Python in the browser

Enhancing the Scientific Python ecosystem's interoperability with Pyodide

The why

• Python projects and their environments are tricky to install and manage
• For maintainers and core developers: difficult to target various platforms and architectures
• Slightly steep learning curve a euphemism for "learn it yourself like we all did"
• Lower barrier to entry to learning Python, especially for new programmers

XKCD 1987

• Your local computer
• your IDE/text editor/etc.
• Online
• Dev Containers
• Literate programming/notebook environments (Google Colab, Binder, Deepnote, etc.)

The why

Environments where you can run Python

• Documentation for scientific software needs to do more
• Some guidelines
  • Explanatory (descriptive)
  • Illustrative (graphs, diagrams, visual aids, and more)
  • Interactive (a boost to user experience)

https://diataxis.fr/

The why

The why

• In come Pyodide and WebAssembly
  • CPython compiled via the Emscripten toolchain using LLVM to run in the browser
  • A paradigm shift towards code that is written once, and runs everywhere
    • Chromium (Google Chrome, Microsoft Edge, Brave Browser, etc.), Mozilla Firefox, Apple Safari, and even inside browsers on Android and iOS (and iPadOS too)!
  • JupyterLite provides a runtime just like standard Jupyter Notebooks

Literate programming environments

• Google Colab
• Binder
• Deepnote

Interactive programming

• Real-time code execution
• Immediate feedback
• Great for learning and experimentation

Pair programming in the browser

• Collaborative coding
• Shared environment
• Ideal for remote work and education

Limitations of Pyodide

• No threading or multiprocessing
• Some modules from Python standard library are unavailable
• Browser security restrictions – the browser is the most secure piece of software on your system, overall!

Browser-based limitations

• Sandboxed environment
• Limited access to system resources
• Restricted file system access (Emscripten FS, etc.)

Interactive Demo

[Placeholder for chosen demo, unsure what to add here]

• Sudoku implementation using pygame-ce
• JupyterLite notebook-based game
• Treasure hunt with Matplotlib plots
• Interactive data analysis pipeline

32-bit architecture

• Pyodide uses 32-bit compilation
• Impacts floating-point precision
• Affects compatibility with some libraries (off-by-one errors, etc.)

Pyodide's ABI

• Application Binary Interface
• Defines how compiled code interacts with the system
• Crucial for compatibility and stability

Emscripten Version Consistency

• Pyodide maintains a consistent Emscripten version per release
• Ensures  ABI compatibility across releases
• Balances stability and new features

The future of the Pyodide build system

• Pyodide will approach a v1 release
• Expanding possibilities for package integration (build farms such as conda-forge)

Interactive documentation

• Run code examples directly in docs
• Test API functionality in real-time
• Improve learning experience for users

The NumPy docs are now interactive

Enhanced Package Examples

• Live, editable code snippets
• Demonstrate real-world use cases
• Allow users to experiment with parameters

Interactive Tutorials

• Step-by-step guides with live code
• Immediate feedback for learners
• Customizable learning paths

Notebooks in the documentation

• Jupyter-style notebooks embedded in docs
• Combine explanations with executable code
• Great for data science and analysis workflows

Improving overall documentation

• Clearer, more engaging content
• Reduced barrier to entry for new users
• Better understanding of package capabilities

JupyterLite kernels

• Browser-based Jupyter environments
• Run notebooks without a server
• Integrate with Pyodide for Python execution

WebAssembly notebooks

• Notebooks running entirely in the browser
• Combine JavaScript, Python, and other languages
• Enable complex computations client-side

In-tree vs Out-of-tree Builds

• Different approaches to Pyodide package building
• In-tree: Built within the Pyodide repository source
• Out-of-tree: Built separately from source (in your own CI)

Benefits of out-of-tree Builds

• Enhanced robustness across public API since changes are tested with CI
• Better runtime support
• Easier maintenance and updates

My work at Quansight

• Improving out-of-tree build process
• Enhancing compatibility with Pyodide
• Developing tools for package authors

Interactive API Documentation

• Live code execution in API refs
• Test methods and classes in real-time
• Improve developer experience

Pyodide build system

• pyodide-build tool
• Automates package compilation for Pyodide
• Handles dependencies and optimizations

Optimising package sizes

• Important for browser-based distributions on throttled networks
• Techniques for reducing binary size: install tags, not building/separating tests
• This balances functionality and download speed

Build backend improvements

• Enhancing existing build tools
• Creating Pyodide-specific optimizations
• Ensuring cross-platform compatibility

Conclusion

• Pyodide enables new possibilities for Python
• Interactive documentation is the future
• Ongoing work to improve package ecosystem

Thank you!

Questions?