This article describes working with multiple windows in Mozilla chrome code (XUL applications and Extensions). It contains tips and example code on opening new windows, finding an already opened window, and passing data between different windows.

Opening windows

From a <script> in a window or an overlay

To open a new window, we usually use a window.open or window.openDialog DOM call, like this:

var win = window.open("chrome://myextension/content/about.xul",
                      "aboutMyExtension", "chrome,centerscreen");

The first parameter to window.open is the URI of the XUL file that describes the window and its contents.

The second parameter is the window's name; the name can be used in links or forms as the target attribute. This is different from the user-visible window title, which is specified using XUL.

The third, and optional, parameter is a list of special window features the window should have.

The window.openDialog function works similarly, but lets you specify optional arguments that can be referenced from the JavaScript code. It also handles window features a little differently, including always assuming the dialog feature is specified.

From XPCOM components and modules

If the Window object is unavailable (for example, when opening a window from XPCOM component code), you might want to use nsIWindowWatcher interface. Its parameters are similar to window.open; in fact, window.open's implementation calls nsIWindowWatcher's methods.

var ww = Components.classes["@mozilla.org/embedcomp/window-watcher;1"]
                   .getService(Components.interfaces.nsIWindowWatcher);
var win = ww.openWindow(null, "chrome://myextension/content/about.xul",
                        "aboutMyExtension", "chrome,centerscreen", null);

Window object

Note the win variable in the above section, which is assigned the return value of window.open. It can be used to access the opened window. The return value of window.open (and similar methods) is a Window object (usually ChromeWindow) – the same type as the window variable.

Technically speaking, it implements a number of interfaces, including nsIDOMJSWindow and nsIDOMWindow, but it also contains the user-defined properties for global variables and functions of the window. So, for example, to access the DOM document corresponding to the window, you can use win.document.

Note however, that the open() call returns before the window is fully loaded, so some calls, like win.document.getElementById() will fail. To overcome this difficulty, you can move the initialization code to a load handler of the window being opened or pass a callback function, as described below.

You can get a Window object from a document using document.defaultView.

Content windows

When a XUL window contains a widget capable of displaying a page, such as <browser> or <iframe>, the document in that widget is, naturally, separate from the document of the chrome window itself. There also is a Window object for each sub-document, although there's no window in a common sense for the sub-document.

The same holds for chrome windows opened inside a tab of <tabbrowser>. The elements above the chrome document opened in the tab are separate from your chrome document.

The following two subsections describe how to cross chrome-content boundaries in either way, i.e. accessing elements which are ancestors of your chrome document, or accessing elements which are descendants of your chrome document (but nevertheless in a different context).

Accessing content documents

Assume you have a document loaded in a <tabbrowser>, <browser>, or <iframe> element inside your document. You can use browser.contentDocument to access that document and browser.contentWindow to access the Window object of that document.

You should be aware of XPCNativeWrappers when working with untrusted content. With XPCNativeWrappers turned on (which is the default in Firefox 1.5+), your extension can safely access the DOM of the content document, but not the content JavaScript. Bypassing XPCNativeWrapper to work with content JavaScript directly can lead to security problems.

See Interaction between privileged and non-privileged pages if you need to interact with the content page.

The content shortcut

In case of <browser type="content-primary" />, you can use the content shortcut property to accesss the Window object of the content document. For example:

// alerts the title of the document displayed in the content-primary widget

alert(content.document.title);

For example, you can use content.document in a browser.xul overlay to access the web page in the selected tab in a Firefox window.

Accessing a document in the sidebar

Firefox has a sidebar, which is implemented as a <browser> element with id="sidebar". To access the elements and variables inside the sidebar, you need to use document.getElementById("sidebar").contentDocument or .contentWindow, like when Accessing content documents.

For more sidebar tips, see Code snippets:Sidebar.

Accessing the elements of the top-level document from a child window

The opposite case is when you want to access the chrome document from a privileged script loaded in a <browser> or an <iframe>.

A typical case when this can be useful is when your code runs in the sidebar in the main Firefox window and you want to access the elements in the main browser window.

The DOM tree, as shown by the DOM Inspector, can look like this:

#document
  window                 main-window
    ...
      browser
        #document
          window         myExtensionWindow

where the child window is where your code runs in.

Your task is to access elements above your chrome document, i.e. to break out of your chrome window and access the ancestors. This can be done using the following statement:

var mainWindow = window.QueryInterface(Components.interfaces.nsIInterfaceRequestor)
                   .getInterface(Components.interfaces.nsIWebNavigation)
                   .QueryInterface(Components.interfaces.nsIDocShellTreeItem)
                   .rootTreeItem
                   .QueryInterface(Components.interfaces.nsIInterfaceRequestor)
                   .getInterface(Components.interfaces.nsIDOMWindow);

This allows you to cross the chrome-content boundaries, and returns the main window object.

