#! /bin/sh # Usage: ./i1-hmmbuild.sh # # Verifies that hmmbuild builds HMMs reproducibly (no stochastic run-to-run variation by default). # if test ! $# -eq 3; then echo "Usage: $0 " exit 1 fi hmmbuild=$1; if test ! -x $hmmbuild; then echo "FAIL: $hmmbuild not executable"; exit 1; fi alifile=$2; if test ! -r $alifile; then echo "FAIL: $alifile not readable"; exit 1; fi hmmfile1=$3.1.hmm hmmfile2=$3.2.hmm outfile1=$3.1.out outfile2=$3.2.out diffile1=$3.1 diffile2=$3.2 # By default: no stochastic variation $hmmbuild $hmmfile1 $alifile > $outfile1; if test $? -ne 0; then echo "FAIL: crash"; exit 1; fi $hmmbuild $hmmfile2 $alifile > $outfile2; if test $? -ne 0; then echo "FAIL: crash"; exit 1; fi cat $outfile1 | grep -v "^#" > $diffile1 cat $outfile2 | grep -v "^#" > $diffile2 diff $diffile1 $diffile2 > /dev/null if test $? -ne 0 then echo "FAIL: output files differ" exit 1 fi cat $hmmfile1| grep -v "^DATE" > $diffile1 cat $hmmfile2| grep -v "^DATE" > $diffile2 diff $diffile1 $diffile2 > /dev/null if test $? -ne 0 then echo "FAIL: HMM files differ" exit 1 fi # With different seeds: HMM files will differ because of statistical fits $hmmbuild --seed 1 $hmmfile1 $alifile > $outfile1; if test $? -ne 0; then echo "FAIL: crash"; exit 1; fi $hmmbuild --seed 2 $hmmfile2 $alifile > $outfile2; if test $? -ne 0; then echo "FAIL: crash"; exit 1; fi cat $hmmfile1| grep -v "^DATE" > $diffile1 cat $hmmfile2| grep -v "^DATE" > $diffile2 diff $diffile1 $diffile2 > /dev/null if test $? -eq 0 then echo "FAIL: HMM files identical, despite different rng seeds" exit 1 fi echo "ok" rm $hmmfile1 $hmmfile2 rm $outfile1 $outfile2 rm $diffile1 $diffile2 exit 0