representations.py 文件源码

def represent_float_as_str(value):
    """Represent a float as a string without losing precision."""
    # In Python 2, calling str() on a float object loses precision:
    #
    # In [1]: 1.23456789012345678
    # Out[1]: 1.2345678901234567
    #
    # In [2]: 1.2345678901234567
    # Out[2]: 1.2345678901234567
    #
    # In [3]: str(1.2345678901234567)
    # Out[3]: '1.23456789012'
    #
    # The best way to ensure precision is not lost is to convert to string via Decimal:
    # https://github.com/mogui/pyorient/pull/226/files
    if not isinstance(value, float):
        raise GraphQLInvalidArgumentError(u'Attempting to represent a non-float as a float: '
                                          u'{}'.format(value))

    with decimal.localcontext() as ctx:
        ctx.prec = 20  # floats are max 80-bits wide = 20 significant digits
        return u'{:f}'.format(decimal.Decimal(value))