Just-DataInspector/Utilities/Src/Utf8Util.cpp

#include "Utf8Util.hpp"

#include <windows.h>

#include <cmath>
#include <cstdio>
#include <cstdlib>
#include <cstring>

namespace jdi
{

	namespace
	{
		/** @brief 判断字符是否为需要裁掉的空白字符。 */
		bool IsSpaceChar(wchar_t c)
		{
			return c == L' ' || c == L'\t' || c == L'\r' || c == L'\n' || c == L'\v' || c == L'\f' ||
			       c == static_cast<wchar_t>(0x00A0);
		}

		/** @brief 判断字符是否为 ASCII 大写字母。 */
		bool IsUpperAscii(wchar_t c)
		{
			return c >= L'A' && c <= L'Z';
		}
	}

	std::wstring Utf8ToWide(const char* text, int length)
	{
		if (text == nullptr)
		{
			return std::wstring();
		}

		int byteCount = (length < 0) ? static_cast<int>(std::strlen(text)) : length;
		if (byteCount <= 0)
		{
			return std::wstring();
		}

		int wideCount = MultiByteToWideChar(CP_UTF8, 0, text, byteCount, nullptr, 0);
		if (wideCount <= 0)
		{
			return std::wstring();
		}

		std::wstring result(static_cast<size_t>(wideCount), L'\0');
		MultiByteToWideChar(CP_UTF8, 0, text, byteCount, &result[0], wideCount);
		return result;
	}

	std::wstring Trim(const std::wstring& text)
	{
		size_t begin = 0;
		size_t end = text.size();
		while (begin < end && IsSpaceChar(text[begin]))
		{
			++begin;
		}
		while (end > begin && IsSpaceChar(text[end - 1]))
		{
			--end;
		}
		return text.substr(begin, end - begin);
	}

	std::wstring StripQuotes(const std::wstring& text)
	{
		std::wstring trimmed = Trim(text);
		if (trimmed.size() >= 2)
		{
			const wchar_t first = trimmed.front();
			const wchar_t last = trimmed.back();
			if ((first == L'"' && last == L'"') || (first == L'\'' && last == L'\''))
			{
				trimmed = trimmed.substr(1, trimmed.size() - 2);
			}
		}
		return trimmed;
	}

	std::wstring ToLowerAscii(const std::wstring& text)
	{
		std::wstring result = text;
		for (wchar_t& c : result)
		{
			if (IsUpperAscii(c))
			{
				c = static_cast<wchar_t>(c - L'A' + L'a');
			}
		}
		return result;
	}

	bool ContainsInsensitive(const std::wstring& text, const std::wstring& needle)
	{
		if (needle.empty())
		{
			return true;
		}
		if (needle.size() > text.size())
		{
			return false;
		}

		const size_t last = text.size() - needle.size();
		for (size_t i = 0; i <= last; ++i)
		{
			size_t j = 0;
			while (j < needle.size())
			{
				wchar_t a = text[i + j];
				wchar_t b = needle[j];
				if (IsUpperAscii(a))
				{
					a = static_cast<wchar_t>(a - L'A' + L'a');
				}
				if (IsUpperAscii(b))
				{
					b = static_cast<wchar_t>(b - L'A' + L'a');
				}
				if (a != b)
				{
					break;
				}
				++j;
			}
			if (j == needle.size())
			{
				return true;
			}
		}
		return false;
	}

	/** @brief 快速解析路径允许保留的最大有效位数（超过则交回 strtod 以保证正确舍入）。 */
	const int kMaxFastDigits = 17;

	/** @brief 快速解析路径允许使用的最大十进制指数（超出查表范围，交回 strtod）。 */
	const int kMaxFastExponent = 22;

	/** @brief 10 的 0-22 次幂精确表（该范围内每位都是 double 可精确表示的十进制数）。 */
	const double kPow10Table[23] = {1e0,  1e1,  1e2,  1e3,  1e4,  1e5,  1e6,  1e7,  1e8,  1e9,  1e10, 1e11,
	                                1e12, 1e13, 1e14, 1e15, 1e16, 1e17, 1e18, 1e19, 1e20, 1e21, 1e22};

