Take the 2-minute tour ×
Unix & Linux Stack Exchange is a question and answer site for users of Linux, FreeBSD and other Un*x-like operating systems.. It's 100% free, no registration required.

I have a csv file formatted as below.

"col1","col2","col3","col4"
"1","text1","<p>big
      html
     text</p>
","4th column"
"2","text2","<p>big2
      html2
     text2</p>
","4th column2"

I want to extract the 4th column using. I think that awk is the best tool for this ( let me know if I am wrong). I tried this

awk -F, '{print $4}' myFile.csv 

but it fails. I think because the 3rd column is multiline one. How can I use awk or any other unix command to extract the 4th column. I am looking for an efficient solution since my real file is big (> 2GB)

share|improve this question
1  
With multi-line CSV files using awk becomes even more difficult. You might want to check out using Perl and Text::CSV. stackoverflow.com/questions/3065095/… –  slm Dec 20 '13 at 15:02
    
What does the rest of the file look line? Is that the only record in the file or are there multiple records, each of which can contain a new line? –  terdon Dec 20 '13 at 15:02
1  
Damn, and the entries are not separated by more than one newline? –  terdon Dec 20 '13 at 15:05
    
@slm one you say complicated , do you mean that the solution should be similar to this? –  agstudy Dec 20 '13 at 15:10
    
@agstudy - perl runs on windows and linux. Also what are you planning on using awk on? Also you tagged this Q as linux so you might want to change that if you need this solution on something else! –  slm Dec 20 '13 at 15:11

5 Answers 5

up vote 9 down vote accepted

UPDATE:

Actually, a much easier way is to set the record separator in gawk:

$ gawk 'BEGIN{RS="\"\n"; FS=","}{print $4}' myFile.csv
"col4
"4th column
"4th column2

However, this will remove the trailing " from the end of each column. To fix that you can print it yourself:

$ gawk 'BEGIN{RS="\"\n"; FS=","}{print $4"\""}' myFile.csv
"col4"
"4th column"
"4th column2"

If you don't want the quotes at all, you can set the field separator to ",":

$ gawk 'BEGIN{RS="\"\n"; FS="\",\""}{print $3}' myFile.csv
col3
4th column
4th column2

The only way I can think of One way of doing this is to first modify the file and then parse it. In your example, the newline that actually separates two records is always following a ":

"col1","col2","col3","col4"   <-- here 
1,"text1","<p>big             <-- no "

If that is the case for the entire file, you can replace all newlines that are not immediately after a " with a placeholder and so have everything in a single line. You can then parse normally with gawk and finally replace the placeholder with the newline again. I will use the string &%& as a placeholder since it is unlikely to exist in your file:

$ perl -pe 's/"\s*\n/"&%&/; s/\n//g; s/&%&/\n/;' myFile.csv | awk -F, '{print $4}'
"col4"
"4th column"
"4th column2"

The -p flag for perl means print each line of the input file after applying the script given by -e. Then there are 3 substitution (s/foo/bar/) commands:

  • s/"\s*\n/"&%&/ : This will find any " which is followed by 0 or more whitespace characters (\s*) and then a newline character (\n). It will replace that with "&%&. The quotes are added to preserve the format and the &%& is just a random placeholder, it could be anything that does not appear in your file.

  • s/\n//g; : since the real newlines have been replaced with the placeholder, we can now safely remove all remaining newlines in this record. This means that all lines of the current record have now been concatenated into the current line.

  • s/&%&/\n/ : This turns the placeholder back into a normal new line.

To understand the output of the command run it without gawk:

$ perl -pe 's/"\s*\n/"&%&/; s/\n//g; s/&%&/\n/;' myFile.csv 
"col1","col2","col3","col4"
1,"text1","<p>big      html     text</p>","4th column"
2,"text2","<p>big2      html2     text2</p>","4th column2"

So, you now have your long records on single lines and this is perfectfood for gawk.

You can also do it directly in Perl:

perl -ne '$/="\"\n"; chomp;@a=split(/,/);print "$a[3]\"\n"' myFile.csv
"col4"
"4th column"
"4th column2"

This is using a bit more Perl magic. The $/special variable is the input record separator. By setting it to "\n we tell Perl to split lines not at \n but only at "\n" so that each record will be treated as a single line. Once that is done, chomp removes the newline from the end of the line (for printing later) and split splits each record (on ,) and saves it in the array @a. Finally, we print the 4th element of the array (arrays are numbered from 0 so that is $a[3]) which is the 4th column.

