2023-02-20 17:59:08 -05:00
|
|
|
// utils.cpp
|
|
|
|
// Copyright (C) 2023 Richard Geldreich, Jr.
|
|
|
|
#include "utils.h"
|
|
|
|
|
|
|
|
std::string combine_strings(std::string a, const std::string& b)
|
|
|
|
{
|
|
|
|
if (!a.size())
|
|
|
|
return b;
|
|
|
|
|
|
|
|
if (!b.size())
|
|
|
|
return a;
|
|
|
|
|
|
|
|
if (a.back() == '-')
|
|
|
|
{
|
|
|
|
if ((a.size() >= 2) && isdigit((uint8_t)a[a.size() - 2]))
|
|
|
|
{
|
|
|
|
}
|
|
|
|
else
|
|
|
|
{
|
|
|
|
a.pop_back();
|
|
|
|
a += b;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
else
|
|
|
|
{
|
|
|
|
if (a.back() != ' ')
|
|
|
|
a += " ";
|
|
|
|
a += b;
|
|
|
|
}
|
|
|
|
|
|
|
|
return a;
|
|
|
|
}
|
|
|
|
|
|
|
|
std::wstring utf8_to_wchar(const std::string& str, UINT code_page)
|
|
|
|
{
|
|
|
|
if (str.empty())
|
|
|
|
return std::wstring();
|
|
|
|
|
|
|
|
int size_needed = MultiByteToWideChar(code_page, 0, &str[0], (int)str.size(), NULL, 0);
|
|
|
|
if (!size_needed)
|
|
|
|
return std::wstring();
|
|
|
|
|
|
|
|
std::wstring wstrTo(size_needed, 0);
|
|
|
|
int res = MultiByteToWideChar(code_page, 0, &str[0], (int)str.size(), &wstrTo[0], size_needed);
|
|
|
|
if (!res)
|
|
|
|
return std::wstring();
|
|
|
|
|
|
|
|
return wstrTo;
|
|
|
|
}
|
|
|
|
|
|
|
|
std::string wchar_to_utf8(const std::wstring& wstr, UINT code_page)
|
|
|
|
{
|
|
|
|
if (wstr.empty())
|
|
|
|
return std::string();
|
|
|
|
|
|
|
|
int size_needed = WideCharToMultiByte(code_page, 0, &wstr[0], (int)wstr.size(), NULL, 0, NULL, NULL);
|
|
|
|
if (!size_needed)
|
|
|
|
return std::string();
|
|
|
|
|
|
|
|
std::string strTo(size_needed, 0);
|
|
|
|
int res = WideCharToMultiByte(code_page, 0, &wstr[0], (int)wstr.size(), &strTo[0], size_needed, NULL, NULL);
|
|
|
|
if (!res)
|
|
|
|
return std::string();
|
|
|
|
|
|
|
|
return strTo;
|
|
|
|
}
|
|
|
|
|
|
|
|
static uint16_t g_codepage_437_to_unicode_0_31[32] =
|
|
|
|
{
|
|
|
|
' ', 0x263A, 0x263B, 0x2665, 0x2666, 0x2663, 0x2660, 0x2022,
|
|
|
|
0x25D8, 0x25CB, 0x25D9, 0x2642, 0x2640, 0x266A, 0x266B, 0x263C,
|
|
|
|
0x25BA, 0x25C4, 0x2195, 0x203C, 0x00B6, 0x00A7, 0x25AC, 0x21A8,
|
|
|
|
0x2191, 0x2193, 0x2192, 0x2190, 0x221F, 0x2194, 0x25B2, 0x25BC
|
|
|
|
};
|
|
|
|
|
|
|
|
static uint16_t g_codepage_437_to_unicode_128_255[129] =
|
|
|
|
{
|
|
|
|
0x2302,
|
|
|
|
0x00C7, 0x00FC, 0x00E9, 0x00E2, 0x00E4, 0x00E0, 0x00E5, 0x00E7,
|
|
|
|
0x00EA, 0x00EB, 0x00E8, 0x00EF, 0x00EE, 0x00EC, 0x00C4, 0x00C5,
|
|
|
|
0x00C9, 0x00E6, 0x00C6, 0x00F4, 0x00F6, 0x00F2, 0x00FB, 0x00F9,
|
|
|
|
0x00FF, 0x00D6, 0x00DC, 0x00A2, 0x00A3, 0x00A5, 0x20A7, 0x0192,
|
|
|
|
0x00C9, 0x00E6, 0x00C6, 0x00F4, 0x00F6, 0x00F2, 0x00FB, 0x00F9,
|
|
|
|
0x00FF, 0x00D6, 0x00DC, 0x00A2, 0x00A3, 0x00A5, 0x20A7, 0x0192,
|
|
|
|
0x2591, 0x2592, 0x2593, 0x2502, 0x2524, 0x2561, 0x2562, 0x2556,
|
|
|
|
0x2555, 0x2563, 0x2551, 0x2557, 0x255D, 0x255C, 0x255B, 0x2510,
|
|
|
|
0x2514, 0x2534, 0x252C, 0x251C, 0x2500, 0x253C, 0x255E, 0x255F,
|
|
|
|
0x255A, 0x2554, 0x2569, 0x2566, 0x2560, 0x2550, 0x256C, 0x2567,
|
|
|
|
0x2568, 0x2564, 0x2565, 0x2559, 0x2558, 0x2552, 0x2553, 0x256B,
|
|
|
|
0x256A, 0x2518, 0x250C, 0x2588, 0x2584, 0x258C, 0x2590, 0x2580,
|
|
|
|
0x03B1, 0x00DF, 0x0393, 0x03C0, 0x03A3, 0x03C3, 0x00B5, 0x03C4,
|
|
|
|
0x03A6, 0x0398, 0x03A9, 0x03B4, 0x221E, 0x03C6, 0x03B5, 0x2229,
|
|
|
|
0x2261, 0x00B1, 0x2265, 0x2264, 0x2320, 0x2321, 0x00F7, 0x2248,
|
|
|
|
0x00B0, 0x2219, 0x00B7, 0x221A, 0x207F, 0x00B2, 0x25A0, 0x00A0
|
|
|
|
};
|
|
|
|
|
|
|
|
// Code page 437 to utf8. WideCharToMultiByte etc. doesn't do the expecting thing for chars<32, and we need them.
