Conda recipe

[hadim]

Hi @ctrueden ,

I need you on this ! And I think it should be easy to solve with your Maven skills :-)

I am creating a conda recipe (see it as a Travis-like mechanism that output a nice package that can be then distributed and installed with conda).

The package is built in an isolated environment with only jupyter, maven and openjdk (I want the conda package to be totally ImageJ-independent since ImageJ integration is made directly within ImageJ).

So my idea is to "install" (copy scijava-jupyter-kernel jar and its dependencies) to a specific folder, probably something like $PREFIX/opt/scijava-jupyter-kernel. And then at the end of the package installation, I run some Java-glue-code to install all the kernels.

Now my question is : how to tell Maven I want to copy all the required jars somewhere ? For this, I don't want to modify the scijava-jupyter-kernel pom.xml (since conda build is very specific) but instead, I am using a settings.xml file in the conda recipe folder.

Note the recipe won't live in the scijava-jupyter-kernel but will live in a separated git repo on the conda-forge GitHub organization.

[hadim]

Here is the recipe : conda-forge/staged-recipes#2841

[hadim]

For now I am using imagej.app.directory, that seems to do the job. Let me know if there is another way.

[hadim]

Recipe submitted conda-forge/staged-recipes#2841

Waiting for approval.

[hadim]

Done

[ctrueden]

You can use mvn dependency:copy-dependencies to place all dependencies into the target/dependency folder. And then from there, copy them wherever you want. This would be a little less hacky than using the imagej-maven-plugin's copy-jars goal. I assume that as things stand, the JARs end up in ${env.PREFIX}/opt/scijava-jupyter-kernel/jars? Is that where you want them to end up? If so, I guess you can just stick with the copy-jars goal.

[hadim]

Great I think I'll use mvn dependency:copy-dependencies and then just move the files where I want.

It doesn't really matter where I put the them. /opt is an arbitrary choice.

The conda repo has been created https://github.com/conda-forge/scijava-jupyter-kernel-feedstock and builds should be generated soon.

[ctrueden]

You might be able to avoid the after-the-fact copy too, by setting the outputDirectory property of the mojo.