Setup the IJava jupyter kernel

In this blog, we will show code snippets and examples to make the best use of ImgLib2, BigDataViewer, and friends. ImgLib2 is written to be fast and we will run code that needs to be compiled, so we cannot use any of the various interpreted scripting languages like Python, Groovy, or Javascript. Instead, we will use the JShell tool that you can use directly in a terminal or through Spencer Park’s IJava jupyter kernel. You can also follow these tutorials in your own Java project and use your preferred IDE, but Jupyter notebooks are a great teaching tool. Since jupyter is written in Python and most popular with the Python community, let’s follow their ways and first thing create a virtual environment with conda. The lack of version controlled dependency management for Python projects makes it necessary that practically every project must run in a container or virtual environment because the dependencies of different projects almost inevitably collide. Conda is the most popular of several attempts to address this situation. Conda cannot currently be installed from the default Ubuntu repositories, so much about that, but the installation instructions are tolerable, there is a PPA. Now let’s create an environment for jupyter:

conda create -n jshell-jupyter python=3
conda init bash
conda activate jshell-jupyter
conda install jupyter

You will also need a modern Java and Maven on your system, so if you have not yet done so, install it:

sudo apt install openjdk-17-jdk maven

The original IJava kernel currently does not build with Java 17 or 18, so we use Philipp Hanslovsky’s fork and build and install both the kernel installer and the IJava Jupyter kernel:

git clone https://github.com/hanslovsky/Jupyter-kernel-installer-gradle.git
cd Jupyter-kernel-installer-gradle/
git checkout try-upgrade-gradle
./gradlew publishToMavenLocal

cd ..
git clone https://github.com/saalfeldlab/IJava.git
cd IJava/
./gradlew installKernel

Now check if the kernel is installed, this should print something like this

jupyter kernelspec list

Available kernels:
  java       /home/saalfeld/.local/share/jupyter/kernels/java
  python3    /home/saalfeld/anaconda3/envs/jshell-jupyter/share/jupyter/kernels/python3

You can now start the jupyter notebook server

jupyter notebook --kernel=java

And experiment with the examples. Spencer Park’s IJava jupyter kernel makes it very easy to include dependencies. You can include the relevant snippets from a Maven POM into a tagged code block, e.g.

%%loadFromPOM
<repository>
    <id>scijava.public</id>
    <url>https://maven.scijava.org/content/groups/public</url>
</repository>
<dependency>
    <groupId>sc.fiji</groupId>
    <artifactId>bigdataviewer-vistools</artifactId>
    <version>1.0.0-beta-29</version>
</dependency>

or in gradle short notation

%mavenRepo scijava.public https://maven.scijava.org/content/groups/public
%maven sc.fiji:bigdataviewer-vistools:1.0.0-beta-29

If you prefer to run JShell directly, you can pull in the dependencies from a complete Maven POM with John Pooth’s Maven Jshell plugin

mvn compile com.github.johnpoth:jshell-maven-plugin:1.3:run

Happy JShelling!

P.S.:

The original IJava kernel currently does not build with Java 17 or 18, so if you prefer this over the above fork, or if you do not care about the latest greatest language features in Java 17, then the easiest at this time is to use OpenJDK-11. If you don’t have it yet, install it via conda:

conda install openjdk
conda install -c conda-forge maven

However, this may take a day of solving environments, so you can also install it globally:

sudo apt install openjdk-11-jdk maven

If you have other versions installed, you can switch between them with the alternatives tool:

sudo update-alternatives --config java
sudo update-alternatives --config javac

Now check out the original IJava and build and install the kernel IJava Jupyter kernel following the installation instructions or:

git clone https://github.com/SpencerPark/IJava.git
cd IJava/
./gradlew installKernel
jupyter kernelspec list

Done.