|
|
|
|
std::string dos_to_utf8(const std::string& str)
|
|
|
|
{
|
|
|
|
std::wstring wstr;
|
|
|
|
|
|
|
|
for (uint8_t c : str)
|
|
|
|
{
|
|
|
|
if (c < 32)
|
|
|
|
wstr.push_back(g_codepage_437_to_unicode_0_31[c]);
|
|
|
|
else if (c >= 127)
|
|
|
|
wstr.push_back(g_codepage_437_to_unicode_128_255[c - 127]);
|
|
|
|
else
|
|
|
|
wstr.push_back(c);
|
|
|
|
}
|
|
|
|
|
|
|
|
return wchar_to_utf8(wstr);
|
|
|
|
}
|
|
|
|
|
|
|
|
bool vformat(std::vector<char>& buf, const char* pFmt, va_list args)
|
|
|
|
{
|
|
|
|
uint32_t buf_size = 8192;
|
|
|
|
|
|
|
|
for (; ; )
|
|
|
|
{
|
|
|
|
buf.resize(buf_size);
|
|
|
|
|
|
|
|
int res = vsnprintf(&buf[0], buf.size(), pFmt, args);
|
|
|
|
if (res == -1)
|
|
|
|
{
|
|
|
|
assert(false);
|
|
|
|
return false;
|
|
|
|
}
|
|
|
|
|
|
|
|
if (res <= buf.size() - 1)
|
|
|
|
break;
|
|
|
|
|
|
|
|
buf_size *= 2;
|
|
|
|
if (buf_size > 16 * 1024 * 1024)
|
|
|
|
{
|
|
|
|
assert(false);
|
|
|
|
return false;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
return true;
|
|
|
|
}
|
|
|
|
|
|
|
|
void ufprintf(FILE* pFile, const char* pFmt, ...)
|
|
|
|
{
|
|
|
|
std::vector<char> buf;
|
|
|
|
|
|
|
|
va_list args;
|
|
|
|
va_start(args, pFmt);
|
|
|
|
if (!vformat(buf, pFmt, args))
|
|
|
|
return;
|
|
|
|
va_end(args);
|
|
|
|
|
|
|
|
std::wstring wbuf(utf8_to_wchar(std::string(&buf[0])));
|
|
|
|
|
|
|
|
// Not thread safe, but we don't care
|
|
|
|
_setmode(_fileno(pFile), _O_U16TEXT);
|
|
|
|
fputws(&wbuf[0], pFile);
|
|
|
|
_setmode(_fileno(pFile), _O_TEXT);
|
|
|
|
}
|
|
|
|
|
|
|
|
void uprintf(const char* pFmt, ...)
|
|
|
|
{
|
|
|
|
std::vector<char> buf;
|
|
|
|
|
|
|
|
va_list args;
|
|
|
|
va_start(args, pFmt);
|
|
|
|
if (!vformat(buf, pFmt, args))
|
|
|
|
return;
|
|
|
|
va_end(args);
|
|
|
|
|
|
|
|
std::wstring wbuf(utf8_to_wchar(std::string(&buf[0])));
|
|
|
|
|
|
|
|
// Not thread safe, but we don't care
|
|
|
|
_setmode(_fileno(stdout), _O_U16TEXT);
|
|
|
|
fputws(&wbuf[0], stdout);
|
|
|
|
_setmode(_fileno(stdout), _O_TEXT);
|
|
|
|
}
|
|
|
|
|
|
|
|
std::string string_format(const char* pMsg, ...)
|
|
|
|
{
|
|
|
|
std::vector<char> buf;
|
|
|
|
|
|
|
|
va_list args;
|
|
|
|
va_start(args, pMsg);
|
|
|
|
if (!vformat(buf, pMsg, args))
|
|
|
|
return "";
|
|
|
|
va_end(args);
|
|
|
|
|
|
|
|
std::string res;
|
|
|
|
if (buf.size())
|
|
|
|
res.assign(&buf[0]);
|
|
|
|
|
|
|
|
return res;
|
|
|
|
}
|
|
|
|
|
|
|
|
void panic(const char* pMsg, ...)
