to optionally include parameter and return type annotations.
ArrowFormalParameters: ( Modified )
CallSignature
The descriptions of function declarations provided in chapter 6 apply to arrow functions as well, except that arrow functions do not support overloading.
The type of an arrow function is determined in the same manner as a function expression (section 4.10). Likewise, parameters of an arrow function and return statements in the body of an arrow function are contextually typed in the same manner as for function expressions.
When an arrow function with an expression body and no return type annotation is contextually typed (section 4.23) by a function type T and a contextual signature S can be extracted from T, the expression body is contextually typed by the return type of S.
An arrow function expression of the form
( ... ) => expr
is exactly equivalent to
( ... ) => { return expr ; }
Furthermore, arrow function expressions of the forms
id => { ... }
id => expr
are exactly equivalent to
( id ) => { ... }
( id ) => expr
Thus, the following examples are all equivalent:
(x) => { return Math.sin(x); }
(x) => Math.sin(x)
x => { return Math.sin(x); }
x => Math.sin(x)
A function expression introduces a new dynamically bound this
, whereas an arrow function expression preserves the this
of its enclosing context. Arrow function expressions are particularly useful for writing callbacks, which otherwise often have an undefined or unexpected this
.
In the example
class Messenger {
message = "Hello World";
start() {
setTimeout(() => alert(this.message), 3000);
}
};
var messenger = new Messenger();
messenger.start();
the use of an arrow function expression causes the callback to have the same this
as the surrounding ‘start’ method. Writing the callback as a standard function expression it becomes necessary to manually arrange access to the surrounding this
, for example by copying it into a local variable:
class Messenger {
message = "Hello World";
start() {
var _this = this;
setTimeout(function() { alert(_this.message); }, 3000);
}
};
var messenger = new Messenger();
messenger.start();
The TypeScript compiler applies this type of transformation to rewrite arrow function expressions into standard function expressions.
A construct of the form
< T > ( ... ) => { ... }
could be parsed as an arrow function expression with a type parameter or a type assertion applied to an arrow function with no type parameter. It is resolved as the former, but parentheses can be used to select the latter meaning:
< T > ( ( ... ) => { ... } )
TODO: Document class expressions.
A property access uses either dot notation or bracket notation. A property access expression is always classified as a reference.
A dot notation property access of the form
object . name
where object is an expression and name is an identifier (including, possibly, a reserved word), is used to access the property with the given name on the given object. A dot notation property access is processed as follows at compile-time:
A bracket notation property access of the form
object [ index ]
where object and index are expressions, is used to access the property with the name computed by the index expression on the given object. A bracket notation property access is processed as follows at compile-time:
TODO: Indexing with symbols.
The rules above mean that properties are strongly typed when accessed using bracket notation with the literal representation of their name. For example:
var type = {
name: "boolean",
primitive: true
};
var s = type["name"]; // string
var b = type["primitive"]; // boolean
Tuple types assign numeric names to each of their elements and elements are therefore strongly typed when accessed using bracket notation with a numeric literal:
var data: [string, number] = ["five", 5];
var s = data[0]; // string
var n = data[1]; // number
A new
operation has one of the following forms:
new C
new C ( ... )
new C < ... > ( ... )
where C is an expression. The first form is equivalent to supplying an empty argument list. C must be of type Any or of an object type with one or more construct or call signatures. The operation is processed as follows at compile-time:
Function calls are extended from JavaScript to support optional type arguments.
Arguments: ( Modified )
TypeArgumentsopt (
ArgumentListopt )
A function call takes one of the forms
func ( ... )
func < ... > ( ... )
where func is an expression of a function type or of type Any. The function expression is followed by an optional type argument list (section 3.6.2) and an argument list.
If func is of type Any, or of an object type that has no call or construct signatures but is a subtype of the Function interface, the call is an untyped function call. In an untyped function call no type arguments are permitted, argument expressions can be of any type and number, no contextual types are provided for the argument expressions, and the result is always of type Any.
If func has apparent call signatures (section 3.11.1) the call is a typed function call. TypeScript employs overload resolution in typed function calls in order to support functions with multiple call signatures. Furthermore, TypeScript may perform type argument inference to automatically determine type arguments in generic function calls.
The purpose of overload resolution in a function call is to ensure that at least one signature is applicable, to provide contextual types for the arguments, and to determine the result type of the function call, which could differ between the multiple applicable signatures. Overload resolution has no impact on the run-time behavior of a function call. Since JavaScript doesn’t support function overloading, all that matters at run-time is the name of the function.
TODO: Describe use of wildcard function types in overload resolution.
The compile-time processing of a typed function call consists of the following steps:
extends
clause order.
A signature is said to be an applicable signature with respect to an argument list when
TODO: Spread operator in function calls and spreading an iterator into a function call.
Given a signature < T1 , T2 , … , Tn > ( p1 : P1 , p2 : P2 , … , pm : Pm ), where each parameter type P references zero or more of the type parameters T, and an argument list ( e1 , e2 , … , em ), the task of type argument inference is to find a set of type arguments A1…An to substitute for T1…Tn such that the argument list becomes an applicable signature.
TODO: Update type argument inference and overload resolution rules.
Type argument inference produces a set of candidate types for each type parameter. Given a type parameter T and set of candidate types, the actual inferred type argument is determined as follows:
In order to compute candidate types, the argument list is processed as follows:
The process of inferentially typing an expression e by a type T is the same as that of contextually typing e by T, with the following exceptions:
An example:
function choose<T>(x: T, y: T): T {
return Math.random() < 0.5 ? x : y;
}
var x = choose(10, 20); // Ok, x of type number
var y = choose("Five", 5); // Error
In the first call to ‘choose’, two inferences are made from ’number’ to ‘T’, one for each parameter. Thus, ’number’ is inferred for ‘T’ and the call is equivalent to
var x = choose<number>(10, 20);
In the second call to ‘choose’, an inference is made from type ‘string’ to ‘T’ for the first parameter and an inference is made from type ’number’ to ‘T’ for the second parameter. Since neither ‘string’ nor ’number’ is a supertype of the other, type inference fails. That in turn means there are no applicable signatures and the function call is an error.
In the example
function map<T, U>(a: T[], f: (x: T) => U): U[] {
var result: U[] = [];
for (var i = 0; i < a.length; i++) result.push(f(a[i]));
return result;
}
var names = ["Peter", "Paul", "Mary"];
var lengths = map(names, s => s.length);
inferences for ‘T’ and ‘U’ in the call to ‘map’ are made as follows: For the first parameter, inferences are made from the type ‘string[]’ (the type of ’names’) to the type ‘T[]’, inferring ‘string’ for ‘T’. For the second parameter, inferential typing of the arrow expression ’s => s.length’ causes ‘T’ to become fixed such that the inferred type ‘string’ can be used for the parameter ’s’. The return type of the arrow expression can then be determined, and inferences are made from the type ‘(s: string) => number’ to the type ‘(x: T) => U’, inferring ’number’ for ‘U’. Thus the call to ‘map’ is equivalent to
var lengths = map<string, number>(names, s => s.length);
and the resulting type of ’lengths’ is therefore ’number[]’.
In the example
function zip<S, T, U>(x: S[], y: T[], combine: (x: S) => (y: T) => U): U[] {
var len = Math.max(x.length, y.length);
var result: U[] = [];
for (var i = 0; i < len; i++) result.push(combine(x[i])(y[i]));
return result;
}
var names = ["Peter", "Paul", "Mary"];
var ages = [7, 9, 12];
var pairs = zip(names, ages, s => n => ({ name: s, age: n }));
inferences for ‘S’, ‘T’ and ‘U’ in the call to ‘zip’ are made as follows: Using the first two parameters, inferences of ‘string’ for ‘S’ and ’number’ for ‘T’ are made. For the third parameter, inferential typing of the outer arrow expression causes ‘S’ to become fixed such that the inferred type ‘string’ can be used for the parameter ’s’. When a function expression is inferentially typed, its return expression(s) are also inferentially typed. Thus, the inner arrow function is inferentially typed, causing ‘T’ to become fixed such that the inferred type ’number’ can be used for the parameter ’n’. The return type of the inner arrow function can then be determined, which in turn determines the return type of the function returned from the outer arrow function, and inferences are made from the type ‘(s: string) => (n: number) => { name: string; age: number }’ to the type ‘(x: S) => (y: T) => R’, inferring ‘{ name: string; age: number }’ for ‘R’. Thus the call to ‘zip’ is equivalent to
var pairs = zip<string, number, { name: string; age: number }>(
names, ages, s => n => ({ name: s, age: n }));
and the resulting type of ‘pairs’ is therefore ‘{ name: string; age: number }[]’.
The inclusion of type arguments in the Arguments production (section 4.15) gives rise to certain ambiguities in the grammar for expressions. For example, the statement
f(g<A, B>(7));
could be interpreted as a call to ‘f’ with two arguments, ‘g < A’ and ‘B > (7)’. Alternatively, it could be interpreted as a call to ‘f’ with one argument, which is a call to a generic function ‘g’ with two type arguments and one regular argument.
The grammar ambiguity is resolved as follows: In a context where one possible interpretation of a sequence of tokens is an Arguments production, if the initial sequence of tokens forms a syntactically correct TypeArguments production and is followed by a ‘(
’ token, then the sequence of tokens is processed an Arguments production, and any other possible interpretation is discarded. Otherwise, the sequence of tokens is not considered an Arguments production.
This rule means that the call to ‘f’ above is interpreted as a call with one argument, which is a call to a generic function ‘g’ with two type arguments and one regular argument. However, the statements
f(g < A, B > 7);
f(g < A, B > +(7));
are both interpreted as calls to ‘f’ with two arguments.
TypeScript extends the JavaScript expression grammar with the ability to assert a type for an expression:
UnaryExpression: ( Modified )
…
<
Type >
UnaryExpression
A type assertion expression consists of a type enclosed in <
and >
followed by a unary expression. Type assertion expressions are purely a compile-time construct. Type assertions are not checked at run-time and have no impact on the emitted JavaScript (and therefore no run-time cost). The type and the enclosing <
and >
are simply removed from the generated code.
In a type assertion expression of the form < T > e, e is contextually typed (section 4.23) by T and the resulting type of* e* is required to be assignable to T, or T is required to be assignable to the widened form of the resulting type of e, or otherwise a compile-time error occurs. The type of the result is T.
Type assertions check for assignment compatibility in both directions. Thus, type assertions allow type conversions that might be correct, but aren’t known to be correct. In the example
class Shape { ... }
class Circle extends Shape { ... }
function createShape(kind: string): Shape {
if (kind === "circle") return new Circle();
...
}
var circle = <Circle> createShape("circle");
the type annotations indicate that the ‘createShape’ function might return a ‘Circle’ (because ‘Circle’ is a subtype of ‘Shape’), but isn’t known to do so (because its return type is ‘Shape’). Therefore, a type assertion is needed to treat the result as a ‘Circle’.
As mentioned above, type assertions are not checked at run-time and it is up to the programmer to guard against errors, for example using the instanceof
operator:
var shape = createShape(shapeKind);
if (shape instanceof Circle) {
var circle = <Circle> shape;
...
}
TODO: Document as operator.
TODO: Document JSX expressions.
The subsections that follow specify the compile-time processing rules of the unary operators. In general, if the operand of a unary operator does not meet the stated requirements, a compile-time error occurs and the result of the operation defaults to type Any in further processing.
These operators, in prefix or postfix form, require their operand to be of type Any, the Number primitive type, or an enum type, and classified as a reference (section 4.1). They produce a result of the Number primitive type.
These operators permit their operand to be of any type and produce a result of the Number primitive type.
The unary + operator can conveniently be used to convert a value of any type to the Number primitive type:
function getValue() { ... }
var n = +getValue();
The example above converts the result of ‘getValue()’ to a number if it isn’t a number already. The type inferred for ’n’ is the Number primitive type regardless of the return type of ‘getValue’.
The ! operator permits its operand to be of any type and produces a result of the Boolean primitive type.
Two unary ! operators in sequence can conveniently be used to convert a value of any type to the Boolean primitive type:
function getValue() { ... }
var b = !!getValue();
The example above converts the result of ‘getValue()’ to a Boolean if it isn’t a Boolean already. The type inferred for ‘b’ is the Boolean primitive type regardless of the return type of ‘getValue’.
The ‘delete’ operator takes an operand of any type and produces a result of the Boolean primitive type.
The ‘void’ operator takes an operand of any type and produces the value ‘undefined’. The type of the result is the Undefined type (3.2.7).
The ’typeof’ operator takes an operand of any type and produces a value of the String primitive type. In positions where a type is expected, ’typeof’ can also be used in a type query (section 3.8.10) to produce the type of an expression.
var x = 5;
var y = typeof x; // Use in an expression
var z: typeof x; // Use in a type query
In the example above, ‘x’ is of type ’number’, ‘y’ is of type ‘string’ because when used in an expression, ’typeof’ produces a value of type string (in this case the string “number”), and ‘z’ is of type ’number’ because when used in a type query, ’typeof’ obtains the type of an expression.
The subsections that follow specify the compile-time processing rules of the binary operators. In general, if the operands of a binary operator do not meet the stated requirements, a compile-time error occurs and the result of the operation defaults to type any in further processing. Tables that summarize the compile-time processing rules for operands of the Any type, the Boolean, Number, and String primitive types, and all other types (the Other column in the tables) are provided.
These operators require their operands to be of type Any, the Number primitive type, or an enum type. Operands of an enum type are treated as having the primitive type Number. If one operand is the null
or undefined
value, it is treated as having the type of the other operand. The result is always of the Number primitive type.
Any | Boolean | Number | String | Other | |
---|---|---|---|---|---|
Any | Number | Number | |||
Boolean | |||||
Number | Number | Number | |||
String | |||||
Other |
TODO: Document the exponentation operator.
The binary + operator requires both operands to be of the Number primitive type or an enum type, or at least one of the operands to be of type Any or the String primitive type. Operands of an enum type are treated as having the primitive type Number. If one operand is the null
or undefined
value, it is treated as having the type of the other operand. If both operands are of the Number primitive type, the result is of the Number primitive type. If one or both operands are of the String primitive type, the result is of the String primitive type. Otherwise, the result is of type Any.
Any | Boolean | Number | String | Other | |
---|---|---|---|---|---|
Any | Any | Any | Any | String | Any |
Boolean | Any | String | |||
Number | Any | Number | String | ||
String | String | String | String | String | String |
Other | Any | String |
A value of any type can converted to the String primitive type by adding an empty string:
function getValue() { ... }
var s = getValue() + "";
The example above converts the result of ‘getValue()’ to a string if it isn’t a string already. The type inferred for ’s’ is the String primitive type regardless of the return type of ‘getValue’.
