Display certain line in a file

rsashok · 07-11-2007, 08:04 PM

I know line number in a file which I want to output. Is there a command to accomplish this? For example:
> cmd -n 20 my_file
Would display only line number 20 in 'my_file'

Thanks

pixellany · 07-11-2007, 08:39 PM

What's "cmd"? (Not on my machine)

You would need to state what language you are using.

In bash, you can use sed for this

gilead · 07-11-2007, 08:41 PM

There are at least a couple of ways to do that. To get the 20th line of my_file:

Code:

sed -n '20p' my_file
head -n 20 my_file | tail -n 1

I'm sure others have more...
EDIT: I'm assuming shell scripting

zhangmaike · 07-11-2007, 08:47 PM

Quote:

Originally Posted by pixellany

In bash, you can use sed for this

You can use sed in any shell. It's just another program, like head and tail.

rsashok · 07-11-2007, 10:37 PM

Both 'head' and 'sed' work fine in my Bash scripting file. Much more elegant way then I had before with awkward 'for' loop.

Thanks.

gilead · 07-11-2007, 11:58 PM

Quote:

Originally Posted by zhangmaike

You can use sed in any shell. It's just another program, like head and tail.

I think his point was that you could do this in other languages (perl, python, etc.), not just shells:

Code:

#!/usr/bin/perl -w

use strict;

my $line;

open (MYFILE, "/path/to/my_file") or die "Can't open data file: $!\n";

my $count = 0;
while ($line = <MYFILE>) {
  $count++;
  if ( $count == 20 ) {
    print $line;
    last;
  }
}

close (MYFILE);

theNbomr · 07-13-2007, 04:06 AM

If the file is not huge, perl can do it as a one-liner:

Code:

perl -e '@file=<>;print $file[19];' yourFileName.ext

I assumed that your line numbers start at 1.
--- rod.

syg00 · 07-13-2007, 04:49 AM

Much as I like perl, for one line, sed has this shot. It'll read the entire file, but that generally isn't an issue for a (semi-)reasonable sized file.

The ability to "short circuit" is handy for *really* big files - perl is useful here. I recently did some testing on printing blocks of lines from a file of 100 million lines. Short circuiting with Perl won - big time. Of course that was in scalar context.

ghostdog74 · 07-13-2007, 05:19 AM

Quote:

Originally Posted by rsashok

I know line number in a file which I want to output. Is there a command to accomplish this? For example:
> cmd -n 20 my_file
Would display only line number 20 in 'my_file'

Thanks

Code:

awk 'NR==20' file

theNbomr · 07-13-2007, 12:27 PM

Quote:

Originally Posted by syg00

Much as I like perl, for one line, sed has this shot. It'll read the entire file, but that generally isn't an issue for a (semi-)reasonable sized file.

The ability to "short circuit" is handy for *really* big files - perl is useful here. I recently did some testing on printing blocks of lines from a file of 100 million lines. Short circuiting with Perl won - big time. Of course that was in scalar context.

By "short-circuit", are you referring to the '@array=<FILE>;' idiom? Most day-to-day work is well within that size. If I have files larger than that, I hope I know about it before processing, and I take the large size into account.

--- rod.

rsashok · 07-13-2007, 06:33 PM

I found another method of printing using 'sed' which is good especially for big files since it stops when it reaches desired line number without reading the whole file.

sed '20q;d' my_file

I don't quite understand why it works though: 20q - tells to quit after first 20 lines; d - delete these lines. But it works. If somebody knows 'sed' good enough to explain, then please.

jschiwal · 07-13-2007, 06:44 PM

I would use "sed -n '21q;20p' file". For GNU sed, you can have two sed commands separated with a semicolon. Otherwise the command could be "sed -n -e '21q' -e '20p'"

This isn't hard to understand. The -n suppresses the output of every line read. The '21q' tells it to quit if it has read line 21. The '20p' says to print the line if it is line 20.

syg00 · 07-13-2007, 09:36 PM

Cute - thanks jschiwal. This; "sed -n '20p;20q' file" also works, and for me is more (human) "readable".
As does this: "sed -n '20,23p;23q' file".
So now I can toss my perl script - thanks jschiwal ....

This is what I meant by "short circuit" (instead of needlessly continuing to read the entire file).

jschiwal · 07-14-2007, 09:56 PM

Head prints the head of a file; tail reads the end of a file. We need a "tummy" command to read the middle of a file!

Code:

#!/bin/bash
if [[ ${#@} -ne 3 ]]; then
   cat <<-EOF
      usage: $0 filename startnumber endlinenumber
EOF
   exit 1
fi

sed -n "$2,$3p;$(($3+1))q" "$1"

syg00 · 07-15-2007, 09:49 PM

Just for entertainment I decided to run some tests on a file (5.7 million+ text records) on my laptop. Reboot before every run to make sure no cache issues.

Test 1: print records 100000-100005
1.390 secs - my perl (quits after the print; i.e. short circuits)
1 minute 14.323 secs - sed (without quit)
1.516 secs - sed with quit.

Test 2: print records 5700000-5700005
50.752 secs - my perl (quits after the print; i.e. short circuits)
1 minute 3.887 secs - sed with quit.

For me makes it worthwhile using the "short circuit" option on decent sized files.
And yes, I was surprised the perl ran faster - for that difference, you would just use sed. Not an exhaustive test, but interesting.