mirror of
https://github.com/dolphin-emu/dolphin.git
synced 2024-11-15 22:09:19 -07:00
e764723832
git-svn-id: https://dolphin-emu.googlecode.com/svn/trunk@366 8ced0084-cf51-0410-be5f-012b33b47a6e
461 lines
8.2 KiB
C++
461 lines
8.2 KiB
C++
// Copyright (C) 2003-2008 Dolphin Project.
|
|
|
|
// This program 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 2.0.
|
|
|
|
// This program 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 2.0 for more details.
|
|
|
|
// A copy of the GPL 2.0 should have been included with the program.
|
|
// If not, see http://www.gnu.org/licenses/
|
|
|
|
// Official SVN repository and contact information can be found at
|
|
// http://code.google.com/p/dolphin-emu/
|
|
|
|
#include <stdlib.h>
|
|
#include <stdio.h>
|
|
|
|
#include "StringUtil.h"
|
|
#include "TestFramework.h"
|
|
|
|
// faster than sscanf
|
|
bool AsciiToHex(const char* _szValue, u32& result)
|
|
{
|
|
u32 value = 0;
|
|
size_t finish = strlen(_szValue);
|
|
|
|
if (finish > 8)
|
|
{
|
|
finish = 8;
|
|
}
|
|
|
|
for (size_t count = 0; count < finish; count++)
|
|
{
|
|
value <<= 4;
|
|
|
|
switch (_szValue[count])
|
|
{
|
|
case '0':
|
|
break;
|
|
|
|
case '1':
|
|
value += 1;
|
|
break;
|
|
|
|
case '2':
|
|
value += 2;
|
|
break;
|
|
|
|
case '3':
|
|
value += 3;
|
|
break;
|
|
|
|
case '4':
|
|
value += 4;
|
|
break;
|
|
|
|
case '5':
|
|
value += 5;
|
|
break;
|
|
|
|
case '6':
|
|
value += 6;
|
|
break;
|
|
|
|
case '7':
|
|
value += 7;
|
|
break;
|
|
|
|
case '8':
|
|
value += 8;
|
|
break;
|
|
|
|
case '9':
|
|
value += 9;
|
|
break;
|
|
|
|
case 'A':
|
|
value += 10;
|
|
break;
|
|
|
|
case 'a':
|
|
value += 10;
|
|
break;
|
|
|
|
case 'B':
|
|
value += 11;
|
|
break;
|
|
|
|
case 'b':
|
|
value += 11;
|
|
break;
|
|
|
|
case 'C':
|
|
value += 12;
|
|
break;
|
|
|
|
case 'c':
|
|
value += 12;
|
|
break;
|
|
|
|
case 'D':
|
|
value += 13;
|
|
break;
|
|
|
|
case 'd':
|
|
value += 13;
|
|
break;
|
|
|
|
case 'E':
|
|
value += 14;
|
|
break;
|
|
|
|
case 'e':
|
|
value += 14;
|
|
break;
|
|
|
|
case 'F':
|
|
value += 15;
|
|
break;
|
|
|
|
case 'f':
|
|
value += 15;
|
|
break;
|
|
|
|
default:
|
|
return false;
|
|
break;
|
|
}
|
|
}
|
|
|
|
result = value;
|
|
return (true);
|
|
}
|
|
|
|
|
|
bool CharArrayFromFormatV(char* out, int outsize, const char* format, va_list args)
|
|
{
|
|
int writtenCount = vsnprintf(out, outsize, format, args);
|
|
|
|
if (writtenCount > 0)
|
|
{
|
|
out[writtenCount] = '\0';
|
|
return(true);
|
|
}
|
|
else
|
|
{
|
|
out[outsize - 1] = '\0';
|
|
return(false);
|
|
}
|
|
}
|
|
|
|
|
|
// Expensive!
|
|
void StringFromFormatV(std::string* out, const char* format, va_list args)
|
|
{
|
|
int writtenCount = -1;
|
|
size_t newSize = strlen(format) + 16;
|
|
char* buf = 0;
|
|
|
|
while (writtenCount < 0)
|
|
{
|
|
delete [] buf;
|
|
buf = new char[newSize + 1];
|
|
writtenCount = vsnprintf(buf, newSize, format, args);
|
|
newSize *= 2;
|
|
}
|
|
|
|
buf[writtenCount] = '\0';
|
|
*out = buf;
|
|
delete[] buf;
|
|
}
|
|
|
|
|
|
std::string StringFromFormat(const char* format, ...)
