Summary
In this chapter, we learned how to manipulate NumPy arrays and how to write fast mathematical expressions using array broadcasting. This knowledge will help you write more concise, expressive code and, at the same time, to obtain substantial performance gains. We also introduced the numexpr library to further speed up NumPy calculations with minimal effort.
Pandas implements efficient data structures that are useful when analyzing large datasets. In particular, Pandas shines when the data is indexed by non-integer keys and provides very fast hashing algorithms.
NumPy and Pandas work well when handling large, homogenous inputs, but they are not suitable when the expressions grow complex and the operations cannot be expressed using the tools provided by these libraries. In such cases, we can leverage Python capabilities as a glue language by interfacing it with C using the Cython package.
