The m_breaks vector may look a little odd, but that's mostly because the code is compact. Our goal is to produce a vector of breakpoints for our month labels over the course of the year. So we start with a sequence of thirteen monthly start-dates (January Ist 2016, February 1st 2016, etc, all the way to January Ist 2017). Then we calculate the difference between those dates, in integer days. Taking the cumulative sum of those differences gives us the day of the year each of those first-of-the-month days falls on. Finally we subtract sixteen from those numbers to move our label breaks to approximately the middle of the month. These are fine details. But this is why this final chapter is called "Refine Your Plots" and not "Don't Bother Doing Things Properly".
https://files.mastodon.social/media_attachments/files/116/098/280/527/538/336/original/d8c49e950d471c81.jpeg