C++十进制角度转DMS精度异常，求更优解决方案

问题背景 #

我编写了以下C++代码实现十进制角度与度分秒(DMS)的互转：

#include <cmath>
#include <iostream>

void toDMS(double decimalDegrees, int& degrees, unsigned int& minutes, double& seconds)
{
    degrees = static_cast<int>(decimalDegrees);
    double remainingMinutes = fabs(decimalDegrees - degrees) * 60.0;
    minutes = static_cast<unsigned int>(remainingMinutes);
    seconds = (remainingMinutes - minutes) * 60.0;
}
double toDecimalDegrees(int degrees, unsigned int minutes, double seconds)
{
    if (degrees >= 0.0)
        return fabs(degrees) + (minutes / 60.0) + (seconds / 3600.0);
    else
        return -(fabs(degrees) + (minutes / 60.0) + (seconds / 3600.0));
}

运行测试代码时：

int main(int argc, char** argv)
{
    double decimalDegrees = 45.666666666666664;
    int degrees; unsigned int minutes; double seconds;
    toDMS(decimalDegrees, degrees, minutes, seconds);
    std::cout << "DMS = " << degrees << ", " << minutes << ", " << seconds << std::endl;
}

得到结果45, 39, 60，而非预期的45, 40, 0。经排查是浮点数精度问题导致remainingMinutes实际为39.9999999...，而非精确的40.0。

我曾尝试通过设定精度阈值的临时方案规避：

double remainingMinutes = fabs(decimalDegrees - degrees) * 60.0;
double ceil = std::ceil(remainingMinutes);
if (ceil - remainingMinutes <= 1.0e-10)
    remainingMinutes = ceil;

但该方案依赖手动设定的精度阈值，仍可能在其他数值场景下出现问题。

更优解决方案：基于总秒数的精确拆分 #

核心思路是先将十进制角度转换为总秒数，通过四舍五入到整数秒来消除浮点数精度误差，再拆分度、分、秒并处理进位，从根源上避免中间步骤的精度损失。

优化后的代码：

#include <cmath>
#include <iostream>

void toDMS(double decimalDegrees, int& degrees, unsigned int& minutes, double& seconds)
{
    // 计算总秒数并四舍五入到整数，消除精度误差
    double totalSeconds = fabs(decimalDegrees) * 3600.0;
    long long totalSecondsRounded = llround(totalSeconds);

    // 确定角度的符号与整数度部分
    degrees = static_cast<int>(decimalDegrees >= 0 ? floor(decimalDegrees) : ceil(decimalDegrees));
    long long remainingSeconds = totalSecondsRounded;

    // 拆分分和秒
    minutes = static_cast<unsigned int>(remainingSeconds / 60);
    seconds = remainingSeconds % 60;

    // 处理进位逻辑：秒满60进1分，分满60进1度
    if (seconds >= 60.0) {
        minutes += 1;
        seconds -= 60.0;
    }
    if (minutes >= 60) {
        degrees += decimalDegrees >= 0 ? 1 : -1;
        minutes -= 60;
    }
}

double toDecimalDegrees(int degrees, unsigned int minutes, double seconds)
{
    double sign = degrees >= 0 ? 1.0 : -1.0;
    return sign * (fabs(degrees) + minutes / 60.0 + seconds / 3600.0);
}

方案优势 #

1. 精度可控：通过llround将总秒数四舍五入到整数，直接规避了浮点数运算的累积误差；
2. 自动进位：显式处理秒→分、分→度的进位逻辑，确保结果符合DMS的格式规范；
3. 通用性强：无需手动设定精度阈值，适用于所有合法的十进制角度转换场景。

测试原数值45.666666666666664时，会得到预期结果45, 40, 0。

内容的提问来源于stack exchange，提问作者ConanLord