Home

Awesome

Julia for Numerical Computation in MIT Courses

Several MIT courses involving numerical computation, including 18.06 / 18.C06, 18.065/18.0651, 18.303, 18.330, 18.335/6.7310, 18.337/6.7320, and 18.338, are beginning to use Julia, a fairly new language for technical computing. This page is intended to supplement the Julia documentation with some simple tutorials on installing and using Julia targeted at MIT students. See also our Julia cheatsheet listing a few basic commands, and various Julia tutorials online.

In particular, we will be using Julia in the Jupyter browser-based enviroment (via the IJulia plug-in), which leverages your web browser and to provide a rich environment combining code, graphics, formatted text, and even equations, with sophisticated plots via Matplotlib.

You can also look at the Jupyter notebook from the fall 2020 tutorial, as well as the tutorial video (MIT only).

Why Julia?

Julia is relatively new high-level free/open-source language for numerical computing in the same spirit, with a rich set of built-in types and libraries for working with linear algebra and other types of computations, with a syntax that is superficially reminiscent of Matlab's. Basically, we are using Julia because, unlike Matlab or Python or R, it scales better to real computational problems — you can write performance-critical "inner loops" in Julia, whereas similar tasks in other high-level languages often require one to drop down to C or similar low-level languages. (See e.g. this 6.172 lecture on performance in Julia vs. Python.) Because of this, we are using Julia more and more in our own research, and we want to teach using software tools that we really employ ourselves.

Running Julia in the cloud

The easiest way to get started with Julia is to run it in the cloud on mybinder.org, which is as easy as clicking this link:

Binder

That link opens up a default MIT-math Julia + Python environment that we set up, but you can also easily set up your own environments. Although the link above gives you access to our tutorial notebook here, you can create alternate links (e.g. for particular MIT courses) using nbgitpuller.

There are two major drawbacks to using the free mybinder.org service:

Eventually you'll probably want to install Julia on your own computer to eliminate these frustrations. Fortunately, this is usually relatively easy:

Installing Julia and IJulia on your own computer

First, download the latest release of Julia and run the installer. Then run the Julia application (double-click on it, or run julia in a terminal); a window with a julia> prompt will appear. At the julia> prompt, type a ] (close square bracket) to get a Julia package prompt pkg>, where you can type

(v1.10) pkg> add IJulia

You may also want to install these packages, which we tend to use in a lot of the lecture materials:

(v1.10) pkg> add Interact PyPlot Plots

(You can install other packages later as you need them using the same interface, of course. Thousands of other packages can be found on JuliaHub.)

Running Julia through the JupyterLab app

Download the jupyterlab-desktop program, and launch it. Click the "Julia" button (or choose "New > Notebook" from the file menu and select the "Julia" kernel):

JupyterLab desktop

You should now have an interactive Julia notebook, whose usage we describe below.

If you have problems printing or exporting PDF from the JupyterLab Desktop (on some systems this fails if you don't have LaTeX installed), a workaround is to export as HTML (from File > Save and Export Notebook As… > HTML), open the resulting .html in your browser (double-click on it) and print to PDF from your browser.

Running Julia through Jupyter in your browser

You can alternatively use Julia itself to install the Jupyter software and have it run its interface through your web browser.

Switch back to the julia> prompt by hitting backspace or ctrl-C, and then you can launch the notebook by running

julia> using IJulia

julia> notebook()

and type "y" if you are asked to install Jupyter. A "dashboard" window like this should open in your web browser (at address localhost:8888, which you can return to at any time as long as the notebook() server is running; I usually keep it running all the time):

Jupyter dashboard

Now, click on the New button and select the Julia option to start a new "notebook".

(You will have to leave this Julia command-line window open in order to keep the Jupyter process running. Alternatively, you can run notebook(detached=true) if you want to run the Jupyter server as a background process, at which point you can close the Julia command line, but then if you ever want to restart the Jupyter server you will need to kill it manually.

Troubleshooting:

Other Julia environments

Pluto.jl

A different interactive-computing environment for Julia is Pluto.jl, which runs in the browser like Jupyter but is more oriented towards "live" interaction where updating one piece of code automatically re-runs anything affected by that change. Running Julia is as easy as:

pkg> add Pluto

julia> using Pluto

julia> Pluto.run()

VSCode

For writing larger programs, modules, and packages (as opposed to little interactive snippets), you'll want to start putting code into files and modules, and use a more full-featured code-editing environment. A popular choice is the free/cross-platform Visual Studio Code (VSCode) editor, which has a Julia VSCode plugin to provide a full-featured integrated development environment (IDE).

Other Editors

Of course, there is also good support for editing Julia in many other programs, such as Emacs, Vim, Atom, and so forth.

Updating Julia and IJulia

Julia is improving rapidly, so it won't be long before you want to update to a more recent version. The same is true of Julia add-on packages like PyPlot. To update the packages only, keeping Julia itself the same, just run:

(v1.10) pkg> update

at the Julia pkg> prompt after typing ]; you can also run ] update in IJulia.

If you download and install a new version of Julia from the Julia web site, you will also probably want to update the packages with update (in case newer versions of the packages are required for the most recent Julia). In any case, if you install a new Julia binary (or do anything that changes the location of Julia on your computer), you must update the IJulia installation (to tell IPython where to find the new Julia) by running build at the Julia pkg> prompt line (not in IJulia).

Running Julia in the Jupyter Notebook

Once you have followed the installation steps above, then you will want to open the Jupyter notebook interface. As explained above, you can either launch the standalone JupyterLab Desktop app (which you download and install separately), or you can install Jupyter via Julia and run it via your web browser.

Either way, a notebook will combine code, computed results, formatted text, and images; for example, you might use one notebook for each problem set. The notebook window that opens will look something like:

Jupyter notebook

In the browser can click the "Untitled" at the top to change the name, e.g. to "My first Julia notebook"; in JupyterLab you click the "Rename" option in the "File" menu. You can enter Julia code at the In[ ] prompt, and hit shift-return to execute it and see the results. If you hit return without the shift key, it will add additional lines to a single input cell. For example, we can define a variable x (using the built-in constant pi and the built-in function sin), and then evaluate a polynomial 3x^2 + 2x - 5 in terms of x (note that, unlike Matlab or Python, we don't have to type 3*x^2 if we don't want to: a number followed by a variable is automatically interpreted as multiplication without having to type *):

Jupyter notebook

The result that is printed (in Out[1]) is the last expression from the input cell, i.e. the polynomial. If you want to see the value of x, for example, you could simply type x at the second In[ ] prompt and hit shift-return.

See, for example, the mathematical operations in the Julia manual for many more basic math functions.

Plotting

There are several plotting packages available for Julia. If you followed the installation instructions, above, you already have one full-featured Matlab-like plotting package installed: PyPlot, which is simply a wrapper around Python's amazing Matplotlib library.

To start using PyPlot to make plots in Julia, first type using PyPlot at an input prompt and hit shift-enter. using is the Julia command to load an external module (which must usually be installed first, e.g. by the ] add PyPlot command from the installation instructions above). The very first time you do using PyPlot, it will take some time: the module and its dependencies will be "precompiled" so that in subsequent Julia sessions it will load quickly.

Then, you can type any of the commands from Matplotlib, which includes equivalents for most of the Matlab plotting functions. For example:

Jupyter notebook

Printing/exporting Notebooks

Currently, printing a notebook from the browser's Print command <a href="https://github.com/ipython/ipython/issues/4196">can be somewhat problematic</a>. There are four solutions: