mirror of
https://github.com/DarkflameUniverse/DarkflameServer.git
synced 2024-11-08 17:28:20 +00:00
ef6f2f133e
* cast and code cleanup * cast cleanup * bug fixes and improvements * no getBoolField method exists * fixes * unbroke sg cannon scoring * removing comments * Remove the c-style cast warning I added from CMakeLists now that they're gone (it triggers on 3rd party dependencies and slows down compilation) * (Hopefully) fix MacOS compilation error * partially-implemented feedback * more updates to account for feedback * change bool default --------- Co-authored-by: jadebenn <jonahebenn@yahoo.com>
326 lines
9.4 KiB
C++
326 lines
9.4 KiB
C++
#include "GeneralUtils.h"
|
|
|
|
// C++
|
|
#include <cstdint>
|
|
#include <cassert>
|
|
#include <algorithm>
|
|
#include <filesystem>
|
|
#include <map>
|
|
|
|
template <typename T>
|
|
inline size_t MinSize(size_t size, const std::basic_string_view<T>& string) {
|
|
if (size == size_t(-1) || size > string.size()) {
|
|
return string.size();
|
|
} else {
|
|
return size;
|
|
}
|
|
}
|
|
|
|
inline bool IsLeadSurrogate(char16_t c) {
|
|
return (0xD800 <= c) && (c <= 0xDBFF);
|
|
}
|
|
|
|
inline bool IsTrailSurrogate(char16_t c) {
|
|
return (0xDC00 <= c) && (c <= 0xDFFF);
|
|
}
|
|
|
|
inline void PushUTF8CodePoint(std::string& ret, char32_t cp) {
|
|
if (cp <= 0x007F) {
|
|
ret.push_back(static_cast<uint8_t>(cp));
|
|
} else if (cp <= 0x07FF) {
|
|
ret.push_back(0xC0 | (cp >> 6));
|
|
ret.push_back(0x80 | (cp & 0x3F));
|
|
} else if (cp <= 0xFFFF) {
|
|
ret.push_back(0xE0 | (cp >> 12));
|
|
ret.push_back(0x80 | ((cp >> 6) & 0x3F));
|
|
ret.push_back(0x80 | (cp & 0x3F));
|
|
} else if (cp <= 0x10FFFF) {
|
|
ret.push_back(0xF0 | (cp >> 18));
|
|
ret.push_back(0x80 | ((cp >> 12) & 0x3F));
|
|
ret.push_back(0x80 | ((cp >> 6) & 0x3F));
|
|
ret.push_back(0x80 | (cp & 0x3F));
|
|
} else {
|
|
assert(false);
|
|
}
|
|
}
|
|
|
|
constexpr const char16_t REPLACEMENT_CHARACTER = 0xFFFD;
|
|
|
|
bool _IsSuffixChar(uint8_t c) {
|
|
return (c & 0xC0) == 0x80;
|
|
}
|
|
|
|
bool GeneralUtils::_NextUTF8Char(std::string_view& slice, uint32_t& out) {
|
|
size_t rem = slice.length();
|
|
if (slice.empty()) return false;
|
|
const uint8_t* bytes = reinterpret_cast<const uint8_t*>(&slice.front());
|
|
if (rem > 0) {
|
|
uint8_t first = bytes[0];
|
|
if (first < 0x80) { // 1 byte character
|
|
out = static_cast<uint32_t>(first & 0x7F);
|
|
slice.remove_prefix(1);
|
|
return true;
|
|
} else if (first < 0xC0) {
|
|
// middle byte, not valid at start, fall through
|
|
} else if (first < 0xE0) { // two byte character
|
|
if (rem > 1) {
|
|
uint8_t second = bytes[1];
|
|
if (_IsSuffixChar(second)) {
|
|
out = (static_cast<uint32_t>(first & 0x1F) << 6)
|
|
+ static_cast<uint32_t>(second & 0x3F);
|
|
slice.remove_prefix(2);
|
|
return true;
|
|
}
|
|
}
|
|
} else if (first < 0xF0) { // three byte character
|
|
if (rem > 2) {
|
|
uint8_t second = bytes[1];
|
|
uint8_t third = bytes[2];
|
|
if (_IsSuffixChar(second) && _IsSuffixChar(third)) {
|
|
out = (static_cast<uint32_t>(first & 0x0F) << 12)
|
|
+ (static_cast<uint32_t>(second & 0x3F) << 6)
|
|
+ static_cast<uint32_t>(third & 0x3F);
|
|
slice.remove_prefix(3);
|
|
return true;
|
|
}
|
|
}
|
|
} else if (first < 0xF8) { // four byte character
|
|
if (rem > 3) {
|
|
uint8_t second = bytes[1];
|
|
uint8_t third = bytes[2];
|
|
uint8_t fourth = bytes[3];
|
|
if (_IsSuffixChar(second) && _IsSuffixChar(third) && _IsSuffixChar(fourth)) {
|
|
out = (static_cast<uint32_t>(first & 0x07) << 18)
|
|
+ (static_cast<uint32_t>(second & 0x3F) << 12)
|
|
+ (static_cast<uint32_t>(third & 0x3F) << 6)
|
|
+ static_cast<uint32_t>(fourth & 0x3F);
|
|
slice.remove_prefix(4);
|
|
return true;
|
|
}
|
|
}
|
|
}
|
|
out = static_cast<uint32_t>(REPLACEMENT_CHARACTER);
|
|
slice.remove_prefix(1);
|
|
return true;
|
|
}
|
|
return false;
|
|
}
|
|
|
|
/// See <https://www.ietf.org/rfc/rfc2781.html#section-2.1>
|
|
bool PushUTF16CodePoint(std::u16string& output, uint32_t U, size_t size) {
|
|
if (output.length() >= size) return false;
|
|
if (U < 0x10000) {
|
|
// If U < 0x10000, encode U as a 16-bit unsigned integer and terminate.
