I would like to get the number from rating as output from this

# nc localhost 9571 
session-with-name:Ubuntu Classic (No effects):gnome-session --session=2d-gnome
session-with-name:Ubuntu (Safe Mode):gnome-session -f --session=2d-gnome
session-with-name:Ubuntu Classic:gnome-session --session=classic-gnome

I can do it like this

# nc localhost 9571 | grep rating | cut -d: -f2

but could awk be used instead for a simpler solution?

  • This question appears to be off-topic because it is about programming in the Awk interface and is more relevant on StackOverflow.com
    – Magellan
    Oct 11 '13 at 19:43
$ nc localhost 9571 | awk -F: '/rating/ { print $2 }'
  • 7
    Can give wrong results if any of the field values or names contain the string "rating", i.e. one of the session names has the word "rating" in it. Better: awk -F: '$1 == "rating" {print $2}' Oct 4 '11 at 0:39
  • 3
    Or maybe awk -F: '/^rating:/ { print $2 }'?
    – user
    Oct 4 '11 at 12:45
  • I know that, but I convert based on the OP's command.
    – quanta
    Oct 4 '11 at 12:47
  • @IngmarHupp I thought exactly the same thing. It is in fact so much better to match on the exact record in question that I've edited the answer with this change. This will hopefully help people learn to use awk more effectively. Apr 1 '15 at 20:25

Quanta beat me to it, but I'll include a sed variant if you're that way inclined:

nc localhost 9571 | sed -ne 's/^rating://p'

Ditto what MadHatter said, though. Your current solution is perfectly sound. (Although I'd grep for "^rating:" rather than just the word to ensure you only get the line you want.)

  • I love sed! It is so underutilized and underappreciated...
    – Mei
    Oct 3 '11 at 15:59

You can also just use the shell:

nc localhost 9571 | 
while IFS=: read key val; do [[ $key = "rating" ]] && echo "$val"; done
  • This is the quickest way - and prevents spawning any processes beyond nc. Nice!
    – Mei
    Oct 3 '11 at 15:50

yes, you can (and should) use (one) awk instead of (two) grep and cut:

$ nc localhost 9571 | awk -F: '/^rating:/ { print $2 }'

Be sure to match you line as good as you can to avoid ugly bugs.

  • /rating/ works,
  • /^rating/ is better (safer),
  • /^rating:/ is best (in this case).

It can:

nc localhost 9571  | awk -F: '{ if ($1 == "rating") print $2 }' 

(I don't know what you're doing with localhost above, so used your output as input to my awk command, then replaced the "cat" with "nc ..." - but it should be fine.)

But why would you do this? The UNIX way is to have lots of small tools, each of which does one thing well, strung together via pipelines, rather than using larger multifunction tools. You need to select a single matching line, the select one field therefrom: grep selects lines with matches, and cut selects fields from input lines; they're perfect for the task. But if you really do want to do it all-in-one with awk, well, there you go.

  • One reason is that using grep and cut requires spawning two processes, and using awk (or sed) requires only one. Spawning a process is computationally expensive. Also, unlike using perl or ruby (et al), sed and awk are much more lightweight up front.
    – Mei
    Oct 3 '11 at 15:48
  • 2
    Fair enough, though I note that the size of the awk binary on my (F15) system is 360948 bytes, while the grep and cut binaries together are 170824. So less cycles with only one fork+exec, but more memory, and more IO to get the binary off disc. Honestly, I doubt that any of these considerations matter on modern systems, but if you're minded to minimise, there you go!
    – MadHatter
    Oct 3 '11 at 16:46
  • Foo Bah: thanks for your edit proposal, but quanta made a later post than mine which is even more streamlined than your suggestion. I preferred mine because it's easier for an entry-level awk user to see how it works, and therefore rejected your edit. Thank you for the thought, though!
    – MadHatter
    Oct 3 '11 at 16:50

While quanta's answer is the easiest and the one I would write too, I bet you didn't know GNU awk (gawk) can do networking too? You can write the entire thing with awk if you really want:

    FS = ":"
    f = "/inet/tcp/0/"
    while ((f |& getline) > 0)
        if ($1 ~ /^rating$/) {print $2}

This might be useful if a server doesn't have nc installed, nc has restricted perms, or if you just really like awk.

  • Oooh. Nice. Networking and a coprocessing example. Jun 29 '12 at 0:04

You can also use sed,

nc localhost 9571 | sed -n "s/^rating:\([\d]*\)/\1/p"

This will make sure that "rating" starts the line, and that digits follow the rating:


If Perl is an option:

nc localhost 9571 | perl -F: -lane 'print $F[1] if /rating/'

-a autosplits each line into the @F array
-F: uses : as the field separator, instead of the default of whitespace
/rating/ returns true if the regex is found
$F[1] is the 2nd element of the @F array.
Perl arrays start with element 0, whereas awk starts with $1

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.