Hi HN!

Last year, we shared marimo [1], an open-source reactive notebook for Python with support for execution through WebAssembly [2].

We wanted to share something new: you can now run marimo and Jupyter notebooks directly from GitHub in a Wasm-powered, codespace-like environment. What makes this powerful is that we mount the GitHub repository's contents as a filesystem in the notebook, making it really easy to share notebooks with data.

All you need to do is prepend 'marimo.app' to any Python notebook on GitHub. Some examples:

- Jupyter Notebook: https://marimo.app/github.com/jakevdp/PythonDataScienceHandb...

- marimo notebook: https://marimo.app/github.com/marimo-team/marimo/blob/07e8d1...

Jupyter notebooks are automatically converted into marimo notebooks using basic static analysis and source code transformations. Our conversion logic assumes the notebook was meant to be run top-down, which is usually but not always true [3]. It can convert many notebooks, but there are still some edge cases.

We implemented the filesystem mount using our own FUSE-like adapter that links the GitHub repository’s contents to the Python filesystem, leveraging Emscripten’s filesystem API. The file tree is loaded on startup to avoid waterfall requests when reading many directories deep, but loading the file contents is lazy. For example, when you write Python that looks like

```python

with open("./data/cars.csv") as f: print(f.read())

# or

import pandas as pd pd.read_csv("./data/cars.csv")

```

behind the scenes, you make a request [4] to https://raw.githubusercontent.com/<org>/<repo>/main/data/car....

Docs: https://docs.marimo.io/guides/publishing/playground/#open-no...

[1] https://github.com/marimo-team/marimo

[2] https://news.ycombinator.com/item?id=39552882

[3] https://blog.jetbrains.com/datalore/2020/12/17/we-downloaded...

[4] We technically proxy it through the playground https://marimo.app to fix CORS issues and GitHub rate-limiting.

> CORS and GitHub

The Godot docs mention coi-serviceworker; https://github.com/orgs/community/discussions/13309 :

gzuidhof/coi-serviceworker: https://github.com/gzuidhof/coi-serviceworker :

> Cross-origin isolation (COOP and COEP) through a service worker for situations in which you can't control the headers (e.g. GH pages)

CF Pages' free unlimited bandwidth and gitops-style deploy might solve for apps that require more than the 100GB software cap of free bandwidth GH has for open source projects.

I absolutely love that this can be hosted on Github Pages. Am I correct in understanding that these notebooks will run independently, and will not need to proxy through marimo.app (in case the app goes down), or is that what the CORS thing is about in note 4, and it will still need to go through this domain?

Nice ! Is it possible to connect to an in browser DB like WASM DuckDB https://duckdb.org/docs/api/wasm/overview.html or https://github.com/babycommando/entity-db ?

That would be most useful imho !

Super cool to see a real use-case of WASM outside of just game dev and nerding out.

This is an awesome, and as a bonus I learned about a mature reactive notebook for python. Great stuff.

The data sharing is awesome. I previously used Google Colab to share runnable code with non-dev coworkers, but their file support requires some kludges to get it working decently.

I know I should just RTFM, but are you all working on tools to embed/cross-compile/emulate non-python binaries in here? I know this is not a good approach, but as a researcher I would love to shut down my server infrastructure and just use 3-4 crusty old binaries I rely on directly in the browser.

I love seeing projects like this. When Pyiodide came out I was excited but it was a bit difficult to use, this looks and feels fantastic.

I really like Observable as well, but I've found it difficult to find robust and broad numerical libraries in javascript like what Python has.

I would love for this type of tool to redefine how we do science. It would be amazing if many scientific papers included both their data and the code in an interactive environment with zero installs and configuration. Plus when discussing a paper you could "fork" it and explore different analysis options live which for many fields would be totally feasible to do in the browser.

> [ FUSE to GitHub FS ]

> Notebooks created from GitHub links have the entire contents of the repository mounted into the notebook's filesystem. This lets you work with files using regular Python file I/O!

Could BusyBox sh compiled to WASM (maybe on emscripten-forge) work with files on this same filesystem?

"Opening a GitHub remote with vscode.dev requires GitHub login? #237371" ... but it works with Marimo and JupyterLite: https://github.com/microsoft/vscode/issues/237371

Does Marimo support local file system access?

jupyterlab-filesystem-access only works with Chrome?: https://github.com/jupyterlab-contrib/jupyterlab-filesystem-...

vscode-marimo: https://github.com/marimo-team/vscode-marimo

"Normalize and make Content frontends and backends extensible #315" https://github.com/jupyterlite/jupyterlite/issues/315

"ENH: Pluggable Cloud Storage provider API; git, jupyter/rtc" https://github.com/jupyterlite/jupyterlite/issues/464

Jupyterlite has read only access to GitHub repos without login, but vscode.dev does not.

Anyways, nbreproduce wraps repo2docker and there's also a repo2jupyterlite.

nbreproduce builds a container to run an .ipynb with: https://github.com/econ-ark/nbreproduce

container2wasm wraps vscode-container-wasm: https://github.com/ktock/vscode-container-wasm

container2wasm: https://github.com/ktock/container2wasm

This is really cool -- going to show it off to my team. I love the fact that you opened it up so that it will work with Jupyter notebooks as well.

The future is awesome. Thanks for building this!

wow a python interpreter is "only" 100MB not sure if that's what's happening here