Pandas is a key element in our dataviz toolchain, as we will use it for both cleaning and exploring our recently scraped dataset (see Chapter 6). The last chapter introduced NumPy, the Python array processing library that is the foundation of Pandas. Before we move on to applying Pandas, this chapter will introduce its key concepts and show how it interacts with existing data files and database tables. The rest of your Pandas learning will be on the job over the next couple of chapters.
Take any dataviz, whether web-based or in print, and chances are that the data visualized was at one point stored in row-columnar form in a spreadsheet like Excel, a CSV file, or HDF5. There are certainly visualizations, like network graphs, for which row-columnar data is not the best form, but they are in the minority. Pandas is tailor-made to manipulate row-columnar data tables with its core datatype, the
DataFrame, which is best thought of as a very fast, programmatic spreadsheet.
First revealed by Wes Kinney in 2008, Pandas was built to solve a particular problem—namely, that while Python was great for manipulating data, munging it, and preparing it, it was weak in the area of data analysis and modeling, certainly compared with big hitters like R.
Pandas is designed to work with heterogeneous data like that found in row-columnar spreadsheets, but cleverly manages to leverage some of the speed ...