Unfortunately, gnuplot just isn't well suited to handle data processing tasks such as these. You probably could come up with a solution, but it would be extremely messy and super hard to use. Fortunately, gnuplot can read from pipes from other programs -- So the easiest solution is to write a simple script to process the input data and write it to standard output. I'd choose python:
import time
from datetime import datetime
from collections import defaultdict
import sys
def datetime_2_epoch(dt):
return int(time.mktime(dt.timetuple()))
def epoch_2_datetime(epoch):
return datetime.fromtimestamp(epoch)
data = defaultdict(list)
with open(sys.argv[1]) as fin:
for line in fin: #Parse file 1 line at a time
timestr,datastr = line.rsplit(None,1)
try:
dt = datetime.strptime(timestr,"%Y%m%d %H:%M:%S")
val = float(datastr)
except ValueError: #couldn't read this line. must be a comment or something.
continue
bin = datetime_2_epoch(dt)//300 #300 = 60*5 -- 5 minute bin size
data[bin].append(val)
for bin,lst in sorted(data.items()):
cum_sum = sum(lst)
avg = cum_sum/len(lst)
print epoch_2_datetime(bin*300),avg,cum_sum
This will format your datafile (run on your sample data) as:
2013-02-06 11:45:00 5029.5 30177.0
2013-02-06 11:55:00 5029.5 30177.0
which can be plotted with boxes in gnuplot:
set xdata time
set timefmt '%Y-%m-%d %H:%M:%S'
set yrange [0:*]
plot '<python test.py test.dat' u 1:3 w boxes title "5 minute average"
or
plot '<python test.py test.dat' u 1:4 w boxes title "5 minute sum"
