This adds a formatter for plurals that works for all locales.

chrome/common/plural_formatter.cc

Index: chrome/common/plural_formatter.cc
diff --git a/chrome/common/plural_formatter.cc b/chrome/common/plural_formatter.cc
new file mode 100644
index 0000000000000000000000000000000000000000..a92388676bb0c71a39d5056556876f33ada84e82
--- /dev/null
+++ b/chrome/common/plural_formatter.cc
@@ -0,0 +1,168 @@
+// Copyright (c) 2011 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/common/plural_formatter.h"
+
+#include <vector>

[tfarina, 2011/03/26 02:05:15]

is this necessary/used in this source file?

[Greg Spencer (Chromium), 2011/03/28 23:13:37]

Nope, not any more.  Removed.

+
+#include "base/scoped_ptr.h"
+#include "base/stl_util-inl.h"
+#include "base/string16.h"

[tfarina, 2011/03/26 02:05:15]

you can remove this include.

[Greg Spencer (Chromium), 2011/03/28 23:13:37]

Done.

+#include "base/utf_string_conversions.h"
+#include "grit/generated_resources.h"
+#include "ui/base/l10n/l10n_util.h"
+#include "unicode/locid.h"
+#include "unicode/plurfmt.h"
+#include "unicode/plurrule.h"
+
+namespace {
+
+// This is the functor for mapping message IDs to strings.  We reset
+// it for tests so that we can avoid using grit for the tests.
+static PluralFormatter::StringSourceFunction g_string_source =

[tfarina, 2011/03/26 02:05:15]

the static keyword is not necessary. Please, could you remove it here, and
below?

[Greg Spencer (Chromium), 2011/03/28 23:13:37]

Done.

+    l10n_util::GetStringUTF8;
+
+// This is used to override which locale is used for the translation.
+// It is used by the tests to test different locales.  If it is set to
+// NULL, then the default locale is used.
+static const char* g_override_locale = NULL;
+
+}  // anonymous namespace

[tfarina, 2011/03/26 02:05:15]

Although I see in many places, and it's misleading (the correct is unnamed), I
think you can stick with just: // namespace (nothing is also ok).

[Greg Spencer (Chromium), 2011/03/28 23:13:37]

I hate that convention: it looks like someone forgot to type the name of the
namespace.  But you're correct: it is in the style guide, so I changed it.

+
+// An implementation class is necessary here because icu does some
+// funky things with namespaces: it declares its stuff under a
+// macro-constructed namespace name and then aliases it to 'icu',
+// which means you have to include the icu header before any forward
+// declaration (which defeats the purpose of forward declarations) in
+// order to use the namespace "icu" or the macro.
+
+class PluralFormatterImpl {
+ public:
+  PluralFormatterImpl();
+  bool Init(const int message_ids[]);
+  string16 GetPluralString(int number) const;
+ private:
+  scoped_ptr<icu::PluralFormat> formatter_;
+};
+
+PluralFormatterImpl::PluralFormatterImpl() : formatter_(NULL) {}
+
+bool PluralFormatterImpl::Init(const int message_ids[]) {
+  static const icu::UnicodeString kKeywords[] = {
+    UNICODE_STRING_SIMPLE("zero"), UNICODE_STRING_SIMPLE("one"),
+    UNICODE_STRING_SIMPLE("two"),  UNICODE_STRING_SIMPLE("few"),
+    UNICODE_STRING_SIMPLE("many"), UNICODE_STRING_SIMPLE("other"),
+  };
+
+  UErrorCode err = U_ZERO_ERROR;
+  scoped_ptr<icu::PluralRules> rules;
+  if (g_override_locale) {
+    icu::Locale overridden_locale(g_override_locale);
+    rules.reset(icu::PluralRules::forLocale(overridden_locale, err));
+  } else {
+    rules.reset(icu::PluralRules::forLocale(icu::Locale::getDefault(), err));
+  }
+
+  // If we fail to create the rules for the given locale, we use
+  // fallback rules.
+  if (U_FAILURE(err)) {
+    err = U_ZERO_ERROR;
+    icu::UnicodeString fallback_rules("one: n is 1", -1, US_INV);
+    rules.reset(icu::PluralRules::createRules(fallback_rules, err));
+    DCHECK(U_SUCCESS(err));
+  }
+
+  icu::UnicodeString pattern;
+  for (size_t j = 0; j < arraysize(kKeywords); ++j) {
+    int msg_id = message_ids[j];
+    std::string sub_pattern = (*g_string_source)(msg_id);
+    // NA means this keyword is not used in the current locale.  Even
+    // if a translator translated for this keyword, we do not use it
+    // unless it's 'other' or it's defined in the rules for the
+    // current locale. Special-casing of 'other' (j = 5) will be
+    // removed once ICU's isKeyword is fixed to return true for
+    // isKeyword('other').
+    if (sub_pattern.compare("NA") != 0 &&
+        (j == 5 || rules->isKeyword(kKeywords[j]))) {
+      pattern += kKeywords[j];
+      pattern += UNICODE_STRING_SIMPLE("{");
+      pattern += icu::UnicodeString(sub_pattern.c_str(), "UTF-8");
+      pattern += UNICODE_STRING_SIMPLE("}");
+    }
+  }
+
+  scoped_ptr<icu::PluralFormat> format(
+      new icu::PluralFormat(*rules, pattern, err));
+
+  // If format creation fails, we fail.
+  if (U_FAILURE(err)) {
+    formatter_.reset(NULL);
+    return false;
+  }
+
+  formatter_.swap(format);
+  return true;
+}
+
+string16 PluralFormatterImpl::GetPluralString(int number) const {
+  DCHECK(formatter_.get() != NULL)
+      << "GetPluralString called before initializing formatter.";
+  icu::UnicodeString plural_string;
+  UErrorCode error = U_ZERO_ERROR;
+  plural_string = formatter_->format(number, error);
+  DCHECK(U_SUCCESS(error));

