Move DOM-inspecting language detection logic to Blink.

chrome/renderer/chrome_render_frame_observer.cc

Index: chrome/renderer/chrome_render_frame_observer.cc
diff --git a/chrome/renderer/chrome_render_frame_observer.cc b/chrome/renderer/chrome_render_frame_observer.cc
index 9fd19893f6daa9b2a071ad61e512ce34570c44f0..1a210f5ce2b3c1f1a38914e7b4b158f4caf4f188 100644
--- a/chrome/renderer/chrome_render_frame_observer.cc
+++ b/chrome/renderer/chrome_render_frame_observer.cc
@@ -28,6 +28,7 @@
 #include "net/base/net_util.h"
 #include "skia/ext/image_operations.h"
 #include "third_party/WebKit/public/platform/WebImage.h"
+#include "third_party/WebKit/public/platform/WebLanguageDetectionDetails.h"
 #include "third_party/WebKit/public/platform/modules/app_banner/WebAppBannerPromptReply.h"
 #include "third_party/WebKit/public/web/WebDataSource.h"
 #include "third_party/WebKit/public/web/WebDocument.h"
@@ -47,16 +48,13 @@
 
 using blink::WebDataSource;
 using blink::WebElement;
+using blink::WebLanguageDetectionDetails;
 using blink::WebLocalFrame;
 using blink::WebNode;
 using blink::WebString;
 using content::SSLStatus;
 using content::RenderFrame;
 
-// Maximum number of characters in the document to index.
-// Any text beyond this point will be clipped.
-static const size_t kMaxIndexChars = 65535;
-
 // Constants for UMA statistic collection.
 static const char kTranslateCaptureText[] = "Translate.CaptureText";
 
@@ -352,21 +350,28 @@ void ChromeRenderFrameObserver::CapturePageText(TextCaptureType capture_type) {
 
   base::TimeTicks capture_begin_time = base::TimeTicks::Now();
 
-  // Retrieve the frame's full text (up to kMaxIndexChars), and pass it to the
-  // translate helper for language detection and possible translation.
-  base::string16 contents = frame->contentAsText(kMaxIndexChars);
+  WebLanguageDetectionDetails incoming_details =
+      frame->document().languageDetectionDetails();
 
   UMA_HISTOGRAM_TIMES(kTranslateCaptureText,
                       base::TimeTicks::Now() - capture_begin_time);
 
   // We should run language detection only once. Parsing finishes before
   // the page loads, so let's pick that timing.
-  if (translate_helper_ && capture_type == PRELIMINARY_CAPTURE)
-    translate_helper_->PageCaptured(contents);
+  if (translate_helper_ && capture_type == PRELIMINARY_CAPTURE) {
+    translate::LanguageDetectionDetails details;
+    details.url = incoming_details.url;
+    details.content_language = incoming_details.contentLanguage.utf8();
+    details.html_root_language = incoming_details.htmlLanguage.utf8();
+    details.has_notranslate = incoming_details.hasNoTranslateMeta;
+    details.contents = incoming_details.content;
+    translate_helper_->LanguageDetectionDetailsCaptured(details);
+  }
 
   TRACE_EVENT0("renderer", "ChromeRenderFrameObserver::CapturePageText");
 
 #if defined(SAFE_BROWSING_CSD)
+  base::string16 contents = incoming_details.content;
   // Will swap out the string.
   if (phishing_classifier_)
     phishing_classifier_->PageCaptured(&contents,