I tried to suggest we should adopt pants for Rikai, but I tried pants’ scala build, for packaging, I find it hard to package the library only.

And for python, it is hard to publish to pypi (migrating from setup.py

I tried pants’ scala build, for packaging, I find it hard to package the library only

There are no java_package or scala_package like go_package

https://github.com/pantsbuild/pants/pull/17251 It seems with the PR, I can make pants work well for Scala project. https://github.com/eto-ai/rikai/pull/711/files And I tried to migrate to rikai to pants, it is a bit hard for us, because there are two many dependencies in rikai:

tensorflow

,

pytorch

,

pyspark

. CI takes a long time. I have to re-organize the code to make tensorflow/pytorch integration in a separated module to make it work. But I guess the owner of the project will not like it.

https://github.com/pantsbuild/pants/pull/17663 Waiting for 2.16.0dev2!