Finding already opened windows

The window mediator XPCOM component (nsIWindowMediator interface) provides information about existing windows. Two of its methods are often used to obtain information about currently open windows: getMostRecentWindow and getEnumerator. Please refer to the nsIWindowMediator page for more information and examples of using nsIWindowMediator. === Example: Opening a window only if it's not opened already === XXX TBD

Passing data between windows

When working with multiple windows, you often need to pass information from one window to another. Since different windows have separate DOM documents and global objects for scripts, you can't just use one global JavaScript variable in scripts from different windows.

There are several techniques of varying power and simplicity that can be used to share data. We'll demonstrate them from the simplest to the most complex in the next few sections.

Note: if you want to pass data between privileged (chrome) and non-privileged (web page) windows, or vice-versa, read this instead.

Example 1: Passing data to window when opening it with openDialog

When you open a window using window.openDialog or nsIWindowWatcher.openWindow, you can pass an arbitrary number of arguments to that window. Arguments are simple JavaScript objects, accessible through window.arguments property in the opened window.

In this example, we're using window.openDialog to open a progress dialog. We pass in the current status text as well as the maximum and current progress values. Note that using nsIWindowWatcher.openWindow is a bit less trivial . TODO: link to How To Pass an XPCOM Object to a New Window when it has a more useful example

Opener code:

window.openDialog("chrome://test/content/progress.xul",
                  "myProgress", "chrome,centerscreen",
                  {status: "Reading remote data", maxProgress: 50, progress: 10} );

progress.xul:

<?xml version="1.0"?>
<?xml-stylesheet href="chrome://global/skin/" type="text/css"?>

<window onload="onLoad();" xmlns="http://www.mozilla.org/keymaster/gatekeeper/there.is.only.xul">
<script><![CDATA[

var gStatus, gProgressMeter;
var maxProgress = 100;

function onLoad() {
  gStatus = document.getElementById("status");
  gProgressMeter = document.getElementById("progressmeter");

  if("arguments" in window && window.arguments.length > 0) {
    maxProgress = window.arguments[0].maxProgress;
    setProgress(window.arguments[0].progress);
    setStatus(window.arguments[0].status);
  }
}

function setProgress(value) {
  gProgressMeter.value = 100 * value / maxProgress;
}

function setStatus(text) {
  gStatus.value = "Status: " + text + "...";
}

]]></script>

<label id="status" value="(No status)" />
<hbox>
  <progressmeter id="progressmeter" mode="determined" />
  <button label="Cancel" oncommand="close();" />
</hbox>

</window>

Example 2: Interacting with the opener

Sometimes an opened window needs to interact with its opener; for example, it might do so in order to give notice that the user has made changes in the window. You can find the window's opener using its window.opener property or via a callback function passed to the window in a way described in the previous section.

Let's add code to the previous example to notify the opener when the user presses Cancel on the progress dialog.

• Using window.opener. The opener property returns its window's opener; that is, the Window object that opened it.

If we're sure the window that opened the progress dialog declares the cancelOperation function, we can use window.opener.cancelOperation() to notify it, like this:

<button label="Cancel" oncommand="opener.cancelOperation(); close();" />

• Using a callback function. Alternatively, the opener window can pass a callback function to the progress dialog in the same way we passed the status string in the previous example:

function onCancel() {
  alert("Operation canceled!");
}

...

window.openDialog(
  "chrome://test/content/progress.xul",
  "myProgress", "chrome,centerscreen",
  {status: "Reading remote data", maxProgress: 50, progress: 10},
  onCancel
);

The progress dialog can then run the callback like this:

<button label="Cancel" oncommand="window.arguments[1](); close();" />

Example 3: Using nsIWindowMediator when opener is not enough

The window.opener property is very easy to use, but it's only useful when you're sure that your window was opened from one of a few well-known places. In more complicated cases you need to use the nsIWindowMediator interface, introduced above.

One case in which you might want to use nsIWindowMediator is in an extension's Options window. Suppose you're developing a browser extension that consists of a browser.xul overlay and an Options window. Suppose the overlay contains a button to open the extension's Options window which needs to read some data from the browser window. As you may remember, Firefox's Extension Manager can also be used to open your Options dialog.

This means the value of window.opener in your Options dialog is not necessarily the browser window -- instead, it might be the Extension Manager window. You could check the location property of the opener and use opener.opener in case it's the Extension Manager window, but a better way is to use nsIWindowMediator:

var wm = Components.classes["@mozilla.org/appshell/window-mediator;1"]
                   .getService(Components.interfaces.nsIWindowMediator);
var browserWindow = wm.getMostRecentWindow("navigator:browser");
// read values from |browserWindow|

You might be tempted to use a similar technique to apply the changes the user made in the Options dialog, but a better way to do that is to use preferences observers.

