Utility for chaining multiple dependent PBS submit scripts

taskchain.sh

# Utility function for chaining multiple PBS submit scripts
# Leif Denby 6/11/2018
#
# usage:
# 
# 1. Create a file containing the individual steps, e.g.
#
# tasklist.sh:
#  WALLCLOCK=00:15:00 ./run.sh example_run
#  WALLCLOCK=00:15:00 ./run.sh analysis
#
# 2. Pipe this list to taskchain.sh or provide with tasklist as first argument
#
# > cat tasklist.sh | taskchain.sh
#
# or
#
# > taskchain.sh tasklist.sh
#
# Each task will the have the JOB id of the previous task provided through the
# environment variables DEPENDSON. This is extracted as the first line of
# output returned from each task command (e.g. what PBS returns when `qsub` is
# invoked)
#
#
# For the above example the dummy script run.sh below can be used to see the
# behaviour
#
# run.sh:
#  echo "42424.job"
#  echo "Simulating PBS job: $@"
#  
#  echo "wc=${WALLCLOCK}"
#  echo "do=${DEPENDSON}"


# https://stackoverflow.com/a/17841619
function join_by { local d=$1; shift; echo -n "$1"; shift; printf "%s" "${@/#/$d}"; }

lines=()
while read line
do
  lines+=("$line")
done < "${1:-/dev/stdin}"

TASKLIST=$(join_by $'\n' "${lines[@]}")

echo "Received task list:"
echo "${TASKLIST}"
echo ""

N_LEFT=$(echo "`echo "${TASKLIST}" | wc -l` - 1" | bc)
NEXTTASK=`echo "${TASKLIST}" | head -n1`
TASKLIST=`echo "${TASKLIST}" | tail -n+2`
JOBID=""

while [ "${NEXTTASK}" != "" ]; do
  echo "taskchain> queuing '${NEXTTASK}', ${N_LEFT} tasks left"
  JOBID=$(eval "DEPENDSON=${JOBID} ${NEXTTASK}")
  echo "taskchain> queued as ${JOBID}"
  JOBID=$(echo "${JOBID}" | head -n1)
  echo ""

  NEXTTASK=`echo "${TASKLIST}" | head -n1`
  TASKLIST=`echo "${TASKLIST}" | tail -n+2`
  N_LEFT=$(echo "`echo "${TASKLIST}" | wc -l` - 1" | bc)
done

echo "taskchain> all tasks completed"