andygrove commented on code in PR #4374:
URL: https://github.com/apache/datafusion-comet/pull/4374#discussion_r3284854997
##########
spark/src/test/resources/tpcds-plan-stability/approved-plans-v1_4-spark4_0/q83.ansi/extended.txt:
##########
@@ -101,4 +127,4 @@ CometNativeColumnarToRow
+- CometFilter
+- CometNativeScan
parquet spark_catalog.default.date_dim
Review Comment:
Tried the narrower form (`originalPlan.canonicalized.output` in the
canonical key, plus `mode.canonicalized` and `equals`/`hashCode` widened to
include `output` and `mode`) and the count-bug regression still reproduces —
`CometAggregateSuite` test returns 1 row vs Spark's 2. So the discriminator
isn't (just) in `output`; the count+1 / count-1 projection lives deeper in the
plan and is structurally indistinguishable at the output level after
canonicalization. Looking into this more to see if there's a tighter key that
recovers q83's reuse without losing the count-bug guard.
--
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]
