json解析Dataframe（insertdf）的格式如下：

+------------------------------------------------+-------------+------------+
Cls_details                                       | prdct_cde   |roll_nbr
+------------------------------------------------+-------------+------------+

{"key1":"value1","key2":"value2","key3":"value3"}|DKLA          |123453  
{"key1":"value1","key2":"value2","key3":"value3"}|GHTD        |123454      
{"key1":"value1","key2":"value2","key3":"value3"}|ILDDA        |123455

我想把这个Dataframe保存到一个cassandra表中。使用的spark scala代码是：

insertDF
      .write.format("org.apache.spark.sql.cassandra")
      .options(Map("keyspace" -> "abc", "table" -> "def"))
      .mode(SaveMode.Append).save()

但我得到了一个错误：

com.datastax.spark.connector.types.TypeConversionException: Cannot convert object {"key1":"value1","key2":"value2","key3":"value3"} of type class java.lang.String to Map[AnyRef,AnyRef].

请提供spark scala的解决方案。任何帮助都将不胜感激。