Custom VM project. Should evolve into a full programming environment some day.
After watching this Talk, I started looking through some of the languages mentioned in the talk and realized, that I do not like the syntax of those languages. I wanted something more "traditional" to play around with the ideas in the talk. Also, some of the languages have really "heavy" runtimes you have to use.
So I decided to start building my own dynamic programing environment. I want it to have some of the nice things that smalltalk and lisp have, with their image based approach, but with enough flexibility to do further research on such topics.
I also thought about syntax quite a bit and decided to try a completely different approach. I want to merge the REPL/image-based approach from lisp, with structural/projectional editing.
For that I want to define the basic structures of a program just in code and build a UI to edit this structure directly.
It's also planned to build that UI dynamically based on some sort of templating or ui language, tailored to building projectional editors.
Currently I have the necessary data structures to define functions, variables, constants, signed integers and a few binary operators.
I have enough functionality to implement a recursive version of factorial, which can be seen in serialized format here.
To try it out, you can do this:
cargo run -- run factorial.ron "init=5&n=10"Which should print out the following:
Scalar(I32(120))
Usage: vorm <COMMAND>
Commands:
create Create an example ron file
run Run a ron or vmstate file
debug Deserialize the vm and print the whole vm on stdout
compile Compile a ron to a vmstate file
pretty Prettify a ron file
help Print this message or the help of the given subcommand(s)
Options:
-h, --help Print help
Pretty fast actually. The first benchmark (done with commit 4a87d9fdc3ae6f5cd310460c453af09eb13e7178) shows, we are already in a similar ballpark as python, for a simple benchmark.
I'll probably not benchmark too much now, as there are more important features right now. If someone knows how to improve performance without restricting whats possible too much, feel free to submit a PR.
The benchmark result can be found at bench/README.md.
The python code that I used for the benchmark can be found at bench/py/fac.py.
Here is also a small flamegraph:
Vormbaar means moldable in dutch, which captures the spirit of the project.