python partition_key_python-spark中partitionBy和groupBy的区别是什么

rdds = self.sc.parallelize([(("good", "spark"), 1), (("sood", "hpark"), 1), (("god", "spak"), 1),

(("food", "spark"), 1), (("fggood", "ssspark"), 1), (("xd", "hk"), 1),

(("good", "spark"), 7), (("good", "spark"), 3), (("good", "spark"), 4),

(("sood", "hpark"), 5), (("sood", "hpark"), 7), (("xd", "hk"), 2),

(("xd", "hk"), 1), (("fggood", "ssspark"), 2), (("fggood", "ssspark"), 1)], 6)

rdds.glom().collect()

def inner_map_1(p):

d = defaultdict(int)

for row in p:

d[row[0]] += row[1]

for item in d.items():

yield item

rdd2 = rdds.partitionBy(4, partitionFunc=lambda x: hash(x)).mapPartitions(inner_map_1)

print(rdd2.glom().collect())

def inner_map_2(p):

for row in p:

item = row[0]

sums = sum([num for _, num in row[1]])

yield item, sums

rdd3 = rdds.groupBy(lambda x: x[0]).mapPartitions(inner_map_2)

print(rdd3.glom().collect())