Index: docs/tab_helpers.md |
diff --git a/docs/tab_helpers.md b/docs/tab_helpers.md |
new file mode 100644 |
index 0000000000000000000000000000000000000000..c68218b3370c54ada9a48a83db7310d4e5c24d40 |
--- /dev/null |
+++ b/docs/tab_helpers.md |
@@ -0,0 +1,119 @@ |
+# 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 `WebContents`es are used to build tabs in browser windows. |
+ |
+[TOC] |
+ |
+## 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](https://code.google.com/p/chromium/codesearch#chromium/src/content/public/browser/web_contents_observer.h&q=webcontentsobserver) |
+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](https://code.google.com/p/chromium/codesearch#chromium/src/chrome/browser/ui/tabs/tab_strip_model.cc&q=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`](https://code.google.com/p/chromium/codesearch#chromium/src/base/supports_user_data.h&q=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`](https://code.google.com/p/chromium/codesearch#chromium/src/content/public/browser/web_contents_user_data.h&q=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, bool explicit_set) 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 `WebContents`es that are used for the browser tabs? |
+ |
+## AttachTabHelpers |
+ |
+There is a function called |
+[`AttachTabHelpers()`](https://code.google.com/p/chromium/codesearch#chromium/src/chrome/browser/ui/tab_helpers.cc&q=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 `WebContents`es |
+ |
+Sometimes it's useful to re-use tab helpers for `WebContents`es 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()`](https://code.google.com/p/chromium/codesearch#chromium/src/chrome/browser/ui/apps/chrome_app_delegate.cc&q=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 |
+`WebContents`es 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 `WebContents`es, be aware of their source. It is extremely |
+rare to have to be able to handle arbitrary `WebContents`es. Know where they |
+come from and what tab helpers are on them, and you'll be fine. |