1
0
mirror of https://github.com/vcmi/vcmi.git synced 2024-12-14 10:12:59 +02:00
vcmi/lib/texts/CLegacyConfigParser.cpp
2024-08-12 18:26:30 +00:00

161 lines
3.6 KiB
C++

/*
* CLegacyConfigParser.cpp, 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
*
*/
#include "StdInc.h"
#include "CLegacyConfigParser.h"
#include "TextOperations.h"
#include "Languages.h"
#include "../VCMI_Lib.h"
#include "filesystem/Filesystem.h"
#include "../modding/CModHandler.h"
VCMI_LIB_NAMESPACE_BEGIN
//Helper for string -> float conversion
class LocaleWithComma: public std::numpunct<char>
{
protected:
char do_decimal_point() const override
{
return ',';
}
};
CLegacyConfigParser::CLegacyConfigParser(const TextPath & resource)
{
auto input = CResourceHandler::get()->load(resource);
std::string modName = VLC->modh->findResourceOrigin(resource);
std::string language = VLC->modh->getModLanguage(modName);
fileEncoding = Languages::getLanguageOptions(language).encoding;
data.reset(new char[input->getSize()]);
input->read(reinterpret_cast<uint8_t*>(data.get()), input->getSize());
curr = data.get();
end = curr + input->getSize();
}
std::string CLegacyConfigParser::extractQuotedPart()
{
assert(*curr == '\"');
curr++; // skip quote
const char * begin = curr;
while (curr != end && *curr != '\"' && *curr != '\t')
curr++;
return std::string(begin, curr++); //increment curr to close quote
}
std::string CLegacyConfigParser::extractQuotedString()
{
assert(*curr == '\"');
std::string ret;
while (true)
{
ret += extractQuotedPart();
// double quote - add it to string and continue quoted part
if (curr < end && *curr == '\"')
{
ret += '\"';
}
//extract normal part
else if(curr < end && *curr != '\t' && *curr != '\r')
{
const char * begin = curr;
while (curr < end && *curr != '\t' && *curr != '\r' && *curr != '\"')//find end of string or next quoted part start
curr++;
ret += std::string(begin, curr);
if(curr>=end || *curr != '\"')
return ret;
}
else // end of string
return ret;
}
}
std::string CLegacyConfigParser::extractNormalString()
{
const char * begin = curr;
while (curr < end && *curr != '\t' && *curr != '\r')//find end of string
curr++;
return std::string(begin, curr);
}
std::string CLegacyConfigParser::readRawString()
{
if (curr >= end || *curr == '\n')
return "";
std::string ret;
if (*curr == '\"')
ret = extractQuotedString();// quoted text - find closing quote
else
ret = extractNormalString();//string without quotes - copy till \t or \r
curr++;
return ret;
}
std::string CLegacyConfigParser::readString()
{
// do not convert strings that are already in ASCII - this will only slow down loading process
std::string str = readRawString();
if (TextOperations::isValidASCII(str))
return str;
return TextOperations::toUnicode(str, fileEncoding);
}
float CLegacyConfigParser::readNumber()
{
std::string input = readRawString();
std::istringstream stream(input);
if(input.find(',') != std::string::npos) // code to handle conversion with comma as decimal separator
stream.imbue(std::locale(std::locale(), new LocaleWithComma()));
float result;
if ( !(stream >> result) )
return 0;
return result;
}
bool CLegacyConfigParser::isNextEntryEmpty() const
{
const char * nextSymbol = curr;
while (nextSymbol < end && *nextSymbol == ' ')
nextSymbol++; //find next meaningful symbol
return nextSymbol >= end || *nextSymbol == '\n' || *nextSymbol == '\r' || *nextSymbol == '\t';
}
bool CLegacyConfigParser::endLine()
{
while (curr < end && *curr != '\n')
readString();
curr++;
return curr < end;
}
VCMI_LIB_NAMESPACE_END