|
|
{
|
|
std::string temp;
|
|
va_list args;
|
|
va_start(args, format);
|
|
StringFromFormatV(&temp, format, args);
|
|
va_end(args);
|
|
return(temp);
|
|
}
|
|
|
|
|
|
void ToStringFromFormat(std::string* out, const char* format, ...)
|
|
{
|
|
va_list args;
|
|
va_start(args, format);
|
|
StringFromFormatV(out, format, args);
|
|
va_end(args);
|
|
}
|
|
|
|
|
|
// Turns " hej " into "hej". Also handles tabs.
|
|
std::string StripSpaces(std::string s)
|
|
{
|
|
int i;
|
|
|
|
for (i = 0; i < (int)s.size(); i++)
|
|
{
|
|
if ((s[i] != ' ') && (s[i] != 9))
|
|
{
|
|
break;
|
|
}
|
|
}
|
|
|
|
s = s.substr(i);
|
|
|
|
for (i = (int)s.size() - 1; i > 0; i--)
|
|
{
|
|
if ((s[i] != ' ') && (s[i] != 9))
|
|
{
|
|
break;
|
|
}
|
|
}
|
|
|
|
return(s.substr(0, i + 1));
|
|
}
|
|
|
|
|
|
// "\"hello\"" is turned to "hello"
|
|
// This one assumes that the string has already been space stripped in both
|
|
// ends, as done by StripSpaces above, for example.
|
|
std::string StripQuotes(const std::string& s)
|
|
{
|
|
if ((s[0] == '\"') && (s[s.size() - 1] == '\"'))
|
|
{
|
|
return(s.substr(1, s.size() - 2));
|
|
}
|
|
else
|
|
{
|
|
return(s);
|
|
}
|
|
}
|
|
|
|
|
|
bool TryParseInt(const char* str, int* outVal)
|
|
{
|
|
const char* s = str;
|
|
int value = 0;
|
|
bool negativ = false;
|
|
|
|
if (*s == '-')
|
|
{
|
|
negativ = true;
|
|
s++;
|
|
}
|
|
|
|
while (*s)
|
|
{
|
|
char c = *s++;
|
|
|
|
if ((c < '0') || (c > '9'))
|
|
{
|
|
return(false);
|
|
}
|
|
|
|
value = value * 10 + (c - '0');
|
|
}
|
|
if (negativ)
|
|
value = -value;
|
|
|
|
*outVal = value;
|
|
return(true);
|
|
}
|
|
|
|
|
|
bool TryParseBool(const char* str, bool* output)
|
|
{
|
|
if ((str[0] == '1') || !strcmp(str, "true") || !strcmp(str, "True") || !strcmp(str, "TRUE"))
|
|
{
|
|
*output = true;
|
|
return(true);
|
|
}
|
|
else if (str[0] == '0' || !strcmp(str, "false") || !strcmp(str, "False") || !strcmp(str, "FALSE"))
|
|
{
|
|
*output = false;
|
|
return(true);
|
|
}
|
|
|
|
return(false);
|
|
}
|
|
|
|
|
|
std::string StringFromInt(int value)
|
|
{
|
|
char temp[16];
|
|
sprintf(temp, "%i", value);
|
|
return(std::string(temp));
|
|
}
|
|
|
|
|
|
std::string StringFromBool(bool value)
|
|
{
|
|
return(value ? "True" : "False");
|
|
}
|
|
|
|
|
|
#ifdef _WIN32
|
|
bool SplitPath(const std::string& full_path, std::string* _pPath, std::string* _pFilename, std::string* _pExtension)
|
|
{
|
|
char drive[_MAX_DRIVE];
|
|
char dir[_MAX_DIR];
|
|
char fname[_MAX_FNAME];
|
|
char ext[_MAX_EXT];
|
|
|
|
if (_splitpath_s(full_path.c_str(), drive, _MAX_DRIVE, dir, _MAX_DIR, fname, _MAX_FNAME, ext, _MAX_EXT) == 0)
|
|
{
|
|
if (_pPath)
|
|
{
|
|
*_pPath = std::string(drive) + std::string(dir);
|
|
}
|
|
|
|
if (_pFilename != 0)
|
|
{
|
|
*_pFilename = fname;
|
|
}
|
|
|
|
if (_pExtension != 0)
|
|
{
|
|
*_pExtension = ext;
|
|
}
|
|
|
|
return(true);
