unix logic to merge lines based on search

ravi t · 04-04-2012, 05:40 AM

Hi,

I want to merge lines based on string starts with ALT.
in below example ALT is header record and RMK is child records. child record is dynamic.
input file
ALT 00623053520 20120306041920120320195927
RMK THIS IS A TEST FOR CODE RED LINE 1
RMK THIS IS A TEST FOR CODE RED LINE 2
ALT 00623053520 20120306041920
RMK STILL LOOKING
output file :-
00623053520 20120306041920120320195927THIS IS A TEST FOR CODE RED LINETHIS IS A TEST FOR CODE RED LINE

I want to merge lines till next master record comes. like in above case want to merge line 1,2,3 in one line and in second line next master and child record in one line.

Please let me know how i can do in unix..

thanks in advance..

grail · 04-04-2012, 06:23 AM

What have you tried? My initial suggestion would be something like or awk or Perl, but it it could be done in bash or even with sed. So you will need to give some direction as to what you know about
and where you are stuck.

danielbmartin · 04-04-2012, 06:52 AM

Quote:

Originally Posted by ravi t

I want to merge lines based on string starts with ALT.

Perhaps merge is the wrong term because merge usually means combining two files according to a specific algorithm. It seems you want to catenate records such that no data is lost but every output record starts with ALT and then remove all those leading ALTs.

Try this...

Code:

sed 's/$/~/' $InFile   \
|tr '\n' ' '           \
|sed 's/~ ALT/\nALT/g' \
|sed 's/~//g'          \
|sed 's/ALT //'        \
|sed '$s/$/\n/'

Daniel B. Martin

firstfire · 04-04-2012, 08:30 AM

Hi.

If you can live with two extra empty lines, here is an awk one-liner:

Code:

$ awk -F'\nRMK *' -vRS='\n*ALT *' '{$1=$1} 1' infile.txt 

00623053520 20120306041920120320195927 THIS IS A TEST FOR CODE RED LINE 1 THIS IS A TEST FOR CODE RED LINE 2
00623053520 20120306041920 STILL LOOKING

Here -F sets the Field Separator, -vRS sets the Record Separator. I had to use '{$1=$1}' to force recomputing the record so that OFS (Output Field Separator), which equals to space by default, is used. Maybe there is a more clever way to do this.

To remove empty lines use sed '/^$/d'.

ravi t · 04-12-2012, 04:54 AM

Thank you guys its working fine

grail · 04-12-2012, 08:17 AM

Quote:

Originally Posted by firstfire

Maybe there is a more clever way to do this.

What you have is fine and if you remove the braces and the 1 at the end it will print the same for you as the evaluation of assigning $1 to $1 equates to true and hence a 1 which will result in a print
line.

firstfire · 04-12-2012, 12:32 PM

Quote:

Originally Posted by grail

What you have is fine and if you remove the braces and the 1 at the end it will print the same for you as the evaluation of assigning $1 to $1 equates to true and hence a 1 which will result in a print
line.

Indeed. Thanks.