04 August 2004 13 comments Python

I have a degree in mathematics and computer science, but still I don't know why different programming languages evaluate `Integer/Integer`

differently. On any calculator if you divide one integer with another you get a decimal number (of course not if the numerator is a factor of the denominator).

Python for example returns a integer number:

```
>>> print 3/10
0
>>> print 3/10.0
0.3
```

Perl on the other hand returns a decimal number:

```
print 3/10; print "\n";
print 3/10.0;
...gives...
0.3
0.3
```

PostgreSQL is like Python:

```
database=# SELECT 3/10 AS quotient1, 3/10.0 As quotient2;
quotient1 | quotient2
-----------+------------------------
0 | 0.30000000000000000000
```

And good old MS Excel gives:

```
=3/10 <=> 0.3
```

**Why** is it so?

## Comments

For Python, the reasons are historical. When it was designed, mathematics and arithmetic were not exactly the focus of attention, and Guido decided to follow how things were done in C when he didn't particularly care one way or the other. That is why division in Python works as in C.

In C, the choice is a more natural one, since it was designed to be a low-level language focused on efficiency. In the early days, this efficiency was achieved largely by providing a close mapping between hardware functionality and language functionality. This meant that you could get good efficiency even with a simple compiler, by leveraging the programmer's intuitions and experience regarding hardware efficiency (with integer arithmetic being much more efficient than floating-point, provided that the latter was supported at all.) Following this principle, the compiler should not automatically slow down things without very good reasons, while automatically moving from fast integers to slow floating-point arithmetic could easily be seen as violating that principle. Also, integer division is sufficient in many cases, especially when used for data organization (subdivision of datasets) rather than mathematics. The creators of C were working on operating systems, compilers, and other kinds of system software were this was true to a large extent. Floating-point arithmetic was included in the language, but in later practice it was often made optional, if at all (especially on small micros and embedded systems.)

For a high-level language focused on ease-of-use for the programmer, automatic conversion between integer and floating-point "where required" makes a lot of sense. So it is not strange that the other languages you mention work that way. Python is going in that direction as well (see e.g. http://python.fyxm.net/peps/pep-0238.html)

Hello Rickard,

This is a great piece of information. I very much liked the way you have explained it.

Thanks for answering a question which was very intriguing for me.

Regards,

Vishruth

Well, there is "real division" and "integer division". They are different.

Calculators do real division. They take a real number, divides it by another real number and return a real number.

In number theory we only work with integers. So we have the integer division. When you divide two numbers, you have the quotient and the remainder.

10 div 3 = 3 (quotient)

10 mod 3 = 1 (remainder)

3 * 3 + 1 = 10

The definition of integer division is: given two numbers a and b, the quotient and remainder or the division of a and b are defined by the formulas:

b*q+r=a

and

0<=r<b

(or 0<=|r|<|b| if dealing with negative numbers)

Perl (as far as I know) does not have a integer division (quotient) operator, but it has a remainder operator. C, Pascal and Python have both operations. But they behave diffent when dealing with negative numbers.

In the definition above, there are two possible values of r, one negative and one positive. In math we always take the positive one, but the languages sometimes take the negative.

C and Pascal (gcc and freepascal to be more specific) truncate the division and then calculate the remainder. Python and Perl, on the other hand, use the floor function instead of truncating. This causes differences in the quotient and remainder when dealing with negative integer division. For example, dividing 3 by (-2):

a = 3, b = -2

b*q+r=a

(-2) * q + r = 3

0<=|r|<2

Two possible solutions:

q=-2

r=-1

(-2) * (-2) + (-1) = 3

or

q=-1

r=1

(-2) * (-1) + (1) = 3

Both are correct. In this case, Perl and Python choose the first one (negative remainder) but Pascal and C choose the second one (positive remainder). On the other hand, if you make a=-3 and b=2, Perl and Python choose the positive r, Pascal and C choose the negative r.

Problem dividing variable of integers numbers in Python, for percentage. Solution: product between one variable by 0,1 :

print(... ), SOLA/SOLB * 100 #being SOLA < SOLB Result: 0 #

Solution

print(... ), int(SOLA/(SOLB*1.0) *100, ("%")

Blog: http://flomana.spaces.live.com/

There are more than just 2 division operators: there's floor, ceil, real and and floating point. Floor and ceil division are both integer valued and differ only for negative results. Unsurprisingly, floor division is equivalent to real division rounded down and ceil to real division rounded up. As Rickard says, which division operator(s) a language supports depends on the language's purpose.

These days (from version 2.2), Python supports both floor division & floating point division with the '//' and '/' operators, respectively. In python 2.X, you have to import the division feature from __future__ to get this behavior.

from __future__ import division

# prints '0.5 0'

print 1/2, 1//2

In Python integer division why does -7/10 == -1 and not 0?

That's a really good question. I honestly don't know the answer.

If you do -7/float(10) what's the answer?

That's very different. That's -7/10.0 which is something else.

For beauty and symmetry. Observe:

>>> [n/10 for n in range(-30, 30)]

[-3, -3, -3, -3, -3, -3, -3, -3, -3, -3, -2, -2, -2, -2, -2, -2, -2, -2, -2, -2, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2]

The operation n/10 results, over the integers, in runs of exactly 10 identical results, followed by the next integer. If -7/10 had the terrible result of 0, then the above sequence would have 20 zeros in a row instead of 10, making that single number an anomaly in what is otherwise a perfectly symmetrical sequence.

That's nice but a strangely beautiful and weird explanation.

In general, when an integer n is divided by an integer d to obtain an integer quotient q and an integer remainder r, the numbers are related as follows: n = dq + r. Python requires the integer remainder of an integer division by a positive integer to be positive, which agrees with the mathematical definition of modulus in number theory. When -7 is divided by 10, the only way to achieve this is for the remainder to be 3 and the quotient to be -1 because -7 = (10)(-1) + 3.

Thank you!