Chapter 5: Numbers & Bits (TL;DR)¶

There exist three numeric types in core Python:

• int: a near-perfect model for whole numbers (i.e., $\mathbb{Z}$); inherently precise
• float: the "gold" standard to approximate real numbers (i.e., $\mathbb{R}$); inherently imprecise
• complex: layer on top of the float type to approximate complex numbers (i.e., $\mathbb{C}$); inherently imprecise

Furthermore, the standard library provides two more types that can be used as substitutes for the float type:

• Decimal: similar to float but allows customizing the precision; still inherently imprecise
• Fraction: a near-perfect model for rational numbers (i.e., $\mathbb{Q}$); built on top of the int type and therefore inherently precise

The important takeaways for the data science practitioner are:

1. Do not mix precise and imprecise data types, and
2. actively expect nan results when working with float numbers as there are no loud failures.

The numerical tower is Python's way of implementing various abstract ideas of what numbers are in mathematics.