Classes that handle non-tabular data structures
Data scientists increasingly receive non-tabular data, often in the form of JSON or XML files. The flexibility of JSON and XML allows organizations to capture complicated relationships between data items in one file. A one-to-many relationship stored in two tables in an enterprise data system can be represented well in JSON by a parent node for the one side and child nodes for data on the many side.
When we receive JSON data we often start by trying to normalize it. Indeed, we do that in a couple of recipes in this book. We try to recover the one-to-one and one-to-many relationships in the data obfuscated by the flexibility of JSON. But there is another way to work with such data, one that has many advantages.
Instead of normalizing the data, we can create a class that instantiates objects at the appropriate unit of analysis, and use the methods of the class to navigate the many side of one-to-many relationships. For example...
