steverweber/imatlab

A Jupyter kernel for MATLAB

This kernel requires Jupyter with Python 3.5+, and the MATLAB engine for Python R2016b+ (this release provides a much better completion API), which needs to be installed first.

Install with python -mpip install imatlab (from PyPI) or python -mpip install git+https://github.com/anntzer/imatlab (from Github); then run python -mimatlab install to register the kernel spec. In the absence of administrator rights, the --user flag should be added to any of these commands.

To use it, run one of:

$ jupyter notebook
# In the notebook interface, select Matlab from the 'New' menu
$ jupyter qtconsole --kernel imatlab
$ jupyter console --kernel imatlab

Inline Graphics

To use Plotly inline graphics in the notebook, the Plotly MATLAB API must be installed.

Clone plotly/MATLAB-Online or download the zip.
Recursively add the resulting extracted folders to the MATLAB path: addpath(genpath(<Plotly MATLAB API path>)).
In MATLAB, run: getplotlyoffline('https://cdn.plot.ly/plotly-latest.min.js') to copy the javascript files.
Set the IMATLAB_EXPORT_FIG environment variable to use fig2plotly as described below.

Environment variables

IMATLAB_CONNECT

If this environment variable is set to a valid MATLAB identifier, the kernel will attempt to connect to the shared engine with that name. If it is set to another non-empty value, it will connect to any existing shared engine.

IMATLAB_CD

If this environment variable is set, the engine's working directory will be changed to match the kernel's working directory.

IMATLAB_EXPORT_FIG

This environment variable can be set to a MATLAB expression, representing a function of one argument (either a quoted function name, or a function handle). If it is, and the current frontend is a notebook, then, after each input is evaluated, the given function is evaluated for each figure handle, while the current folder is changed to a temporary folder. Any .html, .png or .jpeg file that is created is displayed using IPython's display_data mechanism (.html files will also trigger the initialization of plotly’s notebook mode.).

For example, set (e.g., from MATLAB, in startup.m):

setenv('IMATLAB_EXPORT_FIG', func2str(@(h) eval([ ...
   'fig2plotly(h, ''filename'', [tempname(''.''), ''.html''], ', ...
                 '''offline'', true, ''open'', false); ', ...
   'close(h);'])));

to export figures as html with plotly; or set:

setenv('IMATLAB_EXPORT_FIG', func2str(@(h) eval([ ...
    'print(h, [tempname(''.''), ''.png''], ''-dpng'')', ...
    'close(h);'])));

to export figures as static png files.

Note the use of eval to squeeze in two statements in the function handle (export the figure, and close it).

It may be helpful to wrap such settings in your own helper functions.

IMATLAB_CONNECT needs to be set outside of MATLAB (as it is checked before the connection to the engine is made). Other environment variables can be set either outside of MATLAB (before starting the kernel) or from within MATLAB (using setenv).

Asynchronous output

A construct such as 1, pause(1), 2 will output 1 and 2 with a one second interval on Linux and OSX, but together after a one second wait on Windows. PRs improving Windows support are welcome.

Asynchronous output using timer objects seem to be completely unsupported by the MATLAB engine for Python.

MATLAB debugger

The MATLAB debugger is cleared (dbclear all) before each execution, as interactive input is not supported by the engine API.

Differences with the Calysto MATLAB Kernel

The completion system is much more robust, by relying on the new API available in MATLAB 2016b.
History is read from and written to MATLAB's own History.xml, and thus shared with standard MATLAB sessions. Note that if the file does not exist (e.g. if the don't save history file option is set, or in a console-only setup), history not be reloaded into later sessions. (A PR for loading history.m instead would be welcome; it would need to properly parse multiline inputs in that file.)
Synchronous output is supported on Linux and OSX (see above).
There is no magics systems, as MATLAB already provides many functions for this purpose (cd, edit, etc.).
Inline graphics can be based on plotly, and thus interactive.

steverweber / imatlab