Added option to deal with large m8 files; checking for remaining tmp folders from previous analyses

wGRR

@@ -7,7 +7,7 @@ trap 'rm -rf "$tmp"' EXIT
 export LC_ALL=C
 SECONDS=0
 
-readonly VERSION=1.4.1
+readonly VERSION=1.4.2
 
 bold=$(tput bold)
 green=$(tput setaf 2)
@@ -66,6 +66,8 @@ display_usage() {
 	echo "   -f                In sbatch mode, will use the fast QOS (jobs limited to 2 hours)"
 	echo "                        ${bold}WARNING:${normal}: be sure the jobs will run under 2 hours otherwise they will fail."
 	echo "                        default: off"
+	echo "   -n                If turned on, the MMseqs output file will be erased to save disk space."
+	echo "                        default: off"
 	echo ""
 	echo "${bold}DESCRIPTION:${normal}"
 	echo "This pipeline will do all proteins pairwise comparisons using the MMseqs2 software and then process "
@@ -121,18 +123,22 @@ IDLIST="N.O.L.I.S.T"		  ## -l
 BATCHFLAG=0               ## Are we in a sbatch job?
 QT=0                      ## Queing time (for Maestro)
 MAXJOBS=0                 ## -m
-MMS_DEF_MAX_SEQS=300
-MIDENT=0
-MMS_MAX_SEQ_PARAM=""
 TESTRUN=0                 ## -T
 FAST=0                    ## -f
 SKIP=0                    ## -s
 JACCARD=0                 ## -j
 COVT=0.5                  ## -C
 IDT=0.35                  ## -I
+NOM8=0                    ## -n
+
+MMS_DEF_MAX_SEQS=300
+MIDENT=0
+MMS_MAX_SEQ_PARAM=""
+EST_M8_SIZE=0
+tmp="wgrrtmp"
 
 ## catch option values
-while getopts :fTsji:p:o:t:a:m:l:C:I: option ; do
+while getopts :fTsji:p:o:t:a:m:l:C:I:c:n: option ; do
    case $option in
    i) PRT="$OPTARG";       if [[ ! -s $PRT ]]; then printf "${bold}${red}%-17s  --  %s\n${normal}"  "[ERROR]" "Fasta file '$PRT' not found or empty (option -i)." ; exit 1 ; fi ;;
    p) MMPATH="$OPTARG"          ;;
@@ -145,6 +151,7 @@ while getopts :fTsji:p:o:t:a:m:l:C:I: option ; do
 	f) FAST=1 ;;
 	s) SKIP=1 ;;
 	j) JACCARD=1 ;;
+   n) NOM8=1 ;;
 	C) COVT="$OPTARG"; if [[ ${COVT} -gt 1 ]] || [[ ${COVT} -lt 0 ]] ; then printf "${bold}${red}%-17s  --  %s\n${normal}"  "[ERROR]" "Coverage threshold must be between 0 and 1 (option -C)." ; exit 1 ; fi ;;
 	I) IDT="$OPTARG"; if [[ ${IDT} -gt 1 ]] || [[ ${IDT} -lt 0 ]] ; then printf "${bold}${red}%-17s  --  %s\n${normal}"  "[ERROR]" "Identity threshold must be between 0 and 1 (option -C)." ; exit 1 ; fi ;;
    :) printf "${bold}${red}%-17s  --  %s\n${normal}" "[ERROR]" "option $OPTARG : missing argument" ; exit 1  ;;
@@ -153,6 +160,11 @@ while getopts :fTsji:p:o:t:a:m:l:C:I: option ; do
 done
 shift "$((OPTIND - 1))"
 
+if [[ ${PRT} == "N.O.P.R.T" ]] ; then
+	printf "${bold}${red}%-17s  --  %s\n${normal}" "[ERROR]" "No input file provided (option -i)."  | tee -a ${OUT}.wgrr.log
+	exit 1
+fi
+
 # Improvement: test if $PRT is a proper fasta file
 
