support common content-language mistakes

chrome/renderer/translate_helper.cc

Index: chrome/renderer/translate_helper.cc
diff --git a/chrome/renderer/translate_helper.cc b/chrome/renderer/translate_helper.cc
index e2d23bf25baec9cc2e0073ddbbb04ff4a782ede6..382b60ce3eba7fe383f389d1a0d7f400c18c4237 100644
--- a/chrome/renderer/translate_helper.cc
+++ b/chrome/renderer/translate_helper.cc
@@ -78,7 +78,21 @@ void TranslateHelper::PageCaptured(const string16& contents) {
     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();