Floating point arithmetic

Last Updated: February 3, 2018 12:20 PM UTC

Last edited by:

Joseph <joseph@josephbanks.me>

Initial Commit

THIS IS FAKE DEMO DATA

You may have noticed that when doing arithmetic with floats in Python you sometimes get strange results, like this:

>>> 0.1 + 0.2
0.30000000000000004

Why this happens Internally your computer stores floats as binary fractions. Many decimal values cannot be stored as exact binary fractions, which means an approximation has to be used.

How you can avoid this You can use math.isclose to check if two floats are close, or to get an exact decimal representation, you can use the decimal or fractions module. Here are some examples:

>>> math.isclose(0.1 + 0.2, 0.3)
True
>>> decimal.Decimal('0.1') + decimal.Decimal('0.2')
Decimal('0.3')

Note that with decimal.Decimal we enter the number we want as a string so we don't pass on the imprecision from the float.

For more details on why this happens check out this page in the python docs or this Computerphile video.