An improved Matlab command line (plus Jupyter Notebooks for Matlab)

The command line is a large part of any Matlab user’s workflow. This vital tool, however, isn’t as user friendly as it should be: it’s cumbersome to recall multi-line commands from the history, there’s no support for Vim key bindings, and there’s no syntax highlighting if using the nodesktop option (on a remote computer, say). Fortunately, there’s an alternative that avoids these problems: IMatlab. This kernel also lets you run Matlab code in Jupyter Notebooks (originally known as IPython Notebooks).

Do I need an alternative?

Using Matlab with the approach laid out below may or may not suit your workflow. Keep reading if you want (i) a customisable, attractive command line, (ii) easy-to-use multi-line history, (iii) Vim key bindings, or (iv) a minimalist Matlab workflow.

Note also that imatlab requires Matlab R2016 or later. (matlab_kernel provides a similar functionality for older versions.)

What is IMatlab?

I’ve been spoilt in the last few years by the Jupyter console (aka the IPython console). This particularly user-friendly tool lead me on a search to find something comparable for Matlab. As it turns out, Matlab can be run as a Jupyter kernel. (You don’t need to know what a ‘Jupyter kernel’ is, by the way).

An IMatlab session is started with the command
jupyter console --kernel imatlab

For me, that session might look something like this

As well as being a command line tool, IMatlab can act as the kernel for a Jupyter Notebook. Let’s redo the session above, but now in a browser, where the code can be combined with documentation:

Using Matlab without opening Matlab

Although IMatlab is a great command line tool, it leads to the obvious question: how do I do everything else I usually do in Matlab?

• To edit m-files, find a better editor
• To open the help browser, type doc
• To open a variable in the variable editor, say x , type openvar('x')
• To show the workspace window, type workspace

Installing and customising IMatlab

Installing IMatlab isn’t too difficult, but there are a few steps. The links below go to a download or a page with the appropriate installation commands.

1. Download and install the Anaconda distribution
   Although its a large download at 300–500MB, the Anaconda distribution greatly simplifies the installation of anything Python related. Be sure to download the Python 3, not 2, version.
   Edit: June 2018
   As of writing, the Matlab engine may not work with Python 3.6, so install with Python 3.5 instead. Also, if you receive errors during step 2 regarding CXXABI versions, see this suggestion.
2. Install the Matlab Engine API for Python
   The instructions on this page include the command
   python setup.py install
   Ensure that the python binary being called is the one associated with the Anaconda distribution, not another installation of python somewhere else on the computer.
3. Install IMatlab itself
   The following worked for me
   pip install imatlab --user
python -mimatlab install --user
   See the IMatlab page for other installation options
4. Customise
   Generate a configuration file with the command
   jupyter console --generate-config
   This will create a file jupyter_console_config.py in ~/.jupyter (Linux and Mac) or %PROGRAMDATA%\jupyter (Windows)
   The newly generated file is full of customisation suggestions, such as choosing Vim bindings over the default Emacs binding.

Author: Ken Hughes

Post-doctoral research scientist in physical oceanography View all posts by Ken Hughes