Example 4: Using nsIWindowWatcher when you don't have a window (primitives only)

For those times when you don't have access to a window object (such as a JavaScript XPCOM object), you can use nsIWindowWatcher.openWindow to open a XUL window. The problem is passing the data. It's a bit harder than using window.openDialog. Here is helper function that will package the data correctly and pass it to the newly opened window:

function openDialog(parentWindow, url, windowName, features) {
    var array = Components.classes["@mozilla.org/array;1"]
                          .createInstance(Components.interfaces.nsIMutableArray);
    for (var i = 4; i < arguments.length; i++) {
        var variant = Components.classes["@mozilla.org/variant;1"]
                                .createInstance(Components.interfaces.nsIWritableVariant);
        variant.setFromVariant(arguments[i]);
        array.appendElement(variant, false);
    }

    var watcher = Components.classes["@mozilla.org/embedcomp/window-watcher;1"]
                            .getService(Components.interfaces.nsIWindowWatcher);
    return watcher.openWindow(parentWindow, url, windowName, features, array);
}

The function almost works the same as window.openDialog but accepts an optional parent window as first parameter. The parent window can be null. The only data types that can be passed to the new window are primitives and arrays. JavaScript objects cannot be passed. All of the arguments are passed by value, that is to say that any changes made to the values in the new window will not affect the calling code.

Example 5: Using nsIWindowWatcher for passing an arbritrary JavaScript object

It is still possible to pass any JavaScript object using nsIWindowWatcher, it just takes a little more effort. In this example a single object is passed, though it is possible to pass multiple objects and primitives using a combination of this and the example above.

// In the calling code
var args = {
  param1: true,
  param2: 42
};

args.wrappedJSObject = args;

var watcher = Components.classes["@mozilla.org/embedcomp/window-watcher;1"]
                            .getService(Components.interfaces.nsIWindowWatcher);

watcher.openWindow(null, url, windowName, features, args);

// In the window code
var args = window.arguments[0].wrappedJSObject;

This uses the wrappedJSObject trick. By passing args to openWindow xpconnect will automatically wrap it as a generic nsISupports. The opened window can then just get at the underlying JavaScript object using wrappedJSObject.

Advanced data sharing

The above code is useful when you need to pass data from one window to another or to a set of windows, but sometimes you just want to share a JavaScript variable in common between different windows. You could declare a local variable in each window along with corresponding setter functions to keep the "instances" of the variable in sync across windows, but fortunately, there's a better way.

To declare a shared variable, we need to find a place that exists while the application is running and is easily accessible from the code in different chrome windows. There are actually a few such places.

Using JavaScript code modules

JavaScript code modules is a simple method for creating shared global singleton objects that can be imported into any other JavaScript scope. The importing scope will have access to the objects and data in the code module. Since the code module is cached, all scopes get the same instance of the code module and can share the data in the module. See Components.utils.import for more information.

• Pros:
  • It's the "right way"
  • Very simple to make and import.
  • No need to build an XPCOM component.
• Cons:
  • Only works in Firefox 3 or newer.
  • The scope is shared between modules and the importing scope, so you have to worry about name collisions.
  • You can't use the window object, its members, like alert and open, and many other objects available from inside a window. The functionality isn't lost, however -- you just have to use the XPCOM components directly instead of using convenient shortcuts. Of course, this doesn't matter if you just store data in the component.

Using an XPCOM singleton component

The cleanest and most powerful way to share data is to define your own XPCOM component (you can write one in JavaScript) and access it from anywhere using a getService call:

Components.classes["@domain.org/mycomponent;1"].getService();

• Pros:
  • It's the "right way".
  • You can store arbitrary JavaScript objects in the component.
  • The scope is not shared between components, so you don't have to worry about name collisions.
  • Works in older versions of Firefox.
• Cons:
  • You can't use the window object, its members, like alert and open, and many other objects available from inside a window. The functionality isn't lost, however -- you just have to use the XPCOM components directly instead of using convenient shortcuts. Of course, this doesn't matter if you just store data in the component.
  • Learning to create XPCOM components takes time.

There are several articles and books about creating XPCOM components online.

Using FUEL Application object

The FUEL JavaScript library has a simple method for sharing data between windows. The Application object supports a storage property which can be used to store data globally. This method is a simplified form of the XPCOM singleton method.

Application.storage.set(keyname, data);

var data = Application.storage.get(keyname, default);

where: keyname is a string used to identify the data
       data is the data
       default is the data value to return if keyname does not exists

• Pros:
  • Its the "right way".
  • You can store arbitrary JavaScript objects in the component.
  • The scope is not shared between components, so you don't have to worry about name collisions.
  • No need to build an XPCOM component.
• Cons:
  • Only works in Firefox 3 or newer.
  • You can't use the window object, its members, like alert and open, and many other objects available from inside a window. The functionality isn't lost, however -- you just have to use the XPCOM components directly instead of using convenient shortcuts. Of course, this doesn't matter if you just store data in the component.

Storing shared data in preferences

If you just need to store a string or a number, writing a whole XPCOM component may be an unnecessary complication. You can use the preferences service in such cases.

• Pros:
  • Quite easy to use for storing simple data.
• Cons:
  • Can't easily be used to store complex data.
  • Abusing the preferences service and not cleaning up after yourself can cause prefs.js to grow large and slow down application startup.

See Code snippets:Preferences for detailed description of the preferences system and example code.

Example:

var prefs = Components.classes["@mozilla.org/preferences-service;1"]
                      .getService(Components.interfaces.nsIPrefService);
var branch = prefs.getBranch("extensions.myext.");
var var1 = branch.getBoolPref("var1"); // get a pref

See also

• Code snippets:Dialogs