A JIT implementation of the Lemur programming language
Based off the Monkey programming language developed by Thorsten Ball for his book Writing an Interpreter in Go.
Currently, the language runtime (as implemented by Ball) is a tree-walking interpreter, but I plan to expand it and add a JIT native code compiler that can be swapped out with different instruction sets and architectures. I plan to implement the lexer and parser according to the book, and create a module for AST walking, but from there I will add my own code to it. Ball does add a bytecode compiler in the sequel book Writing a Compiler, but I will implement my own backend for the language, possibly using LLVM.
The language itself is also extremely limited. Ball only implements if statements/expressions, variable bindings, functions, and basic data types like lists and maps. This would probably be good enough for a simple scripting language, but I plan to add my own other items to it, such as:
- Import statements (for multi-file projects)
- For and while loops (inexplicably not included in Monkey)
- Classes / Structs (for clustered data and binding methods to them)
- Method calls for classes
- Traits / Interfaces (can be used for operator overloading)
- Sets and Tuples
I have a separate programming language project involving a bytecode compiler and a register-based VM here. Depending on how this language turns out, I might switch that project to just the VM, and add it as a target for this language to compile to.
A standard library may or may not be developed, depending on whether this language reaches maturity.
I may or may not be a bit too optimistic, but I do think that I could make this work.