|
|
|
|
{
|
|
|
|
char buf[4096];
|
|
|
|
|
|
|
|
va_list args;
|
|
|
|
va_start(args, pMsg);
|
|
|
|
vsnprintf(buf, sizeof(buf), pMsg, args);
|
|
|
|
va_end(args);
|
|
|
|
|
|
|
|
ufprintf(stderr, "%s", buf);
|
|
|
|
|
|
|
|
exit(EXIT_FAILURE);
|
|
|
|
}
|
|
|
|
|
|
|
|
FILE* ufopen(const char* pFilename, const char* pMode)
|
|
|
|
{
|
|
|
|
std::wstring wfilename(utf8_to_wchar(pFilename));
|
|
|
|
std::wstring wmode(utf8_to_wchar(pMode));
|
|
|
|
|
|
|
|
if (!wfilename.size() || !wmode.size())
|
|
|
|
return nullptr;
|
|
|
|
|
|
|
|
FILE* pRes = nullptr;
|
|
|
|
_wfopen_s(&pRes, &wfilename[0], &wmode[0]);
|
|
|
|
return pRes;
|
|
|
|
}
|
|
|
|
|
|
|
|
std::string& string_trim(std::string& str)
|
|
|
|
{
|
|
|
|
while (str.size() && isspace((uint8_t)str.back()))
|
|
|
|
str.pop_back();
|
|
|
|
|
|
|
|
while (str.size() && isspace((uint8_t)str[0]))
|
|
|
|
str.erase(0, 1);
|
|
|
|
|
|
|
|
return str;
|
|
|
|
}
|
|
|
|
|
|
|
|
std::string& string_trim_end(std::string& str)
|
|
|
|
{
|
|
|
|
while (str.size() && isspace((uint8_t)str.back()))
|
|
|
|
str.pop_back();
|
|
|
|
|
|
|
|
return str;
|
|
|
|
}
|
|
|
|
|
|
|
|
// Case sensitive, returns -1 if can't find
|
|
|
|
int string_find_first(const std::string& str, const char* pPhrase)
|
|
|
|
{
|
|
|
|
size_t res = str.find(pPhrase, 0);
|
|
|
|
if (res == std::string::npos)
|
|
|
|
return -1;
|
|
|
|
return (int)res;
|
|
|
|
}
|
|
|
|
|
|
|
|
int string_icompare(const std::string& a, const char* pB)
|
|
|
|
{
|
|
|
|
const size_t a_len = a.size();
|
|
|
|
const size_t b_len = strlen(pB);
|
|
|
|
|
|
|
|
const size_t min_len = std::min(a_len, b_len);
|
|
|
|
|
|
|
|
for (size_t i = 0; i < min_len; i++)
|
|
|
|
{
|
|
|
|
const int ac = (uint8_t)utolower(a[i]);
|
|
|
|
const int bc = (uint8_t)utolower(pB[i]);
|
|
|
|
|
|
|
|
if (ac != bc)
|
|
|
|
return (ac < bc) ? -1 : 1;
|
|
|
|
}
|
|
|
|
|
|
|
|
if (a_len == b_len)
|
|
|
|
return 0;
|
|
|
|
|
|
|
|
return (a_len < b_len) ? -1 : 1;
|
|
|
|
}
|
|
|
|
|
|
|
|
bool string_begins_with(const std::string& str, const char* pPhrase)
|
|
|
|
{
|
|
|
|
const size_t str_len = str.size();
|
|
|
|
|
|
|
|
const size_t phrase_len = strlen(pPhrase);
|
|
|
|
assert(phrase_len);
|
|
|
|
|
|
|
|
if (str_len >= phrase_len)
|
|
|
|
{
|
|
|
|
if (_strnicmp(pPhrase, str.c_str(), phrase_len) == 0)
|
|
|
|
return true;
|
|
|
|
}
|
|
|
|
|
|
|
|
return false;
|
|
|
|
}
|
|
|
|
|
|
|
|
bool string_ends_in(const std::string& str, const char* pPhrase)
|
|
|
|
{
|
|
|
|
const size_t str_len = str.size();
|
|
|
|
|
|
|
|
const size_t phrase_len = strlen(pPhrase);
|
|
|
|
assert(phrase_len);
|
|
|
|
|
|
|
|
if (str_len >= phrase_len)
|
|
|
|
{
|
|
|
|
if (_stricmp(pPhrase, str.c_str() + str_len - phrase_len) == 0)
|
|
|
|
return true;
|
|
|
|
}
|
|
|
|
|
|
|
|
return false;
|
|
|
|
}
|
|
|
|
|
|
|
|
std::string encode_url(const std::string& url)
|
|
|
|
{
|
|
|
|
//const char* pValid_chars = ";,/?:@&=+$-_.!~*'()#";
|
|
|
|
//const size_t valid_chars_len = strlen(pValid_chars);
|
|
|
|
|
|
|
|
std::string res;
|
|
|
|
for (uint32_t i = 0; i < url.size(); i++)
|
|
|
|
{
|
|
|
|
uint8_t c = (uint8_t)url[i];
|
|
|
|
|
|
|
|
//const bool is_digit = (c >= 0) && (c <= '9');