|
|
output.push_back(static_cast<uint16_t>(U));
|
|
return true;
|
|
} else if (U > 0x10FFFF) {
|
|
output.push_back(REPLACEMENT_CHARACTER);
|
|
return true;
|
|
} else if (output.length() + 1 < size) {
|
|
// Let U' = U - 0x10000. Because U is less than or equal to 0x10FFFF,
|
|
// U' must be less than or equal to 0xFFFFF. That is, U' can be
|
|
// represented in 20 bits.
|
|
uint32_t Ut = U - 0x10000;
|
|
|
|
// Initialize two 16-bit unsigned integers, W1 and W2, to 0xD800 and
|
|
// 0xDC00, respectively. These integers each have 10 bits free to
|
|
// encode the character value, for a total of 20 bits.
|
|
uint16_t W1 = 0xD800;
|
|
uint16_t W2 = 0xDC00;
|
|
|
|
// Assign the 10 high-order bits of the 20-bit U' to the 10 low-order
|
|
// bits of W1 and the 10 low-order bits of U' to the 10 low-order
|
|
// bits of W2.
|
|
W1 += static_cast<uint16_t>((Ut & 0x3FC00) >> 10);
|
|
W2 += static_cast<uint16_t>((Ut & 0x3FF) >> 0);
|
|
|
|
// Terminate.
|
|
output.push_back(W1); // high surrogate
|
|
output.push_back(W2); // low surrogate
|
|
return true;
|
|
} else return false;
|
|
}
|
|
|
|
std::u16string GeneralUtils::UTF8ToUTF16(const std::string_view& string, size_t size) {
|
|
size_t newSize = MinSize(size, string);
|
|
std::u16string output;
|
|
output.reserve(newSize);
|
|
std::string_view iterator = string;
|
|
|
|
uint32_t c;
|
|
while (_NextUTF8Char(iterator, c) && PushUTF16CodePoint(output, c, size)) {}
|
|
return output;
|
|
}
|
|
|
|
//! Converts an std::string (ASCII) to UCS-2 / UTF-16
|
|
std::u16string GeneralUtils::ASCIIToUTF16(const std::string_view& string, size_t size) {
|
|
size_t newSize = MinSize(size, string);
|
|
std::u16string ret;
|
|
ret.reserve(newSize);
|
|
|
|
for (size_t i = 0; i < newSize; i++) {
|
|
char c = string[i];
|
|
// Note: both 7-bit ascii characters and REPLACEMENT_CHARACTER fit in one char16_t
|
|
ret.push_back((c > 0 && c <= 127) ? static_cast<char16_t>(c) : REPLACEMENT_CHARACTER);
|
|
}
|
|
|
|
return ret;
|
|
}
|
|
|
|
//! Converts a (potentially-ill-formed) UTF-16 string to UTF-8
|
|
//! See: <http://simonsapin.github.io/wtf-8/#decoding-ill-formed-utf-16>
|
|
std::string GeneralUtils::UTF16ToWTF8(const std::u16string_view& string, size_t size) {
|
|
size_t newSize = MinSize(size, string);
|
|
std::string ret;
|
|
ret.reserve(newSize);
|
|
|
|
for (size_t i = 0; i < newSize; i++) {
|
|
char16_t u = string[i];
|
|
if (IsLeadSurrogate(u) && (i + 1) < newSize) {
|
|
char16_t next = string[i + 1];
|
|
if (IsTrailSurrogate(next)) {
|
|
i += 1;
|
|
char32_t cp = 0x10000
|
|
+ ((static_cast<char32_t>(u) - 0xD800) << 10)
|
|
+ (static_cast<char32_t>(next) - 0xDC00);
|
|
PushUTF8CodePoint(ret, cp);
|
|
} else {
|
|
PushUTF8CodePoint(ret, u);
|
|
}
|
|
} else {
|
|
PushUTF8CodePoint(ret, u);
|
|
}
|
|
}
|
|
|
|
return ret;
|
|
}
|
|
|
|
bool GeneralUtils::CaseInsensitiveStringCompare(const std::string& a, const std::string& b) {
|
|
return std::equal(a.begin(), a.end(), b.begin(), b.end(), [](char a, char b) { return tolower(a) == tolower(b); });
|
|
}
|
|
|
|
// MARK: Bits
|
|
|
|
//! Sets a specific bit in a signed 64-bit integer
|
|
int64_t GeneralUtils::SetBit(int64_t value, uint32_t index) {
|
|
return value |= 1ULL << index;
|
|
}
|
|
|
|
//! Clears a specific bit in a signed 64-bit integer
|
|
int64_t GeneralUtils::ClearBit(int64_t value, uint32_t index) {
|
|
return value &= ~(1ULL << index);
