support common content-language mistakes

chrome/renderer/translate_helper.cc

 // Copyright (c) 2012 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 "chrome/renderer/translate_helper.h"
 
 #include "base/bind.h"
 #include "base/compiler_specific.h"
 #include "base/logging.h"
 #include "base/message_loop.h"
@@ skipping 60 matching lines @@
   // language of the intended audience (a distinction really only
   // relevant for things like langauge textbooks).  This distinction
   // shouldn't affect translation.
   std::string language = document.contentLanguage().utf8();
   size_t coma_index = language.find(',');
   if (coma_index != std::string::npos) {
     // There are more than 1 language specified, just keep the first one.
     language = language.substr(0, coma_index);
   }
   TrimWhitespaceASCII(language, TRIM_ALL, &language);
-  language = StringToLowerASCII(language);
+
+  // Underscores instead of dashes are a frequent mistake.
+  size_t underscore_index = language.find('_');
+  if (underscore_index != std::string::npos) {
+    language[underscore_index] = '-';

[MAD, 2012/11/06 18:00:20]

No need for {} here...

[bcwhite, 2012/11/06 18:37:49]

Doh!  I am going to do that _every_ time!
Fixed.

+  }

[MAD, 2012/11/06 18:00:20]

Add a DCHECK that there are not other '_' in the string.

[bcwhite, 2012/11/06 18:37:49]

We're parsing user-supplied input here.  Is crashing as a result of a web-page
really a good idea, even if only in debug mode?

The translation code strips everything after the first dash if it doesn't find a
match and tries again.

[MAD, 2012/11/06 18:39:30]

OK, fine...

+
+  // Change everything up to a dash to lower-case and everything after to upper.
+  size_t dash_index = language.find('-');
+  if (dash_index != std::string::npos) {
+    language = StringToLowerASCII(language.substr(0, dash_index)) +

[MAD, 2012/11/06 18:00:20]

Add a DCHECK that there are no other '-' in the string

[bcwhite, 2012/11/06 18:37:49]

Same as above.  I don't think it's a problem as malformed input just leads to
malformed output and will either match only the base language or get ignored for
no match.

+        StringToUpperASCII(language.substr(dash_index));
+  } else {
+    language = StringToLowerASCII(language);
+  }
 
   if (language.empty()) {
     base::TimeTicks begin_time = base::TimeTicks::Now();
     language = DetermineTextLanguage(contents);
     UMA_HISTOGRAM_MEDIUM_TIMES("Renderer4.LanguageDetection",
                                base::TimeTicks::Now() - begin_time);
   } else {
     VLOG(9) << "PageLanguageFromMetaTag: " << language;
   }
 
@@ skipping 334 matching lines @@
   WebView* web_view = render_view()->GetWebView();
   if (!web_view) {
     // When the WebView is going away, the render view should have called
     // CancelPendingTranslation() which should have stopped any pending work, so
     // that case should not happen.
     NOTREACHED();
     return NULL;
   }
   return web_view->mainFrame();
 }