fuzzyfloat

A utility library that provides transparent floating point types with tolerances for equality comparison.

Getting Started

>>>from fuzzyfloat import rel_fp
>>>value = rel_fp(100)
>>>value == 100
True
>>>value == 99.99999999
True
>>>value == 100.00000001
True
>>>value = 1000
>>>value == 1000.0000001
True
>>>value = 10000
>>>value == 10000.000001
True
>>>value = 100000
>>>value == 100000.00001
True

>>>from fuzzyfloat import abs_fp
>>>value = abs_fp(100)
>>>value = 100000
>>>value == 100000.00001
False
>>>value == 100000.00000001
True

Setting different tolerances

from fuzzyfloat import FuzzyFloatMeta

class my_fp(metaclass=FuzzyFloatMeta, rel_tol=1e-05, atol=0.01):
    pass

Using a different underlying type

import numpy as np
from fuzzyfloat import FuzzyFloatMeta

class np_fp(metaclass=FuzzyFloatMeta, ftype=np.float128):
    pass

class c_fp(metaclass=FuzzyFloatMeta, ftype=complex):
    pass

Limitations

Any operations provided by the operators module will propogate the class type (and therefore the tolerances). However, there are many other functions, especially those that touch c extensions, where the type information will be lost, such as math.sqrt().