#!/bin/sh
###
###		cluster3.sh
###
###	Clustering 3' tag sequences.
###
###	usage : % cluster3.sh -dir(in dir) -date(timestamp)
###
###	exit code:
###		0	succeed
###		1	command line error
###		11	previous process is executing
###		12	input sequence does not exist
###
###	ver 0.00	Jul 25 1997	T.Shin'i of NIG
###	ver 0.10	Jun 17 1998	T.Shin'i of NIG
###			  Remove previous sequences when new sequences of
###			 clones which already exist in DB are processed.
###	ver 0.11	Jun 22 2000	T.Shin'i of NIG
###			  Add "no merge" mode.
###


##
##	initial settings
##
LANG=""; export LANG
if [ $BASELINE ];	then
	baseline=$BASELINE
else
	baseline=/usr/local/yk
fi
PATH=`cat $baseline/etc/extbindir.rc`:$PATH; export PATH
projhome=$baseline/clusterf
bindir=$projhome/bin
logdir=$projhome/log
logabb=$logdir/cluster3.log
workdir=$projhome/work
lockfile=$workdir/.lockfile.cluster3
tmpfile=$workdir/cluster3.$$
archivedir=$projhome/Arc
fastadirabb=$projhome/fasta


##
##  data files and parameters
##
clstlistname=celk-cog.list
clstlist=$projhome/data/$clstlistname
clstseq=$projhome/data/r.seq

score_threshold=150
identity_threshold=89
seqperexec=1000

mergemode=0	#  1 : yes,  0 : no



##
##	parse of command line parameters
##
indir=""
datestr=""
for param
do
	case $param in
	-[Dd][Ii][Rr]*)		##  input directory name
		indir=`echo $param | sed -e 's/^-[Dd][Ii][Rr]//'`;;
	-[Dd][Aa][Tt][Ee]*)	##  time stamp of this process
		datestr=`echo $param | sed -e 's/^-[Dd][Aa][Tt][Ee]//'`;;
	esac
done
if [ -z "$indir" ];	then
	echo "usage : $0 -dir(in-dir) -date(timestamp)"
	exit 1
fi
if [ -z "$datestr" ];	then
	datestr=`date '+%Y%m%d'`
fi

logfile=$logabb.$datestr
fastadir=$fastadirabb/result3/$datestr



##
##  parameter or system error check
##
if [ ! -d $indir ];	then
	echo "directory $indir does not exist"
	exit 12
fi
if [ -f $lockfile -o -f $logfile -o -d $fastadir ];	then
	echo "Previous proceess is running."
	echo "Try again after the process ending."
	exit 11
fi
echo $datestr > $lockfile	## make lockfile



##
##	execute clustering
##
tmpclstlist=$clstlist.tmp

ls $indir > $tmpfile.master
if [ `cat $tmpfile.master | wc -l` -lt 1 ];	then
	echo "sequence does not exist in $indir"
	rm $lockfile
	exit 12
fi


##
##  remove old sequences
##
(sed -e 's/\./ /' $tmpfile.master | gawk '{print $1}' ; gawk '{print $2}' $clstlist) | sort | uniq -d > $tmpfile.duplist
if [ `cat $tmpfile.duplist | wc -l` -gt 0 ];	then
	cp $clstseq $clstseq.org
	split -$seqperexec $tmpfile.duplist $tmpfile.split.
	for list in $tmpfile.split.*
	do
		for clone in `cat $list`
		do
			grep -v "	$clone"'$' $clstlist > $clstlist.new
			grep "	$clone"'$' $clstlist | \
			gawk '{printf("remove %s from %s\n", $2, $1)}' \
				>> $logfile
			mv $clstlist.new $clstlist
		done
		findgetfs -s$clstseq `cat $list` >> $tmpfile.rmseq
		findgetfs -v -s$clstseq `cat $list` > $clstseq.new
		mv $clstseq.new $clstseq
	done
	rm $tmpfile.split.*
fi


mkdir $fastadir
cd $workdir
indirname=`basename $indir`
if [ ! -d $indirname ];	then
	ln -s $indir $indirname
fi


newcelk=`gawk '{print $1}' $clstlist | sort -r -u | head -1 | sed -e 's/^CELK//'`
newcelk=`expr $newcelk + 1`