These operators require one or both of the operand types to be assignable to the other. The result is always of the Boolean primitive type.
Any | Boolean | Number | String | Other | |
---|---|---|---|---|---|
Any | Boolean | Boolean | Boolean | Boolean | Boolean |
Boolean | Boolean | Boolean | |||
Number | Boolean | Boolean | |||
String | Boolean | Boolean | |||
Other | Boolean | Boolean |
The instanceof
operator requires the left operand to be of type Any, an object type, or a type parameter type, and the right operand to be of type Any or a subtype of the ‘Function’ interface type. The result is always of the Boolean primitive type.
Note that object types containing one or more call or construct signatures are automatically subtypes of the ‘Function’ interface type, as described in section 3.3.
The in
operator requires the left operand to be of type Any, the String primitive type, or the Number primitive type, and the right operand to be of type Any, an object type, or a type parameter type. The result is always of the Boolean primitive type.
The && operator permits the operands to be of any type and produces a result of the same type as the second operand.
Any | Boolean | Number | String | Other | |
---|---|---|---|---|---|
Any | Any | Boolean | Number | String | Other |
Boolean | Any | Boolean | Number | String | Other |
Number | Any | Boolean | Number | String | Other |
String | Any | Boolean | Number | String | Other |
Other | Any | Boolean | Number | String | Other |
The || operator permits the operands to be of any type.
If the || expression is contextually typed (section 4.23), the operands are contextually typed by the same type. Otherwise, the left operand is not contextually typed and the right operand is contextually typed by the type of the left operand.
The type of the result is the union type of the two operand types.
Any | Boolean | Number | String | Other | |
---|---|---|---|---|---|
Any | Any | Any | Any | Any | Any |
Boolean | Any | Boolean | N | B | S |
Number | Any | N | B | Number | S |
String | Any | S | B | S | N |
Other | Any | B | O | N | O |
In a conditional expression of the form
test ? expr1 : expr2
the test expression may be of any type.
If the conditional expression is contextually typed (section 4.23), expr1 and expr2 are contextually typed by the same type. Otherwise, expr1 and expr2 are not contextually typed.
The type of the result is the union type of the types of expr1 and expr2.
An assignment of the form
v = expr
requires v to be classified as a reference (section 4.1) or as an assignment pattern (section 4.21.1). The expr expression is contextually typed (section 4.23) by the type of v, and the type of expr must be assignable to (section 3.11.4) the type of v, or otherwise a compile-time error occurs. The result is a value with the type of expr.
A compound assignment of the form
v ??= expr
where ??= is one of the compound assignment operators
*= /= %= += -= <<= >>= >>>= &= ^= |=
is subject to the same requirements, and produces a value of the same type, as the corresponding non-compound operation. A compound assignment furthermore requires v to be classified as a reference (section 4.1) and the type of the non-compound operation to be assignable to the type of v. Note that v is not permitted to be an assignment pattern in a compound assignment.
A destructuring assignment is an assignment operation in which the left hand operand is a destructuring assignment pattern as defined by the AssignmentPattern production in the ECMAScript 2015 specification.
In a destructuring assignment expression, the type of the expression on the right must be assignable to the assignment target on the left. An expression of type S is considered assignable to an assignment target V if one of the following is true:
TODO: Update to specify behavior when assignment element E is a rest element.
In an assignment property or element that includes a default value, the type of the default value must be assignable to the target given in the assignment property or element.
When the output target is ECMAScript 2015 or higher, destructuring variable assignments remain unchanged in the emitted JavaScript code. When the output target is ECMAScript 3 or 5, destructuring variable assignments are rewritten to series of simple assignments. For example, the destructuring assignment
var x = 1;
var y = 2;
[x, y] = [y, x];
is rewritten to the simple variable assignments
var x = 1;
var y = 2;
_a = [y, x], x = _a[0], y = _a[1];
var _a;
The comma operator permits the operands to be of any type and produces a result that is of the same type as the second operand.
Type checking of an expression is improved in several contexts by factoring in the type of the destination of the value computed by the expression. In such situations, the expression is said to be contextually typed by the type of the destination. An expression is contextually typed in the following circumstances:
In the following example
interface EventObject {
x: number;
y: number;
}
interface EventHandlers {
mousedown?: (event: EventObject) => void;
mouseup?: (event: EventObject) => void;
mousemove?: (event: EventObject) => void;
}
function setEventHandlers(handlers: EventHandlers) { ... }
setEventHandlers({
mousedown: e => { startTracking(e.x, e.y); },
mouseup: e => { endTracking(); }
});
the object literal passed to ‘setEventHandlers’ is contextually typed to the ‘EventHandlers’ type. This causes the two property assignments to be contextually typed to the unnamed function type ‘(event: EventObject) => void’, which in turn causes the ’e’ parameters in the arrow function expressions to automatically be typed as ‘EventObject’.
Type guards are particular expression patterns involving the ’typeof’ and ‘instanceof’ operators that cause the types of variables or parameters to be narrowed to more specific types. For example, in the code below, knowledge of the static type of ‘x’ in combination with a ’typeof’ check makes it safe to narrow the type of ‘x’ to string in the first branch of the ‘if’ statement and number in the second branch of the ‘if’ statement.
function foo(x: number | string) {
if (typeof x === "string") {
return x.length; // x has type string here
}
else {
return x + 1; // x has type number here
}
}
The type of a variable or parameter is narrowed in the following situations:
A type guard is simply an expression that follows a particular pattern. The process of narrowing the type of a variable x by a type guard when true or when false depends on the type guard as follows:
x instanceof C
, where x is not of type Any, C is of a subtype of the global type ‘Function’, and C has a property named ‘prototype’
typeof x === s
, where s is a string literal with the value ‘string’, ’number’, or ‘boolean’,
typeof x === s
, where s is a string literal with any value but ‘string’, ’number’, or ‘boolean’,
typeof x !== s
, where s is a string literal,
typeof x === s
when false, ortypeof x === s
when true.!expr
expr1 && expr2
expr1 || expr2
In the rules above, when a narrowing operation would remove all constituent types from a union type, the operation has no effect on the union type.
Note that type guards affect types of variables and parameters only and have no effect on members of objects such as properties. Also note that it is possible to defeat a type guard by calling a function that changes the type of the guarded variable.
TODO: Document user defined type guard functions.
In the example
function isLongString(obj: any) {
return typeof obj === "string" && obj.length > 100;
}
the obj
parameter has type string
in the right operand of the && operator.
In the example
function processValue(value: number | (() => number)) {
var x = typeof value !== "number" ? value() : value;
// Process number in x
}
the value parameter has type () => number
in the first conditional expression and type number
in the second conditional expression, and the inferred type of x is number
.
In the example
function f(x: string | number | boolean) {
if (typeof x === "string" || typeof x === "number") {
var y = x; // Type of y is string | number
}
else {
var z = x; // Type of z is boolean
}
}
the type of x is string | number | boolean
in the left operand of the || operator, number | boolean
in the right operand of the || operator, string | number
in the first branch of the if statement, and boolean
in the second branch of the if statement.
In the example
class C {
data: string | string[];
getData() {
var data = this.data;
return typeof data === "string" ? data : data.join(" ");
}
}
the type of the data
variable is string
in the first conditional expression and string[]
in the second conditional expression, and the inferred type of getData
is string
. Note that the data
property must be copied to a local variable for the type guard to have an effect.
In the example
class NamedItem {
name: string;
}
function getName(obj: Object) {
return obj instanceof NamedItem ? obj.name : "unknown";
}
the type of obj
is narrowed to NamedItem
in the first conditional expression, and the inferred type of the getName
function is string
.
This chapter describes the static type checking TypeScript provides for JavaScript statements. TypeScript itself does not introduce any new statement constructs, but it does extend the grammar for local declarations to include interface, type alias, and enum declarations.
Blocks are extended to include local interface, type alias, and enum declarations (classes are already included by the ECMAScript 2015 grammar).
Declaration: ( Modified )
…
InterfaceDeclaration
TypeAliasDeclaration
EnumDeclaration
Local class, interface, type alias, and enum declarations are block scoped, similar to let and const declarations.
Variable statements are extended to include optional type annotations.
VariableDeclaration: ( Modified )
SimpleVariableDeclaration
DestructuringVariableDeclaration
A variable declaration is either a simple variable declaration or a destructuring variable declaration.
A simple variable declaration introduces a single named variable and optionally assigns it an initial value.
SimpleVariableDeclaration:
BindingIdentifier TypeAnnotationopt Initializeropt
The type T of a variable introduced by a simple variable declaration is determined as follows:
When a variable declaration specifies both a type annotation and an initializer expression, the type of the initializer expression is required to be assignable to (section 3.11.4) the type given in the type annotation.
Multiple declarations for the same variable name in the same declaration space are permitted, provided that each declaration associates the same type with the variable.
When a variable declaration has a type annotation, it is an error for that type annotation to use the typeof
operator to reference the variable being declared.
Below are some examples of simple variable declarations and their associated types.
var a; // any
var b: number; // number
var c = 1; // number
var d = { x: 1, y: "hello" }; // { x: number; y: string; }
var e: any = "test"; // any
The following is permitted because all declarations of the single variable ‘x’ associate the same type (Number) with ‘x’.
var x = 1;
var x: number;
if (x == 1) {
var x = 2;
}
In the following example, all five variables are of the same type, ‘{ x: number; y: number; }’.
interface Point { x: number; y: number; }
var a = { x: 0, y: <number> undefined };
var b: Point = { x: 0, y: undefined };
var c = <Point> { x: 0, y: undefined };
var d: { x: number; y: number; } = { x: 0, y: undefined };
var e = <{ x: number; y: number; }> { x: 0, y: undefined };
A destructuring variable declaration introduces zero or more named variables and initializes them with values extracted from properties of an object or elements of an array.
DestructuringVariableDeclaration:
BindingPattern TypeAnnotationopt Initializer
Each binding property or element that specifies an identifier introduces a variable by that name. The type of the variable is the widened form (section 3.12) of the type associated with the binding property or element, as defined in the following.
TODO: Document destructuring an iterator into an array.
The type T associated with a destructuring variable declaration is determined as follows:
The type T associated with a binding property is determined as follows:
The type T associated with a binding element is determined as follows:
When a destructuring variable declaration, binding property, or binding element specifies an initializer expression, the type of the initializer expression is required to be assignable to the widened form of the type associated with the destructuring variable declaration, binding property, or binding element.
TODO: Update rules to reflect improved checking of destructuring with literal initializers.
When the output target is ECMAScript 2015 or higher, except for removing the optional type annotation, destructuring variable declarations remain unchanged in the emitted JavaScript code.
When the output target is ECMAScript 3 or 5, destructuring variable declarations are rewritten to simple variable declarations. For example, an object destructuring declaration of the form
var { x, p: y, q: z = false } = getSomeObject();
is rewritten to the simple variable declarations
var _a = getSomeObject(),
x = _a.x,
y = _a.p,
_b = _a.q,
z = _b === void 0 ? false : _b;
The ‘_a’ and ‘_b’ temporary variables exist to ensure the assigned expression is evaluated only once, and the expression ‘void 0’ simply denotes the JavaScript value ‘undefined’.
Similarly, an array destructuring declaration of the form
var [x, y, z = 10] = getSomeArray();
is rewritten to the simple variable declarations
var _a = getSomeArray(),
x = _a[0],
y = _a[1],
_b = _a[2],
z = _b === void 0 ? 10 : _b;
Combining both forms of destructuring, the example
var { x, p: [y, z = 10] = getSomeArray() } = getSomeObject();
is rewritten to
var _a = getSomeObject(),
x = _a.x,
_b = _a.p,
_c = _b === void 0 ? getSomeArray() : _b,
y = _c[0],
_d = _c[1],
z = _d === void 0 ? 10 : _d;
A variable, parameter, binding property, or binding element declaration that specifies a binding pattern has an implied type which is determined as follows:
The implied type of a binding property or binding element declaration is
In the example
function f({ a, b = "hello", c = 1 }) { ... }
the implied type of the binding pattern in the function’s parameter is ‘{ a: any; b?: string; c?: number; }’. Since the parameter has no type annotation, this becomes the type of the parameter.
In the example
var [a, b, c] = [1, "hello", true];
the array literal initializer expression is contextually typed by the implied type of the binding pattern, specifically the tuple type ‘[any, any, any]’. Because the contextual type is a tuple type, the resulting type of the array literal is the tuple type ‘[number, string, boolean]’, and the destructuring declaration thus gives the types number, string, and boolean to a, b, and c respectively.
Let and const declarations are exended to include optional type annotations.
LexicalBinding: ( Modified )
SimpleLexicalBinding
DestructuringLexicalBinding
SimpleLexicalBinding:
BindingIdentifier TypeAnnotationopt Initializeropt
DestructuringLexicalBinding:
BindingPattern TypeAnnotationopt Initializeropt
TODO: Document scoping and types of let and const declarations.
Expressions controlling ‘if’, ‘do’, and ‘while’ statements can be of any type (and not just type Boolean).
Variable declarations in ‘for’ statements are extended in the same manner as variable declarations in variable statements (section 5.2).
In a ‘for-in’ statement of the form
for (v in expr) statement
v must be an expression classified as a reference of type Any or the String primitive type, and expr must be an expression of type Any, an object type, or a type parameter type.
In a ‘for-in’ statement of the form
for (var v in expr) statement
v must be a variable declaration without a type annotation that declares a variable of type Any, and expr must be an expression of type Any, an object type, or a type parameter type.
TODO: Document for-of statements.
A ‘continue’ statement is required to be nested, directly or indirectly (but not crossing function boundaries), within an iteration (‘do’, ‘while’, ‘for’, or ‘for-in’) statement. When a ‘continue’ statement includes a target label, that target label must appear in the label set of an enclosing (but not crossing function boundaries) iteration statement.
A ‘break’ statement is required to be nested, directly or indirectly (but not crossing function boundaries), within an iteration (‘do’, ‘while’, ‘for’, or ‘for-in’) or ‘switch’ statement. When a ‘break’ statement includes a target label, that target label must appear in the label set of an enclosing (but not crossing function boundaries) statement.
It is an error for a ‘return’ statement to occur outside a function body. Specifically, ‘return’ statements are not permitted at the global level or in namespace bodies.
A ‘return’ statement without an expression returns the value ‘undefined’ and is permitted in the body of any function, regardless of the return type of the function.
When a ‘return’ statement includes an expression, if the containing function includes a return type annotation, the return expression is contextually typed (section 4.23) by that return type and must be of a type that is assignable to the return type. Otherwise, if the containing function is contextually typed by a type T, Expr is contextually typed by T’s return type.
In a function implementation without a return type annotation, the return type is inferred from the ‘return’ statements in the function body, as described in section 6.3.
In the example
function f(): (x: string) => number {
return s => s.length;
}
the arrow expression in the ‘return’ statement is contextually typed by the return type of ‘f’, thus giving type ‘string’ to ’s’.
Use of the ‘with’ statement in TypeScript is an error, as is the case in ECMAScript 5’s strict mode. Furthermore, within the body of a ‘with’ statement, TypeScript considers every identifier occurring in an expression (section 4.3) to be of the Any type regardless of its declared type. Because the ‘with’ statement puts a statically unknown set of identifiers in scope in front of those that are statically known, it is not possible to meaningfully assign a static type to any identifier.
In a ‘switch’ statement, each ‘case’ expression must be of a type that is assignable to or from (section 3.11.4) the type of the ‘switch’ expression.
The expression specified in a ’throw’ statement can be of any type.
The variable introduced by a ‘catch’ clause of a ’try’ statement is always of type Any. It is not possible to include a type annotation in a ‘catch’ clause.
TypeScript extends JavaScript functions to include type parameters, parameter and return type annotations, overloads, default parameter values, and rest parameters.
Function declarations are extended to permit the function body to be omitted in overload declarations.
FunctionDeclaration: ( Modified )
function
BindingIdentifieropt CallSignature {
FunctionBody }
function
BindingIdentifieropt CallSignature ;
A FunctionDeclaration introduces a named value of a function type in the containing declaration space. The BindingIdentifier is optional only when the function declaration occurs in an export default declaration (section 11.3.4.2).
Function declarations that specify a body are called function implementations and function declarations without a body are called function overloads. It is possible to specify multiple overloads for the same function (i.e. for the same name in the same declaration space), but a function can have at most one implementation. All declarations for the same function must specify the same set of modifiers (i.e. the same combination of declare
, export
, and default
).
When a function has overload declarations, the overloads determine the call signatures of the type given to the function object and the function implementation signature (if any) must be assignable to that type. Otherwise, the function implementation itself determines the call signature.
When a function has both overloads and an implementation, the overloads must precede the implementation and all of the declarations must be consecutive with no intervening grammatical elements.
Function overloads allow a more accurate specification of the patterns of invocation supported by a function than is possible with a single signature. The compile-time processing of a call to an overloaded function chooses the best candidate overload for the particular arguments and the return type of that overload becomes the result type the function call expression. Thus, using overloads it is possible to statically describe the manner in which a function’s return type varies based on its arguments. Overload resolution in function calls is described further in section 4.15.
Function overloads are purely a compile-time construct. They have no impact on the emitted JavaScript and thus no run-time cost.
The parameter list of a function overload cannot specify default values for parameters. In other words, an overload may use only the ?
form when specifying optional parameters.
The following is an example of a function with overloads.
function attr(name: string): string;
function attr(name: string, value: string): Accessor;
function attr(map: any): Accessor;
function attr(nameOrMap: any, value?: string): any {
if (nameOrMap && typeof nameOrMap === "string") {
// handle string case
}
else {
// handle map case
}
}
Note that each overload and the final implementation specify the same identifier. The type of the local variable ‘attr’ introduced by this declaration is
var attr: {
(name: string): string;
(name: string, value: string): Accessor;
(map: any): Accessor;
};
Note that the signature of the actual function implementation is not included in the type.
A function implementation without a return type annotation is said to be an implicitly typed function. The return type of an implicitly typed function f is inferred from its function body as follows:
In the example
function f(x: number) {
if (x <= 0) return x;
return g(x);
}
function g(x: number) {
return f(x - 1);
}
the inferred return type for ‘f’ and ‘g’ is Any because the functions reference themselves through a cycle with no return type annotations. Adding an explicit return type ’number’ to either breaks the cycle and causes the return type ’number’ to be inferred for the other.
An explicitly typed function whose return type isn’t the Void type, the Any type, or a union type containing the Void or Any type as a constituent must have at least one return statement somewhere in its body. An exception to this rule is if the function implementation consists of a single ’throw’ statement.
The type of ’this’ in a function implementation is the Any type.
In the signature of a function implementation, a parameter can be marked optional by following it with an initializer. When a parameter declaration includes both a type annotation and an initializer, the initializer expression is contextually typed (section 4.23) by the stated type and must be assignable to the stated type, or otherwise a compile-time error occurs. When a parameter declaration has no type annotation but includes an initializer, the type of the parameter is the widened form (section 3.12) of the type of the initializer expression.
Initializer expressions are evaluated in the scope of the function body but are not permitted to reference local variables and are only permitted to access parameters that are declared to the left of the parameter they initialize, unless the parameter reference occurs in a nested function expression.
When the output target is ECMAScript 3 or 5, for each parameter with an initializer, a statement that substitutes the default value for an omitted argument is included in the generated JavaScript, as described in section 6.6. The example
function strange(x: number, y = x * 2, z = x + y) {
return z;
}
generates JavaScript that is equivalent to
function strange(x, y, z) {
if (y === void 0) { y = x * 2; }
if (z === void 0) { z = x + y; }
return z;
}
In the example
var x = 1;
function f(a = x) {
var x = "hello";
}
the local variable ‘x’ is in scope in the parameter initializer (thus hiding the outer ‘x’), but it is an error to reference it because it will always be uninitialized at the time the parameter initializer is evaluated.
Parameter declarations can specify binding patterns (section 3.9.2.2) and are then called destructuring parameter declarations. Similar to a destructuring variable declaration (section 5.2.2), a destructuring parameter declaration introduces zero or more named locals and initializes them with values extracted from properties or elements of the object or array passed as an argument for the parameter.
The type of local introduced in a destructuring parameter declaration is determined in the same manner as a local introduced by a destructuring variable declaration, except the type T associated with a destructuring parameter declaration is determined as follows:
any[]
.any
.When the output target is ECMAScript 2015 or higher, except for removing the optional type annotation, destructuring parameter declarations remain unchanged in the emitted JavaScript code. When the output target is ECMAScript 3 or 5, destructuring parameter declarations are rewritten to local variable declarations.
The example
function drawText({ text = "", location: [x, y] = [0, 0], bold = false }) {
// Draw text
}
declares a function drawText
that takes a single parameter of the type
{ text?: string; location?: [number, number]; bold?: boolean; }
When the output target is ECMAScript 3 or 5, the function is rewritten to
function drawText(_a) {
var _b = _a.text,
text = _b === void 0 ? "" : _b,
_c = _a.location,
_d = _c === void 0 ? [0, 0] : _c,
x = _d[0],
y = _d[1],
_e = _a.bold,
bold = _e === void 0 ? false : _e;
// Draw text
}
Destructuring parameter declarations do not permit type annotations on the individual binding patterns, as such annotations would conflict with the already established meaning of colons in object literals. Type annotations must instead be written on the top-level parameter declaration. For example
interface DrawTextInfo {
text?: string;
location?: [number, number];
bold?: boolean;
}
function drawText({ text, location: [x, y], bold }: DrawTextInfo) {
// Draw text
}
A function implementation may include type parameters in its signature (section 3.9.2.1) and is then called a generic function. Type parameters provide a mechanism for expressing relationships between parameter and return types in call operations. Type parameters have no run-time representation—they are purely a compile-time construct.
Type parameters declared in the signature of a function implementation are in scope in the signature and body of that function implementation.
The following is an example of a generic function:
interface Comparable {
localeCompare(other: any): number;
}
function compare<T extends Comparable>(x: T, y: T): number {
if (x == null) return y == null ? 0 : -1;
if (y == null) return 1;
return x.localeCompare(y);
}
Note that the ‘x’ and ‘y’ parameters are known to be subtypes of the constraint ‘Comparable’ and therefore have a ‘compareTo’ member. This is described further in section 3.6.1.
The type arguments of a call to a generic function may be explicitly specified in a call operation or may, when possible, be inferred (section 4.15.2) from the types of the regular arguments in the call. In the example
class Person {
name: string;
localeCompare(other: Person) {
return compare(this.name, other.name);
}
}
the type argument to ‘compare’ is automatically inferred to be the String type because the two arguments are strings.
A function declaration generates JavaScript code that is equivalent to:
function <FunctionName>(<FunctionParameters>) {
<DefaultValueAssignments>
<FunctionStatements>
}
FunctionName is the name of the function (or nothing in the case of a function expression).
FunctionParameters is a comma separated list of the function’s parameter names.
DefaultValueAssignments is a sequence of default property value assignments, one for each parameter with a default value, in the order they are declared, of the form
if (<Parameter> === void 0) { <Parameter> = <Default>; }
where Parameter is the parameter name and Default is the default value expression.
FunctionStatements is the code generated for the statements specified in the function body.
TODO: Document generator functions.
TODO: Document asynchronous functions.
TODO: Document type guard functions, including this type predicates.
Interfaces provide the ability to name and parameterize object types and to compose existing named object types into new ones.
Interfaces have no run-time representation—they are purely a compile-time construct. Interfaces are particularly useful for documenting and validating the required shape of properties, objects passed as parameters, and objects returned from functions.
Because TypeScript has a structural type system, an interface type with a particular set of members is considered identical to, and can be substituted for, another interface type or object type literal with an identical set of members (see section 3.11.2).
Class declarations may reference interfaces in their implements clause to validate that they provide an implementation of the interfaces.
An interface declaration declares an interface type.
InterfaceDeclaration:
interface
BindingIdentifier TypeParametersopt InterfaceExtendsClauseopt ObjectType
InterfaceExtendsClause:
extends
ClassOrInterfaceTypeList
ClassOrInterfaceTypeList:
ClassOrInterfaceType
ClassOrInterfaceTypeList ,
ClassOrInterfaceType
ClassOrInterfaceType:
TypeReference
An InterfaceDeclaration introduces a named type (section 3.7) in the containing declaration space. The BindingIdentifier of an interface declaration may not be one of the predefined type names (section 3.8.1).
An interface may optionally have type parameters (section 3.6.1) that serve as placeholders for actual types to be provided when the interface is referenced in type references. An interface with type parameters is called a generic interface. The type parameters of a generic interface declaration are in scope in the entire declaration and may be referenced in the InterfaceExtendsClause and ObjectType body.
An interface can inherit from zero or more base types which are specified in the InterfaceExtendsClause. The base types must be type references to class or interface types.
An interface has the members specified in the ObjectType of its declaration and furthermore inherits all base type members that aren’t hidden by declarations in the interface:
The following constraints must be satisfied by an interface declaration or otherwise a compile-time error occurs:
An interface is permitted to inherit identical members from multiple base types and will in that case only contain one occurrence of each particular member.
Below is an example of two interfaces that contain properties with the same name but different types:
interface Mover {
move(): void;
getStatus(): { speed: number; };
}
interface Shaker {
shake(): void;
getStatus(): { frequency: number; };
}
An interface that extends ‘Mover’ and ‘Shaker’ must declare a new ‘getStatus’ property as it would otherwise inherit two ‘getStatus’ properties with different types. The new ‘getStatus’ property must be declared such that the resulting ‘MoverShaker’ is a subtype of both ‘Mover’ and ‘Shaker’:
interface MoverShaker extends Mover, Shaker {
getStatus(): { speed: number; frequency: number; };
}
Since function and constructor types are just object types containing call and construct signatures, interfaces can be used to declare named function and constructor types. For example:
interface StringComparer { (a: string, b: string): number; }
This declares type ‘StringComparer’ to be a function type taking two strings and returning a number.
Interfaces are “open-ended” and interface declarations with the same qualified name relative to a common root (as defined in section 2.3) contribute to a single interface.
When a generic interface has multiple declarations, all declarations must have identical type parameter lists, i.e. identical type parameter names with identical constraints in identical order.
In an interface with multiple declarations, the extends
clauses are merged into a single set of base types and the bodies of the interface declarations are merged into a single object type. Declaration merging produces a declaration order that corresponds to prepending the members of each interface declaration, in the order the members are written, to the combined list of members in the order of the interface declarations. Thus, members declared in the last interface declaration will appear first in the declaration order of the merged type.
For example, a sequence of declarations in this order:
interface Document {
createElement(tagName: any): Element;
}
interface Document {
createElement(tagName: string): HTMLElement;
}
interface Document {
createElement(tagName: "div"): HTMLDivElement;
createElement(tagName: "span"): HTMLSpanElement;
createElement(tagName: "canvas"): HTMLCanvasElement;
}
is equivalent to the following single declaration:
interface Document {
createElement(tagName: "div"): HTMLDivElement;
createElement(tagName: "span"): HTMLSpanElement;
createElement(tagName: "canvas"): HTMLCanvasElement;
createElement(tagName: string): HTMLElement;
createElement(tagName: any): Element;
}
Note that the members of the last interface declaration appear first in the merged declaration. Also note that the relative order of members declared in the same interface body is preserved.
TODO: Document class and interface declaration merging.
When an interface type extends a class type it inherits the members of the class but not their implementations. It is as if the interface had declared all of the members of the class without providing an implementation. Interfaces inherit even the private and protected members of a base class. When a class containing private or protected members is the base type of an interface type, that interface type can only be implemented by that class or a descendant class. For example:
class Control {
private state: any;
}
interface SelectableControl extends Control {
select(): void;
}
class Button extends Control {
select() { }
}
class TextBox extends Control {
select() { }
}
class Image extends Control {
}
class Location {
select() { }
}
In the above example, ‘SelectableControl’ contains all of the members of ‘Control’, including the private ‘state’ property. Since ‘state’ is a private member it is only possible for descendants of ‘Control’ to implement ‘SelectableControl’. This is because only descendants of ‘Control’ will have a ‘state’ private member that originates in the same declaration, which is a requirement for private members to be compatible (section 3.11).
Within the ‘Control’ class it is possible to access the ‘state’ private member through an instance of ‘SelectableControl’. Effectively, a ‘SelectableControl’ acts like a ‘Control’ that is known to have a ‘select’ method. The ‘Button’ and ‘TextBox’ classes are subtypes of ‘SelectableControl’ (because they both inherit from ‘Control’ and have a ‘select’ method), but the ‘Image’ and ‘Location’ classes are not.
TypeScript does not provide a direct mechanism for dynamically testing whether an object implements a particular interface. Instead, TypeScript code can use the JavaScript technique of checking whether an appropriate set of members are present on the object. For example, given the declarations in section 7.1, the following is a dynamic check for the ‘MoverShaker’ interface:
var obj: any = getSomeObject();
if (obj && obj.move && obj.shake && obj.getStatus) {
var moverShaker = <MoverShaker> obj;
...