|
|
}
|
|
|
|
//! Checks a specific bit in a signed 64-bit integer
|
|
bool GeneralUtils::CheckBit(int64_t value, uint32_t index) {
|
|
return value & (1ULL << index);
|
|
}
|
|
|
|
bool GeneralUtils::ReplaceInString(std::string& str, const std::string& from, const std::string& to) {
|
|
size_t start_pos = str.find(from);
|
|
if (start_pos == std::string::npos)
|
|
return false;
|
|
str.replace(start_pos, from.length(), to);
|
|
return true;
|
|
}
|
|
|
|
std::vector<std::wstring> GeneralUtils::SplitString(std::wstring& str, wchar_t delimiter) {
|
|
std::vector<std::wstring> vector = std::vector<std::wstring>();
|
|
std::wstring current;
|
|
|
|
for (const auto& c : str) {
|
|
if (c == delimiter) {
|
|
vector.push_back(current);
|
|
current = L"";
|
|
} else {
|
|
current += c;
|
|
}
|
|
}
|
|
|
|
vector.push_back(current);
|
|
return vector;
|
|
}
|
|
|
|
std::vector<std::u16string> GeneralUtils::SplitString(const std::u16string& str, char16_t delimiter) {
|
|
std::vector<std::u16string> vector = std::vector<std::u16string>();
|
|
std::u16string current;
|
|
|
|
for (const auto& c : str) {
|
|
if (c == delimiter) {
|
|
vector.push_back(current);
|
|
current = u"";
|
|
} else {
|
|
current += c;
|
|
}
|
|
}
|
|
|
|
vector.push_back(current);
|
|
return vector;
|
|
}
|
|
|
|
std::vector<std::string> GeneralUtils::SplitString(const std::string& str, char delimiter) {
|
|
std::vector<std::string> vector = std::vector<std::string>();
|
|
std::string current = "";
|
|
|
|
for (size_t i = 0; i < str.length(); i++) {
|
|
char c = str[i];
|
|
|
|
if (c == delimiter) {
|
|
vector.push_back(current);
|
|
current = "";
|
|
} else {
|
|
current += c;
|
|
}
|
|
}
|
|
|
|
vector.push_back(current);
|
|
|
|
return vector;
|
|
}
|
|
|
|
std::u16string GeneralUtils::ReadWString(RakNet::BitStream* inStream) {
|
|
uint32_t length;
|
|
inStream->Read<uint32_t>(length);
|
|
|
|
std::u16string string;
|
|
for (auto i = 0; i < length; i++) {
|
|
uint16_t c;
|
|
inStream->Read(c);
|
|
string.push_back(c);
|
|
}
|
|
|
|
return string;
|
|
}
|
|
|
|
std::vector<std::string> GeneralUtils::GetSqlFileNamesFromFolder(const std::string& folder) {
|
|
// Because we dont know how large the initial number before the first _ is we need to make it a map like so.
|
|
std::map<uint32_t, std::string> filenames{};
|
|
for (auto& t : std::filesystem::directory_iterator(folder)) {
|
|
auto filename = t.path().filename().string();
|
|
auto index = std::stoi(GeneralUtils::SplitString(filename, '_').at(0));
|
|
filenames.insert(std::make_pair(index, filename));
|
|
}
|
|
|
|
// Now sort the map by the oldest migration.
|
|
std::vector<std::string> sortedFiles{};
|
|
auto fileIterator = filenames.begin();
|
|
std::map<uint32_t, std::string>::iterator oldest = filenames.begin();
|
|
while (!filenames.empty()) {
|
|
if (fileIterator == filenames.end()) {
|
|
sortedFiles.push_back(oldest->second);
|
|
filenames.erase(oldest);
|
|
fileIterator = filenames.begin();
|
|
oldest = filenames.begin();
|
|
continue;
|
|
}
|
|
if (oldest->first > fileIterator->first) oldest = fileIterator;
|
|
fileIterator++;
|
|
}
|
|
|
|
return sortedFiles;
|
|
}
|
|
|
|
bool GeneralUtils::TryParse(const std::string& x, const std::string& y, const std::string& z, NiPoint3& dst) {
|
|
return TryParse<float>(x.c_str(), dst.x) && TryParse<float>(y.c_str(), dst.y) && TryParse<float>(z.c_str(), dst.z);
|
|
}
|