We’re happy to announce the release of the wasmtime Ruby gem, the official embedding of Wasmtime for Ruby.

Goal

The initial use case for this came from Shopify – a commerce platform built on Ruby and the company sponsoring this work. Shopify is heavily investing in making commerce extensible. Among other things, it uses WebAssembly to run untrusted code on the server. All that Wasm code must run fast and securely.

The goal of the gem is to allow exposing all of Wasmtime’s power with minimal overhead. The gem’s API should have a nice Ruby feel to it – we think it does! – but it intentionally does not introduce new abstractions or DSLs. These can be added by the community through new gems.

To best reach the project’s goal, we’ve built a Rust-based native Ruby extension using Wasmtime’s Rust API.

Wait – Rust?

Why pick Rust? you may ask, knowing Ruby native extensions are normally written in C? That is a fair question, let’s dive into it.

First, Wasmtime’s Rust API is the canonical one. It is more expressive — thanks to Rust — and tends to get the features first. Closing the gap between Rust’s and C’s API would have increased the scope of work greatly.

Second, Rust is gaining in popularity in the Ruby community, increasing the pool of potential contributors. YJIT (Ruby’s latest JIT compiler) uses Rust and is now considered stable as of Ruby 3.2. RubyGems also added support for scaffolding Rust-based native extensions using the Magnus crate, which makes writing Rust-based Ruby native extensions a breeze.

Finally, Rust helps build more secure and robust programs thanks to its type system and the borrow checker. That’s one of the many reasons why Shopify chose Rust as its systems programming language.

The main drawback of using Rust is that the compiler is not widely available as C’s. This drawback is mitigated by shipping precompiled libraries for macOS, Linux and Windows. Most users won’t even need to know it’s written in Rust! All in all, Rust is a better fit to help us reach our goal and lower the long-term maintenance cost.

How to use it?

Assuming you have a supported Ruby version (we recommend 3.0+) and Bundler, you’re good to go. We’ll use Bundler to install the gem:

bundle init && bundle add wasmtime

We can now use Wasmtime from Ruby. Let’s write a simple program as an example.

# Create an engine. Generally, you only need a single engine
# and can re-use it throughout your program.
engine = Wasmtime::Engine.new

# Compile a Wasm module from either Wasm or WAT.
mod = Wasmtime::Module.new(engine, <<~WAT)
  (module
    (func $hello (import "" "hello"))
    (func (export "run") (call $hello))
  )
WAT

# Create a store with an optional arbitrary object.
store_data = { count: 0 }
store = Wasmtime::Store.new(engine, store_data)

# Define a Wasm function from Ruby code.
# The function has no params ([]) and no results ([]).
func = Wasmtime::Func.new(store, [], []) do |caller|
  puts "Hello from Func!"
  caller.store_data[:count] += 1
  puts "Ran #{caller.store_data[:count]} time(s)"
end

# Build the Wasm instance by providing its "hello" import.
instance = Wasmtime::Instance.new(store, mod, [func])

# Run the `run` export.
instance.invoke("run")

Running the above will print:

Hello from Func!
Ran 1 time(s)

If you’ve seen Wasmtime’s Rust API, this should all feel very familiar!

What’s next

While we’re pretty happy with the gem, it is still early. There is still room for reducing the performance overhead of the gem and some missing features.

If you find a missing feature you would want, something slow, a bug, or any other kind of improvements, come and talk to us on GitHub!