}
If such a check is used often it can be abstracted into a function:
function asMoverShaker(obj: any): MoverShaker {
return obj && obj.move && obj.shake && obj.getStatus ? obj : null;
}
TypeScript extends JavaScript classes to include type parameters, implements clauses, accessibility modifiers, member variable declarations, and parameter property declarations in constructors.
TODO: Document abstract classes.
A class declaration declares a class type and a constructor function.
ClassDeclaration: ( Modified )
class
BindingIdentifieropt TypeParametersopt ClassHeritage {
ClassBody }
A ClassDeclaration introduces a named type (the class type) and a named value (the constructor function) in the containing declaration space. The class type is formed from the instance members declared in the class body and the instance members inherited from the base class. The constructor function is given an anonymous type formed from the constructor declaration, the static member declarations in the class body, and the static members inherited from the base class. The constructor function initializes and returns an instance of the class type.
The BindingIdentifier of a class declaration may not be one of the predefined type names (section 3.8.1). The BindingIdentifier is optional only when the class declaration occurs in an export default declaration (section 11.3.4.2).
A class may optionally have type parameters (section 3.6.1) that serve as placeholders for actual types to be provided when the class is referenced in type references. A class with type parameters is called a generic class. The type parameters of a generic class declaration are in scope in the entire declaration and may be referenced in the ClassHeritage and ClassBody.
The following example introduces both a named type called ‘Point’ (the class type) and a named value called ‘Point’ (the constructor function) in the containing declaration space.
class Point {
constructor(public x: number, public y: number) { }
public length() { return Math.sqrt(this.x * this.x + this.y * this.y); }
static origin = new Point(0, 0);
}
The named type ‘Point’ is exactly equivalent to
interface Point {
x: number;
y: number;
length(): number;
}
The named value ‘Point’ is a constructor function whose type corresponds to the declaration
var Point: {
new(x: number, y: number): Point;
origin: Point;
};
The context in which a class is referenced distinguishes between the class type and the constructor function. For example, in the assignment statement
var p: Point = new Point(10, 20);
the identifier ‘Point’ in the type annotation refers to the class type, whereas the identifier ‘Point’ in the new
expression refers to the constructor function object.
TODO: Update this section to reflect expressions in class extends clauses.
The heritage specification of a class consists of optional extends
and implements
clauses. The extends
clause specifies the base class of the class and the implements
clause specifies a set of interfaces for which to validate the class provides an implementation.
ClassHeritage: ( Modified )
ClassExtendsClauseopt ImplementsClauseopt
ClassExtendsClause:
extends
ClassType
ClassType:
TypeReference
ImplementsClause:
implements
ClassOrInterfaceTypeList
A class that includes an extends
clause is called a derived class, and the class specified in the extends
clause is called the base class of the derived class. When a class heritage specification omits the extends
clause, the class does not have a base class. However, as is the case with every object type, type references (section 3.3.1) to the class will appear to have the members of the global interface type named ‘Object’ unless those members are hidden by members with the same name in the class.
The following constraints must be satisfied by the class heritage specification or otherwise a compile-time error occurs:
extends
clause must denote a class type. Furthermore, the TypeName part of the type reference is required to be a reference to the class constructor function when evaluated as an expression.implements
clause.The following example illustrates a situation in which the first rule above would be violated:
class A { a: number; }
namespace Foo {
var A = 1;
class B extends A { b: string; }
}
When evaluated as an expression, the type reference ‘A’ in the extends
clause doesn’t reference the class constructor function of ‘A’ (instead it references the local variable ‘A’).
The only situation in which the last two constraints above are violated is when a class overrides one or more base class members with incompatible new members.
Note that because TypeScript has a structural type system, a class doesn’t need to explicitly state that it implements an interface—it suffices for the class to simply contain the appropriate set of instance members. The implements
clause of a class provides a mechanism to assert and validate that the class contains the appropriate sets of instance members, but otherwise it has no effect on the class type.
The class body consists of zero or more constructor or member declarations. Statements are not allowed in the body of a class—they must be placed in the constructor or in members.
ClassElement: ( Modified )
ConstructorDeclaration
PropertyMemberDeclaration
IndexMemberDeclaration
The body of class may optionally contain a single constructor declaration. Constructor declarations are described in section 8.3.
Member declarations are used to declare instance and static members of the class. Property member declarations are described in section 8.4 and index member declarations are described in section 8.5.
The members of a class consist of the members introduced through member declarations in the class body and the members inherited from the base class.
Members are either instance members or static members.
Instance members are members of the class type (section 8.2.4) and its associated this-type. Within constructors, instance member functions, and instance member accessors, the type of this
is the this-type (section 3.6.3) of the class.
Static members are declared using the static
modifier and are members of the constructor function type (section 8.2.5). Within static member functions and static member accessors, the type of this
is the constructor function type.
Class type parameters cannot be referenced in static member declarations.
Property members have either public, private, or protected accessibility. The default is public accessibility, but property member declarations may include a public
, private
, or protected
modifier to explicitly specify the desired accessibility.
Public property members can be accessed everywhere without restrictions.
Private property members can be accessed only within their declaring class. Specifically, a private member M declared in a class C can be accessed only within the class body of C.
Protected property members can be accessed only within their declaring class and classes derived from their declaring class, and a protected instance property member must be accessed through an instance of the enclosing class or a subclass thereof. Specifically, a protected member M declared in a class C can be accessed only within the class body of C or the class body of a class derived from C. Furthermore, when a protected instance member M is accessed in a property access E.
M within the body of a class D, the type of E is required to be D or a type that directly or indirectly has D as a base type, regardless of type arguments.
Private and protected accessibility is enforced only at compile-time and serves as no more than an indication of intent. Since JavaScript provides no mechanism to create private and protected properties on an object, it is not possible to enforce the private and protected modifiers in dynamic code at run-time. For example, private and protected accessibility can be defeated by changing an object’s static type to Any and accessing the member dynamically.
The following example demonstrates private and protected accessibility:
class A {
private x: number;
protected y: number;
static f(a: A, b: B) {
a.x = 1; // Ok
b.x = 1; // Ok
a.y = 1; // Ok
b.y = 1; // Ok
}
}
class B extends A {
static f(a: A, b: B) {
a.x = 1; // Error, x only accessible within A
b.x = 1; // Error, x only accessible within A
a.y = 1; // Error, y must be accessed through instance of B
b.y = 1; // Ok
}
}
In class ‘A’, the accesses to ‘x’ are permitted because ‘x’ is declared in ‘A’, and the accesses to ‘y’ are permitted because both take place through an instance of ‘A’ or a type derived from ‘A’. In class ‘B’, access to ‘x’ is not permitted, and the first access to ‘y’ is an error because it takes place through an instance of ‘A’, which is not derived from the enclosing class ‘B’.
A derived class inherits all members from its base class it doesn’t override. Inheritance means that a derived class implicitly contains all non-overridden members of the base class. Only public and protected property members can be overridden.
A property member in a derived class is said to override a property member in a base class when the derived class property member has the same name and kind (instance or static) as the base class property member. The type of an overriding property member must be assignable (section 3.11.4) to the type of the overridden property member, or otherwise a compile-time error occurs.
Base class instance member functions can be overridden by derived class instance member functions, but not by other kinds of members.
Base class instance member variables and accessors can be overridden by derived class instance member variables and accessors, but not by other kinds of members.
Base class static property members can be overridden by derived class static property members of any kind as long as the types are compatible, as described above.
An index member in a derived class is said to override an index member in a base class when the derived class index member is of the same index kind (string or numeric) as the base class index member. The type of an overriding index member must be assignable (section 3.11.4) to the type of the overridden index member, or otherwise a compile-time error occurs.
A class declaration declares a new named type (section 3.7) called a class type. Within the constructor and instance member functions of a class, the type of this
is the this-type (section 3.6.3) of that class type. The class type has the following members:
public
, private
, or protected
modifier.All instance property members (including those that are private or protected) of a class must satisfy the constraints implied by the index members of the class as specified in section 3.9.4.
In the example
class A {
public x: number;
public f() { }
public g(a: any) { return undefined; }
static s: string;
}
class B extends A {
public y: number;
public g(b: boolean) { return false; }
}
the class type of ‘A’ is equivalent to
interface A {
x: number;
f: () => void;
g: (a: any) => any;
}
and the class type of ‘B’ is equivalent to
interface B {
x: number;
y: number;
f: () => void;
g: (b: boolean) => boolean;
}
Note that static declarations in a class do not contribute to the class type—rather, static declarations introduce properties on the constructor function object. Also note that the declaration of ‘g’ in ‘B’ overrides the member inherited from ‘A’.
The type of the constructor function introduced by a class declaration is called the constructor function type. The constructor function type has the following members:
Every class automatically contains a static property member named ‘prototype’, the type of which is the containing class with type Any substituted for each type parameter.
The example
class Pair<T1, T2> {
constructor(public item1: T1, public item2: T2) { }
}
class TwoArrays<T> extends Pair<T[], T[]> { }
introduces two named types corresponding to
interface Pair<T1, T2> {
item1: T1;
item2: T2;
}
interface TwoArrays<T> {
item1: T[];
item2: T[];
}
and two constructor functions corresponding to
var Pair: {
new <T1, T2>(item1: T1, item2: T2): Pair<T1, T2>;
}
var TwoArrays: {
new <T>(item1: T[], item2: T[]): TwoArrays<T>;
}
Note that each construct signature in the constructor function types has the same type parameters as its class and returns an instantiation of its class with those type parameters passed as type arguments. Also note that when a derived class doesn’t declare a constructor, type arguments from the base class reference are substituted before construct signatures are propagated from the base constructor function type to the derived constructor function type.
A constructor declaration declares the constructor function of a class.
ConstructorDeclaration:
AccessibilityModifieropt constructor
(
ParameterListopt )
{
FunctionBody }
AccessibilityModifieropt constructor
(
ParameterListopt )
;
Constructor declarations that specify a body are called constructor implementations and constructor declarations without a body are called constructor overloads. It is possible to specify multiple constructor overloads in a class, but a class can have at most one constructor implementation. All constructor declarations in a class must specify the same set of modifiers. Only public constructors are supported and private or protected constructors result in an error.
In a class with no constructor declaration, an automatic constructor is provided, as described in section 8.3.3.
When a class has constructor overloads, the overloads determine the construct signatures of the type given to the constructor function object, and the constructor implementation signature (if any) must be assignable to that type. Otherwise, the constructor implementation itself determines the construct signature. This exactly parallels the way overloads are processed in a function declaration (section 6.2).
When a class has both constructor overloads and a constructor implementation, the overloads must precede the implementation and all of the declarations must be consecutive with no intervening grammatical elements.
The function body of a constructor is permitted to contain return statements. If return statements specify expressions, those expressions must be of types that are assignable to the this-type (section 3.6.3) of the class.
The type parameters of a generic class are in scope and accessible in a constructor declaration.
Similar to functions, only the constructor implementation (and not constructor overloads) can specify default value expressions for optional parameters. It is a compile-time error for such default value expressions to reference this
. When the output target is ECMAScript 3 or 5, for each parameter with a default value, a statement that substitutes the default value for an omitted argument is included in the JavaScript generated for the constructor function.
A parameter of a ConstructorImplementation may be prefixed with a public
, private
, or protected
modifier. This is called a parameter property declaration and is shorthand for declaring a property with the same name as the parameter and initializing it with the value of the parameter. For example, the declaration
class Point {
constructor(public x: number, public y: number) {
// Constructor body
}
}
is equivalent to writing
class Point {
public x: number;
public y: number;
constructor(x: number, y: number) {
this.x = x;
this.y = y;
// Constructor body
}
}
A parameter property declaration may declare an optional parameter (by including a question mark or a default value), but the property introduced by such a declaration is always considered a required property (section 3.3.6).
Super calls (section 4.9.1) are used to call the constructor of the base class. A super call consists of the keyword super
followed by an argument list enclosed in parentheses. For example:
class ColoredPoint extends Point {
constructor(x: number, y: number, public color: string) {
super(x, y);
}
}
Constructors of classes with no extends
clause may not contain super calls, whereas constructors of derived classes must contain at least one super call somewhere in their function body. Super calls are not permitted outside constructors or in local functions inside constructors.
The first statement in the body of a constructor must be a super call if both of the following are true:
In such a required super call, it is a compile-time error for argument expressions to reference this
.
Initialization of parameter properties and instance member variables with initializers takes place immediately at the beginning of the constructor body if the class has no base class, or immediately following the super call if the class is a derived class.
If a class omits a constructor declaration, an automatic constructor is provided.
In a class with no extends
clause, the automatic constructor has no parameters and performs no action other than executing the instance member variable initializers (section 8.4.1), if any.
In a derived class, the automatic constructor has the same parameter list (and possibly overloads) as the base class constructor. The automatically provided constructor first forwards the call to the base class constructor using a call equivalent to
BaseClass.apply(this, arguments);
and then executes the instance member variable initializers, if any.
Property member declarations can be member variable declarations, member function declarations, or member accessor declarations.
PropertyMemberDeclaration:
MemberVariableDeclaration
MemberFunctionDeclaration
MemberAccessorDeclaration
Member declarations without a static
modifier are called instance member declarations. Instance property member declarations declare properties in the class type (section 8.2.4), and must specify names that are unique among all instance property member and parameter property declarations in the containing class, with the exception that instance get and set accessor declarations may pairwise specify the same name.
Member declarations with a static
modifier are called static member declarations. Static property member declarations declare properties in the constructor function type (section 8.2.5), and must specify names that are unique among all static property member declarations in the containing class, with the exception that static get and set accessor declarations may pairwise specify the same name.
Note that the declaration spaces of instance and static property members are separate. Thus, it is possible to have instance and static property members with the same name.
Except for overrides, as described in section 8.2.3, it is an error for a derived class to declare a property member with the same name and kind (instance or static) as a base class member.
Every class automatically contains a static property member named ‘prototype’, the type of which is an instantiation of the class type with type Any supplied as a type argument for each type parameter. It is an error to explicitly declare a static property member with the name ‘prototype’.
