#!/bin/sh
# rsort: sort multi-line records from stdin to stdout
#
# Description:
#	sort the input records;
#	an input record is one or more lines of text;
#	a record begins with a line that does not begin with white space,
#	ie, it is flush with the left margin;
#	other lines in this record begin with white space
#	(spaces and/or tabs), or are blank lines;
#	#if filtering out blanks, a blank line will function as a
#	#record separator;
#
# Options:
#	See 'usage' below.
#
# D.Singer, 12/20/94
#
# Modifications:
#
# 12/01/95, D.Singer
#	added -c flag;
#	added -ns flag;
#	added -r flag;
# 2/2005, D.Singer
#	a little cleanup...

PATH='/bin:/usr/bin:/usr/ucb'
export PATH

prog=`basename "$0"`

# system/OS dependencies
AWK='awk'
SYS=`uname -sr`
case "$SYS" in
  'SunOS '*)
	AWK='nawk'
  esac

COMBINE=
FIELD=
FIELD_SEP="
"
FIELD_SEP_ARG=
FILTER_BLANKS=0
NO_SORT=
REPLACE=0
REPLACE_STR=
SORT_FLAGS="-bf -t$FIELD_SEP"
SORT_FLAGS_MSG="-bf -t^A"

usage="
Usage:  $prog [-b] [-c] [-f fs] [-ns] [-r \"replace\"] [-s \"sort-flags\"]

	Sorts multi-line text records.  A record begins with a line that does
	not begin with white-space, and continues with lines that begin with
	white-space and blank lines.

	-b	filter out blank lines;
	-c	combine record into single line;
	-f	alternate field separator for record collection; each line in
		  a record is a field;
	-ns	no sorting;
	-r	replace beginning spaces and tabs with \"replace\";
	-s	specify alternative sort flags, default are \"$SORT_FLAGS_MSG\";
"

CMD1='echo "$prog: $cmd" >&2'
CMD2='eval "$cmd"'
#CMD='echo "$prog: $cmd" >&2; eval "$cmd"'
CMD="$CMD1; $CMD2"
debug=
syntax="$prog: option syntax error."

#
# process command line options
#
while [ $# != 0 ]; do
	case "$1" in
	#  -d)
	#	debug=1
	#	CMD='echo "$prog: $cmd" >&2'
	#	echo "$prog: debug mode." >&2
	#	;;
	  -b)
		FILTER_BLANKS=1
		;;
	  -c)
		COMBINE=1
		FIELD_SEP=
		;;
	  -ns)
		NO_SORT=1
		;;
	  -f)
		shift
		FIELD=1
		FIELD_SEP_ARG="$1"
		;;
	  -r)
		shift
		REPLACE=1
		REPLACE_STR="$1"
		;;
	  -s)
		shift
		SORT_FLAGS="$1"
		;;
	  -*)
		echo "$syntax" >&2
		echo "$usage" >&2
		exit 1
		;;
	  *)
		echo "$syntax" >&2
		echo "$usage" >&2
		exit 1
	  esac
	shift
  done

# done this way so order of -f and -c don't matter
[ "$FIELD" ] &&
	FIELD_SEP="$FIELD_SEP_ARG"

#
# combine multi-line records into single lines so that they can be sorted; 
# continuation lines start with SPACE or TAB;
# output is a single line for each multi-line records, where the input
# lines become fields separated by a special field separator;
#
$AWK '
  BEGIN {
	IN_REC = 0;		# in the current input record
	CURR_REC = "";		# the current input record
	SEP = "'"$FIELD_SEP"'";	# separator between lines of multi-line record
	FILTER_BLANKS = "'"$FILTER_BLANKS"'";	# filter blanks flag
	REPLACE = "'"$REPLACE"'";	  # replace beginning tabs and blanks
	REPLACE_STR = "'"$REPLACE_STR"'"; # with this string
  }
  {
	# filter out blank lines, including those
	# containing just space characters
	if (FILTER_BLANKS == 1) {
		if ($0 ~ /^[ \t\f\n\r]*$/) {
			# print out the current record?
			##if (IN_REC == 1)
			##	print CURR_REC;
			##IN_REC = 0;
			next;
		}
	}
	# start of a new record?
	# ie, does not begin with white space, and is not a blank line;
	if ($0 !~ /^[ \t]/ && $0 !~ /^$/) {
		# print out the current record?
		if (IN_REC == 1)
			print CURR_REC;
		# begin the new record;
		CURR_REC = $0;	
		IN_REC = 1;
	}
	else {
		# continuation of the current record;
		if (REPLACE == 1)
			sub(/^[ \t][ \t]*/,REPLACE_STR);
		CURR_REC = CURR_REC SEP $0;
		IN_REC = 1;
	}
  }
  END {
	if (IN_REC == 1)
		print CURR_REC;
  }' |

#
# sort the records;
#
if [ "$NO_SORT" ]; then
	cat
  else
	sort $SORT_FLAGS
  fi |

case "$FIELD_SEP" in
  '')
	cat
	;;
  *)
	#
	# now, output the records, separating the fields back out into lines;
	#
	$AWK '
	  BEGIN {
		FS = "'"$FIELD_SEP"'";
		#FILTER_BLANKS = "'"$FILTER_BLANKS"'";
	  }
	  {
		# if blank line, just print it;
		if ($0 ~ /^[ \t\f\n\r]*$/) {
			print $0;
			next;
		}
		# print out each field, ie, each line of the multi-line record;
		for (FNUM=1; FNUM <= NF; ++FNUM)
			print $FNUM;
	  }'
  esac

exit