chunksort

#! /bin/sh
#
###	chunksort - sort multi-line records (separated by single blank lines)
###	Usage: chunksort [-a] [sort options] [files]
##
## chunksort - SORT CHUNKS OF TEXT SEPARATED BY BLANK LINES, LIKE THIS:
##
## Using C on the UNIX System
## By Dave Curry
## 250 pages, ISBN 0-937175-23-4
## 
## Checking C Programs with lint
## By Ian F. Darwin
## 84 pages, ISBN 0-937175-30-7
##
##	THE -a OPTION MAKES sort SORT BY LINES (SETS FIELD SEPARATOR, THE
##	sort -t OPTION, TO BE CTRL-A).  FOR EXAMPLE,
##	** TO SORT STARTING AT THE SECOND LINE, USE:   -a +1
##	** TO SORT STARTING AT THE FIFTH CHARACTER OF THIRD LINE:  -a +2.4
##
# WORKS BY FIRST JOINING EACH CHUNK WITH A CTRL-A, LIKE THIS:
#
# Using C on the UNIX System^ABy Dave Curry^A250 pages, ISBN 0-937175-23-4
# Checking C Programs with lint^ABy Ian F. Darwin^A84 pages, ISBN 0-937175-30-7
#
# THEN SORTING AND SPLITTING AGAIN.
#
# Thanks to Greg Ubben for help on this.

files=   sortopts=

# PARSE COMMAND LINE:
while :
do
	case "$1" in
	"")	# OUT OF ARGUMENTS (WE HOPE...)
		break
		;;
	-a)	# USE CTRL-A AS SORT FIELD SEPARATOR (SO CAN PICK WHICH LINE OF
		# THE RECORD TO SORT ON -- FOR EXAMPLE, +1 TO SORT ON LINE 2):
		ctrla=`echo a | tr a '\001'`
		sortopts="$sortopts -t$ctrla"
		;;
	-o)	# WE CAN'T PASS -o TO sort BECAUSE USER WOULD GET WRONG OUTPUT:
		echo "`basename $0` quitting: I don't have a -o option." 1>&2
		exit 1
		;;
	-[Tyz])
		# THE sort T, y, AND z OPTIONS HAVE AN ARGUMENT AFTER THEM.
		sortopts="$sortopts $1 $2"
		shift	# GET RID OF ONE (WE'LL DO OTHER BELOW)
		;;
	[-+]*)
		# SOME OTHER OPTION OR SORT FIELD:
		sortopts="$sortopts $1"
		;;
	*)	# IT'S A FILENAME (WE HOPE...)
		files="$files $1"
		;;
	esac
	shift
done

# Change newlines into CTRL-a's, print long line with extra CTRL-a at end:
gawk '{
    gsub(/\r\n/,"\1");
    gsub(/\r/,"\1");
	gsub(/\n/,"\1");
	print $0 "\1" }
' RS= $files |
# sort that text with $sortopts from above, change CTRL-a's to newlines:
sort $sortopts |
tr '\1' '\12'