Grep's line numbers parsed into one line of output.

judgex · 08-13-2006, 07:39 AM

Code:

echo `grep -n foo bar.txt | cut -d: -f1` | tr ' ' :

The -n option to grep prefixes each match with a corresponding line number. The cut gets rid of the matches, giving a list of numbers. The echo converts the newlines into spaces, except for the last newline which is removed. Finally tr is used to convert spaces into colons.

If bar.txt contains foo on lines 15, 16, 21 and 26, the final output will be 15:16:21:26.

Can anyone show me a simpler way of doing this?

Thanks in advance.

druuna · 08-13-2006, 08:59 AM

Hi,

If a : at the end of a line is no problem, you can do this with one command:

awk '/foo/ { printf FNR ":" } END { print"" }' bar.txt

Your output will look like this:

15:16:21:26:

Don't know if you have any experience with AWK, so here's a little breakdown:
/foo/ => looks for lines with foo in them, if that is the case:
printf FNR ":" => print the linenumber (FNR) followed by a : (no newline, that's why printf is used instead of print)
The print"" in the END section prints a newline.

But, like I stated before, there will be an extra : at the end.

Hope this helps.

judgex · 08-13-2006, 10:01 AM

Quote:

Originally Posted by druuna

Hi,

If a : at the end of a line is no problem, you can do this with one command:

awk '/foo/ { printf FNR ":" } END { print"" }' bar.txt

Your output will look like this:

15:16:21:26:

Don't know if you have any experience with AWK, so here's a little breakdown:
/foo/ => looks for lines with foo in them, if that is the case:
printf FNR ":" => print the linenumber (FNR) followed by a : (no newline, that's why printf is used instead of print)
The print"" in the END section prints a newline.

But, like I stated before, there will be an extra : at the end.

Hope this helps.

I like that method. I can get rid of the last colon using:

Code:

awk '/foo/ { printf ":" FNR }' bar.txt | cut -c2-

Thanks for your help.

druuna · 08-13-2006, 10:19 AM

Hi,

Yes, getting rid of the last colon isn't hard, but do you really need to?

Starting a shell is one of the more resource intensive actions, when piping something to something else (awk ... | cut ...) you need a shell for all the commands (2 in this example. The first example you gave needs 3 shells) and pipes between them have to be set up.

It sometimes is wiser to leave the output the way it is, depending on what you are going to do next.

burninGpi · 08-13-2006, 12:54 PM

awk '/foo/{print FNR ":"}' will result in:
15:
16:
21:
26:
...

to solve all the problems:

Code:

awk 'BEGIN{ORS="";first=1}/foo/{if (first==1){first=0;print FNR}else{print ":" FNR}}END{print "\n"}' bar.txt

If you want to search for just the word foo, and not food or afoo, you need to use

Code:

awk 'BEGIN{ORS="";first=1}
/([^[:alnum]]|^)foo([^[:alnum:]]|$)/
{if (first==1){first=0;print FNR}else{print ":" FNR}}
END{print"\n"}' bar.txt

and to match foo case-insensitivitely, replace foo in the RE (between the slashes) with [Ff][Oo][Oo]

judgex · 08-13-2006, 06:34 PM

Thank you Burnin. Those commands do everything I need.

Druuna, I'm not sure if I really need to remove the last colon.

druuna · 08-14-2006, 03:54 AM

Hi,

burninGpi already did the extra awk work, go for that example. It's complete, resource friendly and basically elegant

bigearsbilly · 08-14-2006, 04:10 AM

Code:

grep -n foo bar.txt| cut -f1 -d: | xargs

Code:

 perl -ne 'print "$. " if /foo/' bar.txt

druuna · 08-14-2006, 04:22 AM

Hi,

The perl code snippet has the same pitfall as my command: It will print a charachter (a space in your perl example) after the last found entry.

burninGpi's example still seems to be the only one-liner that works without extra pipe(s).