spark计数单词数量与clob

Question

我有一个数据集

+----------+--------+------------+
|        id|    date|       errors|
+----------+--------+------------+
|1         |20170319|      error1|
|1         |20170319|      error2|
|1         |20170319|      error2|
|1         |20170319|      error1|
|2         |20170319|        err6|
|1         |20170319|       error2|

需要数字错误计数天明智

输出

+----------+--------+------------+
|    date|       errors| count
+----------+--------+------------+
 |20170319|      error1|    2
 |20170319|      error2|    3
 |20170319|        err6|    1

    val dataset = spark.read.json(path);
    val c =dataset.groupBy("date").count()

//我如何继续计数错误

我尝试在Spark Scala SQL中过时的日期，但找不到生产力我需要转换为RDD并找到一种方法。？

Answer 1

您只需要groupBy date和errors。

val c =dataset.groupBy("date","errors").count()