jupyterlab-contrib / jupyterlab-filesystem-access

[Experimental] Browse local files using the non-standard File System Access API
https://jupyterlab-filesystem-access.readthedocs.io
BSD 3-Clause "New" or "Revised" License
30 stars 6 forks source link

S3 bucket access #63

Open juangburgos opened 4 months ago

juangburgos commented 4 months ago

Just want to start by saying thank you for this extension, works flawlessly. IMO it should be bundled by default with jupyterlite, because it is such a basic functionality to be able to persist the files outside of the browser.

On the same lines, I'd like to propose a similar extension, but instead of using the Web Broswer File System Access API, use the AWS Javascript SDK to persist files in the cloud. This would open up so many possibilities for remote work and sharing, plus now a days AWS is not the only S3 provider, which makes the S3 standard a universal language for cloud storage.

Thanks again.

krassowski commented 4 months ago

Like this one: https://github.com/jpmorganchase/jupyter-fs ?

martinRenou commented 4 months ago

I guess jupyter-fs requires a running server for it to work?

juangburgos commented 4 months ago

Yes, there are already jupyter extensions that allow to do what I proposed in a regular jupyter instance. But they are coded in python which in turn uses TCP sockets, which of course are not available in the web environment. If you try to use them in jupyterlite, you will get errors like:

ImportError: cannot import name 'ssl' from 'urllib3.util.ssl_' (/lib/python3.12/site-packages/urllib3/util/ssl_.py)

So this needs to be coded in javascript which can actually send the http requests required to comunicate with the cloud.