I am strongly considering buying a copy of your book and giving data-oriented programming a go in my day job. My only hesitation is data immutability. I work in a data-intensive scientific field where we frequently use custom, mutable data structures holding massive amounts of data. Creating persistent versions of these data structures would be an extremely complex undertaking. Do you have any thoughts about using data-oriented programming in situations when data mutability is difficult to avoid?
That’s right. Our data structures are designed for working with large amounts of compressed text (dynamic FM-Index is one example). They are pretty hard to implement so making a persistent version of such a data structure would require a big time commitment. I imagine that my use case is not that unique and there are many situations where mutable data structures are hard to avoid.
I use C++ and Rust. Unfortunately, there are no persistent versions of the data structures I need available in any language (using dynamic FM-index as an example again). It is just interesting to think about “hybrid” systems where most of the code follows principles of data-oriented programming and some of the code deals with any unavoidable mutability.
I wanted to thank all the people from the Clojure community that helped me in the thinking and the writing process.
The book is my attempt to formulate the underlying principles of Clojure approach to data and illustrate how to apply them in another language.
Here are the 4 principles of Data-Oriented Programming:
1. Separate code (behaviour) from data
2. Represent data with generic data structures
3. Data is immutable
4. Separate data schema from data representation