Pyspark UDF(user define functions) 在Pyspark DataFrame 使用UDF,也可以先转化rdd lambda apply

from pyspark.sql import SparkSession
from pyspark.sql import functions
import pyspark.sql.functions as F
from pyspark.sql.functions import udf
from pyspark.sql.types import StringType

df = spark.createDataFrame((
      (1, "a"),
      (2, "b"),
      (3, "c")
    )).toDF("label", "sentence")
df.show()
+-----+--------+
|label|sentence|
+-----+--------+
|    1|       a|
|    2|       b|
|    3|       c|
+-----+--------+

def to_upper(s):
    if s is not None:
        return s.upper()

toDateUDF = udf(to_upper, StringType())  
# 使用自定义函数
df.withColumn('sentence1',toDateUDF('sentence')).show()
+-----+--------+---------+
|label|sentence|sentence1|
+-----+--------+---------+
|    1|       a|        A|
|    2|       b|        B|
|    3|       c|        C|

df = df.withColumn('col_name', F.udf(to_upper, returnType=StringType())('sentence'))
df.show()
+-----+--------+--------+
|label|sentence|col_name|
+-----+--------+--------+
|    1|       a|       A|
|    2|       b|       B|
|    3|       c|       C|
+-----+--------+--------+