|
|
}
|
|
|
|
return(false);
|
|
}
|
|
|
|
|
|
#else
|
|
bool SplitPath(const std::string& full_path, std::string* _pPath, std::string* _pFilename, std::string* _pExtension)
|
|
{
|
|
size_t last_slash = full_path.rfind('/');
|
|
|
|
if (last_slash == std::string::npos)
|
|
{
|
|
return(false);
|
|
}
|
|
|
|
size_t last_dot = full_path.rfind('.');
|
|
|
|
if ((last_dot == std::string::npos) || (last_dot < last_slash))
|
|
{
|
|
return(false);
|
|
}
|
|
|
|
if (_pPath)
|
|
{
|
|
*_pPath = full_path.substr(0, last_slash + 1);
|
|
}
|
|
|
|
if (_pFilename)
|
|
{
|
|
*_pFilename = full_path.substr(last_slash + 1, last_dot - (last_slash + 1));
|
|
}
|
|
|
|
if (_pExtension)
|
|
{
|
|
*_pExtension = full_path.substr(last_dot + 1);
|
|
_pExtension->insert(0, ".");
|
|
}
|
|
else if (_pFilename)
|
|
{
|
|
*_pFilename += full_path.substr(last_dot);
|
|
}
|
|
|
|
return(true);
|
|
}
|
|
|
|
|
|
#endif
|
|
|
|
void BuildCompleteFilename(std::string& _CompleteFilename, const std::string& _Path, const std::string& _Filename)
|
|
{
|
|
_CompleteFilename = _Path;
|
|
|
|
// check for seperator
|
|
if (_CompleteFilename[_CompleteFilename.size() - 1] != '\\')
|
|
{
|
|
_CompleteFilename += _T("\\");
|
|
}
|
|
|
|
// add the filename
|
|
_CompleteFilename += _Filename;
|
|
}
|
|
|
|
|
|
void SplitString(const std::string& str, const std::string& delim, std::vector<std::string>& output)
|
|
{
|
|
output.clear();
|
|
|
|
size_t offset = 0;
|
|
size_t delimIndex = 0;
|
|
|
|
delimIndex = str.find(delim, offset);
|
|
|
|
while (delimIndex != std::string::npos)
|
|
{
|
|
output.push_back(str.substr(offset, delimIndex - offset));
|
|
offset += delimIndex - offset + delim.length();
|
|
delimIndex = str.find(delim, offset);
|
|
}
|
|
|
|
output.push_back(str.substr(offset));
|
|
}
|
|
|
|
|
|
bool TryParseUInt(const std::string& str, u32* output)
|
|
{
|
|
if (!strcmp(str.substr(0, 2).c_str(), "0x") || !strcmp(str.substr(0, 2).c_str(), "0X"))
|
|
{
|
|
return(sscanf(str.c_str() + 2, "%x", output) > 0);
|
|
}
|
|
else
|
|
{
|
|
return(sscanf(str.c_str(), "%d", output) > 0);
|
|
}
|
|
}
|
|
|
|
|
|
int ChooseStringFrom(const char* str, const char* * items)
|
|
{
|
|
int i = 0;
|
|
|
|
while (items[i] != 0)
|
|
{
|
|
if (!strcmp(str, items[i]))
|
|
{
|
|
return(i);
|
|
}
|
|
|
|
i++;
|
|
}
|
|
|
|
return(-1);
|
|
}
|
|
|
|
|
|
// Hmm, sometimes this test doesn't get run :P
|
|
TEST(splitStringTest)
|
|
{
|
|
/*
|
|
std::string orig = "abc:def";
|
|
std::vector<std::string> split;
|
|
SplitString(orig, std::string(":"), split);
|
|
CHECK(split.size() == 2);
|
|
CHECK(!strcmp(split[0].c_str(), "abc"));
|
|
CHECK(!strcmp(split[1].c_str(), "def"));
|
|
|
|
orig = "abc";
|
|
SplitString(orig, std::string(":"), split);
|
|
CHECK(split.size() == 1);
|
|
orig = ":";
|
|
SplitString(orig, std::string(":"), split);
|
|
CHECK(split.size() == 2);*/
|
|
}
|