Dandandan commented on issue #20325:
URL: https://github.com/apache/datafusion/issues/20325#issuecomment-3912873585

   Just thinking out loud: I think what also would help most ClickBench if 
instead of materializing the Parquet data before evaluating the filter, the 
filter could be evaluated _during decoding_.
   
   We could implement Parquet decoders that push down e.g. scalar (in)equality 
that produces only the arrow array for the rows matching this predicate (e.g. 
`<> ''`) and a boolean mask (when required) for the matching rows.
   
   This would
   * save overhead of decoding / copy / converting rows to e.g. string views
   * save time post-filtering
   * save memory overhead to cache the (unfiltered) column
   
   The downside mainly being is that it only works for the predicates 
implemented (e.g. equality / inequality), but it seems it would be general 
enough to be worth it.


-- 
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]


---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]

Reply via email to