# fractions data type — when to reduce?

Here is my Python implementation of a simple fractions class:

class frac:
def __init__(self, a,b):
self.a = a
self.b = b

return frac(self.a*x.b + self.b*x.a, self.b*x.b)

def __mul__(self, x):
return frac(self.a*x.a, self.b*x.b)

def __sub__(self, x):
return frac(self.a*x.b - self.b*x.a, self.b*x.b)

def __div__(self, x):
return frac(self.a*x.b, self.b*x.a)

def __repr__(self):
return "%s/%s"%(self.a,self.b)


x = frac(1,3)
y = frac(1,2)
print x+ y

>> 5/6


However, it can miss the mark:

x = frac(1,3)
y = frac(2,3)
print x+ y

>> 9/9


It seems if I run this library enough, the numerators and denominators will get very large. When should I start simplifying and where should I put that code?

There seems to be a fractions class already in Python, it this was instructive to put together.

My original goal was to allow for square roots of integers as well, but I will ask this in a separate question.

-
You might consider accepcting the most helpful answer by clicking the check mark below the voting buttons. – ojdo Apr 29 '14 at 13:54

I would probably put it right in your __init__ method, which simplifies it every time a fraction is created (which includes after any operation).

I disagree with Jaime's answer that you should do the simplification as part of the operation, as you'll have to copy/paste that code for every new operation.

I’d write something like:

def gcd(x, y):
while y:
x, y = y, x % y
return x

class Fraction(object):
def __init__(self, numer, denom):
cancel_factor = gcd(numer, denom)
self.numer = numer / cancel_factor
self.denom = denom / cancel_factor


• Per PEP 8 on class names, I'm using Fraction instead of frac.

Also from PEP 8, I would put spaces around your multiplication operators.

• New style classes should inherit from object (as I've done above).

• I think numer and denom are better attribute names than a and b.

• The convention is to use other as the variable name in a method that takes two inputs, rather than x. For example:

def __add__(self, other):
new_numer = self.numer * other.denom + self.denom * other.numer
new_denom = self.denom * other.denom
return Fraction(new_numer, new_denom)

• I disagree with your use of __repr__. As a rule of thumb, __repr__ should be something you can eval() to get back to that object; __str__ is the human-readable object. See difference between __str__ and __repr__ for more.

Here's what I'd write:

def __repr__(self):
return "Fraction(%s, %s)" % (self.numer, self.denom)

def __str__(self):
return "%s / %s" % (self.numer, self.denom)

-
That's indeed a much better structuring of the code, +1. – Jaime Apr 23 '14 at 20:50

If you want to produce correct results, you should simplify after each operation. Euclid's algorithm is very fast in finding the GCD, and very simple to code. If you have a function like:

def gcd(a, b):
while b:
a, b = b, a % b
return a


Then you could rewrite, e.g. you __add__ method as:

def __add__(self, other):
num = self.a * other.b + self.b * other.a
den = self.b * other.b
gcf = gcd(num, den)

return frac(num // gcf, den // gcf)

-
This simplification is needed in each function (and maybe even outside) so it should be a function of its own. – Nobody Apr 23 '14 at 18:31
Euclid's algorithm may be fast compared with a naive approach, but if code reduces after every addition, subtraction, multiplication, or division, it will spend more time doing the reductions than it spends doing everything else combined. – supercat Apr 23 '14 at 21:00
If you want correctness, you have to keep fractions in lowest terms. That the implementation in the standard Python library does exactly the same as @alexwlchan proposes, see here, is a good indicator that this is the proper way of going about it. – Jaime Apr 23 '14 at 23:12