如何创建创建新列并修改现有列的 UDF

Question

我有一个这样的数据框：

id | color
---| -----
1  | red-dark
2  | green-light
3  | red-light
4  | blue-sky
5  | green-dark

我想创建一个 UDF，使我的数据框变为：

id | color | shade
---| ----- | -----
1  | red   |  dark
2  | green |  light
3  | red   |  light
4  | blue  |  sky
5  | green |  dark

我为此编写了一个 UDF：

def my_function(data_str):
    return ",".join(data_str.split("-"))

my_function_udf = udf(my_function, StringType())

#apply the UDF

df = df.withColumn("shade", my_function_udf(df['color']))

但是，这并没有按照我的预期转换数据框。相反，它变成了：

id | color      | shade
---| ---------- | -----
1  | red-dark   |  red,dark
2  | green-dark |  green,light
3  | red-light  |  red,light
4  | blue-sky   |  blue,sky
5  | green-dark |  green,dark

如何在 pyspark 中根据需要转换数据框？

根据建议的问题尝试过

schema = ArrayType(StructType([
    StructField("color", StringType(), False),
    StructField("shade", StringType(), False)
]))

color_shade_udf = udf(
    lambda s: [tuple(s.split("-"))],
    schema
)

df = df.withColumn("colorshade", color_shade_udf(df['color']))

#Gives the following

id | color      | colorshade
---| ---------- | -----
1  | red-dark   |  [{"color":"red","shade":"dark"}]
2  | green-dark |  [{"color":"green","shade":"dark"}]
3  | red-light  |  [{"color":"red","shade":"light"}]
4  | blue-sky   |  [{"color":"blue","shade":"sky"}]
5  | green-dark |  [{"color":"green","shade":"dark"}]

感觉离我越来越近了

Answer 1

可以使用内置函数split()：

from pyspark.sql.functions import split, col

df.withColumn("arr", split(df.color, "\\-")) \
  .select("id", 
          col("arr")[0].alias("color"),
          col("arr")[1].alias("shade")) \
  .drop("arr") \
  .show()
+---+-----+-----+
| id|color|shade|
+---+-----+-----+
|  1|  red| dark|
|  2|green|light|
|  3|  red|light|
|  4| blue|  sky|
|  5|green| dark|
+---+-----+-----+