Hard coded finch config for Popular site

components/ntp_tiles/field_trial.cc

Index: components/ntp_tiles/field_trial.cc
diff --git a/components/ntp_tiles/field_trial.cc b/components/ntp_tiles/field_trial.cc
new file mode 100644
index 0000000000000000000000000000000000000000..779b5b85953135f38dcbd633fe17a5e56fb5ccef
--- /dev/null
+++ b/components/ntp_tiles/field_trial.cc
@@ -0,0 +1,82 @@
+// Copyright 2016 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.
+
+#include "base/metrics/field_trial.h"
+
+#if defined(OS_ANDROID)
+#include <jni.h>
+#endif
+
+#include "base/command_line.h"
+#include "base/metrics/field_trial.h"
+#include "base/strings/string_util.h"
+#include "components/ntp_tiles/constants.h"
+#include "components/ntp_tiles/field_trial.h"
+#include "components/ntp_tiles/switches.h"
+
+#if defined(OS_ANDROID)
+#include "base/android/jni_android.h"
+#include "jni/MostVisitedSites_jni.h"
+#endif
+
+const char kPopularSiteDefaultGroup[] = "Default";
+const char kPopularSiteControlGroup[] = "Control";
+const char kPopularSiteEnabledGroup[] = "Enabled";
+const char kPopularSiteEnabledCommandLineSwitchGroup[] =
+    "EnabledWithCommandLineSwitch";
+const char kPopularSiteDisabledCommandLineSwitchGroup[] =
+    "DisabledWithCommandLineSwitch";
+
+void ntp_tiles::SetUpFirstLaunchFieldTrial(bool is_stable_channel) {

[rkaplow, 2016/11/10 22:56:46]

can you comment this is done as a workaround since ios first launch experiments
are not available

[noyau (Ping after 24h), 2016/11/14 17:08:08]

Done.

+  // Stable channels will run with 10% probability.

[rkaplow, 2016/11/10 22:56:46]

this is higher a % that we usually experiment on stable, 1 is much more common

[noyau (Ping after 24h), 2016/11/14 17:08:08]

Not on iOS, were volume are low compared to desktop and android. Plus this
particular experiement will only touch new users, not all users, as users who
have used Chrome before will already have the slots on the NTP filed with most
likely or most visited.

[rkaplow, 2016/11/14 23:42:51]

Ok.. in either case this will need to get approval by L-R - and I would
highlight this is 10%.

[noyau (Ping after 24h), 2016/11/15 09:05:50]

The highlight is already there: The first sentence of the launch bug mention the
10%. This value already has approval from Chrome Mobile leads, all the involved
PMs and representatives from Search. Trust me, this was not set on a whim :)

+  // Non-stable channels will run with 50% probability.
+  const base::FieldTrial::Probability kTotalProbability = 100;
+  const base::FieldTrial::Probability kEnabledAndControlProbability =
+      is_stable_channel ? 10 : 50;
+
+  // Experiment enabled until March 15, 2017. By default, disabled.

[noyau (Ping after 24h), 2016/11/09 10:56:39]

Before creating the new experiment should I check that if it has been overriden
by a server side config? Or should I plan for two experiment, this one hardcoded
and an emergency one to turn everything off from the server side?

[rkaplow, 2016/11/10 22:56:46]

I think it would make sense to check in case here the study has been setup - you
will get bad behavior if it already is. 
If it is not, then you can set it up manually as you are doing here

[noyau (Ping after 24h), 2016/11/14 17:08:08]

Is Find the correct way to test for this without triggering being put in a
group? This is too early to know if the user meets the conditions of the
experiment, and if I call FindFullName(kPopularSiteFieldTrialName) it will leads
to incorrect numbers as everyone will be counted.  

[rkaplow, 2016/11/14 23:42:51]

I think TrialExists should work since we don't need to know the specifics of
what the group is.

[noyau (Ping after 24h), 2016/11/15 09:05:50]

Done.

+  scoped_refptr<base::FieldTrial> trial(
+      base::FieldTrialList::FactoryGetFieldTrial(
+          kPopularSitesFieldTrialName, kTotalProbability,
+          kPopularSiteDefaultGroup, 2017, 3, 15,  // Mar 15, 2017
+          base::FieldTrial::ONE_TIME_RANDOMIZED, NULL));
+
+  base::CommandLine* command_line = base::CommandLine::ForCurrentProcess();
+  if (command_line->HasSwitch(ntp_tiles::switches::kEnableNTPPopularSites)) {
+    trial->AppendGroup(kPopularSiteEnabledCommandLineSwitchGroup,
+                       kTotalProbability);
+  } else if (command_line->HasSwitch(
+                 ntp_tiles::switches::kDisableNTPPopularSites)) {
+    trial->AppendGroup(kPopularSiteDisabledCommandLineSwitchGroup,
+                       kTotalProbability);
+  } else {
+    trial->AppendGroup(kPopularSiteControlGroup, kEnabledAndControlProbability);
+    trial->AppendGroup(kPopularSiteEnabledGroup, kEnabledAndControlProbability);
+  }
+  trial->group();
+}
+
+bool ntp_tiles::ShouldShowPopularSites() {
+  // Note: It's important to query the field trial state first, to ensure that
+  // UMA reports the correct group.
+  const std::string group_name =
+      base::FieldTrialList::FindFullName(kPopularSitesFieldTrialName);
+
+  base::CommandLine* cmd_line = base::CommandLine::ForCurrentProcess();
+  if (cmd_line->HasSwitch(switches::kDisableNTPPopularSites))
+    return false;
+
+  if (cmd_line->HasSwitch(switches::kEnableNTPPopularSites))
+    return true;
+
+#if defined(OS_ANDROID)
+  if (Java_MostVisitedSites_isPopularSitesForceEnabled(
+          base::android::AttachCurrentThread())) {
+    return true;
+  }
+#endif
+
+  return base::StartsWith(group_name, "Enabled",
+                          base::CompareCase::INSENSITIVE_ASCII);
+}