PysparkNote006---rdd的flatMap

import pyspark.sql.functions as F
from pyspark.sql import SparkSession
from pyspark.sql.types import *
from pyspark.sql.types import Row

# 创建SparkSession对象，调用.builder类
# .appName("testapp")方法给应用程序一个名字；.getOrCreate()方法创建或着获取一个已经创建的SparkSession
spark = SparkSession.builder.appName("pysaprk").getOrCreate()

from pyspark.sql import functions as F

import pandas as pd

df_pd = pd.DataFrame({'id': ['A', 'B'], 'index': ['1', '2'], 'index1': ['2', '4']})
df_pd

	id	index	index1
0	A	1	2
1	B	2	4

df = spark.createDataFrame(df_pd)
df.show()

+---+-----+------+
| id|index|index1|
+---+-----+------+
|  A|    1|     2|
|  B|    2|     4|
+---+-----+------+

df.printSchema()

root
 |-- id: string (nullable = true)
 |-- index: string (nullable = true)
 |-- index1: string (nullable = true)

map

def ff(row):
    # print([row['id'], row.index, row.index1])
    print([row['id'], row['index'], row['index1']])
    return [[row['id'], row['index'], row['index1']]]

map_df = spark.createDataFrame(df.rdd.map(lambda x: ff(x)))
map_df.show()

+---------+
|       _1|
+---------+
|[A, 1, 2]|
|[B, 2, 4]|
+---------+

map_df.printSchema()

root
 |-- _1: array (nullable = true)
 |    |-- element: string (containsNull = true)

map()是将函数用于RDD中的每个元素，将返回值构成新的RDD
对于map来说，dataframe的一行是元素，
flatmap()是将函数应用于RDD中的每个元素，将返回的迭代器的所有内容构成新的RDD

df.rdd.map(lambda x: ff(x)).toDF().show()

+---------+
|       _1|
+---------+
|[A, 1, 2]|
|[B, 2, 4]|
+---------+

flatMap

schema = StructType([StructField(cn, StringType()) for cn in
                     ['id', 'index', 'index1']])

flatMap_df = spark.createDataFrame(df.rdd.flatMap(lambda x: ff(x)))
flatMap_df.show()

+---+---+---+
| _1| _2| _3|
+---+---+---+
|  A|  1|  2|
|  B|  2|  4|
+---+---+---+

flatMap_df.printSchema()

root
 |-- _1: string (nullable = true)
 |-- _2: long (nullable = true)
 |-- _3: long (nullable = true)

df.rdd.flatMap(lambda x: ff(x)).toDF().show()

+---+---+---+
| _1| _2| _3|
+---+---+---+
|  A|  1|  2|
|  B|  2|  4|
+---+---+---+

写的比较抽象，可以看下面第二个blog，有图讲的清晰些
在有些场景下，比如对df做操作，需要保留某些列，并且做些列之间复杂计算时，需要把最终结果再还原成dataframe时，比较好用

Ref

[1] https://blog.csdn.net/ten_sory/article/details/80897648
[2] https://cloud.tencent.com/developer/article/1912787

                                2023-07-26 阴 于南京市江宁区