Detecting form locale

chrome/browser/autofill/autofill_locale_model.cc

Index: chrome/browser/autofill/autofill_locale_model.cc
diff --git a/chrome/browser/autofill/autofill_locale_model.cc b/chrome/browser/autofill/autofill_locale_model.cc
new file mode 100644
index 0000000000000000000000000000000000000000..253744ff55790ab408558aa865d4480558724ae5
--- /dev/null
+++ b/chrome/browser/autofill/autofill_locale_model.cc
@@ -0,0 +1,245 @@
+// Copyright (c) 2010 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/browser/autofill/autofill_locale_model.h"
+
+#include "base/string_util.h"
+#include "chrome/browser/autofill/form_structure.h"
+#include "chrome/browser/tab_contents/language_state.h"
+#include "chrome/browser/tab_contents/tab_contents.h"
+#include "chrome/common/chrome_constants.h"
+#include "net/base/registry_controlled_domain.h"
+#include "third_party/icu/public/common/unicode/locid.h"
+#include "third_party/icu/public/common/unicode/uloc.h"
+
+namespace {
+
+// Associates known registries with their typcial regions and languages.
+struct AutoFillRegistry {
+  const char* const registry;
+  const char* const region;
+  const char* const languages;
+};
+
+// TODO(isherman): Expand this list.
+const AutoFillRegistry kAutoFillRegistries[] = {
+  {"au", "AU", "en"},
+  {"br", "BR", "pt"},
+  {"ca", "CA", "fr, en"},
+  {"co.jp", "JP", "ja"},
+  {"co.il", "IL", "he"},
+  {"co.nz", "NZ", "en"},
+  {"co.uk", "GB", "en"},
+  {"com.au", "AU", "en"},
+  {"com.br", "BR", "pt"},
+  {"com.mx", "MX", "es"},
+  {"il", "IL", "he"},
+  {"it", "IT", "it"},
+  {"jp", "JP", "ja"},
+  {"fr", "FR", "fr"},
+  {"mx", "MX", "es"},
+  {"ru", "RU", "ru"},
+};
+
+// Returns the registry (roughly, top level domain) for |url|.
+std::string ExtractRegistry(const GURL& url) {
+  const size_t registry_length =
+      net::RegistryControlledDomainService::GetRegistryLength(url, false);
+  if (registry_length == 0 || registry_length == std::string::npos)
+    return std::string();
+
+  const std::string host =
+      net::RegistryControlledDomainService::GetDomainAndRegistry(url);
+  DCHECK_LT(registry_length, host.size());
+
+  return StringToLowerASCII(host.substr(host.size() - registry_length));
+}
+
+// Converts |locale_id| into a canonical ICU |Locale|, with a couple of fixups:
+// * An empty |locale_id| returns an empty result, rather than ICU's default
+//   canonicalization to "en-US".
+// * Likewise, a |locale_id| with an unknown language or region returns an empty
+//   result, rather than ICU's default canonicalization to a bogus locale.
+icu::Locale ConvertToICULocale(const std::string& locale_id) {
+  if (locale_id.empty())
+    return icu::Locale("");
+
+  scoped_array<char> sanitized_locale(new char[locale_id.size() + 1]);
+  size_t sanitized_length = 0;
+  for (size_t i = 0; i < locale_id.size(); ++i) {
+    // We aren't interested in locale keywords, so the only valid characters in
+    // the locale string are ASCII letters, hyphen, and underscore.  See
+    // http://userguide.icu-project.org/locale#TOC-Canonicalization
+    if (isalpha(locale_id[i]) || locale_id[i] == '-' || locale_id[i] == '_') {
+      sanitized_locale.get()[sanitized_length++] = locale_id[i];
+    }
+  }
+  sanitized_locale.get()[sanitized_length] = '\0';
+
+  icu::Locale canonical_locale =
+      icu::Locale::createCanonical(sanitized_locale.get());
+
+  std::string language = canonical_locale.getLanguage();
+  if (!AutoFillLocaleModel::IsValidLanguageTag(language))
+    language = std::string();
+
+  std::string region = canonical_locale.getCountry();
+  if (!AutoFillLocaleModel::IsValidRegionTag(region))
+    region = std::string();
+
+  return icu::Locale(language.c_str(), region.c_str());
+}
+
+// Combines |language| and |region| into a single locale code.
+// Returns "en-US" as a default if the |language| is empty.
+std::string LanguageAndRegionToLocaleCode(const std::string& language,
+                                          const std::string& region) {
+  // We should never get a non-empty region with an empty language.
+  DCHECK(!language.empty() || region.empty());
+
+  if (language.empty())
+    return "en-US";
+
+  if (region.empty())
+    return language;
+
+  return (language + "-" + region);
+}
+
+}  // namespace
+
+AutoFillLocaleModel::AutoFillLocaleModel(const TabContents* tab_contents)
+    : tab_contents_(tab_contents),
+      tab_language_determined_(false) {
+  // Populate the registry map.
