An Octave kernel for Jupyter

Prerequisites

Jupyter Notebook and GNU Octave.

Installation

To install using pip:

pip install octave_kernel

To install using conda:

conda config --add channels conda-forge
conda install octave_kernel
conda install texinfo # For the inline documentation (shift-tab) to appear.

We require the octave-cli or octave executable to run the kernel. Add that executable's directory to the PATH environment variable or create the environment variable OCTAVE_EXECUTABLE to point to the executable itself. Note that on Octave 5+ on Windows, the executable is in "Octave-x.x.x.x\mingw64\bin".

We automatically install a Jupyter kernelspec when installing the python package. This location can be found using jupyter kernelspec list. If the default location is not desired, remove the directory for the octave kernel, and install using python -m octave_kernel install. See python -m octave_kernel install --help for available options.

Usage

To use the kernel, run one of:

jupyter notebook  # or ``jupyter lab``, if available
# In the notebook interface, select Octave from the 'New' menu
jupyter qtconsole --kernel octave
jupyter console --kernel octave

This kernel is based on MetaKernel, which means it features a standard set of magics (such as %%html). For a full list of magics, run %lsmagic in a cell.

A sample notebook is available online.

Configuration

The kernel can be configured by adding an octave_kernel_config.py file to the jupyter config path. The OctaveKernel class offers plot_settings, inline_toolkit, kernel_json, and cli_options as configurable traits. The available plot settings are: 'format', 'backend', 'width', 'height', 'resolution', and 'plot_dir'.

cat ~/.jupyter/octave_kernel_config.py
# use Qt as the default backend for plots
c.OctaveKernel.plot_settings = dict(backend='qt')

The path to the Octave kernel JSON file can also be specified by creating an OCTAVE_KERNEL_JSON environment variable.

The command line options to Octave can also be specified with an OCTAVE_CLI_OPTIONS environment variable. The cli options be appended to the default options of --interactive --quiet --no-init-file. Note that the init file is explicitly called after the kernel has set more off to prevent a lockup when the pager is invoked in ~/.octaverc.

The inline toolkit is the graphics_toolkit used to generate plots for the inline backend. It will default to whatever backend octave defaults to. The different backend can be used for inline plotting either by using this configuration or by using the plot magic and putting the backend name after inline:, e.g. plot -b inline:fltk.

Supported Platforms

The octave_kernel supports running on Linux, MacOS, or Windows. On Linux, it supports Octave installed using apt-get, flatpak, or snap. There is no additional configuration required to use flatpak or snap.

Troubleshooting

Kernel Times Out While Starting

If the kernel does not start, run the following command from a terminal:

python -m octave_kernel.check

This can help diagnose problems with setting up integration with Octave. If in doubt, create an issue with the output of that command.

Kernel is Not Listed

If the kernel is not listed as an available kernel, first try the following command:

python -m octave_kernel install --user

If the kernel is still not listed, verify that the following point to the same version of python:

which python  # use "where" if using cmd.exe
which jupyter

Qt Backend Availability

In some cases, the qt graphics toolkit is only available when running with a display enabled.

On a remote system without a display, you can use xvfb-run to provide a virtual framebuffer. For example:

export OCTAVE_EXECUTABLE="xvfb-run octave"

Or in the config file:

c.OctaveKernel.executable = "xvfb-run octave"

JupyterHub with Qt Support

To enable Octave's Qt graphics toolkit in a JupyterHub environment (or any headless server), you need a virtual display. Install the required system packages:

apt-get install -y octave libglu1 xvfb texinfo fonts-freefont-otf ghostscript

Start Xvfb before launching JupyterHub (or in a server startup script):

Xvfb :99 -screen 0 1024x768x24 &
export DISPLAY=:99

Then configure the kernel to use the Qt backend:

# ~/.jupyter/octave_kernel_config.py
c.OctaveKernel.plot_settings = dict(backend="qt")

For Binder-based deployments, place a start script in the binder/ directory to launch Xvfb and export DISPLAY before the server starts, and list the required packages in binder/apt.txt.

Blank Plot

Specify a different format using the %plot -f <backend> magic or using a configuration setting. On some systems, the default 'png' produces a black plot. On other systems 'svg' produces a black plot.

Local Installation

To install from a git checkout run:

pip install -e .