McKinney is the author of pandas and this is his reference book. The extract I have is a hundred pages: the preface, setting up the environment, the chapter on working in an interactive session, and the one on the language's own data structures, which breaks off halfway. In other words, it ends exactly where the thing the book is named after was about to start.
What I took from it
From the preface, a framing line I use when quoting: most analysis work is not analysis, it is cleaning, joining and reshaping data that arrives in a mess. When someone asks for a dashboard, the time is not in the chart; it is in the state of the source data. Since then I ask to see a real sample before giving a date, not a description of the sample.
From the interactive-session chapter, a habit: exploring the data as a conversation, in small chained questions, before writing the final script. It is faster, and it avoids the long script that fails on line forty for something you would have seen on line one.
What I left out
Everything else, and I would rather put it plainly: grouping, time series, joins and cleaning are the body of the book and I did not read them. I have no opinion on what I did not open.
What I do have an opinion on is the format. An extract of a reference book is nearly useless. Books you consult are not read from the beginning; you open them where the problem hurts, and this one stops before the first page you would ever open on a working day.