1.parallelize()创建RDD：

words = sc.parallelize([("hadoop",1),("is",1),("good",1),\
    ("spark",1),("is",1),("fast",1),("spark",1),("is",1),\
        ("better",1)])
wordsres1 = words.groupByKey()
wordsres1.collect()

2.groupByKey()结果：

[('hadoop', <pyspark.resultiterable.ResultIterable at 0x7effad99feb8>),
 ('is', <pyspark.resultiterable.ResultIterable at 0x7effad99f828>),
 ('good', <pyspark.resultiterable.ResultIterable at 0x7effad99fcf8>),
 ('spark', <pyspark.resultiterable.ResultIterable at 0x7effad99fda0>),
 ('fast', <pyspark.resultiterable.ResultIterable at 0x7effad99fbe0>),
 ('better', <pyspark.resultiterable.ResultIterable at 0x7effad99fd68>)]

通过groupByKey，对原始的RDD数据进行分组，结果如下：

("hadoop",1)	("is",(1,1,1))
("spark",(1,1)	("good",1)
("fast",1)	("better",1)

3.reduceByKey()结果：

wordsres2 = words.reduceByKey(lambda a,b:a+b)
wordsres2.collect()


#结果：
[('hadoop', 1),
 ('is', 3),
 ('good', 1),
 ('spark', 2),
 ('fast', 1),
 ('better', 1)]

通过reduceByKey，对原始的RDD数据进行jisuan，结果如下

("hadoop",1)	("is",3)
("spark",2)	("good",1)
("fast",1)	("better",1)