Spark DataFrame中基于List的排序UDF

来源：互联网发布：守望先锋低配优化补丁编辑：程序博客网时间：2024/06/05 05:35

DataFrame中支持的类型如 Struct此类型我们是用对象进行的封装，但是要对strcut里的数据进行排序，个人想到的办法还是把struct转位Json，然后反解析Json

sqlcontext.udf.register("stuSort", (stuList: Seq[String]) => {  val rs = Option(conponList) match {    case None => ""    case Some(c) => {      val couponInfoList = ArrayBuffer[CouponInfo]()      conponList.foreach(x => {        couponInfoList.append(readAsBeanByJson4s[CouponInfo](x))      })

    //基于Ordering的隐式参数      implicit object CouponOrdering extends Ordering[CouponInfo] {        override def compare(s1: CouponInfo, s2: CouponInfo): Int = {          if (s1.coupon_bank == s2.coupon_bank) if (s1.coupon_list == s2.coupon_list) s1.coupon_st.compareTo(s2.coupon_st)          else s1.coupon_list.compareTo(s2.coupon_list)          else s1.coupon_bank.compareTo(s2.coupon_bank)        }      }      couponInfoList.sorted.mkString(",")    }  }  rs.toString})

0 0