Below is an example of a class containing both instance and static property member declarations:
class Point {
constructor(public x: number, public y: number) { }
public distance(p: Point) {
var dx = this.x - p.x;
var dy = this.y - p.y;
return Math.sqrt(dx * dx + dy * dy);
}
static origin = new Point(0, 0);
static distance(p1: Point, p2: Point) { return p1.distance(p2); }
}
The class type ‘Point’ has the members:
interface Point {
x: number;
y: number;
distance(p: Point);
}
and the constructor function ‘Point’ has a type corresponding to the declaration:
var Point: {
new(x: number, y: number): Point;
origin: Point;
distance(p1: Point, p2: Point): number;
}
A member variable declaration declares an instance member variable or a static member variable.
MemberVariableDeclaration:
AccessibilityModifieropt static
opt PropertyName TypeAnnotationopt Initializeropt ;
The type associated with a member variable declaration is determined in the same manner as an ordinary variable declaration (see section 5.2).
An instance member variable declaration introduces a member in the class type and optionally initializes a property on instances of the class. Initializers in instance member variable declarations are executed once for every new instance of the class and are equivalent to assignments to properties of this
in the constructor. In an initializer expression for an instance member variable, this
is of the this-type (section 3.6.3) of the class.
A static member variable declaration introduces a property in the constructor function type and optionally initializes a property on the constructor function object. Initializers in static member variable declarations are executed once when the containing script or module is loaded.
Initializer expressions for instance member variables are evaluated in the scope of the class constructor body but are not permitted to reference parameters or local variables of the constructor. This effectively means that entities from outer scopes by the same name as a constructor parameter or local variable are inaccessible in initializer expressions for instance member variables.
Since instance member variable initializers are equivalent to assignments to properties of this
in the constructor, the example
class Employee {
public name: string;
public address: string;
public retired = false;
public manager: Employee = null;
public reports: Employee[] = [];
}
is equivalent to
class Employee {
public name: string;
public address: string;
public retired: boolean;
public manager: Employee;
public reports: Employee[];
constructor() {
this.retired = false;
this.manager = null;
this.reports = [];
}
}
A member function declaration declares an instance member function or a static member function.
MemberFunctionDeclaration:
AccessibilityModifieropt static
opt PropertyName CallSignature {
FunctionBody }
AccessibilityModifieropt static
opt PropertyName CallSignature ;
A member function declaration is processed in the same manner as an ordinary function declaration (section 6), except that in a member function this
has a known type.
All declarations for the same member function must specify the same accessibility (public, private, or protected) and kind (instance or static).
An instance member function declaration declares a property in the class type and assigns a function object to a property on the prototype object of the class. In the body of an instance member function declaration, this
is of the this-type (section 3.6.3) of the class.
A static member function declaration declares a property in the constructor function type and assigns a function object to a property on the constructor function object. In the body of a static member function declaration, the type of this
is the constructor function type.
A member function can access overridden base class members using a super property access (section 4.9.2). For example
class Point {
constructor(public x: number, public y: number) { }
public toString() {
return "x=" + this.x + " y=" + this.y;
}
}
class ColoredPoint extends Point {
constructor(x: number, y: number, public color: string) {
super(x, y);
}
public toString() {
return super.toString() + " color=" + this.color;
}
}
In a static member function, this
represents the constructor function object on which the static member function was invoked. Thus, a call to ’new this()’ may actually invoke a derived class constructor:
class A {
a = 1;
static create() {
return new this();
}
}
class B extends A {
b = 2;
}
var x = A.create(); // new A()
var y = B.create(); // new B()
Note that TypeScript doesn’t require or verify that derived constructor functions are subtypes of base constructor functions. In other words, changing the declaration of ‘B’ to
class B extends A {
constructor(public b: number) {
super();
}
}
does not cause errors in the example, even though the call to the constructor from the ‘create’ function doesn’t specify an argument (thus giving the value ‘undefined’ to ‘b’).
A member accessor declaration declares an instance member accessor or a static member accessor.
MemberAccessorDeclaration:
AccessibilityModifieropt static
opt GetAccessor
AccessibilityModifieropt static
opt SetAccessor
Get and set accessors are processed in the same manner as in an object literal (section 4.5), except that a contextual type is never available in a member accessor declaration.
Accessors for the same member name must specify the same accessibility.
An instance member accessor declaration declares a property in the class type and defines a property on the prototype object of the class with a get or set accessor. In the body of an instance member accessor declaration, this
is of the this-type (section 3.6.3) of the class.
A static member accessor declaration declares a property in the constructor function type and defines a property on the constructor function object of the class with a get or set accessor. In the body of a static member accessor declaration, the type of this
is the constructor function type.
Get and set accessors are emitted as calls to ‘Object.defineProperty’ in the generated JavaScript, as described in section 8.7.1.
If the PropertyName of a property member declaration is a computed property name that doesn’t denote a well-known symbol (2.2.3), the construct is considered a dynamic property declaration. The following rules apply to dynamic property declarations:
An index member declaration introduces an index signature (section 3.9.4) in the class type.
IndexMemberDeclaration:
IndexSignature ;
Index member declarations have no body and cannot specify an accessibility modifier.
A class declaration can have at most one string index member declaration and one numeric index member declaration. All instance property members of a class must satisfy the constraints implied by the index members of the class as specified in section 3.9.4.
It is not possible to declare index members for the static side of a class.
Note that it is seldom meaningful to include a string index signature in a class because it constrains all instance properties of the class. However, numeric index signatures can be useful to control the element type when a class is used in an array-like manner.
TODO: Document decorators.
When the output target is ECMAScript 2015 or higher, type parameters, implements clauses, accessibility modifiers, and member variable declarations are removed in the emitted code, but otherwise class declarations are emitted as written. When the output target is ECMAScript 3 or 5, more comprehensive rewrites are performed, as described in this section.
A class with no extends
clause generates JavaScript equivalent to the following:
var <ClassName> = (function () {
function <ClassName>(<ConstructorParameters>) {
<DefaultValueAssignments>
<ParameterPropertyAssignments>
<MemberVariableAssignments>
<ConstructorStatements>
}
<MemberFunctionStatements>
<StaticVariableAssignments>
return <ClassName>;
})();
ClassName is the name of the class.
ConstructorParameters is a comma separated list of the constructor’s parameter names.
DefaultValueAssignments is a sequence of default property value assignments corresponding to those generated for a regular function declaration, as described in section 6.6.
ParameterPropertyAssignments is a sequence of assignments, one for each parameter property declaration in the constructor, in order they are declared, of the form
this.<ParameterName> = <ParameterName>;
where ParameterName is the name of a parameter property.
MemberVariableAssignments is a sequence of assignments, one for each instance member variable declaration with an initializer, in the order they are declared, of the form
this.<MemberName> = <InitializerExpression>;
where MemberName is the name of the member variable and InitializerExpression is the code generated for the initializer expression.
ConstructorStatements is the code generated for the statements specified in the constructor body.
MemberFunctionStatements is a sequence of statements, one for each member function declaration or member accessor declaration, in the order they are declared.
An instance member function declaration generates a statement of the form
<ClassName>.prototype.<MemberName> = function (<FunctionParameters>) {
<DefaultValueAssignments>
<FunctionStatements>
}
and static member function declaration generates a statement of the form
<ClassName>.<MemberName> = function (<FunctionParameters>) {
<DefaultValueAssignments>
<FunctionStatements>
}
where MemberName is the name of the member function, and FunctionParameters, DefaultValueAssignments, and FunctionStatements correspond to those generated for a regular function declaration, as described in section 6.6.
A get or set instance member accessor declaration, or a pair of get and set instance member accessor declarations with the same name, generates a statement of the form
Object.defineProperty(<ClassName>.prototype, "<MemberName>", {
get: function () {
<GetAccessorStatements>
},
set: function (<ParameterName>) {
<SetAccessorStatements>
},
enumerable: true,
configurable: true
};
and a get or set static member accessor declaration, or a pair of get and set static member accessor declarations with the same name, generates a statement of the form
Object.defineProperty(<ClassName>, "<MemberName>", {
get: function () {
<GetAccessorStatements>
},
set: function (<ParameterName>) {
<SetAccessorStatements>
},
enumerable: true,
configurable: true
};
where MemberName is the name of the member accessor, GetAccessorStatements is the code generated for the statements in the get acessor’s function body, ParameterName is the name of the set accessor parameter, and SetAccessorStatements is the code generated for the statements in the set accessor’s function body. The ‘get’ property is included only if a get accessor is declared and the ‘set’ property is included only if a set accessor is declared.
StaticVariableAssignments is a sequence of statements, one for each static member variable declaration with an initializer, in the order they are declared, of the form
<ClassName>.<MemberName> = <InitializerExpression>;
where MemberName is the name of the static variable, and InitializerExpression is the code generated for the initializer expression.
A class with an extends
clause generates JavaScript equivalent to the following:
var <ClassName> = (function (_super) {
__extends(<ClassName>, _super);
function <ClassName>(<ConstructorParameters>) {
<DefaultValueAssignments>
<SuperCallStatement>
<ParameterPropertyAssignments>
<MemberVariableAssignments>
<ConstructorStatements>
}
<MemberFunctionStatements>
<StaticVariableAssignments>
return <ClassName>;
})(<BaseClassName>);
In addition, the ‘__extends’ function below is emitted at the beginning of the JavaScript source file. It copies all properties from the base constructor function object to the derived constructor function object (in order to inherit static members), and appropriately establishes the ‘prototype’ property of the derived constructor function object.
var __extends = this.__extends || function(d, b) {
for (var p in b) if (b.hasOwnProperty(p)) d[p] = b[p];
function f() { this.constructor = d; }
f.prototype = b.prototype;
d.prototype = new f();
}
BaseClassName is the class name specified in the extends
clause.
If the class has no explicitly declared constructor, the SuperCallStatement takes the form
_super.apply(this, arguments);
Otherwise the SuperCallStatement is present if the constructor function is required to start with a super call, as discussed in section 8.3.2, and takes the form
_super.call(this, <SuperCallArguments>)
where SuperCallArguments is the argument list specified in the super call. Note that this call precedes the code generated for parameter properties and member variables with initializers. Super calls elsewhere in the constructor generate similar code, but the code generated for such calls will be part of the ConstructorStatements section.
A super property access in the constructor, an instance member function, or an instance member accessor generates JavaScript equivalent to
_super.prototype.<PropertyName>
where PropertyName is the name of the referenced base class property. When the super property access appears in a function call, the generated JavaScript is equivalent to
_super.prototype.<PropertyName>.call(this, <Arguments>)
where Arguments is the code generated for the argument list specified in the function call.
A super property access in a static member function or a static member accessor generates JavaScript equivalent to
_super.<PropertyName>
where PropertyName is the name of the referenced base class property. When the super property access appears in a function call, the generated JavaScript is equivalent to
_super.<PropertyName>.call(this, <Arguments>)
where Arguments is the code generated for the argument list specified in the function call.
An enum type is a distinct subtype of the Number primitive type with an associated set of named constants that define the possible values of the enum type.
An enum declaration declares an enum type and an enum object.
EnumDeclaration:
const
opt enum
BindingIdentifier {
EnumBodyopt }
An EnumDeclaration introduces a named type (the enum type) and a named value (the enum object) in the containing declaration space. The enum type is a distinct subtype of the Number primitive type. The enum object is a value of an anonymous object type containing a set of properties, all of the enum type, corresponding to the values declared for the enum type in the body of the declaration. The enum object’s type furthermore includes a numeric index signature with the signature ‘[x: number]: string’.
The BindingIdentifier of an enum declaration may not be one of the predefined type names (section 3.8.1).
When an enum declaration includes a const
modifier it is said to be a constant enum declaration. The members of a constant enum declaration must all have constant values that can be computed at compile time. Constant enum declarations are discussed in section 9.4.
The example
enum Color { Red, Green, Blue }
declares a subtype of the Number primitive type called ‘Color’ and introduces a variable ‘Color’ with a type that corresponds to the declaration
var Color: {
[x: number]: string;
Red: Color;
Green: Color;
Blue: Color;
};
The numeric index signature reflects a “reverse mapping” that is automatically generated in every enum object, as described in section 9.5. The reverse mapping provides a convenient way to obtain the string representation of an enum value. For example
var c = Color.Red;
console.log(Color[c]); // Outputs "Red"
The body of an enum declaration defines zero or more enum members which are the named values of the enum type. Each enum member has an associated numeric value of the primitive type introduced by the enum declaration.
EnumBody:
EnumMemberList ,
opt
EnumMemberList:
EnumMember
EnumMemberList ,
EnumMember
EnumMember:
PropertyName
PropertyName = EnumValue
EnumValue:
AssignmentExpression
The PropertyName of an enum member cannot be a computed property name (2.2.3).
Enum members are either constant members or computed members. Constant members have known constant values that are substituted in place of references to the members in the generated JavaScript code. Computed members have values that are computed at run-time and not known at compile-time. No substitution is performed for references to computed members.
An enum member is classified as follows:
Enum value expressions must be of type Any, the Number primitive type, or the enum type itself.
A constant enum expression is a subset of the expression grammar that can be evaluated fully at compile time. An expression is considered a constant enum expression if it is one of the following:
In the example
enum Test {
A,
B,
C = Math.floor(Math.random() * 1000),
D = 10,
E
}
‘A’, ‘B’, ‘D’, and ‘E’ are constant members with values 0, 1, 10, and 11 respectively, and ‘C’ is a computed member.
In the example
enum Style {
None = 0,
Bold = 1,
Italic = 2,
Underline = 4,
Emphasis = Bold | Italic,
Hyperlink = Bold | Underline
}
all members are constant members. Note that enum member declarations can reference other enum members without qualification. Also, because enums are subtypes of the Number primitive type, numeric operators, such as the bitwise OR operator, can be used to compute enum values.
Enums are “open-ended” and enum declarations with the same qualified name relative to a common root (as defined in section 2.3) define a single enum type and contribute to a single enum object.
It isn’t possible for one enum declaration to continue the automatic numbering sequence of another, and when an enum type has multiple declarations, only one declaration is permitted to omit a value for the first member.
When enum declarations are merged, they must either all specify a const
modifier or all specify no const
modifier.
An enum declaration that specifies a const
modifier is a constant enum declaration. In a constant enum declaration, all members must have constant values and it is an error for a member declaration to specify an expression that isn’t classified as a constant enum expression.
Unlike regular enum declarations, constant enum declarations are completely erased in the emitted JavaScript code. For this reason, it is an error to reference a constant enum object in any other context than a property access that selects one of the enum’s members. For example:
const enum Comparison {
LessThan = -1,
EqualTo = 0,
GreaterThan = 1
}
var x = Comparison.EqualTo; // Ok, replaced with 0 in emitted code
var y = Comparison[Comparison.EqualTo]; // Error
var z = Comparison; // Error
The entire const enum declaration is erased in the emitted JavaScript code. Thus, the only permitted references to the enum object are those that are replaced with an enum member value.