|
|
|
|
//const bool is_upper = (c >= 'A') && (c <= 'Z');
|
|
|
|
//const bool is_lower = (c >= 'a') && (c <= 'z');
|
|
|
|
|
|
|
|
// Escape some problematic charactes that confuse some Markdown parsers (even after using Markdown '\' escapes)
|
|
|
|
if ((c == ')') || (c == '(') || (c == '_') || (c == '*'))
|
|
|
|
{
|
|
|
|
res.push_back('%');
|
|
|
|
res.push_back(to_hex(c / 16));
|
|
|
|
res.push_back(to_hex(c % 16));
|
|
|
|
continue;
|
|
|
|
}
|
|
|
|
|
|
|
|
res.push_back(c);
|
|
|
|
}
|
|
|
|
|
|
|
|
return res;
|
|
|
|
}
|
|
|
|
|
|
|
|
uint32_t crc32(const uint8_t* pBuf, size_t size, uint32_t init_crc)
|
|
|
|
{
|
|
|
|
uint32_t crc = ~init_crc;
|
|
|
|
|
|
|
|
for (size_t i = 0; i < size; i++)
|
|
|
|
{
|
|
|
|
const uint32_t byte = pBuf[i];
|
|
|
|
|
|
|
|
crc = crc ^ byte;
|
|
|
|
|
|
|
|
for (int j = 7; j >= 0; j--)
|
|
|
|
{
|
|
|
|
uint32_t mask = -((int)(crc & 1));
|
|
|
|
crc = (crc >> 1) ^ (0xEDB88320 & mask);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
return ~crc;
|
|
|
|
}
|
|
|
|
|
|
|
|
bool read_binary_file(const char* pFilename, uint8_vec& buf)
|
|
|
|
{
|
|
|
|
const uint64_t MAX_BINARY_FILE_LEN = 168ULL * 1024ULL * (1024ULL * 1024ULL);
|
|
|
|
|
|
|
|
FILE* pFile = ufopen(pFilename, "rb");
|
|
|
|
if (!pFile)
|
|
|
|
return false;
|
|
|
|
|
|
|
|
_fseeki64(pFile, 0, SEEK_END);
|
|
|
|
int64_t len = _ftelli64(pFile);
|
|
|
|
if (len < 0)
|
|
|
|
{
|
|
|
|
fclose(pFile);
|
|
|
|
return false;
|
|
|
|
}
|
|
|
|
_fseeki64(pFile, 0, SEEK_SET);
|
|
|
|
|
|
|
|
if (len > MAX_BINARY_FILE_LEN)
|
|
|
|
return false;
|
|
|
|
buf.resize(len);
|
|
|
|
|
|
|
|
if (fread(&buf[0], len, 1, pFile) != 1)
|
|
|
|
{
|
|
|
|
fclose(pFile);
|
|
|
|
return false;
|
|
|
|
}
|
|
|
|
|
|
|
|
fclose(pFile);
|
|
|
|
return true;
|
|
|
|
}
|
|
|
|
|
|
|
|
bool read_text_file(const char* pFilename, string_vec& lines, bool trim_lines, bool* pUTF8_flag)
|
|
|
|
{
|
|
|
|
FILE* pFile = ufopen(pFilename, "r");
|
|
|
|
if (!pFile)
|
|
|
|
return false;
|
|
|
|
|
|
|
|
bool first_line = true;
|
|
|
|
|
|
|
|
if (pUTF8_flag)
|
|
|
|
*pUTF8_flag = false;
|
|
|
|
|
|
|
|
while (!feof(pFile))
|
|
|
|
{
|
|
|
|
char buf[16384];
|
|
|
|
|
|
|
|
char* p = fgets(buf, sizeof(buf), pFile);
|
|
|
|
if (!p)
|
|
|
|
{
|
|
|
|
if (feof(pFile))
|
|
|
|
break;
|
|
|
|
|
|
|
|
fclose(pFile);
|
|
|
|
return false;
|
|
|
|
}
|
|
|
|
|
|
|
|
std::string str(p);
|
|
|
|
|
|
|
|
if (first_line)
|
|
|
|
{
|
|
|
|
first_line = false;
|
|
|
|
if ((str.size() >= 3) &&
|
|
|
|
((uint8_t)str[0] == UTF8_BOM0) &&
|
|
|
|
((uint8_t)str[1] == UTF8_BOM1) &&
|
|
|
|
((uint8_t)str[2] == UTF8_BOM2))
|
|
|
|
{
|
|
|
|
if (pUTF8_flag)
|
|
|
|
*pUTF8_flag = true;
|
|
|
|
str.erase(0, 3);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
while (str.size() && ((str.back() == '\n') || (str.back() == '\r')))
|
|
|
|
str.pop_back();
|
|
|
|
|
|
|
|
if (trim_lines)
|
|
|
|
string_trim_end(str);
|
|
|
|
|
|
|
|
lines.push_back(str);
|
|
|
|
}
|
|
|
|
|
|
|
|
fclose(pFile);
|
|
|
|
return true;
|
|
|
|
}
|
|
|
|
|
2023-02-24 14:23:55 -05:00
|
|
|
bool read_text_file(const char* pFilename, std::vector<uint8_t>& buf, bool *pUTF8_flag)
|
2023-02-20 17:59:08 -05:00
