我想创建一个MapReduce程序，其reduce接收按值排序的k-v对。我正在使用mrjob，其SORT_VALUES参数似乎非常适合该任务。将此参数设置为True后，reducer输入未排序，例如，我得到以下内容（考虑A应该在X之前出现）
"ES" ["X", 3]
"ES" ["A", "Spain"]
我正在使用Python 2.7.5，mrjob == 0.6.1和Hadoop。该程序的本地执行给我：
"ES" ["A", "Spain"]
"ES" ["X", 1]
"ES" ["X", 2]
这是正确的。但是hadoop的执行给出了：
"ES" ["X", 3]
"ES" ["A", "Spain"]
我的代码是：
import sys, os, re
from mrjob.job import MRJob
from mrjob.step import MRStep
class MRJoin(MRJob):
SORT_VALUES = True
def mapper(self, _, line):
splits = line.rstrip("\n").split(",")
if len(splits) == 2: # countries
symbol = 'A' # countries before clients
country2digit = splits[1]
yield country2digit, (symbol, splits[0])
else: # clients
symbol = 'X'
country2digit = splits[2]
if splits[1]=='bueno':
yield country2digit,(symbol, 1)
def combiner(self,key, values):
bueno=0
for value in values:
if value[0] == 'A':
yield key, ('A', value[1])
else:
bueno=bueno + 1
if bueno > 0:
yield key, ('X', bueno)
def reducerSimple(self, key, values):
for value in values:
yield key,value
def steps(self):
return [
MRStep(mapper=self.mapper
,combiner=self.combiner
,reducer=self.reducerSimple)
]
if __name__ == '__main__':
MRJoin.run()
我像这样运行上面的代码：
python mrjob-p2.py /media/notebooks/clients.csv /media/notebooks/countries.csv -r hadoop
哪个给：
"ES" ["X", 3]
"ES" ["A", "Spain"]
...
"GN" ["A", "Guinea"]
"GN" ["X", 1]
...
ES键（和其他几个键）的值未排序，但其他键的值已排序。
我期望（如果对值进行排序，则A应该在X之前）
"ES" ["A", "Spain"]
"ES" ["X", 3]
如果我在本地运行：
python mrjob-p2.py /media/notebooks/clients.csv /media/notebooks/countries.csv -r本地
然后我得到：
"ES" ["A", "Spain"]
"ES" ["X", 1]
"ES" ["X", 2]
...
"GN" ["A", "Guinea"]
"GN" ["X", 1]
...
哪个是正确的
有人对如何对值进行排序有想法吗？
谢谢：）