首页
学习
活动
专区
圈层
工具
发布
社区首页 >问答首页 >Python用十进制模块舍入到指定的十进制位

Python用十进制模块舍入到指定的十进制位
EN

Stack Overflow用户
提问于 2021-03-21 03:45:10
回答 1查看 432关注 0票数 1

问题:在计算算术操作时,如何使的十进制模块四舍五入到指定的小数位而不是舍入到指定的精度(显着数字)?

信息

我一直在使用Python中的Decimal模块,使用setcontext方法将值舍入到指定的精度。这是很好的工作,直到我们开始交叉整数和小数,因为重要的数字并没有区分两者。

代码语言:javascript
复制
import decimal as d
from math import pi

decimal_places = 0
d.setcontext(d.Context(prec=decimal_places+1, rounding=d.ROUND_HALF_UP))

# This works fine
num = pi
print(f"Rounding {num} to {decimal_places} decimal places:")
print(f"Traditional rounding (correct): {round(num, decimal_places)}")
print(f"Decimal rounding (correct): {+d.Decimal(num)}")

# This is were issues start to arise
num = pi/10
print(f"\nRounding {num} to {decimal_places} decimal places:")
print(f"Traditional rounding (correct): {round(num, decimal_places)}")
print(f"Decimal rounding (incorrect): {+d.Decimal(num)}")
代码语言:javascript
复制
Rounding 3.141592653589793 to 0 decimal places:
Traditional rounding (correct): 3.0
Decimal rounding (correct): 3

Rounding 0.3141592653589793 to 0 decimal places:
Traditional rounding (correct): 0.0
Decimal rounding (incorrect): 0.3

用例

为什么在Python的圆形函数上使用十进制模块呢?好吧,十进制模块的优点是它将在算术计算(PEMDAS)的所有步骤中应用该精度上限。

例如,如果我想在函数中对x进行求值时循环x,我可以这样做:

代码语言:javascript
复制
function_str = "0.5 * (3*x) ** 2 + 3"
eval(function_str.replace("x", "(+d.Decimal(x))"))

一个更完整(更简单)的例子:

代码语言:javascript
复制
import decimal as d

decimal_places = 0
d.setcontext(d.Context(prec=decimal_places+1, rounding=d.ROUND_HALF_UP))

numerator = 5
denominator = 1.1
num_err = 0.5
new_num = numerator + num_err

print(f"Rounding {numerator}/{denominator} to {decimal_places} decimal places:")
print(f"Traditional rounding (incorrect): {round(new_num, decimal_places)/denominator}")
print(f"Decimal rounding (correct): {+d.Decimal(new_num) / d.Decimal(denominator)}")
代码语言:javascript
复制
Rounding 5/1.1 to 0 decimal places:
Traditional rounding (incorrect): 5.454545454545454
Decimal rounding (correct): 5

在这里,它似乎仍然是一个更简单的解决方案,因为它可以放在输出周围,但是随着函数的复杂性增加,这种方法变得越来越不可行。在用户输入函数的情况下,传统四舍五入的可行性几乎为零,而使用十进制模块则与function_str.replace("x", "(+d.Decimal(x))")一样简单。

注意,quantize方法不是一个可行的选项,因为它只舍入当前数字,而不是所有内容(这就是设置上下文精度的方法)。

EN

回答 1

Stack Overflow用户

回答已采纳

发布于 2021-03-22 03:41:04

为了解决这个问题,我只做了我自己的定点算术库。为了帮助将来遇到这个问题的其他人,我在下面发布了我的定点算术库的代码。

代码语言:javascript
复制
import math


PREC = 0


def no_rounding(x, *args, **kwargs):
    return x


def ceil(x, prec=0):
    mult = 10 ** prec
    return round(math.ceil(x * mult) / mult, prec)


def floor(x, prec=0):
    mult = 10 ** prec
    return round(math.floor(x * mult) / mult, prec)


rounding = {
    None: no_rounding,
    "round": round,
    "ceil": ceil,
    "floor": floor,
}


class Fixed:
    def __init__(self, number, round_function="round", custom_prec=None):
        self.val = float(number)
        self.round_str = round_function
        self.round_func = rounding[round_function]
        self.custom_prec = custom_prec

    def _dup_fixed(self, number):
        return Fixed(number, self.round_str, self.custom_prec)

    def _operation(self, op):
        return self._dup_fixed(self.round_func(op, self.prec))

    @property
    def prec(self):
        return int(self.custom_prec if self.custom_prec is not None else PREC)

    @property
    def num(self):
        return self.round_func(self.val, self.prec)

    @property
    def real(self):
        return self

    @property
    def imag(self):
        return Fixed(0)

    def __setattr__(self, name, value):
        if name == "val":
            value = float(value)
        self.__dict__[name] = value

    def __hash__(self):
        return hash(self.num)

    def __str__(self):
        return str(self.num)

