Home All Posts

Interpreter Planning

2018-06-10

In the past few weeks I’ve been continuing on my “learning plan”, slowly, at that. I’m slowly building the energy to relearn some of the math I once knew. I’ve been working through my old college text book on Algebra, and after that I’ll keep going with the Pre-Calculus and Calculus books. While doing that, before tackling “Concrete Mathematics” by Knuth and others, I’ll also go through a book on Discrete Math that I have.

I’ve also made some progress by building up my version of a Monkey interpreter in C#. A lot of this was simply translating code from Golang to C# from the book, although I handled some things in what I would say is more of a C# way of doing things (which was inspired by the techniques of a blog post by Bob Nystrom and his implementation of a Pratt parser). I still have a bit to do to bring my Monkey interpreter up to spec with the one implemented by Thorsten Ball in his book, but thats just a matter of a little effort and time.

Building my Monkey interpreter and working through the book “Writing an Interpreter in Go” solidified my interest in programming languages and inspired a bit of a book buying. What I quickly found by implementing Monkey was that implementing a programming language and interpreter is no small undertaking. While coding the Monkey interpreter didn’t take too long, it quickly became apparent that it would take a lot of effort to implement anything non trivial. I also discovered that I didn’t want to simply implement a language which was just an interpreted version of what was already out there.

So my impelemtation of Monkey won’t expand too much beyond what Ball has implemented in the book. I do plan on adding classes and a few other things which aren’t present in the book, however this should be fairly trivial. I’m also planning on implementing a byte code interpreted version of Monkey which will be executed by a Stack based VM. Implementing a stack VM will provide a foundation for future language projects as well. I don’t quite think it’s worth the effort to go all out and compile to machine code, however that may be an option in the future.

In the meantime while I’m implemeting a bytecode interpreter for Monkey, I decided I should learn a little bit more about techniques for building interpreters. I also wanted to expand on my knowledge of some of the techniques and practices which are found in functional programming languages. This way, when I finally get around to my next interpreter project, I’ll have more information on what features I would like to put into the language.

So I’ve bought a number of books to start my learning joureny (as I normally do). I think it’s going to take years to get through all these books and for some of them I’m going to have to buckle down and learn some new things (including some mathematical notation). The books I picked up (some a while ago) include:

Reading and working through these books while also solidifying my math skills is certainly going to take some time. My guess is that it’s going to take between 2-3 years if not more doing it at night and on the weekends, especially taking into account my Netflix addiction. But I’m sure that learning this stuff will provide ample amounts of information and ideas for what to include in my interpreter. The only thing that I know for sure right now is that it will be a bytecode interpreter and not a simple tree-walker. The performance gain by using a bytecode interpreter should be substantial even if it is more work to implement.

Based upon what I’ve found so far I think that finding a reason for the language I’ll ultimately design to exist will be important. I don’t think it’s pointless to design and implement something solely for educationaly purposes, however, it would be nice to build something with some direction so that it would continue to be useful. My initial thoughts are to design a language that would be a good language for experimenting with different types of algorithms and data structures. It would also be nice if it could serve some practical purpose.

Despite the fact that it will probably be over a year or so before any serious design is done, on the way I’ll be learning lots of new and useful things. I also think that I’ll take the information I learn on functional programming through these books and projects to learn and write a book about F#.

In the meantime, I’ll keep going on learning new math skills. I’ve already started to read about Standard ML and it’s exciting to learn a non-brackets language like C# or C. I’m also learning Emacs so that I’ll have some environment for writing SML in. It should be a fun project that I’ll try to document here on my blog as I go for anyone interested in following along.