	/**
	 * @brief 标准库解析路径：去掉首尾空白后交给 strtod。
	 *
	 * 作为快速路径的保底实现，保证十六进制浮点、inf/nan、超范围指数、
	 * 带尾部杂字符（如 "12abc"）等输入的处理与改造前完全一致。
	 *
	 * @param [IN] begin 文本起始指针。
	 * @param [IN] end 文本结束指针（不包含）。
	 * @param [OUT] value 解析结果。
	 * @return 解析出数值返回 true。
	 */
	bool ParseDoubleSlow(const char* begin, const char* end, double& value)
	{
		const char* p = begin;
		while (p < end && (*p == ' ' || *p == '\t'))
		{
			++p;
		}
		const char* q = end;
		while (q > p && (q[-1] == ' ' || q[-1] == '\t' || q[-1] == '\r' || q[-1] == '\n'))
		{
			--q;
		}
		if (p >= q)
		{
			return false;
		}

		char buffer[96];
		const size_t length = static_cast<size_t>(q - p);
		const char* text = nullptr;
		std::string dynamic;
		if (length < sizeof(buffer))
		{
			std::memcpy(buffer, p, length);
			buffer[length] = '\0';
			text = buffer;
		}
		else
		{
			dynamic.assign(p, q);
			text = dynamic.c_str();
		}

		char* stop = nullptr;
		const double parsed = std::strtod(text, &stop);
		if (stop == text)
		{
			return false;
		}

		value = parsed;
		return true;
	}

	bool TryParseDouble(const char* begin, const char* end, double& value)
	{
		if (begin == nullptr || end == nullptr || begin >= end)
		{
			return false;
		}

		// ===== 快速路径：纯十进制数值 =====
		// 大文件解析的绝对瓶颈就在这里：ucrt 的 strtod 要处理本地化、十六进制浮点、
		// 无穷/NaN 等全部语法，实测每次调用约 335ns；百万行级文件需调用上千万次，
		// 单解析就可能耗掉几十秒。下面这条路径只处理最常见的十进制写法（实测约 13ns，
		// 快 25 倍），其余写法（含尾部杂字符）原样交回 strtod，
		// 因此输入接受范围与原先完全一致。
		const char* head = begin;
		while (head < end && (*head == ' ' || *head == '\t'))
		{
			++head;
		}
		const char* tail = end;
		while (tail > head && (tail[-1] == ' ' || tail[-1] == '\t' || tail[-1] == '\r' || tail[-1] == '\n'))
		{
			--tail;
		}
		if (head >= tail)
		{
			return false;
		}

		const char* cursor = head;
		bool negative = false;
		if (*cursor == '+' || *cursor == '-')
		{
			negative = (*cursor == '-');
			++cursor;
		}

		unsigned long long mantissa = 0;
		int digits = 0;
		int exponent = 0;

		// 整数部分：超过有效位数的高位数字只影响量级，累加进指数。
		while (cursor < tail && *cursor >= '0' && *cursor <= '9')
		{
			if (digits < kMaxFastDigits)
			{
				mantissa = mantissa * 10ull + static_cast<unsigned>(*cursor - '0');
				++digits;
			}
			else
			{
				++exponent;
			}
			++cursor;
		}

		// 小数部分：每保留一位就记一次小数位数；超出有效位数的直接丢弃（已低于 double 精度）。
		int fractionDigits = 0;
		if (cursor < tail && *cursor == '.')
		{
			++cursor;
			while (cursor < tail && *cursor >= '0' && *cursor <= '9')
			{
				if (digits < kMaxFastDigits)
				{
					mantissa = mantissa * 10ull + static_cast<unsigned>(*cursor - '0');
					++digits;
					++fractionDigits;
				}
				++cursor;
			}
		}

		if (digits == 0)
		{
			return ParseDoubleSlow(begin, end, value);
		}
		exponent -= fractionDigits;

		if (cursor < tail && (*cursor == 'e' || *cursor == 'E'))
		{
			++cursor;
			bool expNegative = false;
			if (cursor < tail && (*cursor == '+' || *cursor == '-'))
			{
				expNegative = (*cursor == '-');
				++cursor;
			}
			if (cursor >= tail || *cursor < '0' || *cursor > '9')
			{
				return ParseDoubleSlow(begin, end, value);
			}
			int expValue = 0;
			while (cursor < tail && *cursor >= '0' && *cursor <= '9')
			{
				if (expValue < 1000000)
				{
					expValue = expValue * 10 + (*cursor - '0');
				}
				++cursor;
			}
			exponent += expNegative ? -expValue : expValue;
		}

