Work Anywhere With Jupyter

Update: Work Anywhere With Jupyter

With the most recent update to this project, it’s now possible to use Jupyter on your local computer or on the Web, with 100% reproducible results.

Furthermore, the data and notebooks bundled in this project are equally available in both environments, in an interactive format.

The reason for this extreme portability is the relatively new technology of “containerization”. That’s when software is wrapped in an abstract version of what would typically be an entire computer system.

For usage instructions, see the README for this project on GitHub.

Container?

As is the case with many modern server technologies, a Jupyter server can be run within a Docker container, so the exact setup and behaviour of the server is completely reproducible and predictable no matter whether you start it up on Windows, macOS, Linux, or any other operating system that supports Docker.

A great number of cloud services operate in Docker or similar container systems.

On the Web at mybinder.org

Access this repository on mybinder.org right now:

If you visit that page, you’ll see the files of the project laid out in a file browser within the web page. You can click a notebook to open it, and if you run any calculations, the results will appear in the notebook.

How It Works

Under the hood, the website at mybinder.org is using a few open source software utilities to make this possible. One of them is called repo2docker, and its purpose is to create a Jupyter server on demand. It does this with the help of containerization by Docker.

The Upsides of Using mybinder.org

URLs are shareable

Because the whole process is based on web technologies, you can give out a URL to any notebook so collaborators or other interested parties can experience the genuine, predictable interaction opportunities you created.

Protective isolation for multiple viewers

Because the project as seen on mybinder.org is downloaded from your online repository at [GitHub.com], anyone can interact with it, and they’ll each get the same environment.

Because the server is in a container, the project workspace cannot be damaged or corrupted by anyone, even when multiple people open the project simultaneously. Each user can edit, save, delete, etc in their own workspace.

The Downside of Using mybinder.org

Changes Aren’t Propagated

Although you can edit your files through the web interface on mybinder.org, the changes won’t automatically become permanent. You’ll have to export changed files if you want to keep them.

On Your Computer

Just as the magic on mybinder.org is orchestrated by repo2docker, so to can you have your own Jupyter server on your computer.

The Downside of the Local Server Approach

Some Installation Required

Prerequisites:

• Docker
• Python 3

If you don’t have the ability to install software on your computer, this option is not for you.

The Upsides of Using a Local Server Container

Create and Destroy at Will

If you can satisfy the prerequisites, you can easily create and destroy your own Jupyter server at will, with no side effects.

When you need to run multiple servers for separate projects, there are no conflicts, and each server gets a unique URL.

Your Project Files, Editable, Saved Automatically

When editing project files, whether they’re spreadsheets, documentation, or Jupyter notebooks, the files are maintained on your computer in realtime. The Jupyter server is merely accessing them on your behalf, to view, edit, or manage them through the web interface. You have the option of working either way.

When you’re satisfied with your edits, you don’t have to export the project from the web server, because it’s already on your computer. This is something the mybinder.org view of your project won’t do.