RulesRegistry for declarative APIs.

chrome/browser/extensions/extension_service.cc

Index: chrome/browser/extensions/extension_service.cc
diff --git a/chrome/browser/extensions/extension_service.cc b/chrome/browser/extensions/extension_service.cc
index 84397982cc4e10874b4d7d1a24dfd29d71d0adef..0c6bfb320752c56c24158c2d11d39ff6b90e1b54 100644
--- a/chrome/browser/extensions/extension_service.cc
+++ b/chrome/browser/extensions/extension_service.cc
@@ -36,6 +36,7 @@
 #include "chrome/browser/extensions/apps_promo.h"
 #include "chrome/browser/extensions/component_loader.h"
 #include "chrome/browser/extensions/crx_installer.h"
+#include "chrome/browser/extensions/api/declarative/rules_registry_service.h"
 #include "chrome/browser/extensions/default_apps_trial.h"
 #include "chrome/browser/extensions/extension_browser_event_router.h"
 #include "chrome/browser/extensions/extension_cookies_api.h"
@@ -1169,6 +1170,9 @@ void ExtensionService::NotifyExtensionUnloaded(
 
   if (plugins_changed || nacl_modules_changed)
     PluginService::GetInstance()->PurgePluginListCache(profile_, false);
+
+  if (rules_registry_service_.get())
+    rules_registry_service_->OnExtensionUnloaded(extension->id());

[not at google - send to devlin, 2012/02/07 03:25:10]

Rather than this, have the RulesRegistryService listen to EXTENSION_UNLOADED
events.

git grep -w NOTIFICATION_EXTENSION_UNLOADED

[battre, 2012/02/07 18:45:33]

I had that before and removed it once I transferred ownership of the rules
registry service from the profile to the extension service. I think that given
the direct ownership I prefer the current implementation over some broadcasting
mechanism. Would you want to subscribe each RulesRegistry to
NOTIFICATION_EXTENSION_UNLOADED?

[not at google - send to devlin, 2012/02/07 23:54:57]

I agree that it's a shame we use a broadcasting mechanism for this sort of
stuff. IMO in general nicer design would be to have some object be responsible
for the installation of extensions (currently that's ExtensionService), then
anything that wants to be notified of that can register listeners on
ExtensionService directly.

I guess one problem with listener architecture in C++ is that lifetimes are more
difficult to express; you rely on ExtensionService to outlive the object which
is listening.

Which is presumably why they used an event bus architecture. It's also easier to
write, and grep for.

In any case, that's what we have. It shouldn't be a concern of ExtensionService
that the RulesRegistry wants to know when extensions are uninstalled.

So yeah, you'd want to subscribe RulesRegistry to
NOTIFICATION_EXTENSION_UNLOADED.

[battre, 2012/02/08 12:49:54]

Done.

 }
 
 void ExtensionService::UpdateExtensionBlacklist(
@@ -2699,3 +2703,9 @@ extensions::SocketController* ExtensionService::socket_controller() {
   }
   return socket_controller_;
 }
+
+extensions::RulesRegistryService* ExtensionService::GetRulesRegistryService() {
+  if (!rules_registry_service_.get())
+    rules_registry_service_.reset(new extensions::RulesRegistryService());
+  return rules_registry_service_.get();
+}