[Paweł Hajdan Jr., 2011/03/28 18:52:29]

Later case handles U_FAILURE, and this only DCHECKs on it. It might be worth it
to signal failure via bool return parameter or some other way, but it's up to
you.

[Greg Spencer (Chromium), 2011/03/28 23:13:37]

OK, I added handling for this to make it consistent.

+  string16 result;
+  int capacity = plural_string.length() + 1;
+  plural_string.extract(static_cast<UChar*>(WriteInto(&result, capacity)),
+                        capacity, error);
+  DCHECK(U_SUCCESS(error));

[Paweł Hajdan Jr., 2011/03/28 18:52:29]

nit: I'd suggest to make this a NOTREACHED() in the "if" below. It's checking
U_SUCCESS(error) and then right after U_FAILURE(error), which is a bit
surprising.

[Greg Spencer (Chromium), 2011/03/28 23:13:37]

NOTREACHED is a good suggestion.  I switched to that.

+  // On error, result will be the empty string.
+  if (U_FAILURE(error))
+    result.clear();
+
+  return result;
+}
+
+/////////////////////////////////////////////////////
+// Implementation of the main class PluralFormatter.
+
+PluralFormatter::PluralFormatter() : impl_(NULL) {}
+
+PluralFormatter::~PluralFormatter() {
+  delete impl_;
+}
+
+bool PluralFormatter::Init(int zero_message_id,
+                           int one_message_id,
+                           int two_message_id,
+                           int few_message_id,
+                           int many_message_id,
+                           int default_message_id) {
+  int message_ids[6];
+  message_ids[0] = zero_message_id;
+  message_ids[1] = one_message_id;
+  message_ids[2] = two_message_id;
+  message_ids[3] = few_message_id;
+  message_ids[4] = many_message_id;
+  message_ids[5] = default_message_id;
+  return Init(message_ids);
+}
+
+bool PluralFormatter::Init(const int message_ids[]) {

[tfarina, 2011/03/26 02:05:15]

this should be implemented above. The order of the definition should match with
the order in the header file.

[Greg Spencer (Chromium), 2011/03/28 23:13:37]

Sorry, missed one. Done.

+  impl_ = new PluralFormatterImpl;
+  return impl_->Init(message_ids);
+}
+
+string16 PluralFormatter::GetPluralString(int number) const {
+  DCHECK(impl_ != NULL) << "GetPluralString called before Init";
+  return impl_->GetPluralString(number);
+}
+
+void PluralFormatter::SetStringSource(StringSourceFunction string_source) {
+  g_string_source = string_source;
+}
+
+void PluralFormatter::SetOverrideLocale(const char* overrride_locale) {
+  g_override_locale = overrride_locale;
+}