1
0
mirror of https://github.com/vcmi/vcmi.git synced 2024-11-24 08:32:34 +02:00
vcmi/lib/Languages.h
2024-03-27 13:10:55 +02:00

173 lines
5.7 KiB
C++

/*
* Languages.h, part of VCMI engine
*
* Authors: listed in file AUTHORS in main folder
*
* License: GNU General Public License v2.0 or later
* Full text of license available in license.txt file, in main folder
*
*/
#pragma once
namespace Languages
{
enum class EPluralForms
{
NONE,
VI_1, // Single plural form, (Vietnamese)
EN_2, // Two forms, singular used for one only (English)
FR_2, // Two forms, singular used for zero and one (French)
UK_3, // Three forms, special cases for numbers ending in 1 and 2, 3, 4, except those ending in 1[1-4] (Ukrainian)
CZ_3, // Three forms, special cases for 1 and 2, 3, 4 (Czech)
PL_3, // Three forms, special case for one and some numbers ending in 2, 3, or 4 (Polish)
};
enum class ELanguages
{
CZECH,
CHINESE,
ENGLISH,
FINNISH,
FRENCH,
GERMAN,
HUNGARIAN,
ITALIAN,
KOREAN,
POLISH,
PORTUGUESE,
RUSSIAN,
SPANISH,
SWEDISH,
TURKISH,
UKRAINIAN,
VIETNAMESE,
// Pseudo-languages, that have no translations but can define H3 encoding to use
OTHER_CP1250,
OTHER_CP1251,
OTHER_CP1252,
COUNT
};
struct Options
{
/// string identifier (ascii, lower-case), e.g. "english"
std::string identifier;
/// human-readable name of language in English
std::string nameEnglish;
/// human-readable name of language in its own language
std::string nameNative;
/// encoding that is used by H3 for this language
std::string encoding;
/// primary IETF language tag
std::string tagIETF;
/// DateTime format
std::string dateTimeFormat;
/// Ruleset for plural forms in this language
EPluralForms pluralForms = EPluralForms::NONE;
/// VCMI supports translations into this language
bool hasTranslation = false;
};
inline const auto & getLanguageList()
{
static const std::array<Options, 20> languages
{ {
{ "czech", "Czech", "Čeština", "CP1250", "cs", "%d.%m.%Y %T", EPluralForms::CZ_3, true },
{ "chinese", "Chinese", "简体中文", "GBK", "zh", "%F %T", EPluralForms::VI_1, true }, // Note: actually Simplified Chinese
{ "english", "English", "English", "CP1252", "en", "%F %T", EPluralForms::EN_2, true }, // English uses international date/time format here
{ "finnish", "Finnish", "Suomi", "CP1252", "fi", "%d.%m.%Y %T", EPluralForms::EN_2, true },
{ "french", "French", "Français", "CP1252", "fr", "%d/%m/%Y %T", EPluralForms::FR_2, true },
{ "german", "German", "Deutsch", "CP1252", "de", "%d.%m.%Y %T", EPluralForms::EN_2, true },
{ "hungarian", "Hungarian", "Magyar", "CP1250", "hu", "%Y. %m. %d. %T", EPluralForms::EN_2, true },
{ "italian", "Italian", "Italiano", "CP1250", "it", "%d/%m/%Y %T", EPluralForms::EN_2, true },
{ "korean", "Korean", "한국어", "CP949", "ko", "%F %T", EPluralForms::VI_1, true },
{ "polish", "Polish", "Polski", "CP1250", "pl", "%d.%m.%Y %T", EPluralForms::PL_3, true },
{ "portuguese", "Portuguese", "Português", "CP1252", "pt", "%d/%m/%Y %T", EPluralForms::EN_2, true }, // Note: actually Brazilian Portuguese
{ "russian", "Russian", "Русский", "CP1251", "ru", "%d.%m.%Y %T", EPluralForms::UK_3, true },
{ "spanish", "Spanish", "Español", "CP1252", "es", "%d/%m/%Y %T", EPluralForms::EN_2, true },
{ "swedish", "Swedish", "Svenska", "CP1252", "sv", "%F %T", EPluralForms::EN_2, true },
{ "turkish", "Turkish", "Türkçe", "CP1254", "tr", "%d.%m.%Y %T", EPluralForms::EN_2, true },
{ "ukrainian", "Ukrainian", "Українська", "CP1251", "uk", "%d.%m.%Y %T", EPluralForms::UK_3, true },
{ "vietnamese", "Vietnamese", "Tiếng Việt", "UTF-8", "vi", "%d/%m/%Y %T", EPluralForms::VI_1, true }, // Fan translation uses special encoding
{ "other_cp1250", "Other (East European)", "", "CP1250", "", "", EPluralForms::NONE, false },
{ "other_cp1251", "Other (Cyrillic Script)", "", "CP1251", "", "", EPluralForms::NONE, false },
{ "other_cp1252", "Other (West European)", "", "CP1252", "", "", EPluralForms::NONE, false }
} };
static_assert(languages.size() == static_cast<size_t>(ELanguages::COUNT), "Languages array is missing a value!");
return languages;
}
inline const Options & getLanguageOptions(ELanguages language)
{
return getLanguageList().at(static_cast<size_t>(language));
}
inline const Options & getLanguageOptions(const std::string & language)
{
for(const auto & entry : getLanguageList())
if(entry.identifier == language)
return entry;
throw std::out_of_range("Language " + language + " does not exists!");
}
template<typename Numeric>
inline constexpr int getPluralFormIndex(EPluralForms form, Numeric value)
{
// Based on https://www.gnu.org/software/gettext/manual/html_node/Plural-forms.html
switch(form)
{
case EPluralForms::NONE:
case EPluralForms::VI_1:
return 0;
case EPluralForms::EN_2:
if (value == 1)
return 1;
return 2;
case EPluralForms::FR_2:
if (value == 1 || value == 0)
return 1;
return 2;
case EPluralForms::UK_3:
if (value % 10 == 1 && value % 100 != 11)
return 1;
if (value%10>=2 && value%10<=4 && (value%100<10 || value%100>=20))
return 2;
return 0;
case EPluralForms::CZ_3:
if (value == 1)
return 1;
if (value>=2 && value<=4)
return 2;
return 0;
case EPluralForms::PL_3:
if (value == 1)
return 1;
if (value%10>=2 && value%10<=4 && (value%100<10 || value%100>=20))
return 2;
return 0;
}
throw std::runtime_error("Invalid plural form enumeration received!");
}
template<typename Numeric>
inline std::string getPluralFormTextID(std::string languageName, Numeric value, std::string textID)
{
int formIndex = getPluralFormIndex(getLanguageOptions(languageName).pluralForms, value);
return textID + '.' + std::to_string(formIndex);
}
}