Github user gatorsmile commented on a diff in the pull request: https://github.com/apache/spark/pull/22731#discussion_r225333576 --- Diff: sql/core/src/main/scala/org/apache/spark/sql/execution/datasources/FileScanRDD.scala --- @@ -106,15 +106,16 @@ class FileScanRDD( // don't need to run this `if` for every record. val preNumRecordsRead = inputMetrics.recordsRead if (nextElement.isInstanceOf[ColumnarBatch]) { + incTaskInputMetricsBytesRead() --- End diff -- 4096 is the default number of the batch reader in both ORC and Parquet. If the users set the conf to a much smaller number, they will face the perf regression due to the the extra overhead in many places. I do not think end users will do this.
--- --------------------------------------------------------------------- To unsubscribe, e-mail: reviews-unsubscr...@spark.apache.org For additional commands, e-mail: reviews-h...@spark.apache.org