将密钥/值对的Pyspark RDD解析为.csv格式

时间:2017-08-02 15:41:59

标签: python parsing apache-spark lambda pyspark

我正在构建一个解析器,它接受“key”=“value”对的原始文本文件,并使用PySpark写入表格/ .csv结构。

我遇到困难的是,我可以在函数中访问它们的键和值来构造每个csv_row,甚至检查键是否等于预期键的列表(col_list),但是我在lambda中调用该函数processCsv,我不知道如何将每个csv_row附加到列表l_of_l的全局列表中,该列表旨在保存.csv的最终列表行。

如何以键/值格式迭代RDD的每个记录并解析为.csv格式?如您所见,我的最终列表列表(l_of_l)为空,但我可以在循环中获取每一行......令人沮丧。

所有建议都赞赏!

原始文本结构(foo.log):

"A"="foo","B"="bar","C"="baz"
"A"="oof","B"="rab","C"="zab"
"A"="aaa","B"="bbb","C"="zzz"

迄今为止的方法:

from pyspark import SparkContext
from pyspark import SQLContext
from pyspark.sql import Row

sc=SparkContext('local','foobar')
sql = SQLContext(sc)

# Read raw text to RDD
lines=sc.textFile('foo.log')
records=lines.map(lambda x: x.replace('"', '').split(","))

print 'Records pre-transform:\n'
print records.take(100)
print '------------------------------\n'

def processRecord(record, col_list):    
    csv_row=[]
    for idx, val in enumerate(record):
        key, value = val.split('=')        
        if(key==col_list[idx]):
            # print 'Col name match'
            # print value
            csv_row.append(value)
        else:
            csv_row.append(None)
            print 'Key-to-Column Mismatch, dropping value.'
    print csv_row
    global l_of_l
    l_of_l.append(csv_row)

l_of_l=[]
colList=['A', 'B', 'C']
records.foreach(lambda x: processRecord(x, col_list=colList))

print 'Final list of lists:\n'
print l_of_l

输出:

Records pre-transform:
[[u'A=foo', u'B=bar', u'C=baz'], [u'A=oof', u'B=rab', u'C=zab'], [u'A=aaa', u'B=bbb', u'C=zzz']]
------------------------------

[u'foo', u'bar', u'baz']
[u'oof', u'rab', u'zab']
[u'aaa', u'bbb', u'zzz']

Final list of lists:
[]

1 个答案:

答案 0 :(得分:1)

尝试此功能:

def processRecord(record, col_list):    
    csv_row=list()
    for idx, val in enumerate(record):
        key, value = val.split('=')        
        if(key==col_list[idx]):
            # print 'Col name match'
            # print value
            csv_row.append(value)
        else:
            csv_row.append(None)
            # print 'Key-to-Column Mismatch, dropping value.'
    return csv_row

然后

colList=['A', 'B', 'C']
l_of_l = records.map(lambda x: processRecord(x, col_list=colList)).collect()

print 'Final list of lists:\n'
print l_of_l

应该给出

Final list of lists: 
[[u'foo', u'bar', u'baz'], [u'oof', u'rab', u'zab'], [u'aaa', u'bbb', u'zzz']]