Missing value handling
Before we continue with more data types, we must step back and talk about how pandas handles missing values. So far, things have been simple (we have only seen pd.NA), but as we explore more types we will see that the way pandas handles missing values is inconsistent, owing mostly to the history of how the library was developed. While it would be great to wave a magic wand and make any inconsistencies go away, in reality, they have existed and will continue to exist in production code bases for years to come. Having a high-level understanding of that evolution will help you write better pandas code, and hopefully convert the unaware to using the idioms we preach in this book.
How to do it
The pandas library was originally built on top of NumPy, whose default data types do not support missing values. As such, pandas had to build its own missing value handling solution from scratch, and, for better or worse, decided that using the np.nan sentinel, which...
