jb - Simple Job Manager

I have a problem where I want to run commands in the background and be able view the status and output from anywhere. Kind of like jobs but not bound to a specific terminal and a little friendlier. Essentially a systemd but I don't want to have to make systemd jobs for every little command I want to run.

My personal use cases include:

run godoc on Go projects to view the documentation
File watcher for hot updating in Svelete development
Long running (but eventually ending) operations e.g. system backups

I looked around and couldn't seem to find anything to suit my needs and so I wrote a small MVP script.

jb is a simple script to manage background jobs. It does four things:

Run jobs in background
Print logs
Lists all recorded jobs
Clear stopped jobs from record

I've been running this for a couple of days without a problem on my laptop running Fedora Linux with Bash 5.1.16. I'm not quite sure how it will go in other systems so give it a try and let me know how it goes!

jb

#!/bin/sh
# A simple script to manage background tasks

JB_DIR=$HOME/.jb
if [ -n "$XDG_DATA_HOME" ]; then
	JB_DIR=$XDG_DATA_HOME/jb
fi

# Stores information about the processes
# Just a TSV file containing three columns: 
# 1. Job ID:  set by this program
# 1. Status:  whether the process is still running
# 2. PID:     process id
# 3. Command: the command used to start the process
JB_FILE=$JB_DIR/data.tsv

usage() {
	NAME=$(basename "$0")
	echo "usage: $NAME <command> [args]

commands:
	run <command>   run command in background
	log <pid>       outputs the contents of the log file for that process
	list            list jobs
	clear           removes stopped jobs from listing and deletes logs

$NAME is a simple script to manage background jobs.
If the variable exists exists, $NAME will store all data in \$XDG_DATA_DIR/jb,
otherwise it will default to \$HOME/.jb/. 

Upon running a command it will write both STDOUT and STDERR to the same file in
the directory as determined above, which can be printed with 

	jb log <job_id>

where <job_id> can be obtained from running

	jb list
"
	exit 2
}

# Checks the $JB_FILE to see if the running jobs have stopped and upates the
# file accordingly
update_job_status() {
	TMP_FILE=$(mktemp)

	awk -F'\t' '{
		if ($2 != "STOPPED") {
			status = system("ps " $3 " > /dev/null 2>&1")
			if (status != 0)
				$2 = "STOPPED"
		}
		print $1 "\t" $2 "\t" $3 "\t" $4
	}' "$JB_FILE" > "$TMP_FILE"

	mv "$TMP_FILE" "$JB_FILE"
}

# Get the smallest number from 1 to {number of jobs + 1}, that is not already
# present
get_job_id() {
	JOB_IDS_FILE=$(mktemp)
	cut "$JB_FILE" -f1 | sort -n > "$JOB_IDS_FILE"

	NUM_JOBS=$(cat "$JOB_IDS_FILE" | wc -l)
	TMP=$(mktemp)
	seq $(( NUM_JOBS = NUM_JOBS + 1 )) > "$TMP"

	# print the first number from 1..#Jobs+1 that isn't in $JOBS_IDS_FILE
	comm -23 "$TMP" "$JOB_IDS_FILE" | head -n1
}

# Ensure files and directories exist
mkdir -p "$JB_DIR"
[ -f "$JB_FILE" ] || touch "$JB_FILE"

case "$1" in
	list) 
		update_job_status

		if [ -s "$JB_FILE" ]; then
			printf "JOB ID\tSTATUS\tPID\tCOMMAND\n"
			cat "$JB_FILE"
		else
			echo no jobs running
		fi
		;;
	run) 
		[ -z "$2" ] && usage
		shift

		JOB_ID=$(get_job_id)
		nohup $@ > "$JB_DIR/$JOB_ID" 2>&1 &
		PID=$!

		# since $@ might have spaces, we need to use a separate echo to
		# handle it
		printf "%s\tRUNNING\t%s\t" "$JOB_ID" "$PID" >> "$JB_FILE"
		echo "$@" >> "$JB_FILE"

		echo "$JOB_ID"
		;;
	clear) 
		update_job_status

		# remove stopped processes from the listing
		STOPPED_JOBS=$(sed -n "/^[0-9]\+\tSTOPPED/p" "$JB_FILE"  \
				| cut -f1 | tr '\n' ' ')

		# intentionally want to allow the STOPPED_JOBS variable to
		# expand to multiple files
		if [ -n "$STOPPED_JOBS" ]; then
			(cd "$JB_DIR" || exit 3; rm $STOPPED_JOBS)
		fi

		sed -i'' '/^[0-9]\+\tSTOPPED/d' "$JB_FILE"
		;;
	log) 
		[ "$#" -ne 2 ] && usage

		LOGFILE="$JB_DIR/$2"

		if [ -f "$LOGFILE" ]; then
			cat "$LOGFILE"
		else
			echo "no job with id: $2"
			exit 1
		fi
		;;
	*) usage
esac