Report frequency of single page app navigations to UMA

third_party/WebKit/Source/core/loader/FrameLoader.cpp

Index: third_party/WebKit/Source/core/loader/FrameLoader.cpp
diff --git a/third_party/WebKit/Source/core/loader/FrameLoader.cpp b/third_party/WebKit/Source/core/loader/FrameLoader.cpp
index 23b058dc70155aed38401d5d15f0860ebd8d6adb..d69ecc288c36b6ee998dd298be862b93e340a059 100644
--- a/third_party/WebKit/Source/core/loader/FrameLoader.cpp
+++ b/third_party/WebKit/Source/core/loader/FrameLoader.cpp
@@ -79,6 +79,7 @@
 #include "core/probe/CoreProbes.h"
 #include "core/svg/graphics/SVGImage.h"
 #include "core/xml/parser/XMLDocumentParser.h"
+#include "platform/Histogram.h"
 #include "platform/InstanceCounters.h"
 #include "platform/PluginScriptForbiddenScope.h"
 #include "platform/ScriptForbiddenScope.h"
@@ -484,6 +485,9 @@ void FrameLoader::UpdateForSameDocumentNavigation(
     HistoryScrollRestorationType scroll_restoration_type,
     FrameLoadType type,
     Document* initiating_document) {
+  UMA_HISTOGRAM_ENUMERATION(
+      "RendererScheduler.UpdateForSameDocumentNavigationCount", type,
+      FrameLoadType::kFrameLoadTypeCount);

[Nate Chapin, 2017/06/14 18:10:50]

FrameLoadTypes might be overloaded. For example, FrameLoadTypeStandard could be
a fragment navigation that adds a history entry, or it could result from a
history.pushState() JS call. Do you want to split those down further?

Or more generally: What are you trying to accomplish with this data? It's not
clear to me what you want to learn from this histogram.

[tdresser, 2017/06/14 18:30:52]

Great question.

The high level question we want to answer is "approximately what fraction of
navigations are within-app navigations."

We just want an approximation, so we figured just counting the number of same
single page app navigations, and comparing to a separate page load metric would
be good enough. 

Since there was already a convenient enum lying around to figure out where same
document navigations were coming from, we figured we might as well split by that
enum.

We could probably also get away with just recording the total number of
navigations with a load type of:
- kStandard
- kReplaceCurrentItem

Does that sound about right? I don't have a ton of context in this space, let us
know if we're completely off base here.

[Liquan (Max) Gu, 2017/06/20 23:35:39]

Done.

   TRACE_EVENT1("blink", "FrameLoader::updateForSameDocumentNavigation", "url",
                new_url.GetString().Ascii().data());