mustafasrepo commented on issue #7518:
URL: 
https://github.com/apache/arrow-datafusion/issues/7518#issuecomment-1713623935

   When a window function with rank is used.  A single pass on the whole data 
is sufficient. What we need is finding partition boundaries in the data. 
However, for window frames (with groups, and range) for each row we find the 
preceding and following boundary (corresponding to each row, determined by 
values of the row.) Hence this calculation should be done for each row(storing 
last boundaries found, helps in decreasing search space) on the subset of the 
table(around row of the interest.) 
   
   I am not sure whether window frame boundary search can be vectorized. That 
being said, I guess instead of `Vec<ScalarValue>` we can use `Row` type. 
   
   


-- 
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