gapdata/nbviewer.js

Client side rendering of Jupyter notebooks

client side rendering of jupyter notebooks

tl;dr: Render Jupyter notebooks straight in the browser, without a back end converter. Can be used as a library.

I often want to read through my Jupyter notebooks, but I rarely have my Jupyter instances running in the right folders. I can't quite use the online nbviewer, because I don't have a public URL for these, so I resort to running dummy Jupyter instances or uploading my file as a one-time gist on Github (one I have to delete thereafter). One last possibility is nbconvert in the command line.

I thought it could be easier and more lightweight. So I hacked together this client side rendering of Jupyter notebooks. All you need is a browser that renders an HTML file (dependent on some JavaScript and CSS, both of which can be inlined). You simply drag and drop an .ipynb file and it renders. It's rather fast and it supports most notebook features. I tried supporting the previous version (v3), since there are tons of examples in this version out there.

Try a live demo

Usage

There are two ways one can use this. You can use the library itself, there is just a single public method, you call nbv.render(data, target), where data is the JSON representation of your Jupyter notebook and target is the node where the notebook is to be rendered.

Or you can use the demo (or a local copy), which is just a simple wrapper of the library, with dropzones and other basic features. There is no data being transferred anywhere, so feel free to bookmark it and use it.

Two immediate use cases come to mind:

1. In Mac OS, you can preview files and this system, Quick Look, supports plugins. It would be rather handy to have a notebook preview one keystroke away.
2. Ever since GitHub introduced notebook rendering last year, Gitlab users have been requesting the same. Gitlab itself recently dropped Python as a dependency, so its reintroduction just for converting notebooks is rather unlikely. A browser-based solution like this could be a good substitute.

Tech details

It's rather simple at this point, all the DOM manipulation is written in vanilla JavaScript, Markdown rendering goes through marked.js, syntax highlighting is administered by Prism.js. The example implementation leverages a few goodies from modern web design, like File API or drag&drops, so a fairly modern browser is necessary.

Showcase

Contact

Drop me an email ([email protected]) or tweet at me (@pndrej) if you have any questions or suggestions. Contributions welcome.