[SOLVED] How to merge lines based on number of columns

shriankur · 03-09-2018, 05:54 PM

keefaz · 03-09-2018, 06:37 PM

From your example data, here is a sed example to get you started

Code:

sed ':a;N;s/\n|/|/;ba' file

Code:

:a       => create a label named a
 N       => append the next input line
s/\n|/|/ => replace a newline followed by a '|' with a '|', ie delete the newline before
ba       => branch to a, ie goto label a

shriankur · 03-09-2018, 06:45 PM

Thanks. Will it work for any number of columns. Meaning this file should have 4 columns. In the 2nd row I got newline character. Hence needed merge.
I will test and come back to you

Thanks four help

shriankur · 03-09-2018, 11:21 PM

Hi Keefaz - I just tested. It is working :-)
You rock sir.

I also wants to remove junk characters (Any unreadable character) from file.
Could you please help Keefaz?

keefaz · 03-10-2018, 06:17 AM

Just tested another sed command without branching

Code:

sed 'N;s/\n|/|/;P;D' file

For non readable characters, maybe try to remove any char that does not belong to [:print:] character class

shriankur · 03-10-2018, 02:05 PM

TB0ne · 03-10-2018, 02:46 PM

Quote:

Originally Posted by shriankur

Keefaz *HAS* helped you a lot...but you also need to read the "Question Guidelines" link in my posting signature. At some point, *YOU* are going to have to actually do/try something of your own. keefaz has given you a HUGE head-start, but now is the time for you to apply what you were told, and accomplish your own work...or at least TRY to.

Come back and post what you've done/tried/written on your own, and what results you get.

shriankur · 03-10-2018, 02:53 PM

Sure TB0ne. I appreciate Keefaz has really helped me. Let me proceed further on this. Thanks for your reply TB0ne :-)
Sorry I am a newbie in the forum just joined yesterday

shriankur · 03-10-2018, 03:02 PM

Quote:

Originally Posted by keefaz

Just tested another sed command without branching

Code:

sed 'N;s/\n|/|/;P;D' file

For non readable characters, maybe try to remove any char that does not belong to [

rint:] character class

keefaz · 03-10-2018, 04:22 PM

shriankur, TB0ne reminded a forum rule. You have to post some code to get help on it.

shriankur · 03-10-2018, 04:54 PM

Ok fair enough Kafeez.

Input file is

123|APAC|Australia|Sydney
456|APAC|New
Zealand|Aukland

I tried below code to make the file as 4 column delimited file.

awk -F"|" '{if (pre!="") {print pre $0; pre=""} else if (NF>=4) print; else pre=$0}' filename.

Output I got and as needed

123|APAC|Australia|Sydney
456|APAC|NewZealand|Aukland

Problem here is I need to pass 4 as an argument since file columns are varying. Is there any generic solution?

syg00 · 03-10-2018, 05:29 PM

awk has command line options - one of which will allow you to set a variable. Have you tested that code on your own data as specified in post #6 ?. Specifically where input data spans more than 2 lines.

shriankur · 03-10-2018, 06:06 PM

Yes syg00 you are right.
I actually used awk -v

awk -F"|" -v var1="$VARIABLE" '{if (pre!="") {print pre $0; pre=""} else if (NF>=var1) print; else pre=$0}' filename

VARIABLE=4 I will pass as an argument from my shell

Thank you syg00 for your reply.
Special thanks to Kafeez for initial heads up.
I am marking this thread as closed and solved.

keefaz · 03-10-2018, 07:16 PM

Glad you had it working, nice to use forum SOLVED thread feature, not all new users do this

I had a perl solution, but your hawk solution seems cool too

Code:

perl -slne '$s.=$_; (split /\|/, $s)==$col and print($s), $s=""' -- -col=4 filename

BudiKusasi · 03-12-2018, 01:34 AM

Code:

sed -r ':x;/true$/I!{N;s/\n//; bx}' file