examples/js/README.md

JavaScript Mojo Example Applications
=====================

hello.js, world.js - A minimal application that connects to another.

wget.js - Uses the network service to load a URL.

cube.js - A JS version of examples/sample_app.

--- Running Mojo Applications --- 

A Mojo application written in JavaScript is launched with mojo_shell like this:

  mojo_shell <js-application-url>

Where js-application-url is a URL understood by the shell. For example
a file or an http URL that names a JS source file. The JS file itself
must begin with a Mojo "shebang" that specifies the Mojo URL of the JS
content handler. In other words, the first line of the JS source file
must be:

  #!mojo:js_content_handler

Following the shebang should be a single AMD module called "main" whose value
is an Application subclass. The JS content handler will create an instance of
the Application and make it the client of the Mojo shell. The JS content handler
is itself a Mojo application and it's responsible for creating an instance of V8
and loading the "main" JS module and all of the modules the main module
depends on. 

This is the overall structure of a JS Mojo application:

  #!mojo:js_content_handler

  define("main", ["mojo/services/public/js/application", 
    <list of other modules that this application depends on>
],
    function(appModule, <one parameter per dependent module>) {
      class MyApplication extends appModule.Application {
        constructor(appShell, url) {
          super(appShell, url); // Initializes this.shell, this.url.
          // MyApplication initializations here. 
        }

        initialize(args) {
        }

        acceptConnection(url, serviceProvider) {
        }
      }

      return MyApplication;
    });

The hello.js example is little more than this basic skeleton.

The JS content handler loads the "main" module and makes an instance of its
value, which must be an Application subclass. The application's constructor is 
passed two arguments:

appShell - a pointer to the Mojo shell. Typically this will be wrapped by a
    Shell object, see below.

url - the URL this application was loaded from as a String.

The (inherited) Application class constructor initializes the shell and url
properties. It's unlikely that you'll want to use the appShell argument
directly.

The initialize() and acceptConnection() methods are defined by application.mojom
and they're needed because the JS content handler makes the JS application the
Mojo shell's client.


--- JavaScript Classes --- 

The JS content handler depends on the ECMAScript6 ("Harmony") classes feature.



--- Mojo Application Structure --- 

Mojo applications can connect to services provided by other applications and
they can provide services of their own. A service is an implementation of a Mojo
interface that was defined as part of a Mojo module in a ".mojom" file.

To implement a service you'll need the JS "bindings" for the Mojo interface. The
bindings are generated by the build system and end up in files whose name is the
same as the '.mojom' file with a '.js' suffix. It's often helpful to look at the
generated 'mojom.js' files.

The JS Shell class simplifies connecting to applications and services. It's a
wrapper for the Application's appShell argument. The Application constructor
creates a Shell and assigns it to |this.shell|.

The Shell's connectToService() method returns a "proxy" to a service provided by
another application.

The JS bindings for a Mojo interface's API are delivered as a JS module whose
name is based on the '.mojom' file's path. For example, to use the Mojo network
service you need the JS module based on network_service.mojom:

  define("main", [
    "mojo/services/network/public/interfaces/network_service.mojom",
    "mojo/services/public/js/application",
  ]
    function(netModule, appModule) {
      class MyApplication extends appModule.Application {
        initialize(args) {
          var netService = this.shell.connectToService(
              "mojo:network_service", netModule.NetworkService);
          // Use netService's NetworkService methods.
        }
        ...
      }

      return MyApplication;
    });

The first connectToService() parameter is the Mojo URL for the network service
application and the second is the JS "interface" object for NetworkService. The
JS interface object's properties identify the (generated) JS bindings classes
used to provide or connect to a service. For example (from
network_service.mojom.js):

  var NetworkService = {
    name: 'mojo::NetworkService', // Fully qualified Mojo interface name.
    proxyClass: NetworkServiceProxy,
    stubClass: NetworkServiceStub,
    // ...
  };

The 'proxyClass' is used to access another application's NetworkService and the
'stubClass' is used to create an implementation of NetworkService. 

In the netService case above the Shell connects to the Mojo application at
"mojo:network_service", then connects to its service called 
'NetworkService.name' with an instance of 'NetworkService.proxyClass'. The proxy
instance is returned. The netService proxy can be used immediately.


--- Mojo Responses are Promises --- 

Mojo functions can return zero or more values called a "response". For example
the EchoString function below returns a string or null.

interface EchoService {
  EchoString(string? value) => (string? value);
};

The response is delivered to the function caller asynchronously. In C++ the 
caller provides a Callback object whose Run() method has one argument for
each response parameter. In JS, Mojo functions that specify a response return
a Promise object. The Promise resolves to an object with one property per
response parameter. In the EchoString case that would be something like 
{value: "foo"}.

Similarly, the implementation of a Mojo interface functions that specify a
response, must return a Promise. The implementation of EchoString() could
be written like this:

  MyEchoStringImpl.prototype.EchoString = function(s) {
    return Promise.resolve({value: s});
  };


- Applications can request and provide services

When an application starts, its initialize() method runs and then its
acceptConnection() method runs. The acceptConnection() method
indicates that another application has connected to this one and it
always runs at least once.

  acceptConnection(initiatorURL, serviceProvider) {
    // provide services to the initiator here
    // request services from the initiator here
  }

The acceptConnection serviceProvider argument can be used to provide
services to the initiator, and to request services from the
initiator. An application can decide exactly what to do based on the
initiator's URL. The serviceProvider argument is-a JS ServiceProvider,
an object that wraps a Mojo ServiceProvider proxy. 

