Tech Tip: Fun With Gawk


When grep and sed aren't enough, gawk may provide the extra horsepower that you need. The following tip contains a sampling of some of the things one might do with gawk.

Extract the last column from a text file, whitespace-separated:

cat myfile | gawk '{print $NF}'


gawk '{print $NF}' myfile

List counts of files owned by each user in the current directory:

/bin/ls -l | \
    gawk 'NR > 1 {counts[$3]++;}
          END {for (s in counts) {
                   printf("  %-15s : % 5d\n",
                          s, counts[s]);}}' | \

Kill your processes (one use is to kill a hung login if you can remotely log in to the workstation from another machine):

ps -elf | \
    gawk -v me="$USER" '$3 == me {print $4}' | \
        egrep -v $$ | \
            xargs -i@@ kill -9 @@; kill -9 $$


Comment viewing options

Select your preferred way to display the comments and click "Save settings" to activate your changes.


James Hinnant's picture

Thanks for the tips, especially pkill (which is new with Solaris 7; 1998! :) ).

The file count routine I use for counting source files in the current directory, so strange filenames are not an issue (also I find find to be kind of slow).

Here is the online manual for gawk:

Here is a little program people might remember coding from FORTRAN class; maybe it can help put the baby to sleep:

echo 100 1 10 0.01 | gawk '{amp = $1; freq = $2; cycles = $3; tdelta = $4; for (i = 0; i <= 2 * 3.14159 * cycles; i += 2 * 3.14159 / 360) {printf("%*s\n", int((amp/2.0) * sin(i * freq) + (amp/2.0)), "*"); system("sleep " tdelta); }}'

What does this have to do

Anonymous's picture

What does this have to do with Fortran? I use Fortran every day, and have no idea what you're talking about. Also, I was born in '84, long after punch cards and FORTRAN IV etc., which may explain my confusion.

don't parse output of ls

chattr's picture

Too many odd things can be in file names, spaces, newlines, etc.

It's a rare case that all you want is a list of file names. In most cases you want to do something to the files based on their names, and if that's the case, 'find' can be a better tool, or maybe ' stat -c %n [glob]'.

Start learning basic unix commands before playing with awk

przemoc's picture

Sorry, but there is no need to use (g)awk here. I really like (g)awk, but your examples are just poor. Ok, maybe first one is a typical case for awk.

You forgot about dot files, but second one can be replaced with:

ls -Al | sed '1d;s/\([^ \t]\+[ \t]\+\)\{2\}//;s/[ \t].*//' | sort | uniq -c

or even better:

find . -printf "%u\n" -mindepth 1 -maxdepth 1 | sort | uniq -c

Third one (really badly and also wrongly implemented) is a perfect case for pkill:

pkill -9 -U "$UID"

I can agree that your second example might be faster than mine, but it's irrelevant here.