/*
* 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 strgen.h Structures related to strgen. */
#ifndef STRGEN_H
#define STRGEN_H
#include "../language.h"
#include "../3rdparty/fmt/format.h"
#include
#include
/** Container for the different cases of a string. */
struct Case {
int caseidx; ///< The index of the case.
std::string string; ///< The translation of the case.
Case(int caseidx, const std::string &string);
};
/** Information about a single string. */
struct LangString {
std::string name; ///< Name of the string.
std::string english; ///< English text.
std::string translated; ///< Translated text.
size_t index; ///< The index in the language file.
int line; ///< Line of string in source-file.
std::vector translated_cases; ///< Cases of the translation.
LangString(const std::string &name, const std::string &english, size_t index, int line);
void FreeTranslation();
};
/** Information about the currently known strings. */
struct StringData {
std::vector> strings; ///< List of all known strings.
std::unordered_map name_to_string; ///< Lookup table for the strings.
size_t tabs; ///< The number of 'tabs' of strings.
size_t max_strings; ///< The maximum number of strings.
size_t next_string_id;///< The next string ID to allocate.
StringData(size_t tabs);
void FreeTranslation();
void Add(std::unique_ptr ls);
LangString *Find(const std::string_view s);
uint VersionHashStr(uint hash, const char *s) const;
uint Version() const;
uint CountInUse(uint tab) const;
};
/** Helper for reading strings. */
struct StringReader {
StringData &data; ///< The data to fill during reading.
const std::string file; ///< The file we are reading.
bool master; ///< Are we reading the master file?
bool translation; ///< Are we reading a translation, implies !master. However, the base translation will have this false.
StringReader(StringData &data, const std::string &file, bool master, bool translation);
virtual ~StringReader() {}
void HandleString(char *str);
/**
* Read a single line from the source of strings.
* @return The line, or std::nullopt if at the end of the file.
*/
virtual std::optional ReadLine() = 0;
/**
* Handle the pragma of the file.
* @param str The pragma string to parse.
*/
virtual void HandlePragma(char *str);
/**
* Start parsing the file.
*/
virtual void ParseFile();
};
/** Base class for writing the header, i.e. the STR_XXX to numeric value. */
struct HeaderWriter {
/**
* Write the string ID.
* @param name The name of the string.
* @param stringid The ID of the string.
*/
virtual void WriteStringID(const std::string &name, int stringid) = 0;
/**
* Finalise writing the file.
* @param data The data about the string.
*/
virtual void Finalise(const StringData &data) = 0;
/** Especially destroy the subclasses. */
virtual ~HeaderWriter() = default;
void WriteHeader(const StringData &data);
};
/** Base class for all language writers. */
struct LanguageWriter {
/**
* Write the header metadata. The multi-byte integers are already converted to
* the little endian format.
* @param header The header to write.
*/
virtual void WriteHeader(const LanguagePackHeader *header) = 0;
/**
* Write a number of bytes.
* @param buffer The buffer to write.
* @param length The amount of byte to write.
*/
virtual void Write(const byte *buffer, size_t length) = 0;
/**
* Finalise writing the file.
*/
virtual void Finalise() = 0;
/** Especially destroy the subclasses. */
virtual ~LanguageWriter() = default;
virtual void WriteLength(uint length);
virtual void WriteLang(const StringData &data);
};
struct CmdStruct;
struct CmdPair {
const CmdStruct *cmd;
std::string param;
};
struct ParsedCommandStruct {
std::vector non_consuming_commands;
std::array consuming_commands{ nullptr }; // ordered by param #
};
const CmdStruct *TranslateCmdForCompare(const CmdStruct *a);
ParsedCommandStruct ExtractCommandString(const char *s, bool warnings);
void StrgenWarningI(const std::string &msg);
void StrgenErrorI(const std::string &msg);
void NORETURN StrgenFatalI(const std::string &msg);
#define StrgenWarning(format_string, ...) StrgenWarningI(fmt::format(FMT_STRING(format_string), ## __VA_ARGS__))
#define StrgenError(format_string, ...) StrgenErrorI(fmt::format(FMT_STRING(format_string), ## __VA_ARGS__))
#define StrgenFatal(format_string, ...) StrgenFatalI(fmt::format(FMT_STRING(format_string), ## __VA_ARGS__))
char *ParseWord(char **buf);
extern const char *_file;
extern int _cur_line;
extern int _errors, _warnings, _show_todo;
extern LanguagePackHeader _lang;
#endif /* STRGEN_H */