/* * This file is part of OpenTTD. * OpenTTD is free software; you can redistribute it and/or modify it under the terms of the GNU General Public License as published by the Free Software Foundation, version 2. * OpenTTD is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. * See the GNU General Public License for more details. You should have received a copy of the GNU General Public License along with OpenTTD. If not, see . */ /** @file strings.cpp Handling of translated strings. */ #include "stdafx.h" #include "currency.h" #include "station_base.h" #include "town.h" #include "waypoint_base.h" #include "depot_base.h" #include "industry.h" #include "newgrf_text.h" #include "fileio_func.h" #include "signs_base.h" #include "fontdetection.h" #include "error.h" #include "error_func.h" #include "strings_func.h" #include "rev.h" #include "core/endian_func.hpp" #include "date_func.h" #include "timer/timer_game_calendar.h" #include "vehicle_base.h" #include "engine_base.h" #include "language.h" #include "townname_func.h" #include "string_func.h" #include "company_base.h" #include "smallmap_gui.h" #include "window_func.h" #include "debug.h" #include "game/game_text.hpp" #include "network/network_content_gui.h" #include "newgrf_engine.h" #include "core/backup_type.hpp" #include #include #include "table/strings.h" #include "table/control_codes.h" #include "safeguards.h" std::string _config_language_file; ///< The file (name) stored in the configuration. LanguageList _languages; ///< The actual list of language meta data. const LanguageMetadata *_current_language = nullptr; ///< The currently loaded language. TextDirection _current_text_dir; ///< Text direction of the currently selected language. #ifdef WITH_ICU_I18N std::unique_ptr _current_collator; ///< Collator for the language currently in use. #endif /* WITH_ICU_I18N */ static uint64 _global_string_params_data[20]; ///< Global array of string parameters. To access, use #SetDParam. static WChar _global_string_params_type[20]; ///< Type of parameters stored in #_global_string_params StringParameters _global_string_params(_global_string_params_data, 20, _global_string_params_type); /** Reset the type array. */ void StringParameters::ClearTypeInformation() { assert(this->type != nullptr); MemSetT(this->type, 0, this->num_param); } /** * Read an int64 from the argument array. The offset is increased * so the next time GetInt64 is called the next value is read. */ int64 StringParameters::GetInt64(WChar type) { if (this->offset >= this->num_param) { Debug(misc, 0, "Trying to read invalid string parameter"); return 0; } if (this->type != nullptr) { if (this->type[this->offset] != 0 && this->type[this->offset] != type) { Debug(misc, 0, "Trying to read string parameter with wrong type"); return 0; } this->type[this->offset] = type; } return this->data[this->offset++]; } /** * Set DParam n to some number that is suitable for string size computations. * @param n Index of the string parameter. * @param max_value The biggest value which shall be displayed. * For the result only the number of digits of \a max_value matter. * @param min_count Minimum number of digits independent of \a max. * @param size Font of the number */ void SetDParamMaxValue(uint n, uint64 max_value, uint min_count, FontSize size) { uint num_digits = 1; while (max_value >= 10) { num_digits++; max_value /= 10; } SetDParamMaxDigits(n, std::max(min_count, num_digits), size); } /** * Set DParam n to some number that is suitable for string size computations. * @param n Index of the string parameter. * @param count Number of digits which shall be displayable. * @param size Font of the number */ void SetDParamMaxDigits(uint n, uint count, FontSize size) { uint front = 0; uint next = 0; GetBroadestDigit(&front, &next, size); uint64 val = count > 1 ? front : next; for (; count > 1; count--) { val = 10 * val + next; } SetDParam(n, val); } /** * Copy \a num string parameters from array \a src into the global string parameter array. * @param offs Index in the global array to copy the first string parameter to. * @param src Source array of string parameters. * @param num Number of string parameters to copy. */ void CopyInDParam(int offs, const uint64 *src, int num) { MemCpyT(_global_string_params.GetPointerToOffset(offs), src, num); } /** * Copy \a num string parameters from the global string parameter array to the \a dst array. * @param dst Destination array of string parameters. * @param offs Index in the global array to copy the first string parameter from. * @param num Number of string parameters to copy. */ void CopyOutDParam(uint64 *dst, int offs, int num) { MemCpyT(dst, _global_string_params.GetPointerToOffset(offs), num); } /** * Copy \a num string parameters from the global string parameter array to the \a dst array. * Furthermore clone raw string parameters into \a strings and amend the data in \a dst. * @param dst Destination array of string parameters. * @param strings Destination array for clone of the raw strings. Must be of same length as dst. Deallocation left to the caller. * @param string The string used to determine where raw strings are and where there are no raw strings. * @param num Number of string parameters to copy. */ void CopyOutDParam(uint64 *dst, const char **strings, StringID string, int num) { char buf[DRAW_STRING_BUFFER]; GetString(buf, string, lastof(buf)); MemCpyT(dst, _global_string_params.GetPointerToOffset(0), num); for (int i = 0; i < num; i++) { if (_global_string_params.HasTypeInformation() && _global_string_params.GetTypeAtOffset(i) == SCC_RAW_STRING_POINTER) { strings[i] = stredup((const char *)(size_t)_global_string_params.GetParam(i)); dst[i] = (size_t)strings[i]; } else { strings[i] = nullptr; } } } static char *StationGetSpecialString(char *buff, int x, const char *last); static char *GetSpecialTownNameString(char *buff, int ind, uint32 seed, const char *last); static char *GetSpecialNameString(char *buff, int ind, StringParameters *args, const char *last); static char *FormatString(char *buff, const char *str, StringParameters *args, const char *last, uint case_index = 0, bool game_script = false, bool dry_run = false); struct LanguagePack : public LanguagePackHeader { char data[]; // list of strings }; struct LanguagePackDeleter { void operator()(LanguagePack *langpack) { /* LanguagePack is in fact reinterpreted char[], we need to reinterpret it back to free it properly. */ delete[] reinterpret_cast(langpack); } }; struct LoadedLanguagePack { std::unique_ptr langpack; std::vector offsets; std::array langtab_num; ///< Offset into langpack offs std::array langtab_start; ///< Offset into langpack offs }; static LoadedLanguagePack _langpack; static bool _scan_for_gender_data = false; ///< Are we scanning for the gender of the current string? (instead of formatting it) const char *GetStringPtr(StringID string) { switch (GetStringTab(string)) { case TEXT_TAB_GAMESCRIPT_START: return GetGameStringPtr(GetStringIndex(string)); /* 0xD0xx and 0xD4xx IDs have been converted earlier. */ case TEXT_TAB_OLD_NEWGRF: NOT_REACHED(); case TEXT_TAB_NEWGRF_START: return GetGRFStringPtr(GetStringIndex(string)); default: return _langpack.offsets[_langpack.langtab_start[GetStringTab(string)] + GetStringIndex(string)]; } } /** * Get a parsed string with most special stringcodes replaced by the string parameters. * @param buffr Pointer to a string buffer where the formatted string should be written to. * @param string * @param args Arguments for the string. * @param last Pointer just past the end of \a buffr. * @param case_index The "case index". This will only be set when FormatString wants to print the string in a different case. * @param game_script The string is coming directly from a game script. * @return Pointer to the final zero byte of the formatted string. */ char *GetStringWithArgs(char *buffr, StringID string, StringParameters *args, const char *last, uint case_index, bool game_script) { if (string == 0) return GetStringWithArgs(buffr, STR_UNDEFINED, args, last); uint index = GetStringIndex(string); StringTab tab = GetStringTab(string); switch (tab) { case TEXT_TAB_TOWN: if (index >= 0xC0 && !game_script) { return GetSpecialTownNameString(buffr, index - 0xC0, args->GetInt32(), last); } break; case TEXT_TAB_SPECIAL: if (index >= 0xE4 && !game_script) { return GetSpecialNameString(buffr, index - 0xE4, args, last); } break; case TEXT_TAB_OLD_CUSTOM: /* Old table for custom names. This is no longer used */ if (!game_script) { FatalError("Incorrect conversion of custom name string."); } break; case TEXT_TAB_GAMESCRIPT_START: return FormatString(buffr, GetGameStringPtr(index), args, last, case_index, true); case TEXT_TAB_OLD_NEWGRF: NOT_REACHED(); case TEXT_TAB_NEWGRF_START: return FormatString(buffr, GetGRFStringPtr(index), args, last, case_index); default: break; } if (index >= _langpack.langtab_num[tab]) { if (game_script) { return GetStringWithArgs(buffr, STR_UNDEFINED, args, last); } FatalError("String 0x{:X} is invalid. You are probably using an old version of the .lng file.