|
|
|
{
|
2023-02-24 14:23:55 -05:00
|
|
|
if (pUTF8_flag)
|
|
|
|
*pUTF8_flag = false;
|
2023-02-20 17:59:08 -05:00
|
|
|
|
|
|
|
FILE* pFile = ufopen(pFilename, "rb");
|
|
|
|
if (!pFile)
|
|
|
|
{
|
|
|
|
ufprintf(stderr, "Failed reading file %s!\n", pFilename);
|
|
|
|
return false;
|
|
|
|
}
|
|
|
|
|
|
|
|
fseek(pFile, 0, SEEK_END);
|
|
|
|
uint64_t filesize = _ftelli64(pFile);
|
|
|
|
fseek(pFile, 0, SEEK_SET);
|
|
|
|
|
|
|
|
buf.resize(filesize + 1);
|
|
|
|
fread(&buf[0], 1, filesize, pFile);
|
|
|
|
|
|
|
|
fclose(pFile);
|
|
|
|
|
|
|
|
if ((buf.size() >= 3) &&
|
|
|
|
((uint8_t)buf[0] == UTF8_BOM0) &&
|
|
|
|
((uint8_t)buf[1] == UTF8_BOM1) &&
|
|
|
|
((uint8_t)buf[2] == UTF8_BOM2))
|
|
|
|
{
|
2023-02-24 14:23:55 -05:00
|
|
|
if (pUTF8_flag)
|
|
|
|
*pUTF8_flag = true;
|
2023-02-20 17:59:08 -05:00
|
|
|
|
|
|
|
buf.erase(buf.begin(), buf.begin() + 3);
|
|
|
|
}
|
|
|
|
|
|
|
|
return true;
|
|
|
|
}
|
|
|
|
|
|
|
|
bool write_text_file(const char* pFilename, string_vec& lines, bool utf8_bom)
|
|
|
|
{
|
|
|
|
FILE* pFile = ufopen(pFilename, "wb");
|
|
|
|
if (!pFile)
|
|
|
|
return false;
|
|
|
|
|
|
|
|
if (utf8_bom)
|
|
|
|
{
|
|
|
|
if ((fputc(UTF8_BOM0, pFile) == EOF) || (fputc(UTF8_BOM1, pFile) == EOF) || (fputc(UTF8_BOM2, pFile) == EOF))
|
|
|
|
{
|
|
|
|
fclose(pFile);
|
|
|
|
return false;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
for (uint32_t i = 0; i < lines.size(); i++)
|
|
|
|
{
|
|
|
|
if ((fwrite(lines[i].c_str(), lines[i].size(), 1, pFile) != 1) || (fwrite("\r\n", 2, 1, pFile) != 1))
|
|
|
|
{
|
|
|
|
fclose(pFile);
|
|
|
|
return false;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
if (fclose(pFile) == EOF)
|
|
|
|
return false;
|
|
|
|
|
|
|
|
return true;
|
|
|
|
}
|
|
|
|
|
|
|
|
bool serialize_to_json_file(const char* pFilename, const json& j, bool utf8_bom)
|
|
|
|
{
|
|
|
|
FILE* pFile = ufopen(pFilename, "wb");
|
|
|
|
if (!pFile)
|
|
|
|
return false;
|
|
|
|
|
|
|
|
if (utf8_bom)
|
|
|
|
{
|
|
|
|
if ((fputc(UTF8_BOM0, pFile) == EOF) || (fputc(UTF8_BOM1, pFile) == EOF) || (fputc(UTF8_BOM2, pFile) == EOF))
|
|
|
|
{
|
|
|
|
fclose(pFile);
|
|
|
|
return false;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
std::string d(j.dump(2));
|
|
|
|
|
|
|
|
if (d.size())
|
|
|
|
{
|
|
|
|
if (fwrite(&d[0], d.size(), 1, pFile) != 1)
|
|
|
|
{
|
|
|
|
fclose(pFile);
|
|
|
|
return false;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
fclose(pFile);
|
|
|
|
|
|
|
|
return true;
|
|
|
|
}
|
|
|
|
|
|
|
|
// Note: This doesn't actually handle utf8. It assumes ANSI (code page 252) text input.
|
|
|
|
static std::string extract_column_text(const std::string& str, uint32_t ofs, uint32_t len)
|
|
|
|
{
|
|
|
|
if (ofs >= str.size())
|
|
|
|
return "";
|
|
|
|
|
|
|
|
const uint32_t max_len = std::min((uint32_t)str.size() - ofs, len);
|
|
|
|
|
|
|
|
std::string res(str);
|
|
|
|
if (ofs)
|
|
|
|
res.erase(0, ofs);
|
|
|
|
|
|
|
|
if (max_len < res.size())
|
|
|
|
res.erase(max_len, res.size());
|
|
|
|
|
|
|
|
string_trim(res);
|
|
|
|
return res;
|
|
|
|
}
|
|
|
|
|
|
|
|
// Note: This doesn't actually handle utf8. It assumes ANSI (code page 252) text input.
