awk help

[Ernie Luzar]

Mike Jeays wrote:
> snip
> That is an amazing difference in performance - I might have expected a 
> five to ten times improvement, but not 300+ times.
> I don't see anything very time-consuming in the script above. Is it 
> possible for you to post the equivalent csh and awk scripts? Either I or 
> someone with more experience with csh might be able to spot the problem.

Here are the 2 scripts

  #! /bin/sh

   # This script is designed to process the o/p from the 2 line hit count

  # Make the input file read a line at a time, not a field at a time.
    IFS=$'\n'
    set -f

    added_date="`date +%Y%m%d`"

    hits_rpt="/etc/ipf_pool_hits_rpt"
    rm $hits_rpt
    touch $hits_rpt

    hits_new="/etc/ipf_pool.hits.yes"
    rm $hits_new
    touch $hits_new

    hits_no="/etc/ipf_pool.hits.no"
    rm $hits_no
    touch $hits_no

    ippool -l -d -m probing_ips > $hits_rpt 2> /dev/null

    for line in `cat $hits_rpt`; do

      # drop the first 3 rpt lines
      poollist_line=""
      poollist_line=`echo -n $line | grep poollist`
      [ -n "${poollist_line}" ] && continue

      role_line=""
      role_line=`echo -n $line | grep Role`
      [ -n "${role_line}" ] && continue

      nodes_line=""
      nodes_line=`echo -n $line | grep Nodes`
      [ -n "${nodes_line}" ] && continue

      in_line1=`echo -n $line | grep Address:`
      [ -n "${in_line1}" ] && save_in_line1="${in_line1}"

      in_line2=`echo -n $line | grep Hits`
      [ -n "${in_line2}" ] && save_in_line2="${in_line2}"

      if [ "${save_in_line1}" -a "${save_in_line2}" ]; then
        build_line1=${save_in_line1##*:}
        build_line1=${build_line1%%/*}
        build_line1="${build_line1};"
        build_line2=${save_in_line2##*Hits }
        # So remove everything to the right of the word Bytes.
        build_line2=${build_line2%%Bytes*}
        if [ ${build_line2} -gt 0 ]; then
          db_rec="$added_date  ${build_line1}"
          echo "${db_rec}" >> $hits_new
        fi
        build_line="${build_line2}  ${build_line1}"
        echo "${build_line}" >> $hits_no
        in_line1=""
        in_line2=""
        save_in_line1=""
        save_in_line2=""
      else
        continue
      fi
    done
    exit 0


  #! /bin/sh

    hits_rpt="/etc/ipf_pool_hits"
  #  ippool -l -d -m probing_ips > $hits_rpt 2> /dev/null

    awk 'BEGIN { "date +%Y%m%d" | getline date
                 hits_yes = "/etc/ipf_pool_awk_hits_yes"
                 hits_no = "/etc/ipf_pool_awk_hits_no"
                 system("rm " hits_yes)
                 system("touch " hits_yes)
                 system("rm " hits_no)
                 system("touch " hits_no)
               }
        /Address/ { address = $2; sub("/32", ";", address); got_address
= 1; }
         /Hits/   { if (got_address)
                     { hits = $2;
                       if (hits > 0)
                         { print date, " ", address > hits_yes;
                           got_address = 0;
                         }
                       else
                         { print "hit no", hits, address > hits_no;
                           got_address = 0;
                         }
                     }
                  }'   $hits_rpt


Note; the
  ippool -l -d -m probing_ips > $hits_rpt 2> /dev/null
command is run in the first script and the file it creates in used as 
input in the second script so both scripts process the same data. This 
ippool command is not the source of the big time difference between the 
2 scripts, where the first one takes 5 minutes and the second takes 1 
second.