熟悉pandas的pythoner 应该知道给dataframe增加一列很容易，直接以字典形式指定就好了，pyspark中就不同了，摸索了一下，可以使用如下方式增加
from pyspark import SparkContext from pyspark import SparkConf from pypsark.sql import SparkSession from pyspark.sql import functions spark = SparkSession.builder.config(conf=SparkConf()).getOrCreate() data = [['Alice', 19, 'blue', '["Alice", 19, "blue"]'], ['Jane', 20, 'green', '["Jane", 20, "green"]'], ['Mary', 21, 'blue', '["Mary", 21, "blue"]'], ] frame = spark.createDataFrame(data, schema=["name", "age", "eye_color", "detail"]) frame.cache() frame.show()
+-----+---+---------+--------------------+
| name|age|eye_color| detail|
+-----+---+---------+--------------------+
|Alice| 19| blue|["Alice", 19, "bl...|
| Jane| 20| green|["Jane", 20, "gre...|
| Mary| 21| blue|["Mary", 21, "blue"]|
+-----+---+---------+--------------------+
1、 增加常数项
frame2 = frame.withColumn("contant", functions.lit(10))
frame2.show()
+-----+---+---------+--------------------+-------+
| name|age|eye_color| detail|contant|
+-----+---+---------+--------------------+-------+
|Alice| 19| blue|["Alice", 19, "bl...| 10|
| Jane| 20| green|["Jane", 20, "gre...| 10|
| Mary| 21| blue|["Mary", 21, "blue"]| 10|
+-----+---+---------+--------------------+-------+
2、简单根据某列进行计算
2.1 使用 withColumn
frame3_1 = frame.withColumn("name_length", functions.length(frame.name))
frame3_1.show()
+-----+---+---------+--------------------+-----------+
| name|age|eye_color| detail|name_length|
+-----+---+---------+--------------------+-----------+
|Alice| 19| blue|["Alice", 19, "bl...| 5|
| Jane| 20| green|["Jane", 20, "gre...| 4|
| Mary| 21| blue|["Mary", 21, "blue"]| 4|
+-----+---+---------+--------------------+-----------+
2.2 使用 select
frame3_2 = frame.select(["name", functions.length(frame.name).alias("name_length")])
frame3_2.show()
+-----+-----------+
| name|name_length|
+-----+-----------+
|Alice| 5|
| Jane| 4|
| Mary| 4|
+-----+-----------+
2.3 使用 selectExpr
frame3_3 = frame.selectExpr(["name", "length(name) as name_length"]) frame3_3.show()
+-----+-----------+
| name|name_length|
+-----+-----------+
|Alice| 5|
| Jane| 4|
| Mary| 4|
+-----+-----------+
3、定制化根据某列进行计算
比如我想对某列做指定操作，但是对应的函数没得咋办，造，自己造~
frame4 = frame.withColumn("detail_length", functions.UserDefinedFunction(lambda obj: len(json.loads(obj)))(frame.detail))
# or
def length_detail(obj):
return len(json.loads(obj))
frame4 = frame.withColumn("detail_length", functions.UserDefinedFunction(length_detail)(frame.detail))
frame4.show()
+-----+---+---------+--------------------+-------------+
| name|age|eye_color| detail|detail_length|
+-----+---+---------+--------------------+-------------+
|Alice| 19| blue|["Alice", 19, "bl...| 3|
| Jane| 20| green|["Jane", 20, "gre...| 3|
| Mary| 21| blue|["Mary", 21, "blue"]| 3|
+-----+---+---------+--------------------+-------------+
