Shell Programming and Scripting

#!/usr/bin/ksh
# INFILE should exists
INFILE=tst

# WORKFILE contains the same entries than INFILE
# but reordered so that brand having the most entries appear first in the file
WORKFILE=tst.w

# TMPFILE & OUTFILE will be generated
TMPFILE=tst.tmp
OUTFILE=tst.out
cat /dev/null>$OUTFILE
# From the initial file, build a list of "brand" (not really in fact, just $1) order by the number of entry it has
# So the brand having the most numerous entries appears at first (here, Telco, Heinz, Schwartz, l'Or\xc3\xa9al...)
#2806 Tesco
# 253 Heinz
# 192 Schwartz
# 174 L'Or\xc3\xa9al
# 146 Cadbury
# 134 Nestle
#  98 Kellogg's
#  94 Cow
#  88 Nivea
#  88 Garnier
#  ...

cat /dev/null>$WORKFILE
awk '{print$1}' $INFILE | sort | uniq -c | sort -k 1n | tail -r | while read a a
do
grep ^"$a " $INFILE | sort -u >>$WORKFILE
done

# if this is the first run of the script, run it with argument 1
# It removes the : since we are going to use it later as a special separator character
# Also the ' are replaced by _ as well as / are replaced with -
# to avoid parsing problem later
if [[ $1 -eq 1 ]];then
sed "s/:/ /g;s/'/_/g;s:/:-:g" $WORKFILE >$TMPFILE
else
sed "s/'/_/g;s:/:-:g" $WORKFILE >$TMPFILE
fi

#================================================
# Main loop  :
#   . read $TMPFILE
#   . ask where it should add the : separator
#   . add the separator at the chosen place for
#      all subsequent lines that start with the
#      same brand name
#   . log those modified entries to $OUTFILE
#================================================

exec 3<$TMPFILE
while read -u3 a
do
        if ( echo "$a" | grep : >/dev/null )
        then
                echo "$a" | sed "s/$B/&:/;s/: /:/" >>$OUTFILE
                continue
        fi

        while :
        do
                if [[ "$f" != x ]];then
                        echo "$a"
                        echo "what is the number of word in this brand [1-9|x] ? (choosing x will set 1 for all remaining entries)"
                        read n
                        case $n in
                                [1-9]) break ;;
                                x) n=1; f="x" ; break;;
                                *) continue ;;
                        esac
                else
                        break
                fi
        done

        B=$(echo "$a" | awk -v V="$n" '{NF=V;$1=$1}1')

        b=$a

        until ! ( echo "$b" | grep ^"$B " >/dev/null )
        do
                echo "$b" | sed "s/$B/&:/;s/: /:/" >>$OUTFILE
                read -u3 b
        done

done
3<&-