An enum declaration generates JavaScript equivalent to the following:
var <EnumName>;
(function (<EnumName>) {
<EnumMemberAssignments>
})(<EnumName>||(<EnumName>={}));
EnumName is the name of the enum.
EnumMemberAssignments is a sequence of assignments, one for each enum member, in order they are declared, of the form
<EnumName>[<EnumName>["<MemberName>"] = <Value>] = "<MemberName>";
where MemberName is the name of the enum member and Value is the assigned constant value or the code generated for the computed value expression.
For example, the ‘Color’ enum example from section 9.1 generates the following JavaScript:
var Color;
(function (Color) {
Color[Color["Red"] = 0] = "Red";
Color[Color["Green"] = 1] = "Green";
Color[Color["Blue"] = 2] = "Blue";
})(Color||(Color={}));
Namespaces provide a mechanism for organizing code and declarations in hierarchies of named containers. Namespaces have named members that each denote a value, a type, or a namespace, or some combination thereof, and those members may be local or exported. The body of a namespace corresponds to a function that is executed once, thereby providing a mechanism for maintaining local state with assured isolation. Namespaces can be thought of as a formalization of the immediately-invoked function expression (IIFE) pattern.
A namespace declaration introduces a name with a namespace meaning and, in the case of an instantiated namespace, a value meaning in the containing declaration space.
NamespaceDeclaration:
namespace
IdentifierPath {
NamespaceBody }
IdentifierPath:
BindingIdentifier
IdentifierPath .
BindingIdentifier
Namespaces are declared using the namespace
keyword, but for backward compatibility of earlier versions of TypeScript a module
keyword can also be used.
Namespaces are either instantiated or non-instantiated. A non-instantiated namespace is a namespace containing only interface types, type aliases, and other non-instantiated namespace. An instantiated namespace is a namespace that doesn’t meet this definition. In intuitive terms, an instantiated namespace is one for which a namespace instance is created, whereas a non-instantiated namespace is one for which no code is generated.
When a namespace identifier is referenced as a NamespaceName (section 3.8.2) it denotes a container of namespace and type names, and when a namespace identifier is referenced as a PrimaryExpression (section 4.3) it denotes the singleton namespace instance. For example:
namespace M {
export interface P { x: number; y: number; }
export var a = 1;
}
var p: M.P; // M used as NamespaceName
var m = M; // M used as PrimaryExpression
var x1 = M.a; // M used as PrimaryExpression
var x2 = m.a; // Same as M.a
var q: m.P; // Error
Above, when ‘M’ is used as a PrimaryExpression it denotes an object instance with a single member ‘a’ and when ‘M’ is used as a NamespaceName it denotes a container with a single type member ‘P’. The final line in the example is an error because ’m’ is a variable which cannot be referenced in a type name.
If the declaration of ‘M’ above had excluded the exported variable ‘a’, ‘M’ would be a non-instantiated namespace and it would be an error to reference ‘M’ as a PrimaryExpression.
A namespace declaration that specifies an IdentifierPath with more than one identifier is equivalent to a series of nested single-identifier namespace declarations where all but the outermost are automatically exported. For example:
namespace A.B.C {
export var x = 1;
}
corresponds to
namespace A {
export namespace B {
export namespace C {
export var x = 1;
}
}
}
The hierarchy formed by namespace and named type names partially mirrors that formed by namespace instances and members. The example
namespace A {
export namespace B {
export class C { }
}
}
introduces a named type with the qualified name ‘A.B.C’ and also introduces a constructor function that can be accessed using the expression ‘A.B.C’. Thus, in the example
var c: A.B.C = new A.B.C();
the two occurrences of ‘A.B.C’ in fact refer to different entities. It is the context of the occurrences that determines whether ‘A.B.C’ is processed as a type name or an expression.
The body of a namespace corresponds to a function that is executed once to initialize the namespace instance.
NamespaceBody:
NamespaceElementsopt
NamespaceElements:
NamespaceElement
NamespaceElements NamespaceElement
NamespaceElement:
Statement
LexicalDeclaration
FunctionDeclaration
GeneratorDeclaration
ClassDeclaration
InterfaceDeclaration
TypeAliasDeclaration
EnumDeclaration
NamespaceDeclaration
AmbientDeclaration
ImportAliasDeclaration
ExportNamespaceElement
ExportNamespaceElement:
export
VariableStatement
export
LexicalDeclaration
export
FunctionDeclaration
export
GeneratorDeclaration
export
ClassDeclaration
export
InterfaceDeclaration
export
TypeAliasDeclaration
export
EnumDeclaration
export
NamespaceDeclaration
export
AmbientDeclaration
export
ImportAliasDeclaration
Import alias declarations are used to create local aliases for entities in other namespaces.
ImportAliasDeclaration:
import
BindingIdentifier =
EntityName ;
EntityName:
NamespaceName
NamespaceName .
IdentifierReference
An EntityName consisting of a single identifier is resolved as a NamespaceName and is thus required to reference a namespace. The resulting local alias references the given namespace and is itself classified as a namespace.
An EntityName consisting of more than one identifier is resolved as a NamespaceName followed by an identifier that names an exported entity in the given namespace. The resulting local alias has all the meanings of the referenced entity. (As many as three distinct meanings are possible for an entity name—value, type, and namespace.) In effect, it is as if the imported entity was declared locally with the local alias name.
In the example
namespace A {
export interface X { s: string }
export var X: X;
}
namespace B {
interface A { n: number }
import Y = A; // Alias for namespace A
import Z = A.X; // Alias for type and value A.X
var v: Z = Z;
}
within ‘B’, ‘Y’ is an alias only for namespace ‘A’ and not the local interface ‘A’, whereas ‘Z’ is an alias for all exported meanings of ‘A.X’, thus denoting both an interface type and a variable.
If the NamespaceName portion of an EntityName references an instantiated namespace, the NamespaceName is required to reference the namespace instance when evaluated as an expression. In the example
namespace A {
export interface X { s: string }
}
namespace B {
var A = 1;
import Y = A;
}
‘Y’ is a local alias for the non-instantiated namespace ‘A’. If the declaration of ‘A’ is changed such that ‘A’ becomes an instantiated namespace, for example by including a variable declaration in ‘A’, the import statement in ‘B’ above would be an error because the expression ‘A’ doesn’t reference the namespace instance of namespace ‘A’.
When an import statement includes an export modifier, all meanings of the local alias are exported.
An export declaration declares an externally accessible namespace member. An export declaration is simply a regular declaration prefixed with the keyword export
.
The members of a namespace’s export declaration space (section 2.3) constitute the namespace’s export member set. A namespace’s instance type is an object type with a property for each member in the namespace’s export member set that denotes a value.
An exported member depends on a (possibly empty) set of named types (section 3.7). Those named types must be at least as accessible as the exported member, or otherwise an error occurs.
The named types upon which a member depends are the named types occurring in the transitive closure of the directly depends on relationship defined as follows:
A named type T having a root namespace R (section 2.3) is said to be at least as accessible as a member M if
In the example
interface A { x: string; }
namespace M {
export interface B { x: A; }
export interface C { x: B; }
export function foo(c: C) { … }
}
the ‘foo’ function depends upon the named types ‘A’, ‘B’, and ‘C’. In order to export ‘foo’ it is necessary to also export ‘B’ and ‘C’ as they otherwise would not be at least as accessible as ‘foo’. The ‘A’ interface is already at least as accessible as ‘foo’ because I t is declared in a parent namespace of foo’s namespace.
Namespaces are “open-ended” and namespace declarations with the same qualified name relative to a common root (as defined in section 2.3) contribute to a single namespace. For example, the following two declarations of a namespace ‘outer’ might be located in separate source files.
File a.ts:
namespace outer {
var local = 1; // Non-exported local variable
export var a = local; // outer.a
export namespace inner {
export var x = 10; // outer.inner.x
}
}
File b.ts:
namespace outer {
var local = 2; // Non-exported local variable
export var b = local; // outer.b
export namespace inner {
export var y = 20; // outer.inner.y
}
}
Assuming the two source files are part of the same program, the two declarations will have the global namespace as their common root and will therefore contribute to the same namespace instance, the instance type of which will be:
{
a: number;
b: number;
inner: {
x: number;
y: number;
};
}
Declaration merging does not apply to local aliases created by import alias declarations. In other words, it is not possible have an import alias declaration and a namespace declaration for the same name within the same namespace body.
TODO: Clarify rules for alias resolution.
Declaration merging also extends to namespace declarations with the same qualified name relative to a common root as a function, class, or enum declaration:
When merging a non-ambient function or class declaration and a non-ambient namespace declaration, the function or class declaration must be located prior to the namespace declaration in the same source file. This ensures that the shared object instance is created as a function object. (While it is possible to add properties to an object after its creation, it is not possible to make an object “callable” after the fact.)
The example
interface Point {
x: number;
y: number;
}
function point(x: number, y: number): Point {
return { x: x, y: y };
}
namespace point {
export var origin = point(0, 0);
export function equals(p1: Point, p2: Point) {
return p1.x == p2.x && p1.y == p2.y;
}
}
var p1 = point(0, 0);
var p2 = point.origin;
var b = point.equals(p1, p2);
declares ‘point’ as a function object with two properties, ‘origin’ and ’equals’. Note that the namespace declaration for ‘point’ is located after the function declaration.
A namespace generates JavaScript code that is equivalent to the following:
var <NamespaceName>;
(function(<NamespaceName>) {
<NamespaceStatements>
})(<NamespaceName>||(<NamespaceName>={}));
where NamespaceName is the name of the namespace and NamespaceStatements is the code generated for the statements in the namespace body. The NamespaceName function parameter may be prefixed with one or more underscore characters to ensure the name is unique within the function body. Note that the entire namespace is emitted as an anonymous function that is immediately executed. This ensures that local variables are in their own lexical environment isolated from the surrounding context. Also note that the generated function doesn’t create and return a namespace instance, but rather it extends the existing instance (which may have just been created in the function call). This ensures that namespaces can extend each other.
An import statement generates code of the form
var <Alias> = <EntityName>;
This code is emitted only if the imported entity is referenced as a PrimaryExpression somewhere in the body of the importing namespace. If an imported entity is referenced only as a TypeName or NamespaceName, nothing is emitted. This ensures that types declared in one namespace can be referenced through an import alias in another namespace with no run-time overhead.
When a variable is exported, all references to the variable in the body of the namespace are replaced with
<NamespaceName>.<VariableName>
This effectively promotes the variable to be a property on the namespace instance and ensures that all references to the variable become references to the property.
When a function, class, enum, or namespace is exported, the code generated for the entity is followed by an assignment statement of the form
<NamespaceName>.<EntityName> = <EntityName>;
This copies a reference to the entity into a property on the namespace instance.
TypeScript implements support for ECMAScript 2015 modules and supports down-level code generation targeting CommonJS, AMD, and other module systems.
A TypeScript program consists of one or more source files.
SourceFile:
ImplementationSourceFile
DeclarationSourceFile
ImplementationSourceFile:
ImplementationScript
ImplementationModule
DeclarationSourceFile:
DeclarationScript
DeclarationModule
Source files with extension ‘.ts’ are implementation source files containing statements and declarations, and source files with extension ‘.d.ts’ are declaration source files containing declarations only.
Declaration source files are a strict subset of implementation source files and are used to declare the static type information associated with existing JavaScript code in an adjunct manner. They are entirely optional but enable the TypeScript compiler and tools to provide better verification and assistance when integrating existing JavaScript code and libraries in a TypeScript application.
When a TypeScript program is compiled, all of the program’s source files are processed together. Statements and declarations in different source files can depend on each other, possibly in a circular fashion. By default, a JavaScript output file is generated for each implementation source file in a compilation, but no output is generated from declaration source files.
The TypeScript compiler automatically determines a source file’s dependencies and includes those dependencies in the program being compiled. The determination is made from “reference comments” and module import declarations as follows:
Any files included as dependencies in turn have their references analyzed in a transitive manner until all dependencies have been determined.
Source files that contain no module import or export declarations are classified as scripts. Scripts form the single global namespace and entities declared in scripts are in scope everywhere in a program.
ImplementationScript:
ImplementationScriptElementsopt
ImplementationScriptElements:
ImplementationScriptElement
ImplementationScriptElements ImplementationScriptElement
ImplementationScriptElement:
ImplementationElement
AmbientModuleDeclaration
ImplementationElement:
Statement
LexicalDeclaration
FunctionDeclaration
GeneratorDeclaration
ClassDeclaration
InterfaceDeclaration
TypeAliasDeclaration
EnumDeclaration
NamespaceDeclaration
AmbientDeclaration
ImportAliasDeclaration
DeclarationScript:
DeclarationScriptElementsopt
DeclarationScriptElements:
DeclarationScriptElement
DeclarationScriptElements DeclarationScriptElement
DeclarationScriptElement:
DeclarationElement
AmbientModuleDeclaration
DeclarationElement:
InterfaceDeclaration
TypeAliasDeclaration
NamespaceDeclaration
AmbientDeclaration
ImportAliasDeclaration
The initialization order of the scripts that make up the global namespace ultimately depends on the order in which the generated JavaScript files are loaded at run-time (which, for example, may be controlled by <script/> tags that reference the generated JavaScript files).
Source files that contain at least one module import or export declaration are considered separate modules. Non-exported entities declared in a module are in scope only in that module, but exported entities can be imported into other modules using import declarations.
ImplementationModule:
ImplementationModuleElementsopt
ImplementationModuleElements:
ImplementationModuleElement
ImplementationModuleElements ImplementationModuleElement
ImplementationModuleElement:
ImplementationElement
ImportDeclaration
ImportAliasDeclaration
ImportRequireDeclaration
ExportImplementationElement
ExportDefaultImplementationElement
ExportListDeclaration
ExportAssignment
DeclarationModule:
DeclarationModuleElementsopt
DeclarationModuleElements:
DeclarationModuleElement
DeclarationModuleElements DeclarationModuleElement
DeclarationModuleElement:
DeclarationElement
ImportDeclaration
ImportAliasDeclaration
ExportDeclarationElement
ExportDefaultDeclarationElement
ExportListDeclaration
ExportAssignment
Initialization order of modules is determined by the module loader being used and is not specified by the TypeScript language. However, it is generally the case that non-circularly dependent modules are automatically loaded and initialized in the correct order.
Modules can additionally be declared using AmbientModuleDeclarations in declaration scripts that directly specify the module names as string literals. This is described further in section 12.2.
