column-histogram

#!/bin/sh
# Read a file in on stdin, use awk to bin column -c into bins of size -b. Optionally use log-size bins with base -L (mutually exclusive with -b)
# E.g. echo -e "log file HTTP status: 200 body_size: 1024\nlog file HTTP status: 200 body_size: 3405" \
# | ./$0 -c 7 -b 1000
#  > 1000: 1
#  > 3000: 1

usage () {
  cat << EOF
Usage: $0 [-c COLUMN] [-F DELIM] <-b BIN|-L LOGBASE>
Reads from stdin and bins column COLUMN (default: 1st column) from a
DELIM-delimited file (e.g. apache log file) into bins of size BIN or
into semi-log bins of size LOGBASE
EOF
}
COL=1
DELIM=" "
while getopts "c:F:b:L:" flag; do
	case $flag in
		"c") COL=$OPTARG;;
		"F") DELIM=$OPTARG;;
		"b") BIN=$OPTARG;;
		"L") LOGBASE=$OPTARG;;
		*) usage; exit;;
	esac
done 

#If both LOGBASE and BIN are set, bail
if [ "x$LOGBASE" != "x" ] && [ "x$BIN" != "x" ]; then
	usage
	exit
fi
#Can't divide by log(1) 
if [ "x$LOGBASE" != "x" ] && [ $LOGBASE -eq 1 ]; then
	echo "LOGBASE cannot be 1, that's division by zero, Homes."
	echo " (log(x) = integral from 1 to x of 1/y dy :. log(1)=0)"
	usage
	exit
fi

if [ "x$BIN" != "x" ]; then
	awk -F"$DELIM" -vCOL=$COL -vBIN=$BIN '{bins[int($COL/BIN)]++} END { for (c in bins) { printf("%s\t%s\n", c*BIN, bins[c]);}}'
elif [ "x$LOGBASE" != "x" ]; then
	awk -F"$DELIM" -vCOL=$COL -vLOGBASE=$LOGBASE '{bins[int(log($COL)/log(LOGBASE))]++} END { for (c in bins) { printf("%s\t%s\n", LOGBASE^c, bins[c]);}}'
else
	usage
	exit
fi