Since last year, I have been trying to piece together my HDR. And I have to admit, it is not easy to write this “short second thesis” (HDR means Habilitation à Diriger des Recherches and is a manuscript that traditionally summarizes one’s work after the PhD). I am simply less motivated to revisit studies that have been published, presented at meetings, and that I consider finalized. But I have to do it regardless of how I feel about it, so I decided to push myself to complete this writing exercise by rewarding myself with a useless, crazy, hobby project: to familiarize myself with a new programming language.
Naturally, two questions need answering: Why would I do that? And which language do I pick? Let me elaborate: for a while now I have been searching for a language that allows for prototyping in few lines of code, yet can also deal with heavy-duty simulations that last a couple of days. Python and Julia fit the first, but I have had some bad experiences with long-running simulations and memory management in Python, and I’ve read about similar memory issues in Julia (see this example). Also, Python I know to a sufficient degree and Julia starts indexing at 1, not 0. In other words, I had enough reasons to look a little further ;-).
In addition to fast prototyping, I would prefer to learn a language that makes it easy to create outreach activities; on top of that, I want to improve my parallel programming skills. That means C with OpenMP and MPI, C++ with HPX, and Rust are not ideal either. They are too low-level or complex. I did consider Chapel for a bit, but decided against that one too. It is very focused on High Performance Computing and therefore not well-suited for outreach activities, though not a bad option for university-level educational purposes.
Making a long story short, I picked Scala, a language developed at EPFL and built on top of the Java Virtual Machine (JVM). Its combination of object-oriented and functional programming should make it suited for rapid prototyping and it supports concurrency and parallelism. The JVM guarantees reasonable bounds on memory use and makes Android a target platform, i.e., outreach activities are possible. Similarly, Scala can be translated to JavaScript, enabling any Internet browser as a platform. According to the Internet its support for functional programming gives it a steeper learning curve, but I appreciate the challenge —for now, that is. Also, I’m aware that it will never be blazingly fast. But if that becomes a requirement, I will talk to the engineers at Inria.
As you can imagine, I have thought of some pilot projects to try in Scala. If all goes well, I hope to share them here in the coming months. Let’s see where these adventures take me.