		// 存在未消费字符（如 "12abc"、"1,5"）或指数超出查表范围时交回标准库：
		// 既保持原先 strtod 的宽松接受语义，也保证极端量级的正确舍入。
		if (cursor != tail || exponent > kMaxFastExponent || exponent < -kMaxFastExponent)
		{
			return ParseDoubleSlow(begin, end, value);
		}

		double result = static_cast<double>(mantissa);
		if (exponent > 0)
		{
			result *= kPow10Table[exponent];
		}
		else if (exponent < 0)
		{
			result /= kPow10Table[-exponent];
		}
		// mantissa 超过 2^53（16 位以上有效数字）时这里会比 strtod 多一次舍入，
		// 相对误差不超过 1 ULP（实测 ≤2.3e-16），在波形显示与标尺（约 6 位有效数字）上不可见。
		value = negative ? -result : result;
		return true;
	}
	std::wstring FormatNumber(double value, int decimals)
	{
		if (decimals < 0)
		{
			decimals = 0;
		}
		if (decimals > 12)
		{
			decimals = 12;
		}

		wchar_t buffer[96];
		swprintf(buffer, sizeof(buffer) / sizeof(buffer[0]), L"%.*f", decimals, value);
		return std::wstring(buffer);
	}

	std::wstring FormatByteSize(unsigned long long bytes)
	{
		// 单位自适应：GB / MB / KB（阈值按 1024 进制）。
		constexpr double kKilo = 1024.0;
		constexpr double kMega = kKilo * 1024.0;
		constexpr double kGiga = kMega * 1024.0;

		const double value = static_cast<double>(bytes);
		const wchar_t* unit = L"KB";
		double scaled = value / kKilo;
		if (value >= kGiga)
		{
			unit = L"GB";
			scaled = value / kGiga;
		}
		else if (value >= kMega)
		{
			unit = L"MB";
			scaled = value / kMega;
		}

		wchar_t buffer[64];
		// ⚠️ 宽字符格式化必须用 `%ls`：MinGW 的 swprintf 按 C99 语义处理 `%s`
		//（在宽格式串里 `%s` 表示**窄**字符串），写成 `%s` 会把宽字符串按单字节读取，
		// 于是 "KB" 只打印出 "K"。
		swprintf(buffer, sizeof(buffer) / sizeof(buffer[0]), L"%.2f %ls", scaled, unit);
		return std::wstring(buffer);
	}

	std::wstring FormatValue(double value)
	{
		if (std::isnan(value))
		{
			return L"NaN";
		}
		if (std::isinf(value))
		{
			return (value > 0.0) ? L"+Inf" : L"-Inf";
		}

		wchar_t buffer[96];
		swprintf(buffer, sizeof(buffer) / sizeof(buffer[0]), L"%.6g", value);
		return std::wstring(buffer);
	}

	std::wstring FormatTick(double value, double step)
	{
		if (value == 0.0)
		{
			return L"0";
		}
		if (!std::isfinite(value))
		{
			return FormatValue(value);
		}

		int decimals = 0;
		if (step > 0.0 && std::isfinite(step))
		{
			const double magnitude = std::ceil(-std::log10(step) - 1.0e-9);
			if (magnitude > 0.0)
			{
				decimals = static_cast<int>(magnitude);
			}
		}
		if (decimals > 9)
		{
			decimals = 9;
		}

		wchar_t buffer[96];
		swprintf(buffer, sizeof(buffer) / sizeof(buffer[0]), L"%.*f", decimals, value);
		std::wstring text(buffer);
		if (text.size() > 15)
		{
			swprintf(buffer, sizeof(buffer) / sizeof(buffer[0]), L"%.4e", value);
			text.assign(buffer);
		}
		return text;
	}