And even more magic, turn on auto spitting (-a) and split on commas (F","). This will split each record into the special @F array and you can print the 4th element of the array:

$ perl -F"," -ane '$/="\"\n";chomp;print "$F[3]"' myFile.csv
"col4"
"4th column"
"4th column2"
share|improve this answer
    
Thanks for your answer. I am really sorry. I edited my question and if you notes I have "1","text1","<p>bi not 1,"text1","<p>bi. Does this solution still work? –  agstudy Dec 20 '13 at 15:20
    
@agstudy yes, as long as the newline at the end of the string is still just after a ". I am writing an explanation of how this thing works, will update in a couple of minutes. –  terdon Dec 20 '13 at 15:22
    
+1!) thank for this great explanation. –  agstudy Dec 20 '13 at 15:39
    
@agstudy see updated answer for a (much) simpler way using gawk. –  terdon Dec 20 '13 at 15:43
1  
+1 for the WOW! –  slm Dec 20 '13 at 16:28

I would recommend using a battle-tested CSV parsing module. For example:

perl -MText::CSV -E '
    $csv = Text::CSV->new({binary=>1}); 
    while ($row = $csv->getline(STDIN)) {say $row->[3]}
' < file.csv
col4
4th column
4th column2

or this produces the same results:

ruby -rcsv -e 'CSV.foreach(ARGV.shift) {|row| puts row[3]}' file.csv
share|improve this answer
    
should I install something to get MText? –  agstudy Dec 20 '13 at 15:34
    
You need to cpan install Text::CSV –  glenn jackman Dec 20 '13 at 15:35
    
I am under windows . cpan don't work. But thanks for your answer.(+1) –  agstudy Dec 20 '13 at 15:38
    
What distribution of Perl do you use in windows? There must be something to install new modules –  glenn jackman Dec 20 '13 at 15:43
    
I have v5.8.8 built for msys –  agstudy Dec 20 '13 at 15:47

Python :

python -c "import csv,sys; print '\n'.join([ r[3] for r in csv.reader(open(sys.argv[1]))])" myfile.csv

A memory-conservative solution for large files that iterates through the file a line at a time unlike the above approach that loads the contents of the file into memory via a list

#!/usr/bin/env python
import sys
import csv
with open(sys.argv[1]) as f:
  for row in csv.reader(f):
    print(row[3])

TEST RESULT of all Solutions:

OS: Ubuntu 12.04

Public CSV Data Download from : http://seanlahman.com/baseball-archive/statistics/

Versions details

root@ubuntu:~# python --version
Python 2.7.3
root@ubuntu:~# ruby --version
ruby 1.8.7 (2011-06-30 patchlevel 352) [i686-linux]
root@ubuntu:~# perl --version

This is perl 5, version 14, subversion 2 (v5.14.2) built for i686-linux-gnu-thread-multi-64int

Result with time

root@ubuntu:~# time python -c "import csv,sys; print '\n'.join([ r[3] for r in csv.reader(open(sys.argv[1]))])" Master.csv > /tmp/python

real    0m1.112s
user    0m0.056s
sys     0m0.316s
root@ubuntu:~# time ruby -rcsv -e 'CSV.foreach(ARGV.shift) {|row| puts row[3]}' Master.csv > /tmp/ruby

real    0m24.582s
user    0m23.397s
sys     0m0.448s
root@ubuntu:~# time perl -MText::CSV -E '
>     $csv = Text::CSV->new({binary=>1});
>     while ($row = $csv->getline(STDIN)) {say $row->[3]}
> ' < Master.csv > /tmp/perl

real    0m7.049s
user    0m5.876s
sys     0m0.468s
share|improve this answer
    
does this solution is efficient? –  agstudy Dec 20 '13 at 16:19
    
file like XML,CSV should be parse using languages like Python,Perl,Ruby.. –  Rahul Patil Dec 20 '13 at 16:26
    
I am an R/python developer. So If I should get a good reason to ask about efficiency here, isn't it? Also here we have some good solutions using unix command. –  agstudy Dec 20 '13 at 16:28
    
Sorry sir, I don't have good ans.. –  Rahul Patil Dec 20 '13 at 16:31
1  
:) come on! you have a python good answer( I already upvote). I am just asking if you solution is efficient to deal with a 2GB file.(personally I prefer to use python also, but here is not as efficient as I want). –  agstudy Dec 20 '13 at 16:36

if it is unix style "\n" ended lines

tr -d "\n" < myfile.csv | awk 'BEGIN{RS=","} !(NR % 4)'

Some fields are multi-lines the tr -d "\n"deletes all the newlines character, creating a stream of "," separated values. the awk is told to use "," as line separator and to print every time (line number modulo 4) is 0.

This only works if the 4th field is the last field (as in your sample). If thats not the case :

tr -d "\n" < myfile.csv | awk 'BEGIN{RS=","; last=12} (++c == 4) (c == last) {c=0}'

It count lines, print the line when count is 4 and reset the count when last field is reached.

share|improve this answer
    
+1! can you please explain a little bit the commands here. –  agstudy Dec 20 '13 at 16:19
    
thanks It is more clear now with the command edit. –  agstudy Dec 20 '13 at 16:27
    
@agstudy while editing and simplifying the script I added an error (forgot the "!" in the awk) . –  Emmanuel Dec 20 '13 at 16:30

Try the following:

while IFS=',' read -r a1 a2 a3 a4 
do 
echo "$a4" >> urname.csv
done < input.csv
share|improve this answer

Your Answer

 
discard

By posting your answer, you agree to the privacy policy and terms of service.

Not the answer you're looking for? Browse other questions tagged or ask your own question.