alamb commented on PR #8991:
URL: 
https://github.com/apache/arrow-datafusion/pull/8991#issuecomment-1915509262

   > As I mentioned in the description, adding an additional Projection under 
Join (e.g., SortMergeJoin) doesn't make a lot sense for Spark due to its 
distribution nature.
   
   🤔 
   
   > (note that by adding additional projection before join in Spark it means 
more data to be shuffled/sorted which can be bad for performance)
   
   I don't understand this
   
   If the join is on `lcol_1 + lcol_2` = `rcol_1 + rcol_2` the plan will have 
to materialize four columns `lcol_1`, `lcol_2`, `rcol_1` and `rcol_2`, and 
would have to feed all 4 of those columns to the join
   
   wouldn't it actually make more sense to compute the expressions prior to the 
networked shuffle so only 2 columns of data (`lcol_1 + lcol_2` and `rcol_1 + 
rcol_2`) need to be sent, rather than the 4 original columns 🤔 


-- 
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