如何使用Python清除大型格式错误的CSV文件

时间:2015-02-12 01:13:50

标签: python csv data-cleansing malformed

我尝试使用Python 2.7.5来清理格式错误的CSV文件。 CSV文件相当大(超过1GB)。文件的第一行正确列出了列标题,但之后每个字段都在一个新行上(除非它是空白的),并且某些字段是多行的。多行字段不用引号括起来,但需要在输出中用引号括起来。列数是静态的并且是已知的。提供的样本输入中的模式在整个文件长度内重复。

输入文件(样本):

Hostname,Username,IP Addresses,Timestamp,Test1,Test2,Test3
my_hostname
,my_username
,10.0.0.1
192.168.1.1
,2015-02-11 13:41:54 -0600
,,true
,false
my_2nd_hostname
,my_2nd_username
,10.0.0.2
192.168.1.2
,2015-02-11 14:04:41 -0600
,true
,,false

期望的输出:

Hostname,Username,IP Addresses,Timestamp,Test1,Test2,Test3
my_hostname,my_username,"10.0.0.1 192.168.1.1",2015-02-11 13:41:54 -0600,,true,false
my_2nd_hostname,my_2nd_username,"10.0.0.2 192.168.1.2",2015-02-11 14:04:41 -0600,true,,false

我走了几条路径,解决其中一个问题只是意识到它没有处理格式错误的数据的另一个方面。如果有人能帮我确定清理此文件的有效方法,我将不胜感激。

由于

修改

我有几个代码片段来自不同的路径,但这是当前的迭代。它不是很漂亮,只是一堆黑客试图弄清楚这一点。

import csv

inputfile = open('input.csv', 'r')
outputfile_1 = open('output.csv', 'w')

counter = 1
for line in inputfile:
    #Skip header row
    if counter == 1:
        outputfile_1.write(line)
        counter = counter + 1
    else:
        line = line.replace('\r', '').replace('\n', '')
        outputfile_1.write(line)

inputfile.close()
outputfile_1.close()

with open('output.csv', 'r') as f:
    text = f.read()

    comma_count = text.count(',') #comma_count/6 = total number of rows

    #need to insert a newline after the field contents after every 6th comma
    #unfortunately the last field of the row and the first field of the next row are now rammed up together becaue of the newline replaces above...
    #then process as normal CSV

    #one path I started to go down... but this isn't even functional
    groups = text.split(',')

    counter2 = 1
    while (counter2 <= comma_count/6):
        line = ','.join(groups[:(6*counter2)]), ','.join(groups[(6*counter2):])
        print line
        counter2 = counter2 + 1

编辑2

感谢@DSM和@Ryan Vincent让我走上正轨。使用他们的想法我制作了以下代码,这似乎纠正了我的格式错误的CSV。我确信有很多可以改进的地方,我很乐意接受。

import csv
import re

outputfile_1 = open('output.csv', 'wb')
wr = csv.writer(outputfile_1, quoting=csv.QUOTE_ALL)

with open('input.csv', 'r') as f:
    text = f.read()
    comma_indices = [m.start() for m in re.finditer(',', text)] #Find all the commas - the fields are between them

    cursor = 0
    field_counter = 1
    row_count = 0
    csv_row = []

    for index in comma_indices:
        newrowflag = False

        if "\r" in text[cursor:index]:
            #This chunk has two fields, the last of one row and first of the next
            next_field=text[cursor:index].split('\r')
            next_field_trimmed = next_field[0].replace('\n',' ').rstrip().lstrip()
            csv_row.extend([next_field_trimmed]) #Add the last field of this row

            #Reset the cursor to be in the middle of the chuck (after the last field and before the next)
            #And set a flag that we need to start the next csvrow before we move on to the next comma index
            cursor = cursor+text[cursor:index].index('\r')+1
            newrowflag = True
        else:
            next_field_trimmed = text[cursor:index].replace('\n',' ').rstrip().lstrip()
            csv_row.extend([next_field_trimmed])

            #Advance the cursor to the character after the comma to start the next field
            cursor = index + 1

        #If we've done 7 fields then we've finished the row
        if field_counter%7==0:
            row_count = row_count + 1
            wr.writerow(csv_row)

            #Reset
            csv_row = []

            #If the last chunk had 2 fields in it...
            if newrowflag:
                next_field_trimmed = next_field[1].replace('\n',' ').rstrip().lstrip()
                csv_row.extend([next_field_trimmed])
                field_counter = field_counter + 1

        field_counter = field_counter + 1
    #Write the last row
    wr.writerow(csv_row)

outputfile_1.close()

# Process output.csv as normal CSV file...    

2 个答案:

答案 0 :(得分:1)

这是关于我将如何解决这个问题的评论。

对于每一行:

我可以轻松识别某些群体的开始和结束:

  • 主机名 - 只有一个
  • 用户名 - 阅读这些直到您遇到的内容看起来不像用户名(以逗号分隔)
  • ip地址 - 阅读这些直到您遇到时间戳 - 用模式匹配标识 - 请注意这些是用空格而不是逗号分隔的。组的结尾由尾随逗号标识。
  • 时间戳 - 易于识别模式匹配
  • test1,test2,test3 - 确定以逗号分隔的字段存在

注意:我会使用&#39;模式&#39;匹配使我能够确定我在正确的地方有正确的东西。它可以更快地发现定位错误。

答案 1 :(得分:0)

从您的数据摘录中看来,任何以逗号开头的行都需要连接到前一行,而以逗号以外的任何行开头的任何行都标记为新行。

如果是这种情况,您可以使用以下代码清理CSV文件,以便标准库csv解析器可以处理它。

#!/usr/bin/python
raw_data = 'somefilename.raw'
csv_data = 'somefilename.csv'
with open(raw_data, 'Ur') as inp, open(csv_data, 'wb') as out:
    row = list()
    for line in inp:
        line.rstrip('\n')
        if line.startswith(','):
            row.append(line)
        else:
            out.write(''.join(row)+'\n')
            row = list()
            row.append(line))
    # Don't forget to write the last row!
    out.write(''.join(row)+'\n')

这是一个微型状态机......在每行中累积行,直到我们找到一条不以逗号开头的行,写下前一行等等。