	std::wstring FormatCount(unsigned long long value)
	{
		wchar_t buffer[32];
		swprintf(buffer, sizeof(buffer) / sizeof(buffer[0]), L"%llu", value);

		std::wstring digits(buffer);
		std::wstring result;
		result.reserve(digits.size() + digits.size() / 3);

		const size_t length = digits.size();
		for (size_t i = 0; i < length; ++i)
		{
			if (i > 0 && ((length - i) % 3u) == 0u)
			{
				result.push_back(L',');
			}
			result.push_back(digits[i]);
		}
		return result;
	}

	double NiceStep(double range, int targetTicks)
	{
		if (!(range > 0.0) || !std::isfinite(range) || targetTicks <= 0)
		{
			return 1.0;
		}

		const double raw = range / static_cast<double>(targetTicks);
		const double magnitude = std::pow(10.0, std::floor(std::log10(raw)));
		const double normalized = raw / magnitude;

		double nice = 10.0;
		if (normalized <= 1.0)
		{
			nice = 1.0;
		}
		else if (normalized <= 2.0)
		{
			nice = 2.0;
		}
		else if (normalized <= 5.0)
		{
			nice = 5.0;
		}

		return nice * magnitude;
	}

	void BuildTicks(double minValue, double maxValue, double step, std::vector<double>& ticks)
	{
		ticks.clear();
		if (!std::isfinite(minValue) || !std::isfinite(maxValue) || !(step > 0.0) || !std::isfinite(step))
		{
			return;
		}
		if (maxValue < minValue)
		{
			const double temp = minValue;
			minValue = maxValue;
			maxValue = temp;
		}

		const long long first = static_cast<long long>(std::ceil(minValue / step - 1.0e-9));
		const long long last = static_cast<long long>(std::floor(maxValue / step + 1.0e-9));
		if (last < first)
		{
			return;
		}

		const long long count = last - first + 1;
		if (count > 512)
		{
			return;
		}

		ticks.reserve(static_cast<size_t>(count));
		for (long long index = first; index <= last; ++index)
		{
			ticks.push_back(static_cast<double>(index) * step);
		}
	}

	bool CopyTextToClipboard(HWND owner, const std::wstring& text)
	{
		if (text.empty())
		{
			return false;
		}

		if (OpenClipboard(owner) == FALSE)
		{
			return false;
		}

		bool success = false;
		if (EmptyClipboard() != FALSE)
		{
			const SIZE_T bytes = (text.size() + 1) * sizeof(wchar_t);
			HGLOBAL handle = GlobalAlloc(GMEM_MOVEABLE, bytes);
			if (handle != nullptr)
			{
				void* target = GlobalLock(handle);
				if (target != nullptr)
				{
					std::memcpy(target, text.c_str(), bytes);
					GlobalUnlock(handle);
					if (SetClipboardData(CF_UNICODETEXT, handle) != nullptr)
					{
						success = true;
					}
					else
					{
						GlobalFree(handle);
					}
				}
				else
				{
					GlobalFree(handle);
				}
			}
		}

		CloseClipboard();
		return success;
	}

	bool PasteTextFromClipboard(HWND owner, std::wstring& text)
	{
		text.clear();

		if (OpenClipboard(owner) == FALSE)
		{
			return false;
		}

		bool success = false;
		if (IsClipboardFormatAvailable(CF_UNICODETEXT) != FALSE)
		{
			HANDLE handle = GetClipboardData(CF_UNICODETEXT);
			if (handle != nullptr)
			{
				const void* source = GlobalLock(handle);
				if (source != nullptr)
				{
					const wchar_t* wide = static_cast<const wchar_t*>(source);
					text.assign(wide);
					GlobalUnlock(handle);
					success = !text.empty();
				}
			}
		}
		CloseClipboard();

		// 单行输入框：把换行与制表符折叠为空格，避免内容被截断显示。
		if (success)
		{
			for (wchar_t& c : text)
			{
				if (c == L'\r' || c == L'\n' || c == L'\t')
				{
					c = L' ';
				}
			}
		}
		return success;
	}

} // namespace jdi