\n", string); } return FormatString(buffr, GetStringPtr(string), args, last, case_index); } char *GetString(char *buffr, StringID string, const char *last) { _global_string_params.ClearTypeInformation(); _global_string_params.offset = 0; return GetStringWithArgs(buffr, string, &_global_string_params, last); } /** * Resolve the given StringID into a std::string with all the associated * DParam lookups and formatting. * @param string The unique identifier of the translatable string. * @return The std::string of the translated string. */ std::string GetString(StringID string) { char buffer[DRAW_STRING_BUFFER]; GetString(buffer, string, lastof(buffer)); return buffer; } /** * This function is used to "bind" a C string to a OpenTTD dparam slot. * @param n slot of the string * @param str string to bind */ void SetDParamStr(uint n, const char *str) { SetDParam(n, (uint64)(size_t)str); } /** * This function is used to "bind" the C string of a std::string to a OpenTTD dparam slot. * The caller has to ensure that the std::string reference remains valid while the string is shown. * @param n slot of the string * @param str string to bind */ void SetDParamStr(uint n, const std::string &str) { SetDParamStr(n, str.c_str()); } /** * Format a number into a string. * @param buff the buffer to write to * @param number the number to write down * @param last the last element in the buffer * @param separator the thousands-separator to use * @param zerofill minimum number of digits to print for the integer part. The number will be filled with zeros at the front if necessary. * @param fractional_digits number of fractional digits to display after a decimal separator. The decimal separator is inserted * in front of the \a fractional_digits last digit of \a number. * @return till where we wrote */ static char *FormatNumber(char *buff, int64 number, const char *last, const char *separator, int zerofill = 1, int fractional_digits = 0) { static const int max_digits = 20; uint64 divisor = 10000000000000000000ULL; zerofill += fractional_digits; int thousands_offset = (max_digits - fractional_digits - 1) % 3; if (number < 0) { if (buff != last) *buff++ = '-'; number = -number; } uint64 num = number; uint64 tot = 0; for (int i = 0; i < max_digits; i++) { if (i == max_digits - fractional_digits) { const char *decimal_separator = _settings_game.locale.digit_decimal_separator.c_str(); if (StrEmpty(decimal_separator)) decimal_separator = _langpack.langpack->digit_decimal_separator; buff = strecpy(buff, decimal_separator, last); } uint64 quot = 0; if (num >= divisor) { quot = num / divisor; num = num % divisor; } if ((tot |= quot) || i >= max_digits - zerofill) { if (buff != last) *buff++ = '0' + quot; // quot is a single digit if ((i % 3) == thousands_offset && i < max_digits - 1 - fractional_digits) buff = strecpy(buff, separator, last); } divisor /= 10; } *buff = '\0'; return buff; } static char *FormatCommaNumber(char *buff, int64 number, const char *last, int fractional_digits = 0) { const char *separator = _settings_game.locale.digit_group_separator.c_str(); if (StrEmpty(separator)) separator = _langpack.langpack->digit_group_separator; return FormatNumber(buff, number, last, separator, 1, fractional_digits); } static char *FormatNoCommaNumber(char *buff, int64 number, const char *last) { return FormatNumber(buff, number, last, ""); } static char *FormatZerofillNumber(char *buff, int64 number, int64 count, const char *last) { return FormatNumber(buff, number, last, "", count); } static char *FormatHexNumber(char *buff, uint64 number, const char *last) { return strecpy(buff, fmt::format("0x{:X}", number).c_str(), last); } /** * Format a given number as a number of bytes with the SI prefix. * @param buff the buffer to write to * @param number the number of bytes to write down * @param last the last element in the buffer * @return till where we wrote */ static char *FormatBytes(char *buff, int64 number, const char *last) { assert(number >= 0); /* 1 2^10 2^20 2^30 2^40 2^50 2^60 */ const char * const iec_prefixes[] = {"", "Ki", "Mi", "Gi", "Ti", "Pi", "Ei"}; uint id = 1; while (number >= 1024 * 1024) { number /= 1024; id++; } const char *decimal_separator = _settings_game.locale.digit_decimal_separator.c_str(); if (StrEmpty(decimal_separator)) decimal_separator = _langpack.langpack->digit_decimal_separator; if (number < 1024) { id = 0; buff += seprintf(buff, last, "%i", (int)number); } else if (number < 1024 * 10) { buff += seprintf(buff, last, "%i%s%02i", (int)number / 1024, decimal_separator, (int)(number % 1024) * 100 / 1024); } else if (number < 1024 * 100) { buff += seprintf(buff, last, "%i%s%01i", (int)number / 1024, decimal_separator, (int)(number % 1024) * 10 / 1024); } else { assert(number < 1024 * 1024); buff += seprintf(buff, last, "%i", (int)number / 1024); } assert(id < lengthof(iec_prefixes)); buff += seprintf(buff, last, NBSP "%sB", iec_prefixes[id]); return buff; } static char *FormatYmdString(char *buff, TimerGameCalendar::Date date, const char *last, uint case_index) { YearMonthDay ymd; ConvertDateToYMD(date, &ymd); int64 args[] = {ymd.day + STR_DAY_NUMBER_1ST - 1, STR_MONTH_ABBREV_JAN + ymd.month, ymd.year}; StringParameters tmp_params(args); return FormatString(buff, GetStringPtr(STR_FORMAT_DATE_LONG), &tmp_params, last, case_index); } static char *FormatMonthAndYear(char *buff, TimerGameCalendar::Date date, const char *last, uint case_index) { YearMonthDay ymd; ConvertDateToYMD(date, &ymd); int64 args[] = {STR_MONTH_JAN + ymd.month, ymd.year}; StringParameters tmp_params(args); return FormatString(buff, GetStringPtr(STR_FORMAT_DATE_SHORT), &tmp_params, last, case_index); } static char *FormatTinyOrISODate(char *buff, TimerGameCalendar::Date date, StringID str, const char *last) { YearMonthDay ymd; ConvertDateToYMD(date, &ymd); /* Day and month are zero-padded with ZEROFILL_NUM, hence the two 2s. */ int64 args[] = {ymd.day, 2, ymd.month + 1, 2, ymd.year}; StringParameters tmp_params(args); return FormatString(buff, GetStringPtr(str), &tmp_params, last); } static char *FormatGenericCurrency(char *buff, const CurrencySpec *spec, Money number, bool compact, const char *last) { /* We are going to make number absolute for printing, so * keep this piece of data as we need it later on */ bool negative = number < 0; const char *multiplier = ""; number *= spec->rate; /* convert from negative */ if (number < 0) { if (buff + Utf8CharLen(SCC_PUSH_COLOUR) > last) return buff; buff += Utf8Encode(buff, SCC_PUSH_COLOUR); if (buff + Utf8CharLen(SCC_RED) > last) return buff; buff += Utf8Encode(buff, SCC_RED); buff = strecpy(buff, "-", last); number = -number; } /* Add prefix part, following symbol_pos specification. * Here, it can can be either 0 (prefix) or 2 (both prefix and suffix). * The only remaining value is 1 (suffix), so everything that is not 1 */ if (spec->symbol_pos != 1) buff = strecpy(buff, spec->prefix.c_str(), last); /* for huge numbers, compact the number into k or M */ if (compact) { /* Take care of the 'k' rounding. Having 1 000 000 k * and 1 000 M is inconsistent, so always use 1 000 M. */ if (number >= 1000000000 - 500) { number = (number + 500000) / 1000000; multiplier = NBSP "M"; } else if (number >= 1000000) { number = (number + 500) / 1000; multiplier = NBSP "k"; } } const char *separator = _settings_game.locale.digit_group_separator_currency.c_str(); if (StrEmpty(separator)) separator = _currency->separator.c_str(); if (StrEmpty(separator)) separator = _langpack.langpack->digit_group_separator_currency; buff = FormatNumber(buff, number, last, separator); buff = strecpy(buff, multiplier, last); /* Add suffix part, following symbol_pos specification. * Here, it can can be either 1 (suffix) or 2 (both prefix and suffix). * The only remaining value is 1 (prefix), so everything that is not 0 */ if (spec->symbol_pos != 0) buff = strecpy(buff, spec->suffix.c_str(), last); if (negative) { if (buff + Utf8CharLen(SCC_POP_COLOUR) > last) return buff; buff += Utf8Encode(buff, SCC_POP_COLOUR); *buff = '\0'; } return buff; } /** * Determine the "plural" index given a plural form and a number. * @param count The number to get the plural index of. * @param plural_form The plural form we want an index for. * @return The plural index for the given form. */ static int DeterminePluralForm(int64 count, int plural_form) { /* The absolute value determines plurality */ uint64 n = abs(count); switch (plural_form) { default: NOT_REACHED(); /* Two forms: singular used for one only. * Used in: * Danish, Dutch, English, German, Norwegian, Swedish, Estonian, Finnish, * Greek, Hebrew, Italian, Portuguese, Spanish, Esperanto */ case 0: return n != 1 ? 1 : 0; /* Only one form. * Used in: * Hungarian, Japanese, Turkish */ case 1: return 0; /* Two forms: singular used for 0 and 1. * Used in: * French, Brazilian Portuguese */ case 2: return n > 1 ? 