MR (key,value)排序, hadoop.Spark.sql的join操作