Hi, I have a file constructed as follows: header , N1 1 Contents 1220 SPC 8 SSN 1 Contents 1221 SPC 8 SSN 1 Contents 1222 SPC 8 SSN footer , N2 The values preceeding contents and ssn never change. I have a sed command that inserts a new batch of three 3 rows (contents, spc, ssn) above the line that ends in N2. The problem is that the numbers preceeding SPC must be in ascending numerical order. Sometimes the spc value in the new batch is < the previous spc value giving me something like this: header , N1 1 Contents 1220 SPC 8 SSN 1 Contents 1221 SPC 8 SSN 1 Contents 1222 SPC 8 SSN 1 Contents 1000 SPC 8 SSN footer , N2 I therefore want to sort in batches of 3 lines by spc which would give me this: header , N1 1 Contents 1000 SPC 8 SSN 1 Contents 1220 SPC 8 SSN 1 Contents 1221 SPC 8 SSN 1 Contents 1222 SPC 8 SSN footer , N2 All is can achieve at the moment is: header , N1 1 contents 1 contents 1 contents 1 contents 8 SSN 8 SSN 8 SSN 8 SSN 1000 SPC 1220 SPC 1221 SPC 1222 SPC footer , N1 Does anybody have any ideas how to sort it as I want, or alternatively place the new batch of three lines in the correct order to start with? Much appreciated Andrew Taylor

I see no way of solving this without breaking apart the data file. It's a kludge, but this appears to work: 1) remove the header and footer, first and last lines: sed -e '$d' -e '1d' data.file > newdata.file 2) In the directory where data.file resides create a directory, mytmp, and split the newdata.file into files with 3 lines: split -l 3 newdata.file mytmp/x 3) rename all the files in the mytmp directory to the 'number preceding SPC' #!/bin/ksh find mytmp -type f |while read fname do mv $fname mytmp/$(sed -n '2p' $fname|awk ' { print $1 } ') done 4) By default, the ls command sorts alphabetically so you can put the file back together with this: #!/bin/ksh ls -1 mytmp|while read ffile do cat mytmp/$ffile >> myfile done # end script I'll leave it to you to put the header and footer back into myfile. Finally, the above should work provided your numbers are four characters. If the number is NOT four characters you'll have to do something like this and sort numerically: # untested ls -1 mytmp|sort -n |while read ffile . .

Thanks for your help, I'll give it a go regards Andrew Taylor

This awk solution outputs each set of 3 lines preceded by the desired sort field, which I have allowed to be as large as 6 digits. After sorting, the prepended sort field is taken off with the cut command.  It assumes that the first line is a header line, but you could easily add an if-statement to confirm that. awk 'BEGIN \ {getline  printf "       %s\n",$0} {if ($1!="footer")    {l1=$0     getline     sortkey=$1     l2=$0     getline l3     printf "%6.6d %s\n",sortkey,l1     printf "%6.6d %s\n",sortkey,l2     printf "%6.6d %s\n",sortkey,l3    }  else     printf "zzzzzz %s\n",$0 } ' file.in |   sort -k1,1 -k3,3n |   cut -c-8- > file.out Below is the direct output from awk prior to being piped into sort:       header , N1 001220 1 Contents 001220 1220 SPC 001220 8 SSN 001221 1 Contents 001221 1221 SPC 001221 8 SSN 001222 1 Contents 001222 1222 SPC 001222 8 SSN 001000 1 Contents 001000 1000 SPC 001000 8 SSN zzzzzz footer , N2

It doesn't work on my comp. Ibaul, wbr