gabotechs opened a new issue, #25610:
URL: https://github.com/apache/datafusion/issues/25610

   ## Is your feature request related to a problem or challenge?
   
   TPC-H and TPC-DS SF1 expose estimation gaps at filters, joins, aggregates and
   window operators. Each task below has a standalone SQL reproducer and 
estimated
   versus actual row counts.
   
   ## Describe the solution you'd like
   
   - [ ] Estimate group counts without explicit NDVs
   - [ ] Retain selectivity from supported AND conjuncts
   - [ ] Estimate Decimal filter selectivity
   - [ ] Distinguish string equality, LIKE and NOT LIKE estimates
   - [ ] Propagate aggregate-result statistics into HAVING
   - [ ] Include residual predicates in join estimates
   - [ ] Avoid inferring full anti-join coverage from range bounds
   - [ ] Improve NDV estimates for sparse join keys
   - [ ] Account for additional equi-join keys
   - [ ] Account for partitions in local TopK statistics
   - [ ] Use null counts when estimating IS NULL and IS NOT NULL
   - [ ] Estimate IN and NOT IN selectivity from the listed values
   - [ ] Propagate window ranking statistics into rank filters
   - [ ] Estimate equality to a scalar subquery without a blanket 20% fallback
   
   Start with mixed AND predicates, local TopK, null checks and IN lists. The 
other
   tasks need additional statistics or estimation-policy decisions.
   
   ## Describe alternatives you've considered
   
   Keep the current fallback estimates, leaving these gaps unresolved.
   
   ## Additional context
   
   Reproductions use the session-registry CLI fix in #25570, which is not yet
   merged. Related work: #8227, #20766 and #14237. Multi-key join work is 
already
   tracked in #21583.
   


-- 
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