Python 3.6.1 (default, Sep  7 2017, 16:36:03) 
[GCC 6.3.0 20170406] on linux
Type "help", "copyright", "credits" or "license" for more information.
>>> round(3.5)
4
>>> round(3.55, 1)
3.5

为什么？据说这跟浮点数的精度有关。我们看到的这个3.55实际上在机器上存储的值是3.54999...999

最终, 我手动写了一个round函数:

def round(n, m=0):
    '''round(3.555, 2) => 3.56'''
    # python自带的round函数有时会把5舍掉,如round(3.55, 1) => 3.5
    if m == 0:
        return int(n+0.5)
    n = str(int(n*10**m+0.5))
    m *= -1
    n = '{}.{}'.format(n[:m], n[m:])
    return float(n)

----------------

2019-12-15 先转为str类型，再运用decimal将代码精简如下：

from decimal import Decimal, ROUND_HALF_UP
def precise_round(n, m=0, deci=False):
    d = Decimal(str(n)).quantize(Decimal('0.{}'.format('0'*m)), ROUND_HALF_UP)
    return d if deci else float(d) 

---------------------

Updated_at 2020/07/20

def to_decimal(amount: Union[int, str, float, Decimal]) -> Decimal:
    """将传入的数值，四舍五入转化为两个小数的Decimal对象

    Usage::
        >>> to_decimal(1.565)
        Decimal('1.57')
        >>> to_decimal('1.564')
        Decimal('1.56')
    """
    return Decimal(str(amount)).quantize(Decimal("0.00"), ROUND_HALF_UP)