    __repr__ = __str__

    def __format__(self, spec):
        if spec == "":
            return str(self)
        else:
            return spec % self.num

    def __reduce__(self):
        return (self.__class__, (self.val,))

    def __copy__(self):
        return self.__class__(self.val)

    def __deepcopy__(self, memo):
        return self.__copy__()

    def __pos__(self):
        return self

    def __neg__(self):
        return self._dup_fixed(-self.val)

    def __abs__(self):
        return self._dup_fixed(abs(self.val))

    def __round__(self, n=None):
        return self._dup_fixed(round(self.val, n))

    def __floor__(self):
        return self._dup_fixed(math.floor(self.val))

    def __ceil__(self):
        return self._dup_fixed(math.ceil(self.val))

    def __int__(self):
        return int(self.num)

    def __trunc__(self):
        return math.trunc(self.num)

    def __float__(self):
        return float(self.num)

    def __complex__(self):
        return complex(self.num)

    def conjugate(self):
        return self

    def __eq__(self, other):
        return self.num == float(other)

    def __ne__(self, other):
        return not self == float(other)

    def __gt__(self, other):
        return self.num > float(other)

    def __ge__(self, other):
        return self.num >= float(other)

    def __lt__(self, other):
        return self.num < float(other)

    def __le__(self, other):
        return self.num <= float(other)

    def __bool__(self):
        return self.num != 0

    def __add__(self, other):
        return self._operation(self.num + float(other))

    __radd__ = __add__

    def __sub__(self, other):
        return self + -other

    def __rsub__(self, other):
        return -self + other

    def __mul__(self, other):
        return self._operation(self.num * float(other))

    __rmul__ = __mul__

    def __truediv__(self, other):
        return self._operation(self.num / float(other))

    def __rtruediv__(self, other):
        return self._operation(float(other) / self.num)

    def __floordiv__(self, other):
        return self._operation(self.num // float(other))

    def __rfloordiv__(self, other):
        return self._operation(float(other) // self.num)

    def __mod__(self, other):
        return self._operation(self.num % float(other))

    def __rmod__(self, other):
        return self._operation(float(other) % self.num)

    def __divmod__(self, other):
        result = divmod(self.num, float(other))
        return (self._operation(result[0]), self._operation(result[1]))

    def __rdivmod__(self, other):
        result = divmod(float(other), self.num)
        return (self._operation(result[0]), self._operation(result[1]))

    def __pow__(self, other):
        return self._operation(self.num ** float(other))

    def __rpow__(self, other):
        return self._operation(float(other) ** self.num)

如果你发现任何错误或问题,请在评论中告诉我,我一定会更新我的答案。

用法

通过将数字传递给Fixed函数来创建一个固定的数字。然后,可以将此固定数与正常数类似地处理。

代码语言:javascript
复制
import fixed_point as fp  # Import file

num = 1.6
fixed_num = fp.Fixed(num)  # Default precision is 0
print("Original number:", num)
print("Fixed number:", fixed_num)
print("Fixed number value multiplied by original number:", fixed_num.val * num)
print("Fixed number multiplied by original number:", fixed_num * num)
print("Fixed number multiplied by itself:", fixed_num * fixed_num)
代码语言:javascript
复制
Original number: 1.6
Fixed number: 2.0
Fixed number value multiplied by original number: 2.56
Fixed number multiplied by original number: 3.0
Fixed number multiplied by itself: 4.0

为了设置全局精度,可以修改PREC变量,这不仅会改变所有新的固定精度数的精度(小数位数),而且还会改变现有的精度。在创建过程中还可以设置特定固定数的精度。

代码语言:javascript
复制
num = 3.14159
fixed_num = fp.Fixed(num)
custom_prec_num = fp.Fixed(num, custom_prec=4)
print("Original number:", num)
print("Fixed number (default precision):", fixed_num)
print("Custom precision fixed number (4 decimals):", custom_prec_num)

fp.PREC = 2  # Update global precision
print("\nGlobal precision updated to", fp.PREC)

print("Fixed number (new precision):", fixed_num)
print("Custom precision fixed number (4 decimals):", custom_prec_num)
代码语言:javascript
复制
Original number: 3.14159
Fixed number (default precision): 3.0
Custom precision fixed number (4 decimals): 3.1416

Global precision updated to 2
Fixed number (new precision): 3.14
Custom precision fixed number (4 decimals): 3.1416

请注意,只有使用fixed_num.val才能使用float(fixed_num)获取固定数字的原始值,这将返回四舍五入到指定小数位数的固定数字(除非四舍五入为none)。

票数 1
EN
页面原文内容由Stack Overflow提供。腾讯云小微IT领域专用引擎提供翻译支持
原文链接:

https://stackoverflow.com/questions/66728640

复制
相关文章

相似问题

领券
问题归档专栏文章快讯文章归档关键词归档开发者手册归档开发者手册 Section 归档