|
2023-02-24 14:23:55 -05:00
|
|
|
bool load_column_text(const char* pFilename, std::vector<string_vec>& rows, std::string& title, string_vec& col_titles, bool empty_line_seps, const char* pExtra_col_text)
|
2023-02-20 17:59:08 -05:00
|
|
|
{
|
|
|
|
string_vec lines;
|
|
|
|
bool utf8_flag = false;
|
2023-02-24 14:23:55 -05:00
|
|
|
if (!read_text_file(pFilename, lines, true, &utf8_flag))
|
2023-02-20 17:59:08 -05:00
|
|
|
panic("Failed reading text file %s", pFilename);
|
|
|
|
|
|
|
|
if (utf8_flag)
|
|
|
|
panic("load_column_text() doesn't support utf8 yet");
|
|
|
|
|
|
|
|
if (!lines.size() || !lines[0].size())
|
|
|
|
panic("Expected title");
|
|
|
|
|
2023-02-24 14:23:55 -05:00
|
|
|
if (lines.size() < 5)
|
2023-02-20 17:59:08 -05:00
|
|
|
panic("File too small");
|
|
|
|
|
|
|
|
for (uint32_t i = 0; i < lines.size(); i++)
|
|
|
|
{
|
|
|
|
if (lines[i].find_first_of(9) != std::string::npos)
|
|
|
|
panic("Tab in file");
|
|
|
|
|
|
|
|
string_trim(lines[i]);
|
|
|
|
}
|
|
|
|
|
|
|
|
title = lines[0];
|
|
|
|
|
|
|
|
if (lines[1].size())
|
|
|
|
panic("Expected empty line");
|
|
|
|
|
|
|
|
std::string col_line = lines[2];
|
|
|
|
|
|
|
|
std::string col_seps = lines[3];
|
|
|
|
if ((!col_seps.size()) || (col_seps[0] != '-') || (col_seps.back() != '-'))
|
|
|
|
panic("Invalid column seperator line");
|
|
|
|
|
|
|
|
for (uint32_t i = 0; i < col_seps.size(); i++)
|
|
|
|
{
|
|
|
|
const uint8_t c = col_seps[i];
|
|
|
|
if ((c != ' ') && (c != '-'))
|
|
|
|
panic("Invalid column separator line");
|
|
|
|
}
|
|
|
|
|
|
|
|
int col_sep_start = 0;
|
|
|
|
std::vector< std::pair<uint32_t, uint32_t> > column_info; // start offset and len of each column in chars
|
|
|
|
|
|
|
|
for (uint32_t i = 1; i < col_seps.size(); i++)
|
|
|
|
{
|
|
|
|
const uint8_t c = col_seps[i];
|
|
|
|
if (c == ' ')
|
|
|
|
{
|
|
|
|
if (col_sep_start != -1)
|
|
|
|
{
|
|
|
|
column_info.push_back(std::make_pair(col_sep_start, i - col_sep_start));
|
|
|
|
col_sep_start = -1;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
else
|
|
|
|
{
|
|
|
|
if (col_sep_start == -1)
|
|
|
|
col_sep_start = i;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
if (col_sep_start != -1)
|
|
|
|
{
|
|
|
|
column_info.push_back(std::make_pair(col_sep_start, (uint32_t)col_seps.size() - col_sep_start));
|
|
|
|
col_sep_start = -1;
|
|
|
|
}
|
|
|
|
|
|
|
|
if (!column_info.size())
|
|
|
|
panic("No columns found");
|
|
|
|
|
|
|
|
col_titles.resize(column_info.size());
|
|
|
|
for (uint32_t i = 0; i < column_info.size(); i++)
|
|
|
|
{
|
|
|
|
col_titles[i] = col_line;
|
2023-02-24 14:23:55 -05:00
|
|
|
|
2023-02-20 17:59:08 -05:00
|
|
|
if (column_info[i].first)
|
|
|
|
col_titles[i].erase(0, column_info[i].first);
|
|
|
|
|
2023-02-24 14:23:55 -05:00
|
|
|
if (column_info[i].second > col_titles[i].size())
|
|
|
|
panic("invalid columns");
|
|
|
|
|
2023-02-20 17:59:08 -05:00
|
|
|
col_titles[i].erase(column_info[i].second, col_titles[i].size() - column_info[i].second);
|
|
|
|
string_trim(col_titles[i]);
|
|
|
|
}
|
|
|
|
|
|
|
|
for (uint32_t i = 0; i < column_info.size() - 1; i++)
|
|
|
|
column_info[i].second = column_info[i + 1].first - column_info[i].first;
|
|
|
|
column_info.back().second = 32000;
|
|
|
|
|
|
|
|
uint32_t cur_line = 4;
|
|
|
|
|
|
|
|
uint32_t cur_record_index = 0;
|
|
|
|
|
|
|
|
while (cur_line < lines.size())
|
|
|
|
{
|
|
|
|
string_vec rec_lines;
|
|
|
|
rec_lines.push_back(lines[cur_line++]);
|
|
|
|
|
2023-02-24 14:23:55 -05:00
|
|
|
if (empty_line_seps)
|
2023-02-20 17:59:08 -05:00
|
|
|
{
|
2023-02-24 14:23:55 -05:00
|
|
|
while (cur_line < lines.size())
|
|
|
|
{
|
|
|
|
if (!lines[cur_line].size())
|
|
|
|
break;
|
2023-02-20 17:59:08 -05:00
|
|
|
|
2023-02-24 14:23:55 -05:00
|
|
|
rec_lines.push_back(lines[cur_line++]);
|
|
|
|
}
|
2023-02-20 17:59:08 -05:00
|
|
|
|
2023-02-24 14:23:55 -05:00
|
|
|
// cur_line should be blank, or we're at the end of the file
|
2023-02-20 17:59:08 -05:00
|
|
|
if (cur_line < lines.size())
|
|
|
|
{
|
2023-02-24 14:23:55 -05:00
|
|
|
cur_line++;
|
|
|
|
if (cur_line < lines.size())
