Adds ChineseScriptClassifier to predict zh-Hant or zh-Hans for input detected as zh.

components/translate/core/language_detection/chinese_script_classifier.h

Index: components/translate/core/language_detection/chinese_script_classifier.h
diff --git a/components/translate/core/language_detection/chinese_script_classifier.h b/components/translate/core/language_detection/chinese_script_classifier.h
new file mode 100644
index 0000000000000000000000000000000000000000..c4243c32286c14a08f8eb932997ffcf3d583169f
--- /dev/null
+++ b/components/translate/core/language_detection/chinese_script_classifier.h
@@ -0,0 +1,49 @@
+// Copyright 2017 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.
+
+#ifndef COMPONENTS_TRANSLATE_CORE_LANGUAGE_DETECTION_CHINESE_SCRIPT_CLASSIFIER_H_
+#define COMPONENTS_TRANSLATE_CORE_LANGUAGE_DETECTION_CHINESE_SCRIPT_CLASSIFIER_H_
+
+#include <memory>
+#include <string>
+#include "third_party/icu/source/i18n/unicode/translit.h"
+
+namespace translate {
+
+class ChineseScriptClassifier {
+ public:
+  // Initializes both the Hant-to-Hans ICU transliterator and the
+  // Hans-to-Hant ICU transliterator.
+  ChineseScriptClassifier();
+  ~ChineseScriptClassifier();
+
+  // Given Chinese text as input, returns either zh-Hant or zh-Hans.
+  // When the input is ambiguous, i.e. not completely zh-Hans and not
+  // completely zh-Hant, this function returns the closest language code
+  // matching the input.
+  //
+  // Behavior is undefined for non-Chinese input.
+  std::string Classify(const std::string& input) const;
+
+  // Returns true if the underlying transliterators were properly initialized
+  // by the constructor.
+  bool IsInitialized() const;

[groby-ooo-7-16, 2017/03/08 00:41:55]

Maybe use a factory mode instead? (It's a personal peeve - so if you feel
strongly about ctor/IsInitialized instead, I won't hold up the CL for it)

[riesa, 2017/03/08 01:47:42]

My understanding is that the factory model is not ideal for very frequently
allocated classes since it is allocated on the heap which adds a (slight) time
and memory overhead. Due to the structure of language_detection_utils I think
that is our case right now.

+
+ private:
+  // BCP 47 language code representing Chinese in Han Simplified script.
+  static const char kChineseSimplifiedCode[];

[groby-ooo-7-16, 2017/03/08 00:41:55]

Why keep those as class members, as opposed to anon namespace?

[riesa, 2017/03/08 01:47:42]

Just a style/organization preference. But in a massive codebase like Chromium
maybe I guess there can be some functional benefits to moving these into an
anonymous namespace in the .cc file. Done.

+
+  // BCP 47 language code representing Chinese in Han Traditional script.
+  static const char kChineseTraditionalCode[];
+
+  // ICU Transliterator that does Hans to Hant conversion.
+  std::unique_ptr<icu::Transliterator> hans2hant_;
+
+  // ICU Transliterator that does Hant to Hans conversion.
+  std::unique_ptr<icu::Transliterator> hant2hans_;
+};
+
+}  // namespace translate
+
+#endif  // COMPONENTS_TRANSLATE_CORE_LANGUAGE_DETECTION_CHINESE_SCRIPT_CLASSIFIER_H_