7

我正在使用 Spark v1.6。我有以下两个 DataFrame,我想在左外连接 ResultSet 中将 null 转换为 0。有什么建议么?

数据帧

val x: Array[Int] = Array(1,2,3)
val df_sample_x = sc.parallelize(x).toDF("x")

val y: Array[Int] = Array(3,4,5)
val df_sample_y = sc.parallelize(y).toDF("y")

左外连接

val df_sample_join = df_sample_x
  .join(df_sample_y,df_sample_x("x") === df_sample_y("y"),"left_outer")

结果集

scala> df_sample_join.show

x  |  y
--------
1  |  null

2  |  null

3  |  3

But I want the resultset to be displayed as.
-----------------------------------------------

scala> df_sample_join.show

x  |  y
--------
1  |  0

2  |  0

3  |  3
4

3 回答 3

13

只需使用na.fill

df.na.fill(0, Seq("y"))
于 2016-11-23T20:13:24.973 回答
7

尝试:

val withReplacedNull = df_sample_join.withColumn("y", coalesce('y, lit(0)))

测试:

import org.apache.spark.sql.Row
import org.apache.spark.sql.functions.{col, udf}
import org.apache.spark.sql.types._

val list = List(Row("a", null), Row("b", null), Row("c", 1));
val rdd = sc.parallelize(list);

val schema = StructType(
    StructField("text", StringType, false) ::
    StructField("y", IntegerType, false) :: Nil)

val df = sqlContext.createDataFrame(rdd, schema)
val df1 = df.withColumn("y", coalesce('y, lit(0)));
df1.show()
于 2016-11-23T19:27:05.287 回答
3

您可以像这样修复现有的数据框:

import org.apache.spark.sql.functions.{when,lit}
val correctedDf=df_sample_join.withColumn("y", when($"y".isNull,lit(0)).otherwise($"y"))

虽然 T. Gawęda 的回答也有效,但我认为这更具可读性

于 2016-11-23T19:49:34.813 回答