1 : 0; /* Three forms: special cases for 0, and numbers ending in 1 except when ending in 11. * Note: Cases are out of order for hysterical reasons. '0' is last. * Used in: * Latvian */ case 3: return n % 10 == 1 && n % 100 != 11 ? 0 : n != 0 ? 1 : 2; /* Five forms: special cases for 1, 2, 3 to 6, and 7 to 10. * Used in: * Gaelige (Irish) */ case 4: return n == 1 ? 0 : n == 2 ? 1 : n < 7 ? 2 : n < 11 ? 3 : 4; /* Three forms: special cases for numbers ending in 1 except when ending in 11, and 2 to 9 except when ending in 12 to 19. * Used in: * Lithuanian */ case 5: return n % 10 == 1 && n % 100 != 11 ? 0 : n % 10 >= 2 && (n % 100 < 10 || n % 100 >= 20) ? 1 : 2; /* Three forms: special cases for numbers ending in 1 except when ending in 11, and 2 to 4 except when ending in 12 to 14. * Used in: * Croatian, Russian, Ukrainian */ case 6: return n % 10 == 1 && n % 100 != 11 ? 0 : n % 10 >= 2 && n % 10 <= 4 && (n % 100 < 10 || n % 100 >= 20) ? 1 : 2; /* Three forms: special cases for 1, and numbers ending in 2 to 4 except when ending in 12 to 14. * Used in: * Polish */ case 7: return n == 1 ? 0 : n % 10 >= 2 && n % 10 <= 4 && (n % 100 < 10 || n % 100 >= 20) ? 1 : 2; /* Four forms: special cases for numbers ending in 01, 02, and 03 to 04. * Used in: * Slovenian */ case 8: return n % 100 == 1 ? 0 : n % 100 == 2 ? 1 : n % 100 == 3 || n % 100 == 4 ? 2 : 3; /* Two forms: singular used for numbers ending in 1 except when ending in 11. * Used in: * Icelandic */ case 9: return n % 10 == 1 && n % 100 != 11 ? 0 : 1; /* Three forms: special cases for 1, and 2 to 4 * Used in: * Czech, Slovak */ case 10: return n == 1 ? 0 : n >= 2 && n <= 4 ? 1 : 2; /* Two forms: cases for numbers ending with a consonant, and with a vowel. * Korean doesn't have the concept of plural, but depending on how a * number is pronounced it needs another version of a particle. * As such the plural system is misused to give this distinction. */ case 11: switch (n % 10) { case 0: // yeong case 1: // il case 3: // sam case 6: // yuk case 7: // chil case 8: // pal return 0; case 2: // i case 4: // sa case 5: // o case 9: // gu return 1; default: NOT_REACHED(); } /* Four forms: special cases for 1, 0 and numbers ending in 02 to 10, and numbers ending in 11 to 19. * Used in: * Maltese */ case 12: return (n == 1 ? 0 : n == 0 || (n % 100 > 1 && n % 100 < 11) ? 1 : (n % 100 > 10 && n % 100 < 20) ? 2 : 3); /* Four forms: special cases for 1 and 11, 2 and 12, 3 .. 10 and 13 .. 19, other * Used in: * Scottish Gaelic */ case 13: return ((n == 1 || n == 11) ? 0 : (n == 2 || n == 12) ? 1 : ((n > 2 && n < 11) || (n > 12 && n < 20)) ? 2 : 3); /* Three forms: special cases for 1, 0 and numbers ending in 01 to 19. * Used in: * Romanian */ case 14: return n == 1 ? 0 : (n == 0 || (n % 100 > 0 && n % 100 < 20)) ? 1 : 2; } } static const char *ParseStringChoice(const char *b, uint form, char **dst, const char *last) { /* {Length of each string} {each string} */ uint n = (byte)*b++; uint pos, i, mypos = 0; for (i = pos = 0; i != n; i++) { uint len = (byte)*b++; if (i == form) mypos = pos; pos += len; } *dst += seprintf(*dst, last, "%s", b + mypos); return b + pos; } /** Helper for unit conversion. */ struct UnitConversion { int multiplier; ///< Amount to multiply upon conversion. int shift; ///< Amount to shift upon conversion. /** * Convert value from OpenTTD's internal unit into the displayed value. * @param input The input to convert. * @param round Whether to round the value or not. * @return The converted value. */ int64 ToDisplay(int64 input, bool round = true) const { return ((input * this->multiplier) + (round && this->shift != 0 ? 1 << (this->shift - 1) : 0)) >> this->shift; } /** * Convert the displayed value back into a value of OpenTTD's internal unit. * @param input The input to convert. * @param round Whether to round the value up or not. * @param divider Divide the return value by this. * @return The converted value. */ int64 FromDisplay(int64 input, bool round = true, int64 divider = 1) const { return ((input << this->shift) + (round ? (this->multiplier * divider) - 1 : 0)) / (this->multiplier * divider); } }; /** Information about a specific unit system. */ struct Units { UnitConversion c; ///< Conversion StringID s; ///< String for the unit unsigned int decimal_places; ///< Number of decimal places embedded in the value. For example, 1 if the value is in tenths, and 3 if the value is in thousandths. }; /** Information about a specific unit system with a long variant. */ struct UnitsLong { UnitConversion c; ///< Conversion StringID s; ///< String for the short variant of the unit StringID l; ///< String for the long variant of the unit }; /** Unit conversions for velocity. */ static const Units _units_velocity[] = { { { 1, 0}, STR_UNITS_VELOCITY_IMPERIAL, 0 }, { { 103, 6}, STR_UNITS_VELOCITY_METRIC, 0 }, { { 1831, 12}, STR_UNITS_VELOCITY_SI, 0 }, { { 37888, 16}, STR_UNITS_VELOCITY_GAMEUNITS, 1 }, { { 7289499, 23}, STR_UNITS_VELOCITY_KNOTS, 0 }, }; /** Unit conversions for power. */ static const Units _units_power[] = { { { 1, 0}, STR_UNITS_POWER_IMPERIAL, 0 }, { {4153, 12}, STR_UNITS_POWER_METRIC, 0 }, { {6109, 13}, STR_UNITS_POWER_SI, 0 }, }; /** Unit conversions for power to weight. */ static const Units _units_power_to_weight[] = { { { 29, 5}, STR_UNITS_POWER_IMPERIAL_TO_WEIGHT_IMPERIAL, 1}, { { 1, 0}, STR_UNITS_POWER_IMPERIAL_TO_WEIGHT_METRIC, 1}, { { 1, 0}, STR_UNITS_POWER_IMPERIAL_TO_WEIGHT_SI, 1}, { { 59, 6}, STR_UNITS_POWER_METRIC_TO_WEIGHT_IMPERIAL, 1}, { { 65, 6}, STR_UNITS_POWER_METRIC_TO_WEIGHT_METRIC, 1}, { { 65, 6}, STR_UNITS_POWER_METRIC_TO_WEIGHT_SI, 1}, { { 173, 8}, STR_UNITS_POWER_SI_TO_WEIGHT_IMPERIAL, 1}, { { 3, 2}, STR_UNITS_POWER_SI_TO_WEIGHT_METRIC, 1}, { { 3, 2}, STR_UNITS_POWER_SI_TO_WEIGHT_SI, 1}, }; /** Unit conversions for weight. */ static const UnitsLong _units_weight[] = { { {4515, 12}, STR_UNITS_WEIGHT_SHORT_IMPERIAL, STR_UNITS_WEIGHT_LONG_IMPERIAL }, { { 1, 0}, STR_UNITS_WEIGHT_SHORT_METRIC, STR_UNITS_WEIGHT_LONG_METRIC }, { {1000, 0}, STR_UNITS_WEIGHT_SHORT_SI, STR_UNITS_WEIGHT_LONG_SI }, }; /** Unit conversions for volume. */ static const UnitsLong _units_volume[] = { { {4227, 4}, STR_UNITS_VOLUME_SHORT_IMPERIAL, STR_UNITS_VOLUME_LONG_IMPERIAL }, { {1000, 0}, STR_UNITS_VOLUME_SHORT_METRIC, STR_UNITS_VOLUME_LONG_METRIC }, { { 1, 0}, STR_UNITS_VOLUME_SHORT_SI, STR_UNITS_VOLUME_LONG_SI }, }; /** Unit conversions for force. */ static const Units _units_force[] = { { {3597, 4}, STR_UNITS_FORCE_IMPERIAL, 0 }, { {3263, 5}, STR_UNITS_FORCE_METRIC, 0 }, { { 1, 0}, STR_UNITS_FORCE_SI, 0 }, }; /** Unit conversions for height. */ static const Units _units_height[] = { { { 3, 0}, STR_UNITS_HEIGHT_IMPERIAL, 0 }, // "Wrong" conversion factor for more nicer GUI values { { 1, 0}, STR_UNITS_HEIGHT_METRIC, 0 }, { { 1, 0}, STR_UNITS_HEIGHT_SI, 0 }, }; /** * Get index for velocity conversion units for a vehicle type. * @param type VehicleType to convert velocity for. * @return Index within velocity conversion units for vehicle type. */ static byte GetVelocityUnits(VehicleType type) { if (type == VEH_SHIP || type == VEH_AIRCRAFT) return _settings_game.locale.units_velocity_nautical; return _settings_game.locale.units_velocity; } /** * Convert the given (internal) speed to the display speed. * @param speed the speed to convert * @return the converted speed. */ uint ConvertSpeedToDisplaySpeed(uint speed, VehicleType type) { /* For historical reasons we don't want to mess with the * conversion for speed. So, don't round it and keep the * original conversion factors instead of the real ones. */ return _units_velocity[GetVelocityUnits(type)].c.ToDisplay(speed, false); } /** * Convert the given display speed to the (internal) speed. * @param speed the speed to convert * @return the converted speed. */ uint ConvertDisplaySpeedToSpeed(uint speed, VehicleType type) { return _units_velocity[GetVelocityUnits(type)].c.FromDisplay(speed); } /** * Convert the given km/h-ish speed to the display speed. * @param speed the speed to convert * @return the converted speed. */ uint ConvertKmhishSpeedToDisplaySpeed(uint speed, VehicleType type) { return _units_velocity[GetVelocityUnits(type)].c.ToDisplay(speed * 10, false) / 16; } /** * Convert the given display speed to the km/h-ish speed. * @param speed the speed to convert * @return the converted speed. */ uint ConvertDisplaySpeedToKmhishSpeed(uint speed, VehicleType type) { return _units_velocity[GetVelocityUnits(type)].c.FromDisplay(speed * 16, true, 10); } static std::vector _game_script_raw_strings; /** * Parse most format codes within a string and write the result to a buffer. * @param buff The buffer to write the final string to. * @param str_arg The original string with format codes. * @param args Pointer to extra arguments used by various string codes. * @param last Pointer to just past the end of the buff array. * @param dry_run True when the argt array is not yet initialized. */ static char *FormatString(char *buff, const char *str_arg, StringParameters *args, const char *last, uint case_index, bool game_script, bool