jerry-024 commented on PR #858:
URL: https://github.com/apache/paimon-rust/pull/858#issuecomment-5712006104

   ### Design question: eager planning vs. lazy execution
   
   `PartitionRowCountProvider::scan()` currently computes all manifest counts 
while creating the physical plan. This follows the existing 
`PaimonTableProvider::scan()` model, but on very large tables it also means 
`EXPLAIN` / `create_physical_plan()` performs the full manifest read before 
execution-time cancellation and timeout controls apply.
   
   Moving this work into a lazy `ExecutionPlan::execute()` would avoid that 
planning latency, but it also makes the unknown-DV-cardinality fallback more 
involved: the ordinary scan and aggregate would need to be planned and executed 
lazily without defeating the optimization during planning.
   
   Would maintainers prefer the lazy execution plan in this PR, or keep the 
current provider model here and handle lazy planning in a focused follow-up?
   


-- 
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