Open r3stl355 opened 10 months ago
Seems like we have a few options here:
I'm not really sure what option is best.
My preference would be 2nd or 3rd options as they seem to give move flexibility (e.g. run specific version of Rust)
The one-ring to rule them all approach isn't as nice as the alternatives. Agreed with @r3stl355.
@MrPowers for the alternative option (3).
delta-spark [check]
deltalake python (rust-bindings) - requires python in addition to rust - drop the weight of the spark runtime, java, etc.
Pure rust approach with the addition of the rust jupyter kernel.
for options 2 and 3, are there any other things we should consider in the docker image? arm/7 image? alternative base images?
@newfront - I'm not sure all the considerations for the Docker best practices.
But from my experience, there are lots of weird Spark runtimes and we will need to build something that works well for all users, regardless of where they want to use the code.
A Jupyter notebook with delta-rs Rust example was just released: https://github.com/delta-io/delta-examples/tree/master/notebooks/delta-rs. It uses a containerized Evcxr Jupyter Kernel, could that Dockerfile be added here (or perhaps integrated into the existing Dockerfile)