awk, how to redirect to stderr

[* Tong *]

Hi,

Is there any way to redirect awk output to stderr?

From the man page I can see that awk only allow to redirect to file;
Moreover, I tried the following but failed:

$ awk 'BEGIN {print 3 > STDERR; }'
awk: fatal: expression for `>' redirection has null string value

Please help.
thanks

tong

[Klaus Alexander Seistrup]

Post by * Tong *
$ awk 'BEGIN {print 3 > STDERR; }'

This works for me:

$ awk 'BEGIN {print 3 >"/dev/stderr";}'

but it might be a gnuism

$ awk --version | head -1
GNU Awk 3.1.4
$

Cheers,

[Kenny McCormack]

Post by Klaus Alexander Seistrup

Post by * Tong *
$ awk 'BEGIN {print 3 > STDERR; }'

$ awk 'BEGIN {print 3 >"/dev/stderr";}'
but it might be a gnuism

It's not a GNU-ism, it's a system-ism.

It will work on any system that supports the /dev/stderr construct;
i.e., most modern Unixes. You could also try /dev/fd/2.

The AWK manual gives the "portable" (i.e., ugly) alternative of:

print | "cat >&2"

P.S. In TAWK, you can do: print > stderr
because "stderr" is a magically supplied value (like in C).

[Stephane CHAZELAS]

Post by Kenny McCormack

Post by Klaus Alexander Seistrup

Post by * Tong *
$ awk 'BEGIN {print 3 > STDERR; }'

$ awk 'BEGIN {print 3 >"/dev/stderr";}'
but it might be a gnuism

It's not a GNU-ism, it's a system-ism.

Not necessarily, some awks may consider "> /dev/stderr"
specially just like GNU bash on systems that don't have such
device files. I might be wrong, but I think gawk is one of
those. To be double checked.

See POSIX uuencode for standard tools that follow that
kind of convention (for /dev/stdout only though).

Post by Kenny McCormack
It will work on any system that supports the /dev/stderr construct;
i.e., most modern Unixes. You could also try /dev/fd/2.

I wouldn't say most. Anyone has figures?

Post by Kenny McCormack
print | "cat >&2"
P.S. In TAWK, you can do: print > stderr
because "stderr" is a magically supplied value (like in C).

[Andre Majorel]

Post by Stephane CHAZELAS

Post by Kenny McCormack

Post by Klaus Alexander Seistrup
$ awk 'BEGIN {print 3 >"/dev/stderr";}'
but it might be a gnuism

It's not a GNU-ism, it's a system-ism.

Not necessarily, some awks may consider "> /dev/stderr"
specially just like GNU bash on systems that don't have such
device files. I might be wrong, but I think gawk is one of
those. To be double checked.

I have vague memories of using ">/dev/stderr" on MS-DOS with GNU
awk 2 point something. That was a while ago, though, so don't
quote me on that.

[Ed Morton]

Post by Kenny McCormack

Post by Klaus Alexander Seistrup

Post by * Tong *
$ awk 'BEGIN {print 3 > STDERR; }'

$ awk 'BEGIN {print 3 >"/dev/stderr";}'
but it might be a gnuism

It's not a GNU-ism, it's a system-ism.
It will work on any system that supports the /dev/stderr construct;
i.e., most modern Unixes. You could also try /dev/fd/2.
print | "cat >&2"
P.S. In TAWK, you can do: print > stderr
because "stderr" is a magically supplied value (like in C).

Obviously in any awk you can do this:

awk 'BEGIN{ stderr="cat >&2" }
...
{ print "ERROR: <scriptname>: foo" | stderr }'

though I prefer:

awk 'function prterr(txt) { printf "ERROR: <scriptname>: %s\n",txt |
"cat >&2" }
...
{ prterr("foo") }'

so you have one function for logging errors so it's easy to instrument
it later to do other things (e.g. prefix common error text as above,
dump variables, etc.) if you want.

Ed.