dry_run) { uint orig_offset = args->offset; /* When there is no array with types there is no need to do a dry run. */ if (args->HasTypeInformation() && !dry_run) { if (UsingNewGRFTextStack()) { /* Values from the NewGRF text stack are only copied to the normal * argv array at the time they are encountered. That means that if * another string command references a value later in the string it * would fail. We solve that by running FormatString twice. The first * pass makes sure the argv array is correctly filled and the second * pass can reference later values without problems. */ struct TextRefStack *backup = CreateTextRefStackBackup(); FormatString(buff, str_arg, args, last, case_index, game_script, true); RestoreTextRefStackBackup(backup); } else { FormatString(buff, str_arg, args, last, case_index, game_script, true); } /* We have to restore the original offset here to to read the correct values. */ args->offset = orig_offset; } WChar b = '\0'; uint next_substr_case_index = 0; char *buf_start = buff; std::stack> str_stack; str_stack.push(str_arg); for (;;) { while (!str_stack.empty() && (b = Utf8Consume(&str_stack.top())) == '\0') { str_stack.pop(); } if (str_stack.empty()) break; const char *&str = str_stack.top(); if (SCC_NEWGRF_FIRST <= b && b <= SCC_NEWGRF_LAST) { /* We need to pass some stuff as it might be modified. */ //todo: should argve be passed here too? b = RemapNewGRFStringControlCode(b, buf_start, &buff, &str, (int64 *)args->GetDataPointer(), args->GetDataLeft(), dry_run); if (b == 0) continue; } switch (b) { case SCC_ENCODED: { uint64 sub_args_data[20]; WChar sub_args_type[20]; bool sub_args_need_free[20]; StringParameters sub_args(sub_args_data, 20, sub_args_type); sub_args.ClearTypeInformation(); memset(sub_args_need_free, 0, sizeof(sub_args_need_free)); char *p; uint32 stringid = std::strtoul(str, &p, 16); if (*p != ':' && *p != '\0') { while (*p != '\0') p++; str = p; buff = strecat(buff, "(invalid SCC_ENCODED)", last); break; } if (stringid >= TAB_SIZE_GAMESCRIPT) { while (*p != '\0') p++; str = p; buff = strecat(buff, "(invalid StringID)", last); break; } int i = 0; while (*p != '\0' && i < 20) { uint64 param; const char *s = ++p; /* Find the next value */ bool instring = false; bool escape = false; for (;; p++) { if (*p == '\\') { escape = true; continue; } if (*p == '"' && escape) { escape = false; continue; } escape = false; if (*p == '"') { instring = !instring; continue; } if (instring) { continue; } if (*p == ':') break; if (*p == '\0') break; } if (*s != '"') { /* Check if we want to look up another string */ WChar l; size_t len = Utf8Decode(&l, s); bool lookup = (l == SCC_ENCODED); if (lookup) s += len; param = std::strtoull(s, &p, 16); if (lookup) { if (param >= TAB_SIZE_GAMESCRIPT) { while (*p != '\0') p++; str = p; buff = strecat(buff, "(invalid sub-StringID)", last); break; } param = MakeStringID(TEXT_TAB_GAMESCRIPT_START, param); } sub_args.SetParam(i++, param); } else { s++; // skip the leading \" char *g = stredup(s); g[p - s - 1] = '\0'; // skip the trailing \" sub_args_need_free[i] = true; sub_args.SetParam(i++, (uint64)(size_t)g); _game_script_raw_strings.push_back(g); } } /* If we didn't error out, we can actually print the string. */ if (*str != '\0') { str = p; buff = GetStringWithArgs(buff, MakeStringID(TEXT_TAB_GAMESCRIPT_START, stringid), &sub_args, last, true); } for (i = 0; i < 20; i++) { if (sub_args_need_free[i]) { free((void *)sub_args.GetParam(i)); _game_script_raw_strings.pop_back(); } } break; } case SCC_NEWGRF_STRINL: { StringID substr = Utf8Consume(&str); str_stack.push(GetStringPtr(substr)); break; } case SCC_NEWGRF_PRINT_WORD_STRING_ID: { StringID substr = args->GetInt32(SCC_NEWGRF_PRINT_WORD_STRING_ID); str_stack.push(GetStringPtr(substr)); case_index = next_substr_case_index; next_substr_case_index = 0; break; } case SCC_GENDER_LIST: { // {G 0 Der Die Das} /* First read the meta data from the language file. */ uint offset = orig_offset + (byte)*str++; int gender = 0; if (!dry_run && args->GetTypeAtOffset(offset) != 0) { /* Now we need to figure out what text to resolve, i.e. * what do we need to draw? So get the actual raw string * first using the control code to get said string. */ char input[4 + 1]; char *p = input + Utf8Encode(input, args->GetTypeAtOffset(offset)); *p = '\0'; /* Now do the string formatting. */ char buf[256]; bool old_sgd = _scan_for_gender_data; _scan_for_gender_data = true; StringParameters tmp_params(args->GetPointerToOffset(offset), args->num_param - offset, nullptr); p = FormatString(buf, input, &tmp_params, lastof(buf)); _scan_for_gender_data = old_sgd; *p = '\0'; /* And determine the string. */ const char *s = buf; WChar c = Utf8Consume(&s); /* Does this string have a gender, if so, set it */ if (c == SCC_GENDER_INDEX) gender = (byte)s[0]; } str = ParseStringChoice(str, gender, &buff, last); break; } /* This sets up the gender for the string. * We just ignore this one. It's used in {G 0 Der Die Das} to determine the case. */ case SCC_GENDER_INDEX: // {GENDER 0} if (_scan_for_gender_data) { buff += Utf8Encode(buff, SCC_GENDER_INDEX); *buff++ = *str++; } else { str++; } break; case SCC_PLURAL_LIST: { // {P} int plural_form = *str++; // contains the plural form for this string uint offset = orig_offset + (byte)*str++; int64 v = *args->GetPointerToOffset(offset); // contains the number that determines plural str = ParseStringChoice(str, DeterminePluralForm(v, plural_form), &buff, last); break; } case SCC_ARG_INDEX: { // Move argument pointer args->offset = orig_offset + (byte)*str++; break; } case SCC_SET_CASE: { // {SET_CASE} /* This is a pseudo command, it's outputted when someone does {STRING.ack} * The modifier is added to all subsequent GetStringWithArgs that accept the modifier. */ next_substr_case_index = (byte)*str++; break; } case SCC_SWITCH_CASE: { // {Used to implement case switching} /* <0x9E> * Each LEN is printed using 2 bytes in big endian order. */ uint num = (byte)*str++; while (num) { if ((byte)str[0] == case_index) { /* Found the case, adjust str pointer and continue */ str += 3; break; } /* Otherwise skip to the next case */ str += 3 + (str[1] << 8) + str[2]; num--; } break; } case SCC_REVISION: // {REV} buff = strecpy(buff, _openttd_revision, last); break; case SCC_RAW_STRING_POINTER: { // {RAW_STRING} const char *raw_string = (const char *)(size_t)args->GetInt64(SCC_RAW_STRING_POINTER); if (game_script && std::find(_game_script_raw_strings.begin(), _game_script_raw_strings.end(), raw_string) == _game_script_raw_strings.end()) { buff = strecat(buff, "(invalid RAW_STRING parameter)", last); break; } buff = FormatString(buff, raw_string, args, last); break; } case SCC_STRING: {// {STRING} StringID string_id = args->GetInt32(SCC_STRING); if (game_script && GetStringTab(string_id) != TEXT_TAB_GAMESCRIPT_START) break; /* WARNING. It's prohibited for the included string to consume any arguments. * For included strings that consume argument, you should use STRING1, STRING2 etc. * To debug stuff you can set argv to nullptr and it will tell you */ StringParameters tmp_params(args->GetDataPointer(), args->GetDataLeft(), nullptr); buff = GetStringWithArgs(buff, string_id, &tmp_params, last, next_substr_case_index, game_script); next_substr_case_index = 0; break; } case SCC_STRING1: case SCC_STRING2: case SCC_STRING3: case SCC_STRING4: case SCC_STRING5: case SCC_STRING6: case SCC_STRING7: { // {STRING1..7} /* Strings that consume arguments */ StringID string_id = args->GetInt32(b); if (game_script && GetStringTab(string_id) != TEXT_TAB_GAMESCRIPT_START) break; uint size = b - SCC_STRING1 + 1; if (game_script && size > args->GetDataLeft()) { buff = strecat(buff, "(too many parameters)", last); } else { StringParameters sub_args(*args, size); buff = GetStringWithArgs(buff, string_id, &sub_args, last, next_substr_case_index, game_script); } next_substr_case_index = 0; break; } case SCC_COMMA: // {COMMA} buff = FormatCommaNumber(buff, args->GetInt64(SCC_COMMA), last); break; case SCC_DECIMAL: {// {DECIMAL} int64 number = args->GetInt64(SCC_DECIMAL); int digits = args->GetInt32(SCC_DECIMAL); buff = FormatCommaNumber(buff, number, last, digits); break; } case SCC_NUM: // {NUM} buff = FormatNoCommaNumber(buff, args->GetInt64(SCC_NUM), last); break; case SCC_ZEROFILL_NUM: { // {ZEROFILL_NUM} int64 num = args->GetInt64(); buff = FormatZerofillNumber(buff, num, args->GetInt64(), last); break; } case SCC_HEX: // {HEX} buff = FormatHexNumber(buff, (uint64)args->GetInt64(SCC_HEX), last); break; case SCC_BYTES: // {BYTES} buff = FormatBytes(buff, args->GetInt64(), last); break; case SCC_CARGO_TINY: { // {CARGO_TINY} /* Tiny description of cargotypes. Layout: * param 1: cargo type * param 2: cargo count */ CargoID cargo = args->GetInt32(SCC_CARGO_TINY); if (cargo >= CargoSpec::GetArraySize()) break; StringID cargo_str = CargoSpec::Get(cargo)->units_volume; int64 amount = 0; switch (cargo_str) { case STR_TONS: amount = _units_weight[_settings_game.locale.units_weight].c.ToDisplay(args->GetInt64()); break; case