#!/bin/sh

prog=`basename $0`
use() {
cat <<END
Highlights selected text from stdin; lines w/o matches are optionally deleted.
    $prog [-hboruB] [-e] pattern [-e pattern] ... [file ...]

where pattern's are valid sed patterns.
The default highlighting method is to use the "standout" escape sequence.
    
Hilighting Options:
    -b	-- bold hilighting, if available.
    -r	-- reverse-video hilighting, if available.
    -u	-- underline hilighting, if available.
    -B	-- underscore-backspace hilighting: precede each character with _\\b.
	   This man page convention is interpreted by "less" to mean underline.
    -o	-- overstrike hilighting: replace each character x with x\\bx.
	   This man page convention is interpreted by "less" to mean boldface.

Other Options:
    -e	-- The next word is a pattern.  This is optional before the first
		pattern, but required before the other patterns, else the
		patterns will be interpreted as filenames.  (The name "-e"
		is used by analogy with sed's "-e", although "e" has no
		mnemonic value in "$prog").
    -d  -- delete lines without any hilighted text (default: print all lines).
    -h	-- print this message, then exit.
    -D  -- debug mode (print sed hilighting command before executing).

Note: options must be entered separately; don't run them together.
END
}

# Need "5bin" on sunos to get tput...
case "$OSNAME$VERSION" in
    sunos4*) PATH=${PATH}:/usr/5bin
esac

findcommand() {
    type "$1" 2>/dev/null | awk '$2 != "not" {print $NF}'
}

AWK=`findcommand mawk`
if [ "$AWK" = "" ] ; then AWK=`findcommand gawk` ; fi
if [ "$AWK" = "" ] ; then AWK=`findcommand nawk` ; fi
if [ "$AWK" = "" ] ; then AWK=`findcommand awk` ; fi

# getopt loses quotes around whitespace, so we won't use it.

debug=no
underscore_style=none
enter_hl=smso
exit_hl=rmso
delete_lines=no
### word_left=
### word_right=
while [ $# -gt 0 ] ; do
    case "$1" in
    -h ) use ; exit 0 ;;
    -D ) debug=yes ; shift ;;
    -e ) break ;;	# reached a sed expression 

    -d ) delete_lines=yes ; shift ;;

    -b ) enter_hl=bold ; exit_hl=rmso ; shift ;;
    -r ) enter_hl=rev ; exit_hl=rmso ; shift ;;
    -u ) enter_hl=smul ; exit_hl=rmul ; shift ;;
    -B ) underscore=backspace ; shift ;;
    -o ) underscore=overstrike ; shift ;;
    ### -w ) word_left='[^_a-zA-Z]' ; word_right='[^_a-zA-Z0-9]' ; shift ;;

    -- ) shift ; break ;;
    -* ) echo "Unknown option \`$1'" ; exit 1 ;;
    * ) break ;;
    esac
done

X=`tput $enter_hl`
Y=`tput $exit_hl`
test "$X" = "" -o "$Y" = "" && { X=`tput smso` ; Y=`tput rmso` ; }

# Build up the appropriate sed args and append them to the command line,
# so as to take advantage of the quote-preserving done with "$@".
if [ $# -gt 0 ] ; then
    test "$1" = "-e" && shift	# Set up to work w/ or w/o initial -e.
    have_e=yes
    nargs=$#
    done_w_pats=no
    while [ $nargs -gt 0 ] ; do
	word=$1 ; shift ; nargs=`expr $nargs - 1`
	### echo "word is $word"
	### echo "arglist is $@"

	anchor_l=
	anchor_r=
	# special-case leading ^, trailing $
	case "$word" in
	"^"* ) word=`expr "$word" : '.\(.*\)'` ; anchor_l="^" ;;
	esac
	case "$word" in
	*\\$ ) ;;
	*$ ) word=`expr "$word" : '\(.*\).'` ; anchor_r="$" ;;
	esac

	# If done with all patterns (only have filenames left), append filename
	# to end of arg list being built and continue.
	case "$done_w_pats" in
	    yes ) set -- ${1+"$@"} "$word" ; continue ;;
	esac

	# Not done with options and pattern processing...
	case "$have_e,$word" in

	no,-e ) # This is the "-e" that introduces a pattern as the next arg.
		have_e=yes
		continue
		;;

	no,* )	# Previous word wasn't -e, and this word isn't -e, either.
		# Therefore it must be a filename.
		done_w_pats=yes

		# Append filename to the arglist, but first check if we must
		# add the sed code for deleting unmatched lines.
		case $delete_lines in
		    yes )
			set -- ${1+"$@"} -e t -e d "$word"
			# Turn off the delete_lines flag so we don't
			# add that sed code again:
			delete_lines=no
			;;
		    no )
			set -- ${1+"$@"} "$word"
			;;
		esac
		;;

	yes,* )	# Previous word was -e; this word must be a pattern.
		have_e=no
		pat=`echo "$word" | sed 's/\//\\\/'`
		set -- ${1+"$@"} -e \
		"s/$anchor_l\($pat\)$anchor_r/$X\1$Y/g"
		### Word selection doesn't work because the following
		### regexp won't match at beginning or end of line.
		### set -- ${1+"$@"} -e \
		### "s/$anchor_l\($word_left\)\($pat\)\($word_right\)$anchor_r/\1$X\2$Y\3/g"
		;;

	esac
    done

    # Now check if $delete_lines is still pending:
    case $delete_lines in
	yes ) set -- ${1+"$@"} -e t -e d ;;
    esac
fi

test "$debug" = yes && echo "	#	sed " ${1+"$@"}
case "$underscore" in
    overstrike ) sed ${1+"$@"} | $AWK '
		NR == 1 { lX = length(X) ; lY = length(Y) }
		{
		    iY = 3000
		    while ( (iX=index($0, X)) > 0 ) {
			if ((iY = index($0, Y)) == 0)
			    iY = length($0) + 1
			printf("%s", substr($0, 1, iX-1))
			for (i = iX+lX; i < iY; i++)
			    printf("%s\b%s", substr($0, i, 1), substr($0, i, 1))
			$0 = substr($0, iY+lY, 3000)
		    }
		    print $0
		}' X="$X" Y="$Y" -
	;;
    backspace ) sed ${1+"$@"} | $AWK '
		NR == 1 { lX = length(X) ; lY = length(Y) }
		{
		    iY = 3000
		    while ( (iX=index($0, X)) > 0 ) {
			if ((iY = index($0, Y)) == 0)
			    iY = length($0) + 1
			printf("%s", substr($0, 1, iX-1))
			for (i = iX+lX; i < iY; i++)
			    printf("_\b%s", substr($0, i, 1))
			$0 = substr($0, iY+lY, 3000)
		    }
		    print $0
		}' X="$X" Y="$Y" -
	;;
    * ) sed ${1+"$@"} ;;
esac