Tab Helpers

The content/ layer of Chromium has a class called WebContents, which is one of the most basic building blocks of all of Chromium. This document describes how WebContentses are used to build tabs in browser windows.

Introduction

What is a “tab helper”? It is a WebContentsObserver owned by the WebContents itself. Let's break that down.

WebContentsObserver

WebContentsObserver is a simple interface that allows an object to observe events in the life of a WebContents. As an example, if we look at the TabStripModel, there are times when it need to watch out for WebContents being deleted. So it creates a DeletionObserver. The DeletionObserver overrides WebContentsDestroyed(), and when a WebContents gets destroyed, the callback is called and the DeletionObserver processes the message. Note that DeletionObserver is not owned by the WebContents. It is owned indirectly by the TabStripModel.

SupportsUserData and WebContentsUserData

There is a mechanism used in Chromium called SupportsUserData that allows attaching of arbitrary objects to an object. The mechanism is simple: host objects derive from SupportsUserData, and owned objects derive from SupportsUserData::Data. There are three calls to attach and detach the data.

WebContents derives from SupportsUserData, so that mechanism works for attaching objects to a WebContents, but the SupportsUserData mechanism is a bit low-level. A higher level abstraction is WebContentsUserData, which is easy to derive from and has easy-to-use functionality in CreateForWebContents() and FromWebContents().

Adding a feature to a browser tab

Let's combine WebContentsObserver and WebContentsUserData together, to log whenever the title of a tab changes.

class TitleLoggerTabHelper
    : public content::WebContentsObserver,
      public content::WebContentsUserData<TitleLoggerTabHelper> {
 public:
  ~TitleLoggerTabHelper() override;

  // content::WebContentsObserver
  void TitleWasSet(NavigationEntry* entry) override {
      LOG(INFO) << "Title: " << entry->GetTitle();
  }

 private:
  explicit TitleLoggerTabHelper(content::WebContents* web_contents);
  friend class content::WebContentsUserData<TitleLoggerTabHelper>;

  DISALLOW_COPY_AND_ASSIGN(TitleLoggerTabHelper);
};

DEFINE_WEB_CONTENTS_USER_DATA_KEY(TitleLoggerTabHelper);

We want each tab to have this WebContentsObserver attached to it, so that it will properly handle the events it's looking for, and when the tab goes away, then this tab helper will go away too.

But how do you hook in to browser tab creation? How can we attach this tab helper to the WebContentses that are used for the browser tabs?

AttachTabHelpers

There is a function called AttachTabHelpers(). Whenever a WebContents is created for use as a browser tab, AttachTabHelpers() is called. Every tab helper from around Chromium, from ContentSettings to Favicons to History to Prefs, all take this opportunity to hook into those WebContents used as tabs.

If you are writing a feature that needs to deal with browser tabs, this is where you go. Create a tab helper, and add it (in alphabetical order, please!) to AttachTabHelpers(). Note, though, that you are never allowed to call AttachTabHelpers() yourself. AttachTabHelpers() is only for WebContents that are in browser tabs, and all of those code paths are already written.

Reusing tab helpers with non-browser tab WebContentses

Sometimes it‘s useful to re-use tab helpers for WebContentses that aren’t browser tabs. For example, the Chrome Apps code wants to be able to print, and wants to use the printing code that browser tabs use. So in ChromeAppDelegate::InitWebContents() we see that whenever the Apps code creates a new WebContents, it attaches a carefully-chosen subset of tab helpers, including two printing ones.

You can do that too. If you are creating a WebContents, make a very deliberate decision about which tab helpers you need. Chances are, you don't need them all; you probably only need a handful. In fact, most tab helpers assume they are attached to browser tabs, so only add the bare minimum.

Not every WebContents has every tab helper

The other consequence of this design is that you can't make the assumption that an arbitrary WebContents will have an arbitrary tab helper. The WebContentses used as browser tabs likely will have most tab helpers (though not necessarily all of them!) but a WebContents only has a tab helper if it is installed on it.

The deeper (false and dangerous) assumption is that every WebContents is a browser tab. Do not assume that either!

If your code handles WebContentses, be aware of their source. It is extremely rare to have to be able to handle arbitrary WebContentses. Know where they come from and what tab helpers are on them, and you'll be fine.