Dandandan commented on code in PR #9414:
URL: https://github.com/apache/arrow-rs/pull/9414#discussion_r2966107445
##########
parquet/src/arrow/arrow_reader/mod.rs:
##########
@@ -430,6 +434,32 @@ impl<T> ArrowReaderBuilder<T> {
..self
}
}
+
+ /// Set a scatter threshold for filter deferral.
Review Comment:
Yeah - I also think this is quite a bit better - I think it's OK if a filter
"only" selects 50% of the rows, if it is nicely packed that will probably be
able to skip almost 50% of the pages and lead to efficient
But (especially when combining multiple filters), chances are the filter
gets too fragmented, a 50% packed one is better than a 25% selective one that
just does select 1 / skip 1 / select 1 ...
I think you could go even go further and check which combination of filtera
are giving the best of both worlds (selectivity and fragmentation)
--
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]
