Wasmer Python

A complete and mature WebAssembly runtime for Python based on Wasmer.

Features

• Secure by default. No file, network, or environment access, unless explicitly enabled.
• Fast. Run WebAssembly at near-native speeds.
• Compliant with latest WebAssembly Proposals (SIMD, Reference Types, Threads, ...)

Documentation: browse the detailed API documentation full of examples.

Examples and tutorials: browse the examples/ directory, it's the best place for a complete introduction!

Install

To install the wasmer Python package, and let's say the wasmer_compiler_cranelift compiler, just run those commands in your shell:

$ pip install wasmer wasmer_compiler_cranelift

Usage

from wasmer import engine, Store, Module, Instance

store = Store()

# Let's compile the module to be able to execute it!
module = Module(store, """
(module
  (type (func (param i32 i32) (result i32)))
  (func (export "sum") (type 0) (param i32) (param i32) (result i32)
    local.get 0
    local.get 1
    i32.add))
""")

# Now the module is compiled, we can instantiate it.
instance = Instance(module)

# Call the exported `sum` function.
result = instance.exports.sum(5, 37)

print(result) # 42!

And then, finally, enjoy by running:

$ python examples/appendices/simple.py

We highly recommend to read the examples/ directory, which contains a sequence of examples/tutorials. It's the best place to learn by reading examples.

Quick Introduction

The wasmer package brings the required API to execute WebAssembly modules. In a nutshell, wasmer compiles the WebAssembly module into compiled code, and then executes it. wasmer is designed to work in various environments and platforms: From nano single-board computers to large and powerful servers, including more exotic ones. To address those requirements, Wasmer provides 2 engines and 3 compilers.

Succinctly, an engine is responsible to drive the compilation and the execution of a WebAssembly module. By extension, a headless engine can only execute a WebAssembly module, i.e. a module that has previously been compiled, or compiled, serialized and deserialized. By default, the wasmer package comes with 2 headless engines:

1. wasmer.engine.Universal, the compiled machine code lives in memory,
2. wasmer.engine.Native, the compiled machine code lives in a shared object file (.so, .dylib, or .dll), and is natively executed.

Because wasmer does not embed compilers in its package, engines are headless, i.e. they can't compile WebAssembly module; they can only execute them. Compilers live in their own standalone packages. Let's briefly introduce them:

Compiler package	Description	PyPi
wasmer_compiler_singlepass	Super fast compilation times, slower execution times. Not prone to JIT-bombs. Ideal for blockchains
wasmer_compiler_cranelift	Fast compilation times, fast execution times. Ideal for development
wasmer_compiler_llvm	Slow compilation times, very fast execution times (close to native, sometimes faster). Ideal for Production

We generally recommend wasmer_compiler_cranelift for development purposes and wasmer_compiler_llvm in production.

Learn more by reading the documentation of the wasmer.engine submodule.

Supported platforms

We try to provide wheels for as many platforms and architectures as possible. For the moment, here are the supported platforms and architectures:

Platform	Architecture	Triple	Packages
Linux	amd64	x86_64-unknown-linux-gnu	wasmer	✅
			wasmer_compiler_singlepass	✅
			wasmer_compiler_cranelift	✅
			wasmer_compiler_llvm	✅
	aarch64	aarch64-unknown-linux-gnu	wasmer	✅
			wasmer_compiler_singlepass	❌ 1
			wasmer_compiler_cranelift	✅
			wasmer_compiler_llvm	✅
Darwin	amd64	x86_64-apple-darwin	wasmer	✅
			wasmer_compiler_singlepass	✅
			wasmer_compiler_cranelift	✅
			wasmer_compiler_llvm	✅
Windows	amd64	x86_64-pc-windows-msvc	wasmer	✅
			wasmer_compiler_singlepass	✅
			wasmer_compiler_cranelift	✅
			wasmer_compiler_llvm	❌ 2

Notes:

• ¹ wasmer_compiler_singlepass does not support aarch64 for the moment
• ² wasmer_compiler_llvm is not packaging properly on Windows for the moment

Wheels are all built for the following Python versions:

• Python 3.7,
• Python 3.8.
• Python 3.9.
• Python 3.10,

Learn about the “fallback” py3-none-any wheel

py3-none-any.whl

A special wasmer-$(version)-py3-none-any wheel is built as a fallback. The wasmer library will be installable, but it will raise an ImportError exception saying that “Wasmer is not available on this system”.

This wheel will be installed if none matches before (learn more by reading the PEP 425, Compatibility Tags for Built Distributions).

Development

The Python extension is written in Rust, with pyo3 and maturin.

First, you need to install Rust and Python. We will not make you the affront to explain to you how to install Python (if you really need, check pyenv). For Rust though, we advise to use rustup, then:

$ rustup install stable

To set up your environment, you'll need just, and then, install the prelude of this project:

$ cargo install just
$ just --list # to learn about all the available recipes
$ just prelude

It will install pyo3 and maturin for Python and for Rust. It will also install virtualenv.

Then, simply run:

$ source .env/bin/activate
$ just build api
$ just build compiler-cranelift
$ python examples/appendices/simple.py

Testing

To build all tests you'll need LLVM 12.0 in your system. We recommend either installing prepackaged libraries with llvm.sh or building it yourself with llvmenv.

Build all the packages and run the tests:

$ just build-all
$ just test

What is WebAssembly?

Quoting the WebAssembly site:

WebAssembly (abbreviated Wasm) is a binary instruction format for a stack-based virtual machine. Wasm is designed as a portable target for compilation of high-level languages like C/C++/Rust, enabling deployment on the web for client and server applications.

About speed:

WebAssembly aims to execute at native speed by taking advantage of common hardware capabilities available on a wide range of platforms.

About safety:

WebAssembly describes a memory-safe, sandboxed execution environment […].

License

The entire project is under the MIT License. Please read the LICENSE file.