The ServiceProvider requestService() method gets a proxy for a service
from the initator and optionally provides a client implementation.

The ServiceProvider provideService() method registers an interface
implementation factory for a Mojo interface. The factory function is
provided with an proxy for the interface's client, if it has one.

An application can also connect to other applications and their
services using its shell's connectToApplication() and
connectToService() methods. The shell's connectToApplication() returns
a ServiceProvider. The shell's connectToService() method is just a
convenience, it's defined like this:

  connectToService(url, service, client) {
    return this.connectToApplication(url).requestService(service, clientImpl);
  };

The value of service is an interface object that identifies a Mojo
interface that the application at url implements.

The usage examples that follow are based on the following trivial Mojo
interface:

interface EchoService {
  EchoString(string? value) => (string? value);
};

-- Requesting a service using the Application's Shell

Given the URL of a Mojo application that implements the EchoService we
can use the application's shell to get an EchoService proxy. Here's a
complete application:

  #!mojo:js_content_handler

  define("main", [
    "console",
    "mojo/services/public/js/application",
    "services/js/test/echo_service.mojom"
  ], function(console, appModule, echoModule) {

    class EchoShellRequest extends appModule.Application {
      initialize(args) {
        var url = "file:/foo/bar/echo.js";
        var echoService = this.shell.connectToService(url, echoModule.EchoService);
        echoService.echoString("foo").then(function(result) {
          console.log("echoString(foo) => " + result.value);
        });
      }
    }
    return EchoShellRequest;
  });


-- Providing a service

A complete application that unconditionally provides the EchoService
looks like this:

  #!mojo:js_content_handler

  define("main", [
    "mojo/services/public/js/application",
    "services/js/test/echo_service.mojom"
  ], function(appModule, echoModule) {

    class EchoService extends appModule.Application {
      acceptConnection(initiatorURL, serviceProvider) {
        function EchoServiceImpl(client) {
          this.echoString = function(s) {
            return Promise.resolve({value: s});
          };
        }
        serviceProvider.provideService(echoModule.EchoService, EchoServiceImpl);
      }
    }
    return EchoService;
  });

As you can see, EchoServiceImpl is just a function that returns an
object that implements the methods in the Mojo EchoService
interface. If the EchoService defined a client interface, the factory
function's client parameter would be a proxy for the initiator's
client service. EchoService doesn't have a client so we could have
omitted this parameter.

Each time another application connects to this one, the EchoServiceImpl
function will be called. The caller will be able to run the
echoString() method and will get its response via a Promise.


-- Final note

An initiator's serviceProvider object can be retained and used to
request or provide services at any time, not just from within
application's acceptConnection() method.


--- Interface Parameters --- 

The caller and callee use cases that follow are  in terms of the following mojom:

  [Client=Bar]
  interface Foo {
  }

  [Client=Foo] // Redundant but always implicitly true.
  interface Bar {
  }

  interface I {
    provideFoo(Foo foo);
    requestFoo(Foo& foo); // effectively: provideFoo(Bar bar)
  } 

-- In General

From a user's point of view, the bindings are in terms of the (remote)
proxy class and the (local) stub class's implementation delegate
(internally, that's the stub class's delegate$ property).  The
bindings will add/use a local$ property on proxy objects which points
to the stub class's implementation delegate. They also manage remote$
property on the implementation delegate whose value is the proxy.

All that implies:

fooImpl.remote$.local$ == fooImpl (the stub class's delegate)
fooProxy.local$.remote$ == fooProxy


-- Callers

Assuming that we have a proxy for interface I, iProxy.

An iProxy.provideFoo() call implies that we have an implementation of
Foo, and want a proxy for Bar (Foo's client).

  var myFooImpl;
  provideFoo(myFooImpl);
  myFooImpl.remote$; // A Bar proxy initialized by provideFoo(), undefined if Foo has no client.

An iProxy.requestFoo() call implies that we have an implementation of
Bar and want a proxy for Foo (Bar's client).

  var myBarImpl; // If Foo has no client then this is just {}.
  requestFoo(myBarImpl);
  myBarImpl.remote$; // A Foo proxy initialized by requestFoo.

The wget.js example includes a request for the URLLoader service.

-- Callees

An implementation of provideFoo(Foo foo) implies that we have an
implementation of Bar (Foo's client) and want a proxy to the Foo
that has been passed to us.

  void provideFoo(fooProxy) {
    fooProxy.local$ = myBarImpl; // sets myFooImpl.remote$ = fooProxy
  }

An implementation of requestFoo(Foo& foo) implies that we have an
implementation of Foo and want a proxy for the Bar (Foo's client)
that's been passed to us.

  void requestFoo(barProxy) {
    barProxy.local$ = myFooImpl; // sets myFooImpl.remote$ = barProxy
  }