Github user cloud-fan commented on the issue: https://github.com/apache/spark/pull/19471 waiting for more feedbacks before moving forward :) Another thing I wanna point out: for `sql("create table t using parquet options(skipHiveMetadata=true) location '/tmp/t'")`, it works in Spark 2.0, and the created table has a schema that the partition column is at the beginning. In Spark 2.1, it also works, and `DESC TABLE` also shows the table schema has partition column at the beginning. However, if you query the table, the output schema has partition column at the end. It's been a long time since Spark 2.1 was released and no one reports this behavior change. It seems this is really a corner case and makes me feel we should not compilcate our code too much for it.
--- --------------------------------------------------------------------- To unsubscribe, e-mail: reviews-unsubscr...@spark.apache.org For additional commands, e-mail: reviews-h...@spark.apache.org