STR_LITERS: amount = _units_volume[_settings_game.locale.units_volume].c.ToDisplay(args->GetInt64()); break; default: { amount = args->GetInt64(); break; } } buff = FormatCommaNumber(buff, amount, last); break; } case SCC_CARGO_SHORT: { // {CARGO_SHORT} /* Short description of cargotypes. Layout: * param 1: cargo type * param 2: cargo count */ CargoID cargo = args->GetInt32(SCC_CARGO_SHORT); if (cargo >= CargoSpec::GetArraySize()) break; StringID cargo_str = CargoSpec::Get(cargo)->units_volume; switch (cargo_str) { case STR_TONS: { assert(_settings_game.locale.units_weight < lengthof(_units_weight)); int64 args_array[] = {_units_weight[_settings_game.locale.units_weight].c.ToDisplay(args->GetInt64())}; StringParameters tmp_params(args_array); buff = FormatString(buff, GetStringPtr(_units_weight[_settings_game.locale.units_weight].l), &tmp_params, last); break; } case STR_LITERS: { assert(_settings_game.locale.units_volume < lengthof(_units_volume)); int64 args_array[] = {_units_volume[_settings_game.locale.units_volume].c.ToDisplay(args->GetInt64())}; StringParameters tmp_params(args_array); buff = FormatString(buff, GetStringPtr(_units_volume[_settings_game.locale.units_volume].l), &tmp_params, last); break; } default: { StringParameters tmp_params(*args, 1); buff = GetStringWithArgs(buff, cargo_str, &tmp_params, last); break; } } break; } case SCC_CARGO_LONG: { // {CARGO_LONG} /* First parameter is cargo type, second parameter is cargo count */ CargoID cargo = args->GetInt32(SCC_CARGO_LONG); if (cargo != CT_INVALID && cargo >= CargoSpec::GetArraySize()) break; StringID cargo_str = (cargo == CT_INVALID) ? STR_QUANTITY_N_A : CargoSpec::Get(cargo)->quantifier; StringParameters tmp_args(*args, 1); buff = GetStringWithArgs(buff, cargo_str, &tmp_args, last); break; } case SCC_CARGO_LIST: { // {CARGO_LIST} CargoTypes cmask = args->GetInt64(SCC_CARGO_LIST); bool first = true; for (const auto &cs : _sorted_cargo_specs) { if (!HasBit(cmask, cs->Index())) continue; if (buff >= last - 2) break; // ',' and ' ' if (first) { first = false; } else { /* Add a comma if this is not the first item */ *buff++ = ','; *buff++ = ' '; } buff = GetStringWithArgs(buff, cs->name, args, last, next_substr_case_index, game_script); } /* If first is still true then no cargo is accepted */ if (first) buff = GetStringWithArgs(buff, STR_JUST_NOTHING, args, last, next_substr_case_index, game_script); *buff = '\0'; next_substr_case_index = 0; /* Make sure we detect any buffer overflow */ assert(buff < last); break; } case SCC_CURRENCY_SHORT: // {CURRENCY_SHORT} buff = FormatGenericCurrency(buff, _currency, args->GetInt64(), true, last); break; case SCC_CURRENCY_LONG: // {CURRENCY_LONG} buff = FormatGenericCurrency(buff, _currency, args->GetInt64(SCC_CURRENCY_LONG), false, last); break; case SCC_DATE_TINY: // {DATE_TINY} buff = FormatTinyOrISODate(buff, args->GetInt32(SCC_DATE_TINY), STR_FORMAT_DATE_TINY, last); break; case SCC_DATE_SHORT: // {DATE_SHORT} buff = FormatMonthAndYear(buff, args->GetInt32(SCC_DATE_SHORT), last, next_substr_case_index); next_substr_case_index = 0; break; case SCC_DATE_LONG: // {DATE_LONG} buff = FormatYmdString(buff, args->GetInt32(SCC_DATE_LONG), last, next_substr_case_index); next_substr_case_index = 0; break; case SCC_DATE_ISO: // {DATE_ISO} buff = FormatTinyOrISODate(buff, args->GetInt32(), STR_FORMAT_DATE_ISO, last); break; case SCC_FORCE: { // {FORCE} assert(_settings_game.locale.units_force < lengthof(_units_force)); int64 args_array[1] = {_units_force[_settings_game.locale.units_force].c.ToDisplay(args->GetInt64())}; StringParameters tmp_params(args_array); buff = FormatString(buff, GetStringPtr(_units_force[_settings_game.locale.units_force].s), &tmp_params, last); break; } case SCC_HEIGHT: { // {HEIGHT} assert(_settings_game.locale.units_height < lengthof(_units_height)); int64 args_array[] = {_units_height[_settings_game.locale.units_height].c.ToDisplay(args->GetInt64())}; StringParameters tmp_params(args_array); buff = FormatString(buff, GetStringPtr(_units_height[_settings_game.locale.units_height].s), &tmp_params, last); break; } case SCC_POWER: { // {POWER} assert(_settings_game.locale.units_power < lengthof(_units_power)); int64 args_array[1] = {_units_power[_settings_game.locale.units_power].c.ToDisplay(args->GetInt64())}; StringParameters tmp_params(args_array); buff = FormatString(buff, GetStringPtr(_units_power[_settings_game.locale.units_power].s), &tmp_params, last); break; } case SCC_POWER_TO_WEIGHT: { // {POWER_TO_WEIGHT} auto setting = _settings_game.locale.units_power * 3u + _settings_game.locale.units_weight; assert(setting < lengthof(_units_power_to_weight)); auto const &x = _units_power_to_weight[setting]; int64 args_array[] = {x.c.ToDisplay(args->GetInt64()), x.decimal_places}; StringParameters tmp_params(args_array); buff = FormatString(buff, GetStringPtr(x.s), &tmp_params, last); break; } case SCC_VELOCITY: { // {VELOCITY} int64 arg = args->GetInt64(SCC_VELOCITY); // Unpack vehicle type from packed argument to get desired units. VehicleType vt = static_cast(GB(arg, 56, 8)); byte units = GetVelocityUnits(vt); assert(units < lengthof(_units_velocity)); unsigned int decimal_places = _units_velocity[units].decimal_places; uint64 args_array[] = {ConvertKmhishSpeedToDisplaySpeed(GB(arg, 0, 56), vt), decimal_places}; StringParameters tmp_params(args_array, decimal_places ? 2 : 1, nullptr); buff = FormatString(buff, GetStringPtr(_units_velocity[units].s), &tmp_params, last); break; } case SCC_VOLUME_SHORT: { // {VOLUME_SHORT} assert(_settings_game.locale.units_volume < lengthof(_units_volume)); int64 args_array[1] = {_units_volume[_settings_game.locale.units_volume].c.ToDisplay(args->GetInt64())}; StringParameters tmp_params(args_array); buff = FormatString(buff, GetStringPtr(_units_volume[_settings_game.locale.units_volume].s), &tmp_params, last); break; } case SCC_VOLUME_LONG: { // {VOLUME_LONG} assert(_settings_game.locale.units_volume < lengthof(_units_volume)); int64 args_array[1] = {_units_volume[_settings_game.locale.units_volume].c.ToDisplay(args->GetInt64(SCC_VOLUME_LONG))}; StringParameters tmp_params(args_array); buff = FormatString(buff, GetStringPtr(_units_volume[_settings_game.locale.units_volume].l), &tmp_params, last); break; } case SCC_WEIGHT_SHORT: { // {WEIGHT_SHORT} assert(_settings_game.locale.units_weight < lengthof(_units_weight)); int64 args_array[1] = {_units_weight[_settings_game.locale.units_weight].c.ToDisplay(args->GetInt64())}; StringParameters tmp_params(args_array); buff = FormatString(buff, GetStringPtr(_units_weight[_settings_game.locale.units_weight].s), &tmp_params, last); break; } case SCC_WEIGHT_LONG: { // {WEIGHT_LONG} assert(_settings_game.locale.units_weight < lengthof(_units_weight)); int64 args_array[1] = {_units_weight[_settings_game.locale.units_weight].c.ToDisplay(args->GetInt64(SCC_WEIGHT_LONG))}; StringParameters tmp_params(args_array); buff = FormatString(buff, GetStringPtr(_units_weight[_settings_game.locale.units_weight].l), &tmp_params, last); break; } case SCC_COMPANY_NAME: { // {COMPANY} const Company *c = Company::GetIfValid(args->GetInt32()); if (c == nullptr) break; if (!c->name.empty()) { int64 args_array[] = {(int64)(size_t)c->name.c_str()}; StringParameters tmp_params(args_array); buff = GetStringWithArgs(buff, STR_JUST_RAW_STRING, &tmp_params, last); } else { int64 args_array[] = {c->name_2}; StringParameters tmp_params(args_array); buff = GetStringWithArgs(buff, c->name_1, &tmp_params, last); } break; } case SCC_COMPANY_NUM: { // {COMPANY_NUM} CompanyID company = (CompanyID)args->GetInt32(); /* Nothing is added for AI or inactive companies */ if (Company::IsValidHumanID(company)) { int64 args_array[] = {company + 1}; StringParameters tmp_params(args_array); buff = GetStringWithArgs(buff, STR_FORMAT_COMPANY_NUM, &tmp_params, last); } break; } case SCC_DEPOT_NAME: { // {DEPOT} VehicleType vt = (VehicleType)args->GetInt32(SCC_DEPOT_NAME); if (vt == VEH_AIRCRAFT) { uint64 args_array[] = {(uint64)args->GetInt32()}; WChar types_array[] = {SCC_STATION_NAME}; StringParameters tmp_params(args_array, 1, types_array); buff = GetStringWithArgs(buff, STR_FORMAT_DEPOT_NAME_AIRCRAFT, &tmp_params, last); break; } const Depot *d = Depot::Get(args->GetInt32()); if (!d->name.empty()) { int64 args_array[] = {(int64)(size_t)d->name.c_str()}; StringParameters tmp_params(args_array); buff = GetStringWithArgs(buff, STR_JUST_RAW_STRING, &tmp_params, last); } else { int64 args_array[] = {d->town->index, d->town_cn + 1}; StringParameters tmp_params(args_array); buff = GetStringWithArgs(buff, STR_FORMAT_DEPOT_NAME_TRAIN + 2 * vt + (d->town_cn == 0 ? 