Rich-T-kid commented on PR #24227:
URL: https://github.com/apache/datafusion/pull/24227#issuecomment-5370432706

   TLDR: the current state of this PR is that it enables direct reads of 
RLE_DICTIONARY-encoded Parquet columns into dictionary arrays. The main issue 
right now is a performance regression on some queries, since the flag currently 
promotes every variable-length column to a dictionary type regardless of 
cardinality, causing high cardinality columns (5-15 million) to degrade query 
performance. I still think it's worth keeping the flag: it lets us benchmark 
dictionary-encoding-related changes, as @alamb pointed out 
[here](https://github.com/apache/datafusion/pull/24418#pullrequestreview-4965416771),
 and this PR lays the groundwork for the follow-up PRs in #24111 that will add 
fine-grained, per-column tuning. The flag will also be turned off by default
   
   
   @alamb @adriangb 
   the PR isn't ready for review I'm more so checking that this is the correct 
direction to be moving in


-- 
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