andygrove commented on issue #2709:
URL: 
https://github.com/apache/arrow-datafusion/issues/2709#issuecomment-1158922350

   > The main issue a few months ago was that the datafusion codebase was not 
split like it is now, so it created conflicts in the various places where it 
wasn't possible to confine differences between arrow and arrow2::io to 
extension traits. So my question is, can a contrib crate exist since arrow2 is 
essentially required at the core ?
   
   Regardless of where the code lives, I would view us as having two main 
options:
   
   1. Continue to maintain a copy of DataFusion "core" with modifications to 
allow building with arrow or arrow2. This will be an ongoing maintenance burden 
and IMO only worth the effort if we are working towards having this merged into 
DataFusion master at some point and I am not sure how much appetite there is 
for this.
   
   2. Fork `datafusion-core` to create a new query engine based on `arrow2` 
that can leverage the new DataFusion crates for logical plan building, SQL 
query planning, logical plan optimizations etc. Over time maybe more 
functionality can be moved out of core and re-used by other query engines. This 
would decouple the development work and allow those interested in arrow2 to 
move faster and does not put any burden on the core DataFusiuon development. 
That would be my choice if I were working on this.
   
   
   


-- 
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.

To unsubscribe, e-mail: [email protected]

For queries about this service, please contact Infrastructure at:
[email protected]

Reply via email to