##
##  the beginning of loop
##
for infile in `cat $tmpfile.master`
do
	abbrev=`echo $infile | sed -e 's/\./ /' | gawk '{print $1}'`
	fasta $indirname/$infile $clstseq > $fastadir/$abbrev
	$bindir/postfasta -s$score_threshold -i$identity_threshold \
		$fastadir/$abbrev > $tmpfile.fsout

	if [ `cat $tmpfile.fsout | wc -l` -lt 1 ];	then
		##
		##  This is a new type  --  create a new CELK group
		##
		echo "$newcelk $abbrev" | \
			gawk '{printf("CELK%05d\t%s\n", $1, $2)}' >> $clstlist
		cat $indirname/$infile >> $clstseq
		echo "$abbrev $newcelk" | \
			gawk '{printf("add %s to new (CELK%05d)\n", $1, $2)}' \
			>> $logfile
		newcelk=`expr $newcelk + 1`
	else
		##
		##  find cluster ID for each hit
		##
		for hit in `gawk '{print $2}' $tmpfile.fsout`
		do
			grep "	$hit"'$' $clstlist >> $tmpfile.clst
		done

		##
		##  find cluster ID for primary hit
		##
		priclst=`head -1 $tmpfile.clst | gawk '{print $1}'`

		if [ $mergemode = 1 ];	then
			echo $priclst > $tmpfile.merge
			##
			##  find cluster ID to be merged to the primary cluster
			##
			tail +2 $tmpfile.clst | grep -v "^$priclst	" | \
				gawk '{print $1}' | sort -u > $tmpfile.clstlist
			for clst in `cat $tmpfile.clstlist`
			do
				grep "^$clst	" $clstlist \
					| gawk '{print $2}' | \
					sort -u > $tmpfile.mchk
				grep "^$clst	" $tmpfile.clst \
					| gawk '{print $2}' \
					| sort -u >> $tmpfile.mchk
				if [ `sort $tmpfile.mchk | uniq -u | wc -l` -lt 1 ]; then
					echo $clst >> $tmpfile.merge
				fi
			done

			##
			##  merge one or more CELK clusters
			##
			priclst=`sort $tmpfile.merge | head -1`
			for clst in `cat $tmpfile.merge`
			do
				if [ $clst != $priclst ];	then
					echo "s/$clst	/$priclst	/" \
						>> $tmpfile.sed
					echo "merge cluster $clst to $priclst" \
						>> $logfile
				fi
			done
			if [ -f $tmpfile.sed ];	then
				mv $clstlist $tmpclstlist
				sed -f $tmpfile.sed $tmpclstlist > $clstlist
				rm $tmpclstlist $tmpfile.sed
			fi
		fi

		##
		##  add new clone
		##
		echo "$priclst	$abbrev" >> $clstlist
		cat $indirname/$infile >> $clstseq
		echo "add $abbrev to $priclst" >> $logfile
		rm $tmpfile.clst
	fi
done
##
##  The end of the loop
##


##
##	make backup file
##
cp $clstlist $archivedir/$clstlistname.$datestr
gzip $archivedir/$clstlistname.$datestr
if [ -f $tmpfile.rmseq ];	then
	cp $tmpfile.rmseq $archivedir/rrmseq.$datestr
	mv $clstseq.org $archivedir/rbackup.seq.$datestr
	gzip $archivedir/rrmseq.$datestr $archivedir/rbackup.seq.$datestr
fi


rm $tmpfile.* $lockfile
if [ ! -d $indirname ];	then
	rm $indirname
fi

exit 0