This post is a difficult one for me to write. For the past year I have spent hundreds of hours working on a programming language project that, as of this post, I will permanently shelve.
The goal of the project was simple: create a lightweight language for prototyping and embedded scripting that could run without a garbage collector. My hope was that the lack of garbage collection would lead to some interesting design decisions, even if the expressive power of the resulting language and interpreter would be lacking in comparison to languages such as Lua or Python. I figured that automatic reference counting with all data being assigned by value would provide a mechanism for automatic memory management while avoiding the issue of circular references. Data organized in this manner naturally forms a tree which, if implemented as a persistent data structure, would allow for copy-on-write to be used as a way avoid cloning an entire sub-tree with every assignment.
This all sounded pretty good on paper, but when it came to producing a working prototype, days turned into weeks turned into months, and after a year of development, five unsuccessful implementation attempts, and dozens of iterations on the core data model, I could not find an elegant and efficient way to implement my vision of the language. To clarify - it is possible to create a dynamic language without garbage collection: Bone-Lisp and Dyon are just two of the many examples of such languages. But the trade-offs that these languages make, the very same trade-offs that I had to consider in my own language, always left me unsatisfied. Perhaps there is an elegant, simple, and performant way to deliver a zenful Pythonic experience in a language without garbage collection, but if so then I do not think I will be the one to provide that experience.
I have obsessed over this problem for more than twelve months and in the end all I have to show for my efforts is thousands upon thousands of lines of abandoned code. This failure isn't the end of the world, and I feel that it certainly wont be the end of my involvement with programming language design. It just really sucks to put so much work into a project only to feel like it was all for nothing.