Add ContentViewDelegate

android_webview/native/aw_contents_io_thread_delegate.h

Index: android_webview/native/aw_contents_io_thread_delegate.h
diff --git a/android_webview/native/aw_contents_io_thread_delegate.h b/android_webview/native/aw_contents_io_thread_delegate.h
new file mode 100644
index 0000000000000000000000000000000000000000..e266773f68086517dc4c62b61eb63d06fb6dc008
--- /dev/null
+++ b/android_webview/native/aw_contents_io_thread_delegate.h
@@ -0,0 +1,73 @@
+// Copyright (c) 2012 The Chromium Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style license that can be
+// found in the LICENSE file.
+
+#ifndef ANDROID_WEBVIEW_NATIVE_AW_CONTENTS_IO_THREAD_DELEGATE_H_
+#define ANDROID_WEBVIEW_NATIVE_AW_CONTENTS_IO_THREAD_DELEGATE_H_
+
+#include "base/android/scoped_java_ref.h"
+#include "base/basictypes.h"
+#include "base/memory/scoped_ptr.h"
+
+class GURL;
+class InterceptedRequestData;
+
+namespace content {
+class RenderViewHost;
+struct Referrer;
+class WebContents;
+}
+
+namespace net {
+class URLRequest;
+}
+
+// This lets us implement URLRequest intercepting in a way that lets the
+// embedder associate URLRequests with the WebContents that the URLRequest
+// is made for.
+//
+// It is preferable to use the static getter methods to obtain a new instance
+// of the class rather than holding on to one for prolonged periods of time
+// (see note for more details).
+//
+// Note: The native AwWebContentsIoThreadDelegate instance has a Global ref to
+// the Java object. By keeping the native AwWebContentsIoThreadDelegate
+// instance alive you're also prolonging the lifetime of the Java instance, so
+// don't keep a AwWebContentsIoThreadDelegate if you don't need to.
+// It is also possible that AwWebContentsIoThreadDelegate becomes invalid when
+// you're holding on to it. This could happen if the WebContents is destroyed.

[joth, 2012/09/01 01:24:14]

if the WC is destroyed, the  AwWebContentsIoThreadDelegate instance will still
be bound to the java side class and will still call through to it? 

[mkosiba (inactive), 2012/09/11 01:22:25]

the Java side class will exist, but we only make calls from native to Java and
not the other way around.

+// In that case the instance methods will noop and return default/null values.
+class AwWebContentsIoThreadDelegate {
+ public:
+  // This will attempt to fetch the AwWebContentsIoThreadDelegate for the given
+  // |rvh|.
+  // This method can be called from any thread.
+  // An empty scoped_ptr is a valid return value.
+  static scoped_ptr<AwWebContentsIoThreadDelegate>
+      ForRenderViewHost(content::RenderViewHost* rvh);

[joth, 2012/09/01 01:24:14]

returning scoped_ptr (returning ownership) makes it sound like this should be a
CreateFoo() type method.

[mkosiba (inactive), 2012/09/11 01:22:25]

it's a bit uncommon but the intent behind this is that we're not really creating
a new object, we're only new-ing up the native JNI binding to it.

+
+  // This will attempt to fetch the AwWebContentsIoThreadDelegate for the given
+  // This method can be called from any thread.
+  // |contents|. An empty scoped_ptr is a valid return value.
+  static scoped_ptr<AwWebContentsIoThreadDelegate>
+      ForWebContents(content::WebContents* contents);

[joth, 2012/09/01 01:24:14]

Can't quite follow the usage model for these without seeing the code that uses
them. Questions in my mind are:

- Do we actually need both versions? I'd have thought we always lookup via just
RVH or web contents
- is it really new-ing up a new native object instance on every request? i guess
there's no problem in that, but seems unusual.
- what thread(s) will each these be called on?

[mkosiba (inactive), 2012/09/11 01:22:25]

yeath.. I'll clean this up. We should be getting the Delegate via a (rvh_id,
rview_id) pair.
yes, this is because the native object "pins" the java-side object.
should be possible to call this from any thread.

+
+  static bool RegisterAwWebContentsIoThreadDelegate(JNIEnv* env);
+
+  // This method is called on the IO thread only.
+  scoped_ptr<InterceptedRequestData> ShouldInterceptRequest(
+      const net::URLRequest* request);
+
+  ~AwWebContentsIoThreadDelegate();
+
+  class WebContentsObserver;

[joth, 2012/09/01 01:24:14]

nit: call this something about its job, JavaDelegateMapMaintainer or something.
(it could be an anon class in the .cc I think)

[mkosiba (inactive), 2012/09/11 01:22:25]

Renamed, has to be inner/friend in order to be able to use the private c'tor.

+ private:
+  DISALLOW_COPY_AND_ASSIGN(AwWebContentsIoThreadDelegate);

[joth, 2012/09/01 01:24:14]

should be last.

the WCO forward declaration isn't needed?

d'tor could also be private.

[mkosiba (inactive), 2012/09/11 01:22:25]

Done
it's needed, otherwise the WCO won't be able to instantiate new AWCIOTD's. I
moved the thing to be a friend class to keep the method declarations in the .cc
file shorter.
sadly scoped_ptr insists on calling the destructor so I'll need to keep it
public.

+
+  AwWebContentsIoThreadDelegate(const base::android::JavaRef<jobject>& obj);
+  void OnWebContentsDestroyed();
+
+  base::android::ScopedJavaGlobalRef<jobject> java_object_;
+};
+
+#endif  // ANDROID_WEBVIEW_NATIVE_AW_CONTENTS_IO_THREAD_DELEGATE_H_