 # Remove all aliases
@@ -177,20 +189,29 @@ fi
 
 /usr/bin/mktemp --version > /dev/null 2>&1
 if [[ $? -eq 1 ]] ; then
-	tmp=$(mktemp -d tmp-XXXXXX)
+	tmp=$(mktemp -d wgrrtmp-XXXXXX)
 else
 	if [[ `hostname` == "maestro-"* ]] ; then
 		if [[ ! -d ${APPASCRATCH}/${USER} ]] ; then
 			mkdir ${APPASCRATCH}/${USER}
 		fi
-		tmp=$(mktemp -p ${APPASCRATCH}/${USER} -d tmp-XXXXXXXX)
+		tmp=$(mktemp -p ${APPASCRATCH}/${USER} -d wgrrtmp-XXXXXXXX)
+	else
+		tmp=$(mktemp -d wgrrtmp-XXXXXXXX)
+	fi
+fi
+
+if [[ ${tmp} == "wgrrtmp" ]] ; then
+	if [[ -d ${tmp} ]] ; then
+		printf "${bold}${red}%-17s  --  %s\n${normal}" "[ERROR]" "Could not create temporary directory. Please delete ${tmp} directory and relaunch the analysis."  | tee -a ${OUT}.wgrr.log
+		exit 1
 	else
-		tmp=$(mktemp -d tmp-XXXXXXXX)
+		mkdir ${tmp}
 	fi
 fi
 
 if [[ $OUT == "N.O.O.U.T" ]] ; then
-	OUT=$(basename $tmp)
+	OUT=$(basename $tmp | cut -d'-' -f 2)
 fi
 
 ## Get partition (if sbatch) or check nproc (if local/Check number of threads
@@ -222,6 +243,18 @@ fi
 
 printf "%-17s  --  %s\n" "["$(textifyDuration $SECONDS)"]" "wGRR version ${VERSION}" | tee -a ${OUT}.wgrr.log
 
+NTMP=0
+for f in ${tmp}/../wgrrtmp* ; do
+	if [[ -d ${f} ]] ; then
+		NTMP=$(( NTMP+1 ))
+	fi
+	if [[ ${NTMP} -gt 1 ]] ; then
+		g=$(echo "$(cd "$(dirname $tmp)" ; pwd)")
+		printf "${bold}${yellow}%-17s  --  %s\n${normal}" "[WARNING]" "Found tmp folders from previous analyses in ${g}" | tee -a ${OUT}.wgrr.log
+		printf "${bold}${yellow}%-17s  --  %s\n${normal}" "[WARNING]" "Consider deleting them" | tee -a ${OUT}.wgrr.log
+	fi
+done
+
 if [[ $BATCHFLAG == 0 ]] ; then
 	if [[ $FAST == 1 ]] ; then
 		printf "${bold}${yellow}%-17s  --  %s\n${normal}" "[WARNING]" "The fast flag (-f) is useless when not using sbatch"  | tee -a ${OUT}.wgrr.log
@@ -291,6 +324,16 @@ STATS=($(awk '/^>/{p++;g=substr($1,2);gsub(/_[^_]+$/,"",g);if(!a[g]++){c++};LNR=
 printf "%-17s  --  %s %s %s %s %s\n" "["$(textifyDuration $SECONDS)"]" "Input file has" $STATS[1] "genomes and a total of" $STATS[2] "proteins"  | tee -a ${OUT}.wgrr.log
 printf "%-17s  --  %s %s\n" "["$(textifyDuration $SECONDS)"]" "Mean number of proteins per genome:" $STATS[4]  | tee -a ${OUT}.wgrr.log
 
+EST_M8_SIZE=$((STATS[1] * STATS[1] * STATS[4] / 10000000 ))
+
+if [[ ${EST_M8_SIZE} == 0 ]] ; then
+	printf "${bold}${yellow}%-17s  --  %s\n${normal}" "[WARNING]" "Unable to estimate the MMseqs output file size." | tee -a ${OUT}.wgrr.log
+elif [[ ${EST_M8_SIZE} -gt 10 ]] ; then
+	if [[ ${NOM8} == 0 ]] ; then
+		printf "${bold}${yellow}%-17s  --  %s\n${normal}" "[WARNING]" "MMseqs output file could exceed 10GB. Consider using the -n option to save disk space." | tee -a ${OUT}.wgrr.log
+	fi
+fi
+
 if [[ $ARRAYSIZE == "AUTO" ]] ; then
 	if [[ ${STATS[4]} -le 100 ]] ; then
 		ARRAYSIZE=10000