|
|
|
|
{
|
|
|
|
if (!lines[cur_line].size())
|
|
|
|
panic("Expected non-empty line");
|
|
|
|
}
|
2023-02-20 17:59:08 -05:00
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2023-02-24 14:23:55 -05:00
|
|
|
//uprintf("%u:\n", cur_record_index);
|
2023-02-20 17:59:08 -05:00
|
|
|
//for (uint32_t i = 0; i < rec_lines.size(); i++)
|
|
|
|
// uprintf("%s\n", rec_lines[i].c_str());
|
|
|
|
|
|
|
|
string_vec col_lines(column_info.size());
|
|
|
|
|
|
|
|
for (uint32_t col_index = 0; col_index < column_info.size(); col_index++)
|
|
|
|
{
|
|
|
|
for (uint32_t l = 0; l < rec_lines.size(); l++)
|
|
|
|
{
|
|
|
|
std::string col_text(extract_column_text(rec_lines[l], column_info[col_index].first, column_info[col_index].second));
|
|
|
|
|
|
|
|
if (col_text.size())
|
|
|
|
{
|
|
|
|
if (col_lines[col_index].size())
|
|
|
|
{
|
|
|
|
if ((col_lines[col_index].back() != '-') && ((uint8_t)col_lines[col_index].back() != ANSI_SOFT_HYPHEN))
|
|
|
|
col_lines[col_index].push_back(' ');
|
|
|
|
else
|
|
|
|
{
|
|
|
|
if ((col_lines[col_index].size() >= 2) && (!isdigit((uint8_t)col_lines[col_index][col_lines[col_index].size() - 2])))
|
|
|
|
col_lines[col_index].pop_back();
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
col_lines[col_index] += col_text;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2023-02-24 14:23:55 -05:00
|
|
|
if (pExtra_col_text)
|
|
|
|
col_lines.push_back(pExtra_col_text);
|
|
|
|
|
2023-02-20 17:59:08 -05:00
|
|
|
// Convert from ANSI (code page 252) to UTF8.
|
|
|
|
for (auto& l : col_lines)
|
|
|
|
l = ansi_to_utf8(l);
|
|
|
|
|
2023-02-24 14:23:55 -05:00
|
|
|
rows.push_back(col_lines);
|
|
|
|
|
2023-02-20 17:59:08 -05:00
|
|
|
cur_record_index++;
|
|
|
|
}
|
|
|
|
|
|
|
|
return true;
|
|
|
|
}
|
|
|
|
|
|
|
|
bool invoke_curl(const std::string& args, string_vec& reply)
|
|
|
|
{
|
|
|
|
reply.clear();
|
|
|
|
|
|
|
|
remove("__temp.html");
|
|
|
|
|
|
|
|
// Invoke curl.exe
|
|
|
|
std::string cmd(string_format("curl.exe \"%s\" -o __temp.html", args.c_str()));
|
|
|
|
uprintf("Command: %s\n", cmd.c_str());
|
|
|
|
|
|
|
|
int status = system(cmd.c_str());
|
|
|
|
uprintf("curl returned status %i\n", status);
|
|
|
|
|
|
|
|
if (status != EXIT_SUCCESS)
|
|
|
|
return false;
|
|
|
|
|
|
|
|
// Read output file.
|
|
|
|
|
|
|
|
FILE* pFile = ufopen("__temp.html", "rb");
|
|
|
|
if (!pFile)
|
|
|
|
{
|
|
|
|
Sleep(50);
|
|
|
|
pFile = ufopen("__temp.html", "rb");
|
|
|
|
if (!pFile)
|
|
|
|
return false;
|
|
|
|
}
|
|
|
|
|
|
|
|
uint8_t buf[6] = { 0,0,0,0,0,0 };
|
|
|
|
fread(buf, 5, 1, pFile);
|
|
|
|
fclose(pFile);
|
|
|
|
|
|
|
|
// Try to detect some common binary file types
|
|
|
|
|
|
|
|
// PDF
|
|
|
|
if (memcmp(buf, "%PDF-", 5) == 0)
|
|
|
|
{
|
|
|
|
uprintf("PDF file detected\n");
|
|
|
|
|
|
|
|
std::string filename(args);
|
|
|
|
for (size_t i = filename.size() - 1; i >= 0; i--)
|
|
|
|
{
|
|
|
|
if (filename[i] == '/')
|
|
|
|
{
|
|
|
|
filename.erase(0, i + 1);
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
std::string new_link_deescaped;
|
|
|
|
for (uint32_t i = 0; i < filename.size(); i++)
|
|
|
|
{
|
|
|
|
uint8_t c = filename[i];
|
|
|
|
if ((c == '%') && ((i + 2) < filename.size()))
|
|
|
|
{
|
|
|
|
int da = convert_hex_digit(filename[i + 1]);
|
|
|
|
int db = convert_hex_digit(filename[i + 2]);
|
|
|
|
if (da >= 0 && db >= 0)
|
|
|
|
{
|
|
|
|
int val = da * 16 + db;
|
|
|
|
new_link_deescaped.push_back((uint8_t)val);
|
|
|
|
}
|
|
|
|
|
|
|
|
i += 2;
|
|
|
|
}
|
|
|
|
else
|
|
|
|
new_link_deescaped.push_back(c);
|
|
|
|
}
|
|
|
|
|
|
|
|
rename("__temp.html", new_link_deescaped.c_str());
|
|
|
|
uprintf("Renamed __temp.html to %s\n", new_link_deescaped.c_str());
|
|
|
|
|
|
|
|
return true;
|
|
|
|
}
|
|
|
|
|
|
|
|
// JPEG
|
|
|
|
if (memcmp(buf, "\xFF\xD8\xFF\xE0", 4) == 0)
|
|
|
|
{
|
|
|
|
uprintf("JPEG file detected\n");
|
|
|
|
return true;
|
|
|
|
}
|
|
|
|
|
2023-02-24 14:23:55 -05:00
|
|
|
if (!read_text_file("__temp.html", reply, true, nullptr))
|
2023-02-20 17:59:08 -05:00
|
|
|
{
|
|
|
|
// Wait a bit and try again, rarely needed under Windows.
