将行值转换为带有Spark Scala中另一列的值的列 [英] Convert row values into columns with its value from another column in spark scala

问题描述

我正在尝试将行中的值与另一列中的值转换为不同的列.例如-

I'm trying to convert values from row into different columns with its value from another column. For example -

输入数据框就像-

+-----------+
| X | Y | Z |
+-----------+
| 1 | A | a |
| 2 | A | b |
| 3 | A | c |
| 1 | B | d |
| 3 | B | e |
| 2 | C | f |
+-----------+

输出数据帧应如下所示-

And the output dataframe should be like this -

+------------------------+
| Y | 1    | 2    | 3    |
+------------------------+
| A | a    | b    | c    |
| B | d    | null | e    |
| C | null | f    | null |
+------------------------+

我尝试将基于Y的值和X和Z上的collect_list分组，然后压缩X&Z一起获得某种键值对.但是某些Y值可能会缺少一些X，因此为了用空值填充它们，我交叉联接了X的所有可能值和Y的所有可能值，然后将其联接到原始数据帧中.这种方法效率很低.

I've tried to groupBy the values based on Y and collect_list on X and Z and then zipped X & Z together to get some sort of key-value pairs. But some Xs may be missing for some values of Y so in order to fill them with null values, I cross joined all possible values of X and all possible values of Y and then joined it my original dataframe. This is approach is highly inefficient.

有没有有效的方法来解决此问题?提前致谢.

Is there any efficient method to approach this problem ? Thanks in advance.

推荐答案

您可以简单地将 groupBy 与 pivot 和 first 用作聚合函数

You can simply use groupBy with pivot and first as aggregate function as

import org.apache.spark.sql.functions._
df.groupBy("Y").pivot("X").agg(first("z")) 

输出:

+---+----+----+----+
|Y  |1   |2   |3   |
+---+----+----+----+
|B  |d   |null|e   |
|C  |null|f   |null|
|A  |a   |b   |c   |
+---+----+----+----+

这篇关于将行值转换为带有Spark Scala中另一列的值的列的文章就介绍到这了，希望我们推荐的答案对大家有所帮助，也希望大家多多支持IT屋！