Below is an example of two modules written in separate source files:
// -------- main.ts --------
import { message } from "./log";
message("hello");
// -------- log.ts --------
export function message(s: string) {
console.log(s);
}
The import declaration in the ‘main’ module references the ’log’ module and compiling the ‘main.ts’ file causes the ’log.ts’ file to also be compiled as part of the program.
TypeScript supports multiple patterns of JavaScript code generation for modules:
The desired module code generation pattern is selected through a compiler option and does not affect the TypeScript source code. Indeed, it is possible to author modules that can be compiled for use both on the server side (e.g. using node.js) and on the client side (using an AMD compliant loader) with no changes to the TypeScript source code.
Modules are identified and referenced using module names. The following definition is aligned with that provided in the CommonJS Modules 1.0 specification.
For purposes of resolving module references, TypeScript associates a file path with every module. The file path is simply the path of the module’s source file without the file extension. For example, a module contained in the source file ‘C:\src\lib\io.ts’ has the file path ‘C:/src/lib/io’ and a module contained in the source file ‘C:\src\ui\editor.d.ts’ has the file path ‘C:/src/ui/editor’.
A module name in an import declaration is resolved as follows:
Import declarations are used to import entities from other modules and provide bindings for them in the current module.
An import declaration of the form
import * as m from "mod";
imports the module with the given name and creates a local binding for the module itself. The local binding is classified as a value (representing the module instance) and a namespace (representing a container of types and namespaces).
An import declaration of the form
import { x, y, z } from "mod";
imports a given module and creates local bindings for a specified list of exported members of the module. The specified names must each reference an entity in the export member set (11.3.4.4) of the given module. The local bindings have the same names and classifications as the entities they represent unless as
clauses are used to that specify different local names:
import { x as a, y as b } from "mod";
An import declaration of the form
import d from "mod";
is exactly equivalent to the import declaration
import { default as d } from "mod";
An import declaration of the form
import "mod";
imports the given module without creating any local bindings (this is useful only if the imported module has side effects).
Import require declarations exist for backward compatibility with earlier versions of TypeScript.
ImportRequireDeclaration:
import
BindingIdentifier =
require
(
StringLiteral )
;
An import require declaration introduces a local identifier that references a given module. The string literal specified in an import require declaration is interpreted as a module name (section 11.3.1). The local identifier introduced by the declaration becomes an alias for, and is classified exactly like, the entity exported from the referenced module. Specifically, if the referenced module contains no export assignment the identifier is classified as a value and a namespace, and if the referenced module contains an export assignment the identifier is classified exactly like the entity named in the export assignment.
An import require declaration of the form
import m = require("mod");
is equivalent to the ECMAScript 2015 import declaration
import * as m from "mod";
provided the referenced module contains no export assignment.
An export declaration declares one or more exported module members. The exported members of a module can be imported in other modules using import declarations (11.3.2).
In the body of a module, a declaration can export the declared entity by including an export
modifier.
ExportImplementationElement:
export
VariableStatement
export
LexicalDeclaration
export
FunctionDeclaration
export
GeneratorDeclaration
export
ClassDeclaration
export
InterfaceDeclaration
export
TypeAliasDeclaration
export
EnumDeclaration
export
NamespaceDeclaration
export
AmbientDeclaration
export
ImportAliasDeclaration
ExportDeclarationElement:
export
InterfaceDeclaration
export
TypeAliasDeclaration
export
AmbientDeclaration
export
ImportAliasDeclaration
In addition to introducing a name in the local declaration space of the module, an exported declaration introduces the same name with the same classification in the module’s export declaration space. For example, the declaration
export function point(x: number, y: number) {
return { x, y };
}
introduces a local name point
and an exported name point
that both reference the function.
Export default declarations provide short-hand syntax for exporting an entity named default
.
ExportDefaultImplementationElement:
export
default
FunctionDeclaration
export
default
GeneratorDeclaration
export
default
ClassDeclaration
export
default
AssignmentExpression ;
ExportDefaultDeclarationElement:
export
default
AmbientFunctionDeclaration
export
default
AmbientClassDeclaration
export
default
IdentifierReference ;
An ExportDefaultImplementationElement or ExportDefaultDeclarationElement for a function, generator, or class introduces a value named default
, and in the case of a class, a type named default
, in the containing module’s export declaration space. The declaration may optionally specify a local name for the exported function, generator, or class. For example, the declaration
export default function point(x: number, y: number) {
return { x, y };
}
introduces a local name point
and an exported name default
that both reference the function. The declaration is effectively equivalent to
function point(x: number, y: number) {
return { x, y };
}
export default point;
which again is equivalent to
function point(x: number, y: number) {
return { x, y };
}
export { point as default };
An ExportDefaultImplementationElement or ExportDefaultDeclarationElement for an expression consisting of a single identifier must name an entity declared in the current module or the global namespace. The declaration introduces an entity named default
, with the same classification as the referenced entity, in the containing module’s export declaration space. For example, the declarations
interface Point {
x: number;
y: number;
}
function Point(x: number, y: number): Point {
return { x, y };
}
export default Point;
introduce a local name Point
and an exported name default
, both with a value and a type meaning.
An ExportDefaultImplementationElement for any expression but a single identifier introduces a value named default
in the containing module’s export declaration space. For example, the declaration
export default "hello";
introduces an exported value named default
of type string.
An export list declaration exports one or more entities from the current module or a specified module.
ExportListDeclaration:
export
*
FromClause ;
export
ExportClause FromClause ;
export
ExportClause ;
An ExportListDeclaration without a FromClause exports entities from the current module. In a declaration of the form
export { x };
the name x
must reference an entity declared in the current module or the global namespace, and the declaration introduces an entity with the same name and meaning in the containing module’s export declaration space.
An ExportListDeclaration with a FromClause re-exports entities from a specified module. In a declaration of the form
export { x } from "mod";
the name x
must reference an entity in the export member set of the specified module, and the declaration introduces an entity with the same name and meaning in the containing module’s export declaration space. No local bindings are created for x
.
The ExportClause of an ExportListDeclaration can specify multiple entities and may optionally specify different names to be used for the exported entities. For example, the declaration
export { x, y as b, z as c };
introduces entities named x
, b
, and c
in the containing module’s export declaration space with the same meaning as the local entities named x
, y
, and z
respectively.
An ExportListDeclaration that specifies *
instead of an ExportClause is called an export star declaration. An export star declaration re-exports all members of a specified module.
export * from "mod";
Explicitly exported members take precedence over members re-exported using export star declarations, as described in the following section.
The export member set of a particular module is determined by starting with an empty set of members E and an empty set of processed modules P, and then processing the module as described below to form the full set of exported members in E. Processing a module M consists of these steps:
A module’s instance type is an object type with a property for each member in the module’s export member set that denotes a value.
If a module contains an export assignment it is an error for the module to also contain export declarations. The two types of exports are mutually exclusive.
Export assignments exist for backward compatibility with earlier versions of TypeScript. An export assignment designates a module member as the entity to be exported in place of the module itself.
ExportAssignment:
export
=
IdentifierReference ;
A module containing an export assignment can be imported using an import require declaration (11.3.3), and the local alias introduced by the import require declaration then takes on all meanings of the identifier named in the export assignment.
A module containing an export assignment can also be imported using a regular import declaration (11.3.2) provided the entity referenced in the export assignment is declared as a namespace or as a variable with a type annotation.
Assume the following example resides in the file ‘point.ts’:
export = Point;
class Point {
constructor(public x: number, public y: number) { }
static origin = new Point(0, 0);
}
When ‘point.ts’ is imported in another module, the import alias references the exported class and can be used both as a type and as a constructor function:
import Pt = require("./point");
var p1 = new Pt(10, 20);
var p2 = Pt.origin;
Note that there is no requirement that the import alias use the same name as the exported entity.
The CommonJS Modules definition specifies a methodology for writing JavaScript modules with implied privacy, the ability to import other modules, and the ability to explicitly export members. A CommonJS compliant system provides a ‘require’ function that can be used to synchronously load other modules to obtain their singleton module instance, as well as an ’exports’ variable to which a module can add properties to define its external API.
The ‘main’ and ’log’ example from section 11.3 above generates the following JavaScript code when compiled for the CommonJS Modules pattern:
File main.js:
var log_1 = require("./log");
log_1.message("hello");
File log.js:
function message(s) {
console.log(s);
}
exports.message = message;
A module import declaration is represented in the generated JavaScript as a variable initialized by a call to the ‘require’ function provided by the module system host. A variable declaration and ‘require’ call is emitted for a particular imported module only if the imported module, or a local alias (section 10.3) that references the imported module, is referenced as a PrimaryExpression somewhere in the body of the importing module. If an imported module is referenced only as a NamespaceName or TypeQueryExpression, nothing is emitted.
An example:
File geometry.ts:
export interface Point { x: number; y: number };
export function point(x: number, y: number): Point {
return { x, y };
}
File game.ts:
import * as g from "./geometry";
let p = g.point(10, 20);
The ‘game’ module references the imported ‘geometry’ module in an expression (through its alias ‘g’) and a ‘require’ call is therefore included in the emitted JavaScript:
var g = require("./geometry");
var p = g.point(10, 20);
Had the ‘game’ module instead been written to only reference ‘geometry’ in a type position
import * as g from "./geometry";
let p: g.Point = { x: 10, y: 20 };
the emitted JavaScript would have no dependency on the ‘geometry’ module and would simply be
var p = { x: 10, y: 20 };
The Asynchronous Module Definition (AMD) specification extends the CommonJS Modules specification with a pattern for authoring asynchronously loadable modules with associated dependencies. Using the AMD pattern, modules are emitted as calls to a global ‘define’ function taking an array of dependencies, specified as module names, and a callback function containing the module body. The global ‘define’ function is provided by including an AMD compliant loader in the application. The loader arranges to asynchronously load the module’s dependencies and, upon completion, calls the callback function passing resolved module instances as arguments in the order they were listed in the dependency array.
The “main” and “log” example from above generates the following JavaScript code when compiled for the AMD pattern.
File main.js:
define(["require", "exports", "./log"], function(require, exports, log_1) {
log_1.message("hello");
}
File log.js:
define(["require", "exports"], function(require, exports) {
function message(s) {
console.log(s);
}
exports.message = message;
}
The special ‘require’ and ’exports’ dependencies are always present. Additional entries are added to the dependencies array and the parameter list as required to represent imported modules. Similar to the code generation for CommonJS Modules, a dependency entry is generated for a particular imported module only if the imported module is referenced as a PrimaryExpression somewhere in the body of the importing module. If an imported module is referenced only as a NamespaceName, no dependency is generated for that module.
Ambient declarations are used to provide static typing over existing JavaScript code. Ambient declarations differ from regular declarations in that no JavaScript code is emitted for them. Instead of introducing new variables, functions, classes, enums, or namespaces, ambient declarations provide type information for entities that exist “ambiently” and are included in a program by external means, for example by referencing a JavaScript library in a <script/> tag.
Ambient declarations are written using the declare
keyword and can declare variables, functions, classes, enums, namespaces, or modules.
AmbientDeclaration:
declare
AmbientVariableDeclaration
declare
AmbientFunctionDeclaration
declare
AmbientClassDeclaration
declare
AmbientEnumDeclaration
declare
AmbientNamespaceDeclaration
An ambient variable declaration introduces a variable in the containing declaration space.
AmbientVariableDeclaration:
var
AmbientBindingList ;
let
AmbientBindingList ;
const
AmbientBindingList ;
AmbientBindingList:
AmbientBinding
AmbientBindingList ,
AmbientBinding
AmbientBinding:
BindingIdentifier TypeAnnotationopt
An ambient variable declaration may optionally include a type annotation. If no type annotation is present, the variable is assumed to have type Any.
An ambient variable declaration does not permit an initializer expression to be present.
An ambient function declaration introduces a function in the containing declaration space.
AmbientFunctionDeclaration:
function
BindingIdentifier CallSignature ;
Ambient functions may be overloaded by specifying multiple ambient function declarations with the same name, but it is an error to declare multiple overloads that are considered identical (section 3.11.2) or differ only in their return types.
Ambient function declarations cannot specify a function bodies and do not permit default parameter values.
An ambient class declaration declares a class type and a constructor function in the containing declaration space.
AmbientClassDeclaration:
class
BindingIdentifier TypeParametersopt ClassHeritage {
AmbientClassBody }
AmbientClassBody:
AmbientClassBodyElementsopt
AmbientClassBodyElements:
AmbientClassBodyElement
AmbientClassBodyElements AmbientClassBodyElement
AmbientClassBodyElement:
AmbientConstructorDeclaration
AmbientPropertyMemberDeclaration
IndexSignature
AmbientConstructorDeclaration:
constructor
(
ParameterListopt )
;
AmbientPropertyMemberDeclaration:
AccessibilityModifieropt static
opt PropertyName TypeAnnotationopt ;
AccessibilityModifieropt static
opt PropertyName CallSignature ;
An ambient enum is grammatically equivalent to a non-ambient enum declaration.
AmbientEnumDeclaration:
EnumDeclaration
Ambient enum declarations differ from non-ambient enum declarations in two ways:
const
modifier, enum member declarations that omit a value are considered computed members (as opposed to having auto-incremented values assigned).Ambient enum declarations are otherwise processed in the same manner as non-ambient enum declarations.
An ambient namespace declaration declares a namespace.
AmbientNamespaceDeclaration:
namespace
IdentifierPath {
AmbientNamespaceBody }
AmbientNamespaceBody:
AmbientNamespaceElementsopt
AmbientNamespaceElements:
AmbientNamespaceElement
AmbientNamespaceElements AmbientNamespaceElement
AmbientNamespaceElement:
export
opt AmbientVariableDeclaration
export
opt AmbientLexicalDeclaration
export
opt AmbientFunctionDeclaration
export
opt AmbientClassDeclaration
export
opt InterfaceDeclaration
export
opt AmbientEnumDeclaration
export
opt AmbientNamespaceDeclaration
export
opt ImportAliasDeclaration
Except for ImportAliasDeclarations, AmbientNamespaceElements always declare exported entities regardless of whether they include the optional export
modifier.
An AmbientModuleDeclaration declares a module. This type of declaration is permitted only at the top level in a source file that contributes to the global namespace (section 11.1). The StringLiteral must specify a top-level module name. Relative module names are not permitted.
AmbientModuleDeclaration:
declare
module
StringLiteral {
DeclarationModule }
An ImportRequireDeclaration in an AmbientModuleDeclaration may reference other modules only through top-level module names. Relative module names are not permitted.
If an ambient module declaration includes an export assignment, it is an error for any of the declarations within the module to specify an export
modifier. If an ambient module declaration contains no export assignment, entities declared in the module are exported regardless of whether their declarations include the optional export
modifier.
