UNIX for Dummies Questions & Answers

What I want to do is delete everything upto the last underscore (_) in column 2.

Code:

 awk '{ $2=$(echo $2 | sed 's/.*_//'); print $0}'

Sed works fine if I echo the string into it, this doesnt work inside awk. What am I doing wrong?

Similarly, how do I store the substring starting with a keyword KEY,

So if my 2nd column is

Code:

WGGG_TTGSGGDS_KEY12-34

then I want to output

Code:

KEY12-34

Code:

In
23 WGGG_TTGSGGDS_KEY12-34 45
26 WGGG_TTGSG67S_KEY14-34 47

Out
23 KEY12-34 45
26 KEY14-34 47

Why should it work inside awk?
1) your construct can't work; you can't use command substitution inside awk.
2) did you consider the sub/gsub functions in awk?

RudiC is correct, here is the long version about process substitution. It is a shell construct which works like that: "$( <command> )" is taken, then "command" is executed in a separate subshell and any output it produces is placed instead of the construct prior to evaluating the rest of the line it is part of, very similar to expanding a variable. The following two examples illustrate this and the similarity of the process (showing the steps in which the shell decodes the respective lines):

Code:

var="my text"
echo "$var"
   |
   V
echo my text
   |
   V
my text

Code:

echo $(echo "my text")
   +---> echo "my text"
   |        |
   |        V
   |     my text
   |        |
   | <------+
   |
   V
echo my text
   |
   V
my text

It is clear that such a syntactic device of the shell will not work outside the shell - like inside awk. For the same reasons PASCALs "BEGIN"-statements do not work in C and curly braces not in PASCAL.

Like RudiC rightfully said, awk can perform "global substitutions" (->gsub()) which do about the same as the "s/.../.../" (substitute) command in sed. It even uses the same regexps so the conversion of your sed statement should be pretty straightforward.

I hope this helps.

bakunin

Thank you for great explainations Rudi and Bakunin.

Hello Senhia83,

Following awk solutions may help you in same.

Code:

i-  awk '{gsub(/.*\_/,X,$2);print $0}' Input_file
ii- awk '{match($2,/KEY.*/);$2=substr($2,RSTART,RLENGTH)} 1' Input_file
####Considering that all your data will have word KEY present in 2nd column.

Output will be as follows in both conditions.

Code:

23 KEY12-34 45
26 KEY14-34 47

Thanks,
R. Singh

For the RE used in this example, the RE used by sed and the RE used by awk sub(...) are the same. In general, they aren't quite the same. The REs used by sed are defined to be basic regular expressions and the REs used by awk are defined to be extended regular expressions; so in some cases you can't directly use a sed substitute command BRE as an awk sub() or gsub() ERE.

There is no need for the backslash in this ERE since underscore is not a special character in an ERE (nor in a BRE). And, there can't be more than one match for the ERE .*_ so gsub() isn't needed here; sub() will work just as well. And print $0 is longhand for print, so this can be simplified to:

Code:

awk '{sub(/.*_/,X,$2);print}' Input_file
        or even
awk '{sub(/.*_/,X,$2)}1' Input_file