Warning

THESE PAGES HAVE MOVED TO swarm.magneticearth.org

About Jupyter

Note

This page provides some guidance regarding Jupyter. For a full guide and reference, you should take a look at the JupyterLab docs (though by using the Swarm VRE, you can ignore the sections about installing Jupyter yourself).

Terminology

Jupyter can be confusing! So here is a handy list of terms:

Project Jupyter

The broader open source project and community.

https://jupyter.org/

Jupyter Notebook

A notebook document (file extension .ipynb) is a document that can be rendered in a web browser that can contain text, mathematics, images, live code, and other interactive elements. “Jupyter Notebook” can also refer to the web application that builds and runs such notebooks.

https://jupyter-notebook.readthedocs.io/

JupterLab

This is the evolution of the Jupyter Notebook software providing a richer interface. It is the default interface for the VRE when you first login.

https://jupyterlab.readthedocs.io/

JupyterHub

This software runs on servers to provide a multi-user system which automatically spawns and manages instances of JupyterLab/Notebook

https://jupyterhub.readthedocs.io/

Jupyter kernel
When used with reference to Jupyter, a “kernel” refers to the program which executes code cells within a notebook. Only one kernel at a time can be connected to a notebook. In the VRE we currently provide two kernels: Python 3.7 (with a number of packages installed), and Octave.
containers, images, and Docker

A container image is a unit of software that holds everything needed to run some particular code. An image is loaded into a container, which is an isolated environment where your code can run with a dependable environment based on the image - this is a bit like a virtual machine. Docker is a software system for working with such containers. The environment that a particular image provides is specified by a Dockerfile (e.g. jupyter/scipy-notebook).

We employ our own Docker image which is built on top of jupyter/scipy-notebook, providing access to JupyterLab and a selection of software relevant to Swarm.

Quick tips

There are some quirks about the behaviour of Jupyter that can be surprising and confusing:

  • There is often effectively two right-click menus. To access the other one, hold down shift and then right-click

    • This might vary between browsers and systems (?)
    • This is most noticeable when trying to copy and paste

  • When using the terminal, Ctrl+c and Ctrl+v will not copy and paste. The keyboard shortcuts for copy and paste are Ctrl+Insert and Shift+Insert (this is a normal Unix thing)

  • When using notebooks:

    • Press Ctrl+Enter to execute a cell
    • Press Shift+Enter to execute a cell and advance to the next one
    • There are many more keyboard shortcuts to learn that can help to manipulate notebooks faster
    • Be aware that the execution state of your notebook (i.e. the variables created/changed etc.) is affected by all the cell executions you have made. It is good practice to occasionally refactor and re-run cells in the correct order, and from a fresh un-executed state, as you are developing a notebook, so as to ensure that the notebook outputs will be the same when it is run from scratch. A common way to achieve this is to go to the menu, to Kernel / Restart Kernel and Run All Cells
    • You can switch which kernel the notebook is using to run code, by clicking on the kernel name e.g. “Python 3” at the top right of the notebook.

  • The default view when logging in is to load the “JupyterLab” interface. To get the old single-notebook view, go to the menu, to Help / Launch Classic Notebook. (This can be necessary for compatibility with some extensions which have not yet been ported to JupyterLab) This is simply a different view for interacting with the system - the software is otherwise identical.

  • Installing things, e.g. using pip or conda won’t behave the same as on a normal machine. See Software available for more.

  • If you find other things, or otherwise struggling to achieve some task, please contact us.