Spark学习实例(Python)：RDD、DataFrame、DataSet相互转换

在学习转换之前先了解以下它们的基本概念

RDD：弹性分布式数据集，是一个只读分区集合 DataFrame：以命名列方式组织的分布式数据集，概念上和关系型数据库的一张表一样 DataSet：分布式数据集合，Python暂时不支持

了解了基本的概念之后，接下来我们通过代码编写三种数据集的形成

RDD的形成

from pyspark.sql import SparkSession

if __name__ == __main__:
    spark = SparkSession 
        .builder 
        .appName("rddData") 
        .master("local[*]") 
        .getOrCreate()
    # 方式一：
    data = [1, 2, 3, 4, 5]
    rdd1 = spark.sparkContext.parallelize(data)
    print(rdd1.collect())
    # [1, 2, 3, 4, 5]
    # 方式二：
    rdd2 = spark.sparkContext.textFile("/home/llh/data/people.txt")
    print(rdd2.collect())
    # [Jack 27, Rose 24, Andy 32]
    spark.stop()

DataFrame的形成

from pyspark.sql import SparkSession

if __name__ == __main__:
    spark = SparkSession 
        .builder 
        .appName("rddDataFrame") 
        .master("local[*]") 
        .getOrCreate()
    df = spark.read.text("/home/llh/data/people.txt")
    df.show()
    # +---+----+
    # |age|name|
    # +---+----+
    # | 27|Jack|
    # | 24|Rose|
    # | 32|Andy|
    # +---+----+
    spark.stop()

RDD转成DataFrame

from pyspark.sql import SparkSession
from pyspark.sql import Row

if __name__ == __main__:
    spark = SparkSession 
        .builder 
        .appName("rddRDD") 
        .master("local[*]") 
        .getOrCreate()
    data = [1, 2, 3]
    rdd1 = spark.sparkContext.parallelize(data)
    print(rdd1.collect())
    # [1, 2, 3]
    # rdd ->  dataframe
    rdd2 = rdd1.map(lambda x: Row(x))
    df = spark.createDataFrame(rdd2, schema=[num])
    df.show()
    # +---+
    # |num|
    # +---+
    # | 1 |
    # | 2 |
    # | 3 |
    # +---+
    spark.stop()

DataFrame转成RDD

from pyspark.sql import SparkSession

if __name__ == __main__:
    spark = SparkSession 
        .builder 
        .appName("rddDataFrame") 
        .master("local[*]") 
        .getOrCreate()
    df = spark.read.text("/home/llh/data/people.txt")
    rdd = df.rdd
    print(rdd.collect())
    # [Row(value=Jack 27), Row(value=Rose 24), Row(value=Andy 32)]
    spark.stop()

以上就是RDD与DataFrame形成与相互转换

Spark学习目录：