Shell Programming and Scripting

Code:

sed 's/|[^|]*$//' file

If all you want to do is to get rid of the last field and none of your input lines contain a single field, MadeInGermany's suggestion using sed is probably better than using awk. If you want to do this using any awk (/usr/xpg4/bin/awk, /usr/xpg6/bin/awk, or nawk on Solaris/SunOS systems), try:

Code:

nawk '
BEGIN {	FS = OFS = "|" }
NF < 2 {print ""; next }
{	NF -= 1; $1 = $1; print }
' file

Code:

perl -wlnaF\| -e '@F = splice @F, 0, -1; $" = "|"; print "@F"' file

Code:

perl -wlnaF\| -e '$" = "|"; print "@F[0..$#F-1]";' file

Empty the last field then remove its delimiter

Code:

sed 's/[^|]*$//;s/|$//' file

The trouble with manipulation of NF is that its effect is not specified by POSIX, so if there is an effect, it is implementation-specific.
So I think in awk the regex route is the most reliable way, or removing the last field and field separator, so:

Code:

awk '{$NF=x; sub(/\|$/,x)}1' FS=\| OFS=\| file

(which is analogous to MadeinGermany's sed suggestion) or

Code:

awk '{sub(/\|?[^|]*$/,x)}1' file

With sed:

Code:

sed 's/|\{0,1\}[^|]*$//' file

GNU sed -r or BSD sed -E :

Code:

sed -r 's/\|?[^|]*$//' file

--
Hi Don, that works with nawk, but it does not appear to work with /usr/xpg4/bin/awk .
I have not come across /usr/xpg6/bin/awk is that part of Solaris 11 or is that available as part of an additional package?

Hi Scrutinizer,
Thanks for the information. I no longer have access to a Solaris system to test these types of things. I didn't realize there was a difference in the way NF was handled in nawk and /usr/xpg[46]/awk. (One key difference between the three is which shell is used when you use the awk system() function (/bin/sh, /usr/xpg4/bin/sh, or /usr/xpg6/bin/sh, respectively)). The Solaris nawk comes straight from the AT&T new awk from the 1980's. I don't remember where the base source for /usr/xpg[46]/bin/awk came from.

The /usr/xpg6 hierarchy should be available on any Solaris system released since about 2004 or 2005, but it may have been an install time option rather than part of the default install.

The code I suggested works on OS X, so it probably works on BSD derived versions of awk. (And, this isn't surprising since the new awk came out of AT&T's Research UNIX group rather than AT&T's Programmer's Workbench UNIX group.)

Code:

nawk -F"|"  'NR>0{NF-=1}$1=$1' OFS="|" file