Bash command

android70 · 08-05-2012, 03:25 PM

Hi all,

I have a file that has the follow data

#cat file.txt

+2012_01
ferrari-honda-chevrolet
audi-bmw
mercedes
+2012_02
google-android
yahoo
twitter
+2012_03
ubuntu
fedora
suse

I want a command ou script that saves the output to other file like this:

+2012_01
ferrari-honda-chevrolet audi-bmwmercedes
+2012_02
google-androidyahootwitter
+2012_03
ubuntufedorasuse

Basically, I just want the lines that don't start with "+" to merge in one line

Thanks in advanced

pixellany · 08-05-2012, 03:39 PM

Do you really want them merged with no spaces or punctuation?
Regardless, SED will do it. Go here and read up on SED: http://www.grymoire.com/Unix/Sed.html

Basically, the approach is to read a line into the working register, move to the hold register, and then get the next line and test it for "+". If it does NOT have +, then append it to the hold register. When you come to the next line starting with +, then switch it with the hold register (which now contains all the previous lines), remove the newlines, and then print the combination line.

This seems like it might be homework---but regardless, tell us more about what commands and utilities you are familiar with.

android70 · 08-05-2012, 04:07 PM

I used sed and tr command

sed -n '/+/,+3p' file.txt | tr -d '\n' > NewFile.txt

and the output was something like this

+2012_01ferrari-honda-chevroletaudi-bmwmercedes+2012_02
google-androidyahootwitter+2012_03ubuntufedorasuse

I also used small script like this

Code:

input=File.txt

while 
read line

do 

if [[ $line = *'+'* ]];
then

echo $line >> NewFile.txt
 

elif [[ $linha != *'+'* ]]; then
echo $line | tr -d '\n' >> NewFile.txt

fi

done < $input

pixellany · 08-05-2012, 04:22 PM

We can see how the first example would not work (tr removes all newlines).

I cannot immediately see how the 2nd example would work...

grail · 08-05-2012, 08:08 PM

Assuming you guarantee 4 lines per entry:

Code:

awk 'ORS=NR%4?" ":"\n"' file

android70 · 08-06-2012, 03:00 AM

Quote:

Originally Posted by grail

Assuming you guarantee 4 lines per entry:

Code:

awk 'ORS=NR%4?" ":"\n"' file

It's almost that. The output was

+2012_01 ferrari-honda-chevrolet audi-bmw mercedes
+2012_02 google-android yahoo twitter
+2012_03 ubuntu fedora suse

but I wanted lines with plus sign in their own line

+2012_01
ferrari-honda-chevrolet audi-bmw mercedes
+2012_02
google-android yahoo twitter
+2012_03
ubuntu fedora suse

amboxer21 · 08-06-2012, 03:08 AM

Quote:

Originally Posted by grail

Assuming you guarantee 4 lines per entry:

Code:

awk 'ORS=NR%4?" ":"\n"' file

Grail can you break down your solution for me please? I was not aware that you could assign NR to ORS like that. Thanks

grail · 08-06-2012, 09:45 AM

@ amboxer21 - the trick here is that we are not assigning anything prior to the '?'. As a combination, ?:, is equivalent to a shortened form of and if/else. Everything after '=' and prior to '?'
is an expression to be tested and when true it will return the portion between '?' and ':', and when false it will return the portion after ':'.

So:

ORS = " ", when NR % 4 is any non-zero value, ie anything not exactly divisible by 4

ORS = "\n", when the above is false, ie. NR == 4 or 8 or 12 ...

@OP - Sorry, misread the output data, try:

Code:

awk 'ORS = (NR%4 && /^[^+]/)?" ":"\n"' file

pixellany · 08-06-2012, 01:12 PM

For completeness, here is a SED solution:

Code:

[mherring@herring_lap play]$ sed -n '$b1;/^+/!{H;b2};:1;x;s/\n/ /g;p;:2' linem

+2012_01 ferrari-honda-chevrolet audi-bmw mercedes
+2012_02 google-android yahoo twitter
+2012_03 ubuntu fedora
[mherring@herring_lap play]$

It has 2 issues to be fixed--the extra line at the beginning, and the deletion of the newline after the date.

pixellany · 08-06-2012, 01:26 PM

Fixed one of the issues:

Code:

sed -n '$b1;/^+/!{H;b2};:1;x;s/\n/ /2g;p;:2' filename > newfilename

the "2g" tells SED to do the substitution on the 2nd instance and every one after.

whizje · 08-06-2012, 01:33 PM

Not the most beautiful solution but it's get the job done.

Code:

bash-4.2$ cat file.txt |tr "\n" " "| tr "+" "\n+";echo

2012_01 ferrari-honda-chevrolet audi-bmw mercedes 
2012_02 google-android yahoo twitter 
2012_03 ubuntu fedora suse

whizje · 08-06-2012, 01:56 PM

The + disappeared

Code:

bash-4.2$ cat file.txt |tr "\n" " "| sed 's/\+/\n+/g';echo

+2012_01 ferrari-honda-chevrolet audi-bmw mercedes 
+2012_02 google-android yahoo twitter 
+2012_03 ubuntu fedora suse