<@UQ8NE6A86> did you come to any decision on publishing the Kotlin Kernel as a package? I am current...

# datascience

[Florian Magin]

@Ilya Muradyan did you come to any decision on publishing the Kotlin Kernel as a package? I am currently setting up automatic builds of my plugin that depends on the Kotlin kernel. I can just use the workaround I have been using before and grab the libraries from the Python package, but that is a bit inelegant.

[Ilya Muradyan]

Do you need a package with kernel in its current form or some API changes are required?

[Florian Magin]

current Form should be fine. Currently I am installing the python packaging and just copying the jar files from .local/lib/python3.9/site-packages/run_kotlin_kernel/jars/* to the location that the buildscript of my plugin expects them to be

[Ilya Muradyan]

Ok, I may add a publication for kernel package then. I'll publish it in Central along with other Kotlin kernel packages:
https://search.maven.org/search?q=kotlin-jupyter

[Ilya Muradyan]

ETA is today-tomorrow

[Florian Magin]

Thank you very much!

[Ilya Muradyan]

https://search.maven.org/artifact/org.jetbrains.kotlinx/kotlin-jupyter-kernel/0.8.3.231/jar

[Ilya Muradyan]

Please try it out and let me know if everything is fine

[Florian Magin]

It seems like just using maven isn't enough because the kernel requires the khttp library which isn't on maven yet. But khttp was never on the run_kotlin_kernel/jars folder, so I am not sure if it is required?

[Ilya Muradyan]

It is required. In kernel installation khttp is shadowed into the fat JAR.
Just add jcenter() into your repositories section. It seems that we need to contact khttp author and ask him to publish it on Central

[Florian Magin]

Ah, that works too. The documentation at https://github.com/Kotlin/kotlin-jupyter/blob/3c0d1cf891b6c8ea5606768fc3e9ce4bb5ef34a3/jupyter-lib/shared-compiler/build.gradle.kts#L43 implies this is just for remote library dependency resolving which I think is something that isn't required in my use case anyway. But might as well include it for now

[Florian Magin]

Huh, what is the reason the package names differ? i.e. when simply coping the libs from the python package the package path for the kernel entry is org.jetbrains.kotlin.jupyter.IkotlinKt but the libraries downloaded via maven have the class at the path org/jetbrains/kotlinx/jupyter/IkotlinKt.class

Is this intentional and something I need to change in my code that uses the kernel or an indication that I am doing something else wrong?

[Florian Magin]

Ah, the current code as published on maven is simply newer than the compiled code included in the Python package. So I just need to update my code accordingly

[Florian Magin]

Okay, I think this is the last issue, but I am somewhat stumped:

My call to the embedKernel method currently looks like this:
https://github.com/GhidraJupyter/ghidra-jupyter-kotlin/blob/main/GhidraJupyterKotl[…]in/src/main/java/GhidraJupyterKotlin/KotlinQtConsoleThread.java
i.e. passes null to what used to be the libraryFactory . You refactored this as part of https://github.com/Kotlin/kotlin-jupyter/commit/782f0882ed5f9e89ad378b4a986591397027721d#diff-3c1c4b1ddbb3338fe28601f[…]95dafb1c3ad4b60a8ffeedc62R85

Now the value isn't nullable anymore, and the obvious solution would be to pass EmptyResolutionInfoProvider . But for reasons that I don't fully understand I can't import EmptyResolutionInfoProvider
from java code.
My best guess is that this is because it is actually a singleton object and this can't be simply imported like a class in java https://github.com/Kotlin/kotlin-jupyter/blob/dd04fadbe693447c854b4b7668d9d1c4c30c[…]g/jetbrains/kotlinx/jupyter/libraries/ResolutionInfoProvider.kt

is there some way to properly import this in Java, or would the signature of embedKernel need to be changed?

[Ilya Muradyan]

@Florian Magin I believe you need to refer the object like

EmptyResolutionInfoProvider.INSTANCE

[Florian Magin]

ah, that does seem to work. Couldn't find any documentation on it, despite around 10 minutes of searching around

[Florian Magin]

Can confirm that this is enough, and I can now build my application by just declaring the kotlin kernel as a dependency. Thank you!

[Florian Magin]

@Ilya Muradyan is there some way to correlate the Maven releases with git commits? I assume the versions 0.8.3.241 and above that appeared yesterday are the the new changes I needed, but this seems somewhat hard to track. How do releases to maven currently work (roughly, mostly if you are doing them manually or if they just run for every new commit to master)

[Ilya Muradyan]

They run for every commit in master, every commit has the corresponding version, but version tags are only pushed for "release" versions (which are published to stable Conda and pip channels).
The list of tags is here:
https://github.com/Kotlin/kotlin-jupyter/tags

Last release version is 0.8.3.122