|
|
|
|
Sleep(50);
|
2023-02-24 14:23:55 -05:00
|
|
|
if (!read_text_file("__temp.html", reply, true, nullptr))
|
2023-02-20 17:59:08 -05:00
|
|
|
return false;
|
|
|
|
}
|
|
|
|
|
|
|
|
return true;
|
|
|
|
}
|
|
|
|
|
|
|
|
void convert_args_to_utf8(string_vec& args, int argc, wchar_t* argv[])
|
|
|
|
{
|
|
|
|
args.resize(argc);
|
|
|
|
|
|
|
|
for (int i = 0; i < argc; i++)
|
|
|
|
args[i] = wchar_to_utf8(argv[i]);
|
|
|
|
}
|
2023-02-24 14:23:55 -05:00
|
|
|
|
|
|
|
std::string string_slice(const std::string& str, size_t ofs, size_t len)
|
|
|
|
{
|
|
|
|
if (!len)
|
|
|
|
return "";
|
|
|
|
|
|
|
|
if (ofs > str.size())
|
|
|
|
{
|
|
|
|
assert(0);
|
|
|
|
return "";
|
|
|
|
}
|
|
|
|
|
|
|
|
const size_t max_len = str.size() - ofs;
|
|
|
|
|
|
|
|
len = std::min(len, max_len);
|
|
|
|
|
|
|
|
std::string res(str);
|
|
|
|
if (ofs)
|
|
|
|
res.erase(0, ofs);
|
|
|
|
|
|
|
|
if (len)
|
|
|
|
res.resize(len);
|
|
|
|
|
|
|
|
return res;
|
|
|
|
}
|
|
|
|
|
|
|
|
bool invoke_openai(const std::string& prompt, std::string& reply)
|
|
|
|
{
|
|
|
|
reply.clear();
|
|
|
|
|
|
|
|
// Write prompt to i.txt
|
|
|
|
FILE* pFile = ufopen("i.txt", "wb");
|
|
|
|
fwrite(prompt.c_str(), prompt.size(), 1, pFile);
|
|
|
|
fclose(pFile);
|
|
|
|
|
|
|
|
// Invoke openai.exe
|
|
|
|
int status = system("openai.exe i.txt o.txt");
|
|
|
|
if (status != EXIT_SUCCESS)
|
|
|
|
return false;
|
|
|
|
|
|
|
|
// Read output file.
|
|
|
|
string_vec lines;
|
|
|
|
if (!read_text_file("o.txt", lines, true, nullptr))
|
|
|
|
{
|
|
|
|
// Wait a bit and try again, rarely needed under Windows.
|
|
|
|
Sleep(50);
|
|
|
|
if (!read_text_file("o.txt", lines, true, nullptr))
|
|
|
|
return false;
|
|
|
|
}
|
|
|
|
|
|
|
|
// Skip any blank lines at the beginning of the reply.
|
|
|
|
uint32_t i;
|
|
|
|
for (i = 0; i < lines.size(); i++)
|
|
|
|
{
|
|
|
|
std::string s(lines[i]);
|
|
|
|
string_trim(s);
|
|
|
|
if (s.size())
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
|
|
|
|
for (; i < lines.size(); i++)
|
|
|
|
reply += lines[i];
|
|
|
|
|
|
|
|
return true;
|
|
|
|
}
|
|
|
|
|
|
|
|
std::string get_deg_to_dms(double deg)
|
|
|
|
{
|
|
|
|
deg = std::round(fabs(deg) * 3600.0f);
|
|
|
|
|
|
|
|
int min_secs = (int)fmod(deg, 3600.0f);
|
|
|
|
|
|
|
|
deg = std::floor((deg - (double)min_secs) / 3600.0f);
|
|
|
|
|
|
|
|
int minutes = min_secs / 60;
|
|
|
|
int secs = min_secs % 60;
|
|
|
|
|
|
|
|
return string_format("%02i%:%02i:%02i", (int)deg, minutes, secs);
|
|
|
|
}
|