llazzaro
llazzaro

Reputation: 4188

awk save command ouput to variable

I need to execute a command per line of some file. For example:

file1.txt 100 4
file2.txt 19 8

So my awk script need to execute something like

command $1 $2 $3

and save the output of command $1 $2 $3, so system() will not work and neither will getline. (I can't pipe the output if I do something like this.)

The restriction to this problem is to use only awk. (i already had a solution with bashscriot + awk...but I only want awk...just to know more about this)

Upvotes: 4

Views: 3678

Answers (2)

Peter Cordes
Peter Cordes

Reputation: 364039

Awk's system() function passes the string to /bin/sh, so you can use redirect operators, like ">file.out" if you want.

awk '{system("command " $1 " " $2 " " $3 ">" $1 ".out");}'

Edit: ok, by save, you mean into an awk variable. ephemient is on the right track, then. That's what awk's getline does, like backticks or $(cmd) in shell/perl. In fact, google for awk backticks found this: http://www.omnigroup.com/mailman/archive/macosx-admin/2006-May/054665.html

You say you can't use getline because then you couldn't pipe. But you can work around that with tee and file-descriptor tricks. This works if /bin/sh is bash:

{ "set +o posix; command " $1 " " $2 " "  $3  " | tee >(grep foo)"  | getline var; print toupper(var); } # bash-only, and broken.

set +o posix is necessary because awk runs bash as sh, which makes it go into posix mode after readings its startup files. Hmm, I'm not having any luck getting that to work, and it requires bash anyway.

Ok, this works:


    $ touch foo bar
    $ echo "foo bar" | 
      awk '{ "{ ls " $1 " " $2 " "  $3  " | tee /dev/fd/10 | grep foo > /dev/tty; } 10>&1"  | getline var; print toupper(var); }'
    foo
    BAR

Upvotes: 0

ephemient
ephemient

Reputation: 204688

What's wrong with using getline?

$ ./test.awk test.txt
# ls -F | grep test
test.awk*
test.txt

# cat test.txt | nl
     1  ls -F | grep test
     2  cat test.txt | nl
     3  cat test.awk

# cat test.awk
#!/usr/bin/awk -f
{
        cmd[NR] = $0
        while ($0 | getline line) output[NR] = output[NR] line RS
}
END {
        for (i in cmd) print "# " cmd[i] ORS output[i]
}

Upvotes: 3

Related Questions