Modifying language detection code

extensions/renderer/i18n_custom_bindings.cc

Index: extensions/renderer/i18n_custom_bindings.cc
diff --git a/extensions/renderer/i18n_custom_bindings.cc b/extensions/renderer/i18n_custom_bindings.cc
index 7799d9d01a14917df75658b4d8434e471ba9d662..be20d3829de7199b84da66073cdf359ada1cefc2 100644
--- a/extensions/renderer/i18n_custom_bindings.cc
+++ b/extensions/renderer/i18n_custom_bindings.cc
@@ -10,6 +10,7 @@
 #include <vector>
 
 #include "base/bind.h"
+#include "base/logging.h"
 #include "base/macros.h"
 #include "base/memory/ptr_util.h"
 #include "base/metrics/histogram_macros.h"
@@ -135,6 +136,7 @@ void InitDetectedLanguages(
     LanguageDetectionResult* result) {
   std::vector<std::unique_ptr<DetectedLanguage>>* detected_languages =
       &result->languages;
+  DCHECK(detected_languages->empty());
   bool* is_reliable = &result->is_reliable;
 
   // is_reliable is set to "true", so that the reliability can be calculated by
@@ -144,20 +146,31 @@ void InitDetectedLanguages(
     const chrome_lang_id::NNetLanguageIdentifier::Result& lang_result =
         lang_results.at(i);
     const std::string& language_code = lang_result.language;
+
+    // If the first language is kUnknown, then all languages are kUnknown.
     if (language_code == chrome_lang_id::NNetLanguageIdentifier::kUnknown) {
-      // If the first language is kUnknown, then all languages are kUnknown.
-      // Thus, is_reliable is set to "false".
-      if (i == 0) {
-        *is_reliable = false;
-      }
       break;
     }
 
+    // The list of languages supported by CLD3 is saved in kLanguageNames
+    // in the following file:
+    // //src/third_party/cld_3/src/src/task_context_params.cc
+    // To maintain backwards compatibility, xx-Latn predictions are ignored

[Devlin, 2016/09/27 22:34:28]

We should still provide more detail here.  Why does a -Latn suffix break
existing functionality?  The goal here is to have a reader understand (at least
mostly) the motivation and what would change in the future.

If you prefer, you can also file a bug at crbug.com and link to it here.

[Anton Bakalov, 2016/09/28 04:23:26]

Done.

+    // for now.
+    if (base::EndsWith(language_code, "-Latn",
+                       base::CompareCase::INSENSITIVE_ASCII)) {
+      continue;
+    }
+
     *is_reliable = *is_reliable && lang_result.is_reliable;
     const int percent = static_cast<int>(100 * lang_result.proportion);
     detected_languages->push_back(
         base::MakeUnique<DetectedLanguage>(language_code, percent));
   }
+
+  if (detected_languages->empty()) {
+    *is_reliable = false;
+  }
 }
 #else
 # error "CLD_VERSION must be 2 or 3"