Spark SQL-更新DataFrame行/列值而不转换为RDD

Question

How do i update Spark SQL DataFrame Rows/Columns values without converting as RDD ?

Why we cannot directly update the DataFrame like RDD and returns another one.

Answer 1

RDDs are immutable and a 'transformation' on an RDD can only generates a new RDD. Dataframes are wrappers around RDD and as such suffer from said immutablility for ex:

oldDF.registerTempTable("whatever") 

val newDF = sqlContext.sql("select field1,field2,sum(field3) as times from whatever where substring(field1,1,4)='test' group by field1,field2 having times>100").collect().saveAsParquetFile("xxx.parquet")