0 : 1), &tmp_params, last); } break; } case SCC_ENGINE_NAME: { // {ENGINE} int64 arg = args->GetInt64(SCC_ENGINE_NAME); const Engine *e = Engine::GetIfValid(static_cast(arg)); if (e == nullptr) break; if (!e->name.empty() && e->IsEnabled()) { int64 args_array[] = {(int64)(size_t)e->name.c_str()}; StringParameters tmp_params(args_array); buff = GetStringWithArgs(buff, STR_JUST_RAW_STRING, &tmp_params, last); break; } if (HasBit(e->info.callback_mask, CBM_VEHICLE_NAME)) { uint16 callback = GetVehicleCallback(CBID_VEHICLE_NAME, static_cast(arg >> 32), 0, e->index, nullptr); /* Not calling ErrorUnknownCallbackResult due to being inside string processing. */ if (callback != CALLBACK_FAILED && callback < 0x400) { const GRFFile *grffile = e->GetGRF(); assert(grffile != nullptr); StartTextRefStackUsage(grffile, 6); uint64 tmp_dparam[6] = { 0 }; WChar tmp_type[6] = { 0 }; StringParameters tmp_params(tmp_dparam, 6, tmp_type); buff = GetStringWithArgs(buff, GetGRFStringID(grffile->grfid, 0xD000 + callback), &tmp_params, last); StopTextRefStackUsage(); break; } } StringParameters tmp_params(nullptr, 0, nullptr); buff = GetStringWithArgs(buff, e->info.string_id, &tmp_params, last); break; } case SCC_GROUP_NAME: { // {GROUP} const Group *g = Group::GetIfValid(args->GetInt32()); if (g == nullptr) break; if (!g->name.empty()) { int64 args_array[] = {(int64)(size_t)g->name.c_str()}; StringParameters tmp_params(args_array); buff = GetStringWithArgs(buff, STR_JUST_RAW_STRING, &tmp_params, last); } else { int64 args_array[] = {g->index}; StringParameters tmp_params(args_array); buff = GetStringWithArgs(buff, STR_FORMAT_GROUP_NAME, &tmp_params, last); } break; } case SCC_INDUSTRY_NAME: { // {INDUSTRY} const Industry *i = Industry::GetIfValid(args->GetInt32(SCC_INDUSTRY_NAME)); if (i == nullptr) break; static bool use_cache = true; if (use_cache) { // Use cached version if first call AutoRestoreBackup cache_backup(use_cache, false); buff = strecpy(buff, i->GetCachedName(), last); } else if (_scan_for_gender_data) { /* Gender is defined by the industry type. * STR_FORMAT_INDUSTRY_NAME may have the town first, so it would result in the gender of the town name */ StringParameters tmp_params(nullptr, 0, nullptr); buff = FormatString(buff, GetStringPtr(GetIndustrySpec(i->type)->name), &tmp_params, last, next_substr_case_index); } else { /* First print the town name and the industry type name. */ int64 args_array[2] = {i->town->index, GetIndustrySpec(i->type)->name}; StringParameters tmp_params(args_array); buff = FormatString(buff, GetStringPtr(STR_FORMAT_INDUSTRY_NAME), &tmp_params, last, next_substr_case_index); } next_substr_case_index = 0; break; } case SCC_PRESIDENT_NAME: { // {PRESIDENT_NAME} const Company *c = Company::GetIfValid(args->GetInt32(SCC_PRESIDENT_NAME)); if (c == nullptr) break; if (!c->president_name.empty()) { int64 args_array[] = {(int64)(size_t)c->president_name.c_str()}; StringParameters tmp_params(args_array); buff = GetStringWithArgs(buff, STR_JUST_RAW_STRING, &tmp_params, last); } else { int64 args_array[] = {c->president_name_2}; StringParameters tmp_params(args_array); buff = GetStringWithArgs(buff, c->president_name_1, &tmp_params, last); } break; } case SCC_STATION_NAME: { // {STATION} StationID sid = args->GetInt32(SCC_STATION_NAME); const Station *st = Station::GetIfValid(sid); if (st == nullptr) { /* The station doesn't exist anymore. The only place where we might * be "drawing" an invalid station is in the case of cargo that is * in transit. */ StringParameters tmp_params(nullptr, 0, nullptr); buff = GetStringWithArgs(buff, STR_UNKNOWN_STATION, &tmp_params, last); break; } static bool use_cache = true; if (use_cache) { // Use cached version if first call AutoRestoreBackup cache_backup(use_cache, false); buff = strecpy(buff, st->GetCachedName(), last); } else if (!st->name.empty()) { int64 args_array[] = {(int64)(size_t)st->name.c_str()}; StringParameters tmp_params(args_array); buff = GetStringWithArgs(buff, STR_JUST_RAW_STRING, &tmp_params, last); } else { StringID string_id = st->string_id; if (st->indtype != IT_INVALID) { /* Special case where the industry provides the name for the station */ const IndustrySpec *indsp = GetIndustrySpec(st->indtype); /* Industry GRFs can change which might remove the station name and * thus cause very strange things. Here we check for that before we * actually set the station name. */ if (indsp->station_name != STR_NULL && indsp->station_name != STR_UNDEFINED) { string_id = indsp->station_name; } } uint64 args_array[] = {STR_TOWN_NAME, st->town->index, st->index}; WChar types_array[] = {0, SCC_TOWN_NAME, SCC_NUM}; StringParameters tmp_params(args_array, 3, types_array); buff = GetStringWithArgs(buff, string_id, &tmp_params, last); } break; } case SCC_TOWN_NAME: { // {TOWN} const Town *t = Town::GetIfValid(args->GetInt32(SCC_TOWN_NAME)); if (t == nullptr) break; static bool use_cache = true; if (use_cache) { // Use cached version if first call AutoRestoreBackup cache_backup(use_cache, false); buff = strecpy(buff, t->GetCachedName(), last); } else if (!t->name.empty()) { int64 args_array[] = {(int64)(size_t)t->name.c_str()}; StringParameters tmp_params(args_array); buff = GetStringWithArgs(buff, STR_JUST_RAW_STRING, &tmp_params, last); } else { buff = GetTownName(buff, t, last); } break; } case SCC_WAYPOINT_NAME: { // {WAYPOINT} Waypoint *wp = Waypoint::GetIfValid(args->GetInt32(SCC_WAYPOINT_NAME)); if (wp == nullptr) break; if (!wp->name.empty()) { int64 args_array[] = {(int64)(size_t)wp->name.c_str()}; StringParameters tmp_params(args_array); buff = GetStringWithArgs(buff, STR_JUST_RAW_STRING, &tmp_params, last); } else { int64 args_array[] = {wp->town->index, wp->town_cn + 1}; StringParameters tmp_params(args_array); StringID string_id = ((wp->string_id == STR_SV_STNAME_BUOY) ? STR_FORMAT_BUOY_NAME : STR_FORMAT_WAYPOINT_NAME); if (wp->town_cn != 0) string_id++; buff = GetStringWithArgs(buff, string_id, &tmp_params, last); } break; } case SCC_VEHICLE_NAME: { // {VEHICLE} const Vehicle *v = Vehicle::GetIfValid(args->GetInt32(SCC_VEHICLE_NAME)); if (v == nullptr) break; if (!v->name.empty()) { int64 args_array[] = {(int64)(size_t)v->name.c_str()}; StringParameters tmp_params(args_array); buff = GetStringWithArgs(buff, STR_JUST_RAW_STRING, &tmp_params, last); } else if (v->group_id != DEFAULT_GROUP) { /* The vehicle has no name, but is member of a group, so print group name */ int64 args_array[] = {v->group_id, v->unitnumber}; StringParameters tmp_params(args_array); buff = GetStringWithArgs(buff, STR_FORMAT_GROUP_VEHICLE_NAME, &tmp_params, last); } else { int64 args_array[] = {v->unitnumber}; StringParameters tmp_params(args_array); StringID string_id; switch (v->type) { default: string_id = STR_INVALID_VEHICLE; break; case VEH_TRAIN: string_id = STR_SV_TRAIN_NAME; break; case VEH_ROAD: string_id = STR_SV_ROAD_VEHICLE_NAME; break; case VEH_SHIP: string_id = STR_SV_SHIP_NAME; break; case VEH_AIRCRAFT: string_id = STR_SV_AIRCRAFT_NAME; break; } buff = GetStringWithArgs(buff, string_id, &tmp_params, last); } break; } case SCC_SIGN_NAME: { // {SIGN} const Sign *si = Sign::GetIfValid(args->GetInt32()); if (si == nullptr) break; if (!si->name.empty()) { int64 args_array[] = {(int64)(size_t)si->name.c_str()}; StringParameters tmp_params(args_array); buff = GetStringWithArgs(buff, STR_JUST_RAW_STRING, &tmp_params, last); } else { StringParameters tmp_params(nullptr, 0, nullptr); buff = GetStringWithArgs(buff, STR_DEFAULT_SIGN_NAME, &tmp_params, last); } break; } case SCC_STATION_FEATURES: { // {STATIONFEATURES} buff = StationGetSpecialString(buff, args->GetInt32(SCC_STATION_FEATURES), last); break; } default: if (buff + Utf8CharLen(b) < last) buff += Utf8Encode(buff, b); break; } } *buff = '\0'; return buff; } static char *StationGetSpecialString(char *buff, int x, const char *last) { if ((x & FACIL_TRAIN) && (buff + Utf8CharLen(SCC_TRAIN) < last)) buff += Utf8Encode(buff, SCC_TRAIN); if ((x & FACIL_TRUCK_STOP) && (buff + Utf8CharLen(SCC_LORRY) < last)) buff += Utf8Encode(buff, SCC_LORRY); if ((x & FACIL_BUS_STOP) && (buff + Utf8CharLen(SCC_BUS) < last)) buff += Utf8Encode(buff, SCC_BUS); if ((x & FACIL_DOCK) && (buff + Utf8CharLen(SCC_SHIP) < last)) buff += Utf8Encode(buff, SCC_SHIP); if ((x & FACIL_AIRPORT) && (buff + Utf8CharLen(SCC_PLANE) < last)) buff += Utf8Encode(buff, SCC_PLANE); *buff = '\0'; return buff; } static char *GetSpecialTownNameString(char *buff, int ind, uint32 seed, const char *last) { return GenerateTownNameString(buff, last, ind, seed); } static const char * const _silly_company_names[] = { "Bloggs Brothers", "Tiny Transport Ltd.", "Express Travel", "Comfy-Coach & Co.", "Crush & Bump Ltd.", "Broken & Late Ltd.", "Sam Speedy & Son", "Supersonic Travel", "Mike's Motors", "Lightning International", "Pannik & Loozit Ltd.", "Inter-City Transport", "Getout & Pushit Ltd." }; static const char * const _surname_list[] = { "Adams", "Allan", "Baker", "Bigwig", "Black", "Bloggs", "Brown", "Campbell", "Gordon", "Hamilton", "Hawthorn", "Higgins", "Green", "Gribble", "Jones", "McAlpine", "MacDonald", "McIntosh", "Muir", "Murphy", "Nelson", "O'Donnell", "Parker", "Phillips", "Pilkington", "Quigley", "Sharkey", "Thomson", "Watkins" }; static const char * const _silly_surname_list[] = { "Grumpy", "Dozy", "Speedy", "Nosey", "Dribble", "Mushroom", "Cabbage", "Sniffle", "Fishy", "Swindle", "Sneaky", "Nutkins" }; static const char _initial_name_letters[] = { 'A', 'B', 'C', 'D', 'E', 'F', 'G', 'H', 'I', 'J', 'K', 'L', 'M', 'N', 'P', 'R', 'S', 'T', 'W', }; static char *GenAndCoName(char *buff, uint32 arg, const char *last) { const char * const *base; uint num; if (_settings_game.game_creation.landscape == LT_TOYLAND) { base = _silly_surname_list; num = lengthof(_silly_surname_list); } else { base = _surname_list; num = lengthof(_surname_list); } buff = strecpy(buff, base[num * GB(arg, 16, 8) >> 8], last); buff = strecpy(buff, " & Co.", last); return buff; } static char *GenPresidentName(char *buff, uint32 x, const char *last) { char initial[] = "?. "; const char * const *base; uint num; uint i; initial[0] = _initial_name_letters[sizeof(_initial_name_letters) * GB(x, 0, 8) >> 8]; buff = strecpy(buff, initial, last); i = (sizeof(_initial_name_letters) + 35) * GB(x, 8, 8) >> 8; if (i < sizeof(_initial_name_letters)) { initial[0] = _initial_name_letters[i]; buff = strecpy(buff, initial, last); } if (_settings_game.game_creation.landscape == LT_TOYLAND) { base = _silly_surname_list; num = lengthof(_silly_surname_list); } else { base = _surname_list; num = lengthof(_surname_list); } buff = strecpy(buff, base[num * GB(x, 16, 8) >> 8], last); return buff; } static char *GetSpecialNameString(char *buff, int ind, StringParameters *args, const char *last) { switch (ind) { case 1: // not used return strecpy(buff, _silly_company_names[std::min(args->GetInt32() & 0xFFFF, lengthof(_silly_company_names) - 1)], last); case 2: // used for Foobar & Co company names return GenAndCoName(buff, args->GetInt32(), last); case 3: // President name return GenPresidentName(buff, args->GetInt32(), last); } /* town name? */ if (IsInsideMM(ind - 6, 0, SPECSTR_TOWNNAME_LAST - SPECSTR_TOWNNAME_START + 1)) { buff = GetSpecialTownNameString(buff, ind - 6, args->GetInt32(), last); return strecpy(buff, " Transport", last); } NOT_REACHED(); } /** * Check whether the header is a valid header for OpenTTD. * @return true iff the header is deemed valid. */ bool LanguagePackHeader::IsValid() const { return this->ident == TO_LE32(LanguagePackHeader::IDENT) && this->version == TO_LE32(LANGUAGE_PACK_VERSION) && this->plural_form < LANGUAGE_MAX_PLURAL && this->text_dir <= 1 && this->newgrflangid < MAX_LANG && this->num_genders < MAX_NUM_GENDERS && this->num_cases < MAX_NUM_CASES && StrValid(this->name, lastof(this->name)) && StrValid(this->own_name, lastof(this->own_name)) && StrValid(this->isocode, lastof(this->isocode)) && StrValid(this->digit_group_separator, lastof(this->digit_group_separator)) && StrValid(this->digit_group_separator_currency, lastof(this->digit_group_separator_currency)) && StrValid(this->digit_decimal_separator, lastof(this->digit_decimal_separator)); } /** * Check whether a translation is sufficiently finished to offer it to the public. */ bool LanguagePackHeader::IsReasonablyFinished() const { /* "Less than 25% missing" is "sufficiently finished". */ return 4 * this->missing < LANGUAGE_TOTAL_STRINGS; } /** * Read a particular language. * @param lang The metadata about the language. * @return Whether the loading went okay or not. */ bool ReadLanguagePack(const LanguageMetadata *lang) { /* Current language pack */ size_t len = 0; std::unique_ptr lang_pack(reinterpret_cast(ReadFileToMem(lang->file, len, 1U << 20).release())); if (!lang_pack) return false; /* End of read data (+ terminating zero added in ReadFileToMem()) */ const char *end = (char *)lang_pack.get() + len + 1; /* We need at least one byte of lang_pack->data */ if (end <= lang_pack->data || !lang_pack->IsValid()) { return false; } #if TTD_ENDIAN == TTD_BIG_ENDIAN for (uint i = 0; i < TEXT_TAB_END; i++) { lang_pack->offsets[i] = ReadLE16Aligned(&lang_pack->offsets[i]); } #endif /* TTD_ENDIAN == TTD_BIG_ENDIAN */ std::array tab_start, tab_num; uint count = 0; for (uint i = 0; i < TEXT_TAB_END; i++) { uint16 num = lang_pack->offsets[i]; if (num > TAB_SIZE) return false; tab_start[i] = count; tab_num[i] = num; count += num; } /* Allocate offsets */ std::vector offs(count); /* Fill offsets */ char *s = lang_pack->data; len = (byte)*s++; for (uint i = 0; i < count; i++) { if (s + len >= end) return false; if (len >= 0xC0) { len = ((len & 0x3F) << 8) + (byte)*s++; if (s + len >= end) return false; } offs[i] = s; s += len; len = (byte)*s; *s++ = '\0'; // zero terminate the string } _langpack.langpack = std::move(lang_pack); _langpack.offsets = std::move(offs); _langpack.langtab_num = tab_num; _langpack.langtab_start = tab_start; _current_language = lang; _current_text_dir = (TextDirection)_current_language->text_dir; const char *c_file = strrchr(_current_language->file, PATHSEPCHAR) + 1; _config_language_file = c_file; SetCurrentGrfLangID(_current_language->newgrflangid); #ifdef _WIN32 extern void Win32SetCurrentLocaleName(const char *iso_code); Win32SetCurrentLocaleName(_current_language->isocode); #endif #ifdef WITH_COCOA extern void MacOSSetCurrentLocaleName(const char *iso_code); MacOSSetCurrentLocaleName(_current_language->isocode); #endif #ifdef WITH_ICU_I18N /* Create a collator instance for our current locale. */ UErrorCode status = U_ZERO_ERROR; _current_collator.reset(icu::Collator::createInstance(icu::Locale(_current_language->isocode), status)); /* Sort number substrings by their numerical value. */ if (_current_collator) _current_collator->setAttribute(UCOL_NUMERIC_COLLATION, UCOL_ON, status); /* Avoid using the collator if it is not correctly set. */ if (U_FAILURE(status)) { _current_collator.reset(); } #endif /* WITH_ICU_I18N */ /* Some lists need to be sorted again after a language change. */ ReconsiderGameScriptLanguage(); InitializeSortedCargoSpecs(); SortIndustryTypes(); BuildIndustriesLegend(); BuildContentTypeStringList(); InvalidateWindowClassesData(WC_BUILD_VEHICLE); // Build vehicle window. InvalidateWindowClassesData(WC_TRAINS_LIST); // Train group window. InvalidateWindowClassesData(WC_ROADVEH_LIST); // Road vehicle group window. InvalidateWindowClassesData(WC_SHIPS_LIST); // Ship group window. InvalidateWindowClassesData(WC_AIRCRAFT_LIST); // Aircraft group window. InvalidateWindowClassesData(WC_INDUSTRY_DIRECTORY); // Industry directory window. InvalidateWindowClassesData(WC_STATION_LIST); // Station list window. return true; } /* Win32 implementation in win32.cpp. * OS X implementation in os/macosx/macos.mm. */ #if !(defined(_WIN32) || defined(__APPLE__)) /** * Determine the current charset based on the environment * First check some default values, after this one we passed ourselves * and if none exist return the value for $LANG * @param param environment variable to check conditionally if default ones are not * set. Pass nullptr if you don't want additional checks. * @return return string containing current charset, or nullptr if not-determinable */ const char *GetCurrentLocale(const char *param) { const char *env; env = std::getenv("LANGUAGE"); if (env != nullptr) return env; env = std::getenv("LC_ALL"); if (env != nullptr) return env; if (param != nullptr) { env = std::getenv(param); if (env != nullptr) return env; } return std::getenv("LANG"); } #else const char *GetCurrentLocale(const char *param); #endif /* !(defined(_WIN32) || defined(__APPLE__)) */ bool StringIDSorter(const StringID &a, const StringID &b) { char stra[512]; char strb[512]; GetString(stra, a, lastof(stra)); GetString(strb, b, lastof(strb)); return strnatcmp(stra, strb) < 0; } /** * Get the language with the given NewGRF language ID. * @param newgrflangid NewGRF languages ID to check. * @return The language's metadata, or nullptr if it is not known. */ const LanguageMetadata *GetLanguage(byte newgrflangid) { for (const LanguageMetadata &lang : _languages) { if (newgrflangid == lang.newgrflangid) return ⟨ } return nullptr; } /** * Reads the language file header and checks compatibility. * @param file the file to read * @param hdr the place to write the header information to * @return true if and only if the language file is of a compatible version */ static bool GetLanguageFileHeader(const char *file, LanguagePackHeader *hdr) { FILE *f = fopen(file, "rb"); if (f == nullptr) return false; size_t read = fread(hdr, sizeof(*hdr), 1, f); fclose(f); bool ret = read == 1 && hdr->IsValid(); /* Convert endianness for the windows language ID */ if (ret) { hdr->missing = FROM_LE16(hdr->missing); hdr->winlangid = FROM_LE16(hdr->winlangid); } return ret; } /** * Gets a list of languages from the given directory. * @param path the base directory to search in */ static void GetLanguageList(const char *path) { DIR *dir = ttd_opendir(path); if (dir != nullptr) { struct dirent *dirent; while ((dirent = readdir(dir)) != nullptr) { std::string d_name = FS2OTTD(dirent->d_name); const char *extension = strrchr(d_name.c_str(), '.'); /* Not a language file */ if (extension == nullptr || strcmp(extension, ".lng") != 0) continue; LanguageMetadata