Ambient modules are “open-ended” and ambient module declarations with the same string literal name contribute to a single module. For example, the following two declarations of a module ‘io’ might be located in separate source files.
declare module "io" {
export function readFile(filename: string): string;
}
declare module "io" {
export function writeFile(filename: string, data: string): void;
}
This has the same effect as a single combined declaration:
declare module "io" {
export function readFile(filename: string): string;
export function writeFile(filename: string, data: string): void;
}
This appendix contains a summary of the grammar found in the main document. As described in section 2.1, the TypeScript grammar is a superset of the grammar defined in the ECMAScript 2015 Language Specification (specifically, the ECMA-262 Standard, 6th Edition) and this appendix lists only productions that are new or modified from the ECMAScript grammar.
TypeParameters:
<
TypeParameterList >
TypeParameterList:
TypeParameter
TypeParameterList ,
TypeParameter
TypeParameter:
BindingIdentifier Constraintopt
Constraint:
extends
Type
TypeArguments:
<
TypeArgumentList >
TypeArgumentList:
TypeArgument
TypeArgumentList ,
TypeArgument
TypeArgument:
Type
Type:
UnionOrIntersectionOrPrimaryType
FunctionType
ConstructorType
UnionOrIntersectionOrPrimaryType:
UnionType
IntersectionOrPrimaryType
IntersectionOrPrimaryType:
IntersectionType
PrimaryType
PrimaryType:
ParenthesizedType
PredefinedType
TypeReference
ObjectType
ArrayType
TupleType
TypeQuery
ThisType
ParenthesizedType:
(
Type )
PredefinedType:
any
number
boolean
string
symbol
void
TypeReference:
TypeName [no LineTerminator here] TypeArgumentsopt
TypeName:
IdentifierReference
NamespaceName .
IdentifierReference
NamespaceName:
IdentifierReference
NamespaceName .
IdentifierReference
ObjectType:
{
TypeBodyopt }
TypeBody:
TypeMemberList ;
opt
TypeMemberList ,
opt
TypeMemberList:
TypeMember
TypeMemberList ;
TypeMember
TypeMemberList ,
TypeMember
TypeMember:
PropertySignature
CallSignature
ConstructSignature
IndexSignature
MethodSignature
ArrayType:
PrimaryType [no LineTerminator here] [
]
TupleType:
[
TupleElementTypes ]
TupleElementTypes:
TupleElementType
TupleElementTypes ,
TupleElementType
TupleElementType:
Type
UnionType:
UnionOrIntersectionOrPrimaryType |
IntersectionOrPrimaryType
IntersectionType:
IntersectionOrPrimaryType &
PrimaryType
FunctionType:
TypeParametersopt (
ParameterListopt )
=>
Type
ConstructorType:
new
TypeParametersopt (
ParameterListopt )
=>
Type
TypeQuery:
typeof
TypeQueryExpression
TypeQueryExpression:
IdentifierReference
TypeQueryExpression .
IdentifierName
ThisType:
this
PropertySignature:
PropertyName ?
opt TypeAnnotationopt
PropertyName:
IdentifierName
StringLiteral
NumericLiteral
TypeAnnotation:
:
Type
CallSignature:
TypeParametersopt (
ParameterListopt )
TypeAnnotationopt
ParameterList:
RequiredParameterList
OptionalParameterList
RestParameter
RequiredParameterList ,
OptionalParameterList
RequiredParameterList ,
RestParameter
OptionalParameterList ,
RestParameter
RequiredParameterList ,
OptionalParameterList ,
RestParameter
RequiredParameterList:
RequiredParameter
RequiredParameterList ,
RequiredParameter
RequiredParameter:
AccessibilityModifieropt BindingIdentifierOrPattern TypeAnnotationopt
BindingIdentifier :
StringLiteral
AccessibilityModifier:
public
private
protected
BindingIdentifierOrPattern:
BindingIdentifier
BindingPattern
OptionalParameterList:
OptionalParameter
OptionalParameterList ,
OptionalParameter
OptionalParameter:
AccessibilityModifieropt BindingIdentifierOrPattern ?
TypeAnnotationopt
AccessibilityModifieropt BindingIdentifierOrPattern TypeAnnotationopt Initializer
BindingIdentifier ?
:
StringLiteral
RestParameter:
...
BindingIdentifier TypeAnnotationopt
ConstructSignature:
new
TypeParametersopt (
ParameterListopt )
TypeAnnotationopt
IndexSignature:
[
BindingIdentifier :
string
]
TypeAnnotation
[
BindingIdentifier :
number
]
TypeAnnotation
MethodSignature:
PropertyName ?
opt CallSignature
TypeAliasDeclaration:
type
BindingIdentifier TypeParametersopt =
Type ;
PropertyDefinition: ( Modified )
IdentifierReference
CoverInitializedName
PropertyName :
AssignmentExpression
PropertyName CallSignature {
FunctionBody }
GetAccessor
SetAccessor
GetAccessor:
get
PropertyName (
)
TypeAnnotationopt {
FunctionBody }
SetAccessor:
set
PropertyName (
BindingIdentifierOrPattern TypeAnnotationopt )
{
FunctionBody }
FunctionExpression: ( Modified )
function
BindingIdentifieropt CallSignature {
FunctionBody }
ArrowFormalParameters: ( Modified )
CallSignature
Arguments: ( Modified )
TypeArgumentsopt (
ArgumentListopt )
UnaryExpression: ( Modified )
…
<
Type >
UnaryExpression
Declaration: ( Modified )
…
InterfaceDeclaration
TypeAliasDeclaration
EnumDeclaration
VariableDeclaration: ( Modified )
SimpleVariableDeclaration
DestructuringVariableDeclaration
SimpleVariableDeclaration:
BindingIdentifier TypeAnnotationopt Initializeropt
DestructuringVariableDeclaration:
BindingPattern TypeAnnotationopt Initializer
LexicalBinding: ( Modified )
SimpleLexicalBinding
DestructuringLexicalBinding
SimpleLexicalBinding:
BindingIdentifier TypeAnnotationopt Initializeropt
DestructuringLexicalBinding:
BindingPattern TypeAnnotationopt Initializeropt
FunctionDeclaration: ( Modified )
function
BindingIdentifieropt CallSignature {
FunctionBody }
function
BindingIdentifieropt CallSignature ;
InterfaceDeclaration:
interface
BindingIdentifier TypeParametersopt InterfaceExtendsClauseopt ObjectType
InterfaceExtendsClause:
extends
ClassOrInterfaceTypeList
ClassOrInterfaceTypeList:
ClassOrInterfaceType
ClassOrInterfaceTypeList ,
ClassOrInterfaceType
ClassOrInterfaceType:
TypeReference
ClassDeclaration: ( Modified )
class
BindingIdentifieropt TypeParametersopt ClassHeritage {
ClassBody }
ClassHeritage: ( Modified )
ClassExtendsClauseopt ImplementsClauseopt
ClassExtendsClause:
extends
ClassType
ClassType:
TypeReference
ImplementsClause:
implements
ClassOrInterfaceTypeList
ClassElement: ( Modified )
ConstructorDeclaration
PropertyMemberDeclaration
IndexMemberDeclaration
ConstructorDeclaration:
AccessibilityModifieropt constructor
(
ParameterListopt )
{
FunctionBody }
AccessibilityModifieropt constructor
(
ParameterListopt )
;
PropertyMemberDeclaration:
MemberVariableDeclaration
MemberFunctionDeclaration
MemberAccessorDeclaration
MemberVariableDeclaration:
AccessibilityModifieropt static
opt PropertyName TypeAnnotationopt Initializeropt ;
MemberFunctionDeclaration:
AccessibilityModifieropt static
opt PropertyName CallSignature {
FunctionBody }
AccessibilityModifieropt static
opt PropertyName CallSignature ;
MemberAccessorDeclaration:
AccessibilityModifieropt static
opt GetAccessor
AccessibilityModifieropt static
opt SetAccessor
IndexMemberDeclaration:
IndexSignature ;
EnumDeclaration:
const
opt enum
BindingIdentifier {
EnumBodyopt }
EnumBody:
EnumMemberList ,
opt
EnumMemberList:
EnumMember
EnumMemberList ,
EnumMember
EnumMember:
PropertyName
PropertyName = EnumValue
EnumValue:
AssignmentExpression
NamespaceDeclaration:
namespace
IdentifierPath {
NamespaceBody }
IdentifierPath:
BindingIdentifier
IdentifierPath .
BindingIdentifier
NamespaceBody:
NamespaceElementsopt
NamespaceElements:
NamespaceElement
NamespaceElements NamespaceElement
NamespaceElement:
Statement
LexicalDeclaration
FunctionDeclaration
GeneratorDeclaration
ClassDeclaration
InterfaceDeclaration
TypeAliasDeclaration
EnumDeclaration
NamespaceDeclaration
AmbientDeclaration
ImportAliasDeclaration
ExportNamespaceElement
ExportNamespaceElement:
export
VariableStatement
export
LexicalDeclaration
export
FunctionDeclaration
export
GeneratorDeclaration
export
ClassDeclaration
export
InterfaceDeclaration
export
TypeAliasDeclaration
export
EnumDeclaration
export
NamespaceDeclaration
export
AmbientDeclaration
export
ImportAliasDeclaration
ImportAliasDeclaration:
import
BindingIdentifier =
EntityName ;
EntityName:
NamespaceName
NamespaceName .
IdentifierReference
SourceFile:
ImplementationSourceFile
DeclarationSourceFile
ImplementationSourceFile:
ImplementationScript
ImplementationModule
DeclarationSourceFile:
DeclarationScript
DeclarationModule
ImplementationScript:
ImplementationScriptElementsopt
ImplementationScriptElements:
ImplementationScriptElement
ImplementationScriptElements ImplementationScriptElement
ImplementationScriptElement:
ImplementationElement
AmbientModuleDeclaration
ImplementationElement:
Statement
LexicalDeclaration
FunctionDeclaration
GeneratorDeclaration
ClassDeclaration
InterfaceDeclaration
TypeAliasDeclaration
EnumDeclaration
NamespaceDeclaration
AmbientDeclaration
ImportAliasDeclaration
DeclarationScript:
DeclarationScriptElementsopt
DeclarationScriptElements:
DeclarationScriptElement
DeclarationScriptElements DeclarationScriptElement
DeclarationScriptElement:
DeclarationElement
AmbientModuleDeclaration
DeclarationElement:
InterfaceDeclaration
TypeAliasDeclaration
NamespaceDeclaration
AmbientDeclaration
ImportAliasDeclaration
ImplementationModule:
ImplementationModuleElementsopt
ImplementationModuleElements:
ImplementationModuleElement
ImplementationModuleElements ImplementationModuleElement
ImplementationModuleElement:
ImplementationElement
ImportDeclaration
ImportAliasDeclaration
ImportRequireDeclaration
ExportImplementationElement
ExportDefaultImplementationElement
ExportListDeclaration
ExportAssignment
DeclarationModule:
DeclarationModuleElementsopt
DeclarationModuleElements:
DeclarationModuleElement
DeclarationModuleElements DeclarationModuleElement
DeclarationModuleElement:
DeclarationElement
ImportDeclaration
ImportAliasDeclaration
ExportDeclarationElement
ExportDefaultDeclarationElement
ExportListDeclaration
ExportAssignment
ImportRequireDeclaration:
import
BindingIdentifier =
require
(
StringLiteral )
;
ExportImplementationElement:
export
VariableStatement
export
LexicalDeclaration
export
FunctionDeclaration
export
GeneratorDeclaration
export
ClassDeclaration
export
InterfaceDeclaration
export
TypeAliasDeclaration
export
EnumDeclaration
export
NamespaceDeclaration
export
AmbientDeclaration
export
ImportAliasDeclaration
ExportDeclarationElement:
export
InterfaceDeclaration
export
TypeAliasDeclaration
export
AmbientDeclaration
export
ImportAliasDeclaration
ExportDefaultImplementationElement:
export
default
FunctionDeclaration
export
default
GeneratorDeclaration
export
default
ClassDeclaration
export
default
AssignmentExpression ;
ExportDefaultDeclarationElement:
export
default
AmbientFunctionDeclaration
export
default
AmbientClassDeclaration
export
default
IdentifierReference ;
ExportListDeclaration:
export
*
FromClause ;
export
ExportClause FromClause ;
export
ExportClause ;
ExportAssignment:
export
=
IdentifierReference ;
AmbientDeclaration:
declare
AmbientVariableDeclaration
declare
AmbientFunctionDeclaration
declare
AmbientClassDeclaration
declare
AmbientEnumDeclaration
declare
AmbientNamespaceDeclaration
AmbientVariableDeclaration:
var
AmbientBindingList ;
let
AmbientBindingList ;
const
AmbientBindingList ;
AmbientBindingList:
AmbientBinding
AmbientBindingList ,
AmbientBinding
AmbientBinding:
BindingIdentifier TypeAnnotationopt
AmbientFunctionDeclaration:
function
BindingIdentifier CallSignature ;
AmbientClassDeclaration:
class
BindingIdentifier TypeParametersopt ClassHeritage {
AmbientClassBody }
AmbientClassBody:
AmbientClassBodyElementsopt
AmbientClassBodyElements:
AmbientClassBodyElement
AmbientClassBodyElements AmbientClassBodyElement
AmbientClassBodyElement:
AmbientConstructorDeclaration
AmbientPropertyMemberDeclaration
IndexSignature
AmbientConstructorDeclaration:
constructor
(
ParameterListopt )
;
AmbientPropertyMemberDeclaration:
AccessibilityModifieropt static
opt PropertyName TypeAnnotationopt ;
AccessibilityModifieropt static
opt PropertyName CallSignature ;
AmbientEnumDeclaration:
EnumDeclaration
AmbientNamespaceDeclaration:
namespace
IdentifierPath {
AmbientNamespaceBody }
AmbientNamespaceBody:
AmbientNamespaceElementsopt
AmbientNamespaceElements:
AmbientNamespaceElement
AmbientNamespaceElements AmbientNamespaceElement
AmbientNamespaceElement:
export
opt AmbientVariableDeclaration
export
opt AmbientLexicalDeclaration
export
opt AmbientFunctionDeclaration
export
opt AmbientClassDeclaration
export
opt InterfaceDeclaration
export
opt AmbientEnumDeclaration
export
opt AmbientNamespaceDeclaration
export
opt ImportAliasDeclaration
AmbientModuleDeclaration:
declare
module
StringLiteral {
DeclarationModule }