evolbug
commented
4 years ago Hey, this is a good concept, however I see that what you are currently building is a tree based interpreter. Unfortunately, this by definition will be slower than even unoptimized Lua code. Afterall, you are going up against two decades of professional optimization and architecture. There's LuaJIT too which is one of the top performing scripting language VMs.
The initial concept of Wu was a bytecode interpreter, but we opted to target Lua due to the ecosystem available, as we wanted to target games. The ability to use Love2D is a big sell.
Regardless, the original VM is still available in the vm-version branch, you can try to get that back up and running if you want, it's 99% there, I think we stopped at FFI and some scope bugs, I don't remember exactly :smile:
cedric-h
This PR adds a standalone interpreter to Wu, so that you don't have to have luac to run Wu programs. Because we offer finer grained control that isn't always reflected in Lua, having our own interpreter gives us the opportunity to run Wu code with much higher performance. Also, writing interpreters is just fun.
Interaction with the existing code: I butchered src/main.rs to make the interpreter accessible from the command line. I also added some convenience methods to parser::ast::Expression, so that I can easily convert Expressions into their base values. I was able to leave everything else alone.
Progress so far: The Interpreter isn't entirely usable on its own yet, but you can call external Rust functions and evaluate simple binary operators between floats, create user defined functions and cast floats to strings.
Going forward: Now that a template is in place for the Binary operations and casting, all that needs to be done is for these patterns to be extended with more types. We would probably benefit from having a better way to provide Rust functions to the Wu evaluator as it's running, zesterer provided a very good example with his Forge engine construct. If and While are also crucial additions, as well as Block processing, but the majority of the heavy lifting required for these to happen is already in place.