lmd; seprintf(lmd.file, lastof(lmd.file), "%s%s", path, d_name.c_str()); /* Check whether the file is of the correct version */ if (!GetLanguageFileHeader(lmd.file, &lmd)) { Debug(misc, 3, "{} is not a valid language file", lmd.file); } else if (GetLanguage(lmd.newgrflangid) != nullptr) { Debug(misc, 3, "{}'s language ID is already known", lmd.file); } else { _languages.push_back(lmd); } } closedir(dir); } } /** * Make a list of the available language packs. Put the data in * #_languages list. */ void InitializeLanguagePacks() { for (Searchpath sp : _valid_searchpaths) { std::string path = FioGetDirectory(sp, LANG_DIR); GetLanguageList(path.c_str()); } if (_languages.size() == 0) UserError("No available language packs (invalid versions?)"); /* Acquire the locale of the current system */ const char *lang = GetCurrentLocale("LC_MESSAGES"); if (lang == nullptr) lang = "en_GB"; const LanguageMetadata *chosen_language = nullptr; ///< Matching the language in the configuration file or the current locale const LanguageMetadata *language_fallback = nullptr; ///< Using pt_PT for pt_BR locale when pt_BR is not available const LanguageMetadata *en_GB_fallback = _languages.data(); ///< Fallback when no locale-matching language has been found /* Find a proper language. */ for (const LanguageMetadata &lng : _languages) { /* We are trying to find a default language. The priority is by * configuration file, local environment and last, if nothing found, * English. */ const char *lang_file = strrchr(lng.file, PATHSEPCHAR) + 1; if (_config_language_file == lang_file) { chosen_language = &lng; break; } if (strcmp (lng.isocode, "en_GB") == 0) en_GB_fallback = &lng; /* Only auto-pick finished translations */ if (!lng.IsReasonablyFinished()) continue; if (strncmp(lng.isocode, lang, 5) == 0) chosen_language = &lng; if (strncmp(lng.isocode, lang, 2) == 0) language_fallback = &lng; } /* We haven't found the language in the config nor the one in the locale. * Now we set it to one of the fallback languages */ if (chosen_language == nullptr) { chosen_language = (language_fallback != nullptr) ? language_fallback : en_GB_fallback; } if (!ReadLanguagePack(chosen_language)) UserError("Can't read language pack '{}'", chosen_language->file); } /** * Get the ISO language code of the currently loaded language. * @return the ISO code. */ const char *GetCurrentLanguageIsoCode() { return _langpack.langpack->isocode; } /** * Check whether there are glyphs missing in the current language. * @return If glyphs are missing, return \c true, else return \c false. */ bool MissingGlyphSearcher::FindMissingGlyphs() { InitFontCache(this->Monospace()); const Sprite *question_mark[FS_END]; for (FontSize size = this->Monospace() ? FS_MONO : FS_BEGIN; size < (this->Monospace() ? FS_END : FS_MONO); size++) { question_mark[size] = GetGlyph(size, '?'); } this->Reset(); for (const char *text = this->NextString(); text != nullptr; text = this->NextString()) { FontSize size = this->DefaultSize(); for (WChar c = Utf8Consume(&text); c != '\0'; c = Utf8Consume(&text)) { if (c >= SCC_FIRST_FONT && c <= SCC_LAST_FONT) { size = (FontSize)(c - SCC_FIRST_FONT); } else if (!IsInsideMM(c, SCC_SPRITE_START, SCC_SPRITE_END) && IsPrintable(c) && !IsTextDirectionChar(c) && c != '?' && GetGlyph(size, c) == question_mark[size]) { /* The character is printable, but not in the normal font. This is the case we were testing for. */ std::string size_name; switch (size) { case 0: size_name = "medium"; break; case 1: size_name = "small"; break; case 2: size_name = "large"; break; case 3: size_name = "mono"; break; default: NOT_REACHED(); } Debug(fontcache, 0, "Font is missing glyphs to display char 0x{:X} in {} font size", (int)c, size_name); return true; } } } return false; } /** Helper for searching through the language pack. */ class LanguagePackGlyphSearcher : public MissingGlyphSearcher { uint i; ///< Iterator for the primary language tables. uint j; ///< Iterator for the secondary language tables. void Reset() override { this->i = 0; this->j = 0; } FontSize DefaultSize() override { return FS_NORMAL; } const char *NextString() override { if (this->i >= TEXT_TAB_END) return nullptr; const char *ret = _langpack.offsets[_langpack.langtab_start[this->i] + this->j]; this->j++; while (this->i < TEXT_TAB_END && this->j >= _langpack.langtab_num[this->i]) { this->i++; this->j = 0; } return ret; } bool Monospace() override { return false; } void SetFontNames(FontCacheSettings *settings, const char *font_name, const void *os_data) override { #if defined(WITH_FREETYPE) || defined(_WIN32) || defined(WITH_COCOA) settings->small.font = font_name; settings->medium.font = font_name; settings->large.font = font_name; settings->small.os_handle = os_data; settings->medium.os_handle = os_data; settings->large.os_handle = os_data; #endif } }; /** * Check whether the currently loaded language pack * uses characters that the currently loaded font * does not support. If this is the case an error * message will be shown in English. The error * message will not be localized because that would * mean it might use characters that are not in the * font, which is the whole reason this check has * been added. * @param base_font Whether to look at the base font as well. * @param searcher The methods to use to search for strings to check. * If nullptr the loaded language pack searcher is used. */ void CheckForMissingGlyphs(bool base_font, MissingGlyphSearcher *searcher) { static LanguagePackGlyphSearcher pack_searcher; if (searcher == nullptr) searcher = &pack_searcher; bool bad_font = !base_font || searcher->FindMissingGlyphs(); #if defined(WITH_FREETYPE) || defined(_WIN32) || defined(WITH_COCOA) if (bad_font) { /* We found an unprintable character... lets try whether we can find * a fallback font that can print the characters in the current language. */ bool any_font_configured = !_fcsettings.medium.font.empty(); FontCacheSettings backup = _fcsettings; _fcsettings.mono.os_handle = nullptr; _fcsettings.medium.os_handle = nullptr; bad_font = !SetFallbackFont(&_fcsettings, _langpack.langpack->isocode, _langpack.langpack->winlangid, searcher); _fcsettings = backup; if (!bad_font && any_font_configured) { /* If the user configured a bad font, and we found a better one, * show that we loaded the better font instead of the configured one. * The colour 'character' might change in the * future, so for safety we just Utf8 Encode it into the string, * which takes exactly three characters, so it replaces the "XXX" * with the colour marker. */ static char *err_str = stredup("XXXThe current font is missing some of the characters used in the texts for this language. Using system fallback font instead."); Utf8Encode(err_str, SCC_YELLOW); SetDParamStr(0, err_str); ShowErrorMessage(STR_JUST_RAW_STRING, INVALID_STRING_ID, WL_WARNING); } if (bad_font && base_font) { /* Our fallback font does miss characters too, so keep the * user chosen font as that is more likely to be any good than * the wild guess we made */ InitFontCache(searcher->Monospace()); } } #endif if (bad_font) { /* All attempts have failed. Display an error. As we do not want the string to be translated by * the translators, we 'force' it into the binary and 'load' it via a BindCString. To do this * properly we have to set the colour of the string, otherwise we end up with a lot of artifacts. * The colour 'character' might change in the future, so for safety we just Utf8 Encode it into * the string, which takes exactly three characters, so it replaces the "XXX" with the colour marker. */ static char *err_str = stredup("XXXThe current font is missing some of the characters used in the texts for this language. Read the readme to see how to solve this."); Utf8Encode(err_str, SCC_YELLOW); SetDParamStr(0, err_str); ShowErrorMessage(STR_JUST_RAW_STRING, INVALID_STRING_ID, WL_WARNING); /* Reset the font width */ LoadStringWidthTable(searcher->Monospace()); return; } /* Update the font with cache */ LoadStringWidthTable(searcher->Monospace()); #if !defined(WITH_ICU_LX) && !defined(WITH_UNISCRIBE) && !defined(WITH_COCOA) /* * For right-to-left languages we need the ICU library. If * we do not have support for that library we warn the user * about it with a message. As we do not want the string to * be translated by the translators, we 'force' it into the * binary and 'load' it via a BindCString. To do this * properly we have to set the colour of the string, * otherwise we end up with a lot of artifacts. The colour * 'character' might change in the future, so for safety * we just Utf8 Encode it into the string, which takes * exactly three characters, so it replaces the "XXX" with * the colour marker. */ if (_current_text_dir != TD_LTR) { static char *err_str = stredup("XXXThis version of OpenTTD does not support right-to-left languages. Recompile with icu enabled."); Utf8Encode(err_str, SCC_YELLOW); SetDParamStr(0, err_str); ShowErrorMessage(STR_JUST_RAW_STRING, INVALID_STRING_ID, WL_ERROR); } #endif /* !WITH_ICU_LX */ }