Small experiments with dynamic library loading across different languages.
Compiles a rust project for loading into ruby, python, and javascript.
Build the rust project
$ cargo build --release
Run the python program:
$ python embed.py
Run the ruby program
$ ruby embed.rb
Run the js program
node js/index.js
Rust and it's toolchain are a joy. Add a crate-type=["dylib"] to the config, and you can use the normal cargo toolchain without any other modifications.
Ruby is always magic. Using the 'ffi' gem, you can attach functions from a dynamic library to a module. Pretty simple, didn't have much trouble getting it going.
Python is well known for its compatibility with C and C++ libraries, and it lived up to this reputation. ctypes from the standard library made loading the libraryand calling a function pretty painless.
Node was a pain relative to the others. node-gyp, while a cool tool, spits out a ton of warnings on install. The code that I had also broke when I upgraded node versions, due to some package being compiled for a different version of node. This may just be me being grumpy, but it was annoying that I had to have a package.json, node_modules folder, and package-lock.json in addition to a js file in order to do a fancy console log.
--
All told, loading and running code across a language boundary was actually surprisingly easy. I got all three high-level languages running my dumb rust code relatively quickly and painlessly. It makes me think that if I had something that really needed optimizing, I shouldn't be scared to reach for a lower level language and break out the ffi, so long as I am working on a project where I can own the build process like that.
I hope that this will encourage me to do more FFI stuff in the future. There are times when I want to explore some interesting problem in Rust, but also want the ease of solving parts of it with a high-level language. Maybe FFI can be a bridge, so that I can solve each part with a tool at the appropriate level.