matching and merging lines with awk & printing with Solaris

Question

I want to merge the matching pattern with awk on a Solaris system.

$ luxadm probe 
 No Network Array enclosures found in /dev/es

 Found Fibre Channel device(s): 
 Node WWN:2000000c50f1c8da Device Type:Disk device 
 Logical Path:/dev/rdsk/c1t1d0s2 
 Node WWN:2000000c50f1d9c9 Device Type:Disk device 
 Logical Path:/dev/rdsk/c1t0d0s2

Note - While I am trying with awk, it takes Only one line merging from last.

$ luxadm probe |awk '/Node/{n=$2}/Logical/{l=$2} END {print n,l}' 
 WWN:2000000c50f1d9c9 Path:/dev/rdsk/c1t0d0s2

While the output expected is:

WWN:2000000c50f1d9c9 Path:/dev/rdsk/c1t0d0s2
WWN:2000000c50f1c8da Path:/dev/rdsk/c1t1d0s2 

Answer 1

The short answer which isn't mentioned yet in awk

awk '/Node/{n=$2; getline; print n " "$2; next}' filename

• /Node/ # Look for pattern Node and store the second field in n variable n=$2
• getline # Move to the next line (This is where Logical line starts), print what's stored in n variable and then second field of the next line.
• next # Stop processing current record and move to the next one.

Answer 2

The reason it's faililng is because you are only printing the variables you capture once, in an END{} block. This means that only the last two will be printed. @Costas already gave you a more elegant approach, but you could also use the same logic you were trying with a small change:

awk '{ if(/Node/){
        if(length(n)){print n,l} ## if we have an n, print them
        n=$2
        }
        if(/Logical/){l=$2}
      } 
        END {print n,l}' 

Alternatively, you could use other tools:

1. GNU grep/perl

   luxadm probe |  grep -Po '^\s*(Node|Logical)\s*\K.*?(\s|$)' file | 
       perl -lne '$. % 2 ? printf : print'
   

2. Perl

   luxadm probe |  
     perl -ne 'chomp;if(s/.*(WWN:.*?)\s.*/$1 / || s/.*(Path:.*?)$/$1\n/){print}'
   

Answer 3

You should use sed, in my opinion. In the first place, it will usually outperform awk, and, in this case, it offers a more simple solution overall:

sed -n 'N;s/Node \([^ ]*\).*Logical/\1/p' <infile

...which, given your example data, prints...

 WWN:2000000c50f1c8da Path:/dev/rdsk/c1t1d0s2
 WWN:2000000c50f1d9c9 Path:/dev/rdsk/c1t0d0s2

For each line read, sed also gets the Next appended to pattern space following an inserted \newline delimiter. So the above s///ubstitution spans two lines, and prints the two fields you want separated only by a space and only when a match is found.

The above sed syntax should be fully portable to any POSIX system.

Answer 4

If you want awk

awk -v RS=' Node ' -v FS=' Logical Path:| ' '/WWN/{print $1,$6}'

Other (may be simple)

awk -F":| " '/Node/{printf "%s ", $3":"$4}/Logical Path/{print $4}'

with remain Path:

awk '/Node/{printf "%s ", $2}/Logical Path/{print $2}'

or

awk '/Node/{wwn=$2}/Logical Path/{print wwn,$2}'

And GNU sed at last:

sed -n '/WWN/{s/^\s*\S* *\|\( \S*\)\{3\}\s*$//g;N;s/\n\s*\S*//;p}'