+  for (size_t i = 0; i < arraysize(kAutoFillRegistries); ++i) {
+    AutoFillRegistryInfo info;
+    info.region = kAutoFillRegistries[i].region;
+    SplitString(kAutoFillRegistries[i].languages, ',', &info.languages);
+
+    registries_[kAutoFillRegistries[i].registry] = info;
+  }
+}
+
+AutoFillLocaleModel::~AutoFillLocaleModel() {
+}
+
+// static
+bool AutoFillLocaleModel::IsValidLanguageTag(const std::string& language_tag) {
+  // Special case: "und" is an officially registered language tag for
+  // "Undetermined".
+  if (language_tag == chrome::kUnknownLanguageCode)
+    return false;
+
+  // |languages| is a NULL-terminated array of C-strings
+  for (const char* const* languages = uloc_getISOLanguages(); languages[0];
+       ++languages) {
+    if (languages[0] == language_tag)
+      return true;
+  }
+  return false;
+}
+
+// static
+bool AutoFillLocaleModel::IsValidRegionTag(const std::string& region_tag) {
+  // |regions| is a NULL-terminated array of C-strings
+  for (const char* const* regions = uloc_getISOCountries(); regions[0];
+       ++regions) {
+    if (regions[0] == region_tag)
+      return true;
+  }
+  return false;
+}
+
+// If |guessed_language| is empty or the languages associated with |registry|
+// include |guessed_language|, returns the region associated with |registry|.
+// Otherwise, returns an empty string.
+std::string AutoFillLocaleModel::RegionFromRegistry(
+    const std::string& registry,
+    std::string* guessed_language) const {
+  AutoFillRegistryMap::const_iterator result = registries_.find(registry);
+  if (result == registries_.end())
+    return std::string();
+
+  const AutoFillRegistryInfo& info = result->second;
+  for (size_t i = 0; i < info.languages.size(); ++i) {
+    // For each language spoken in the region, check to see if our current guess
+    // at the language matches.  Since we accept the first match, the order in
+    // which the langauges are listed matters.  If we previously had no guess
+    // for the language, update the guess.
+    if (guessed_language->empty()) {
+      *guessed_language = info.languages[i];
+      return info.region;
+    } else if (*guessed_language == info.languages[i]) {
+      return info.region;
+    }
+  }
+
+  return std::string();
+}
+
+// TODO(isherman): We use the page's original language to detect the form's
+// language.  We might want to also use the page's current language (after
+// translation), e.g. to help us figure out the address's language.
+void AutoFillLocaleModel::UpdateLocale(FormStructure* form) const {
+  DCHECK(tab_language_determined_);
+
+  icu::Locale dom_locale = ConvertToICULocale(form->locale());
+  const std::string detected_locale_str = tab_contents_?
+      tab_contents_->language_state().original_language() : std::string();
+  icu::Locale detected_locale = ConvertToICULocale(detected_locale_str);
+
+  std::string result_language;
+  std::string result_region;
+
+  // If we got a non-empty locale from the DOM, use that.  Moreover, we don't
+  // trust the locales "en" and "en-US" -- these can occur with international
+  // forms due to copy/pasted or otherwise careless HTML.  We'll ignore them
+  // for now, but come back to them if we find no other useful signals.
+  if ((strlen(dom_locale.getLanguage()) != 0) &&
+      dom_locale != icu::Locale::getEnglish() &&
+      dom_locale != icu::Locale::getUS()) {
+    result_language = dom_locale.getLanguage();
+    result_region = dom_locale.getCountry();
+  }
+
+  // If we didn't get a language from the DOM, use the CLD-detected locale.
+  // If we did get a language but not a region form the DOM, try using the
+  // CLD-detected region.
+  if (result_language.empty()) {
+    result_language = detected_locale.getLanguage();
+    result_region = detected_locale.getCountry();
+  } else if (result_language == detected_locale.getLanguage() &&
+             result_region.empty()) {
+    result_region = detected_locale.getCountry();
+  }
+
+  // If we still don't have a region, see if we can guess it based on the url's
+  // registry.  Try the form's source url first, then the target url, then the
+  // tab's registry (which might be different if the form is being framed).
+  if (result_region.empty()) {
+    result_region = RegionFromRegistry(ExtractRegistry(form->source_url()),
+                                       &result_language);
+  }
+  if (result_region.empty()) {
+    result_region = RegionFromRegistry(ExtractRegistry(form->target_url()),
+                                       &result_language);
+  }
+  if (result_region.empty() && tab_contents_) {
+    result_region = RegionFromRegistry(ExtractRegistry(tab_contents_->GetURL()),
+                                       &result_language);
+  }
+
+  // If we _still_ don't have a region, see if we ignored an untrusted region --
+  // e.g. "en-US" -- specified by the DOM.  Only do this if our
+  // current guess at the language matches what the DOM specified.
+  if (result_region.empty() && (result_language.empty() ||
+                                result_language == dom_locale.getLanguage())){
+    result_language = dom_locale.getLanguage();
+    result_region = dom_locale.getCountry();
+  }
+
+  form->set_locale(
+      LanguageAndRegionToLocaleCode(result_language, result_region));
+}