Question

我想修改高级教程的内部连接示例，使其可以使用mapreduce进行稀疏矩阵乘法（由Ullman描述）。因此，我需要第二个map-reduce步骤，将结果矩阵中相等位置的值相加。

不幸的是，我没有设法将类CsvInnerJoin的第一个reduce函数的输出转换为SumJob的map函数。

import sys
sys.path.append("/home/damian/disco/lib/")
from disco.core import Job, result_iterator
from disco.worker.classic.func import chain_reader
import csv, sys


if __name__ == '__main__':
    input_filename = "input.csv"
    output_filename = "output.csv"
    if len(sys.argv) > 1:
        input_filename = sys.argv[1]
        if len(sys.argv) > 2:
            output_filename = sys.argv[2]

    from CsvInnerJoiner import CsvInnerJoiner
    from SumJob import SumJob

    job = CsvInnerJoiner().run(input=[input_filename])
    job = SumJob().run() (******************)

    with open(output_filename, 'w') as fp:
        writer = csv.writer(fp)
        for url_key, descriptors in result_iterator(job.wait(show=True)):
            writer.writerow([url_key] + descriptors)

CsvInnerJoiner.py就是这个文件：

import sys
sys.path.append("/home/damian/disco/lib/")
from disco.core import Job, result_iterator
from disco.worker.classic.func import chain_reader
import csv, sys
class CsvInnerJoiner(Job):
    partitions = 2
    sort = True

    def map(self, row, params):
        yield row[0], row[1:]

    @staticmethod
    def map_reader(fd, size, url, params):
        reader = csv.reader(fd, delimiter=',')
        for row in reader:
            yield row

    #@staticmethod
def reduce(self, rows_iter, out, params):
    from disco.util import kvgroup
    from itertools import chain
    #for url_key, descriptors in kvgroup(sorted(rows_iter)):
    for url_key, descriptors in kvgroup(rows_iter):
        merged_descriptors = list(chain.from_iterable(descriptors))
        print url_key,"_______",merged_descriptors
        if len(merged_descriptors) > 3:
            Alist = merged_descriptors[:merged_descriptors.index("B")]
            Blist = merged_descriptors[merged_descriptors.index("B"):]
            Alistlength = len(Alist)/3
            Blistlength = len(Blist)/3
            for i in range(Alistlength):
                for j in range(Blistlength):
                    container = int(Alist[3*i+2])*int(Blist[3*j+2])
                    yield [Alist[3*i+1],Blist[3*j+1]],container
                    #out.add(Alist[3*i+1],[Blist[3*j+1],container])

SumJob.py就是这样：

import sys
sys.path.append("/home/damian/disco/lib/")
from disco.core import Job, result_iterator
from disco.worker.classic.func import chain_reader
import csv, sys


class SumJob(Job):
    map_reader = staticmethod(chain_reader)

    @staticmethod
    def map(self,key_value, params):
        print "KEY::::::",str(key_value[0])
        print "VAL::::::",str(key_value[1])
        yield key_value[0], key_value[1]
    @staticmethod    
    def reduce(self,key_value,out, params):
        Summe = sum(key_value[1])
        out.add(key_value[0],Summe)

问题是我不知道如何更改（**）行，以便第一个reduce步骤的第二个输出被第二个map-function视为输入。

非常感谢你的帮助！达米安

Answer 1

您可以使用map / reduce阶段的输出作为另一个的输入（从job.wait()返回）。

job1 = SumJob().run(input=[...])
job2 = SumJob().run(input=[...])

output = SomeOtherJob.run(input=[job1.wait(), job2.wait()]).wait(show=True)
for key, value in result_iterator(output):
    print key, value

我不是那种代码适合我的专家（我实现了pagerank算法，它有很多阶段和多次迭代）。

在Disco中链接工作（MapReduce）

1 个答案: