Github user gatorsmile commented on a diff in the pull request: https://github.com/apache/spark/pull/20394#discussion_r164254321 --- Diff: sql/core/src/main/scala/org/apache/spark/sql/execution/columnar/InMemoryRelation.scala --- @@ -73,11 +73,16 @@ case class InMemoryRelation( @transient val partitionStatistics = new PartitionStatistics(output) override def computeStats(): Statistics = { - if (batchStats.value == 0L) { - // Underlying columnar RDD hasn't been materialized, use the stats from the plan to cache - statsOfPlanToCache + if (sizeInBytesStats.value == 0L) { + // Underlying columnar RDD hasn't been materialized, use the stats from the plan to cache. + // Note that we should drop the hint info here. We may cache a plan whose root node is a hint + // node. When we lookup the cache with a semantically same plan without hint info, the plan + // returned by cache lookup should not have hint info. If we lookup the cache with a + // semantically same plan with a different hint info, `CacheManager.useCachedData` will take + // care of it and retain the hint info in the lookup input plan. + statsOfPlanToCache.copy(hints = HintInfo()) --- End diff -- This is a new behavior we introduced in 2.3. I will first keep the behavior unchanged and merge it to 2.3. We can have more discussion in the next release.
--- --------------------------------------------------------------------- To unsubscribe, e-mail: reviews-unsubscr...@spark.apache.org For additional commands, e-mail: reviews-h...@spark.apache.org