Pythonized version of ROOT's hadd program for adding histograms together. It breaks jobs into manageable sized chunks.

hadder.py

#!/usr/bin/env python2.7


""" 
root's hadd gets really inefficiency for large numbers of files - it becomes much more efficienct to do it in small batches. 
This tool hadds some collection of files in smaller batches and gives you the result. maybe in the future ill make it parallelized
"""
import argparse
import os
import glob



def chunks(l, n):
    """Yield successive n-sized chunks from l. - thank you Ned Batchelder : http://stackoverflow.com/users/14343"""
    for i in xrange(0, len(l), n):
        yield l[i:i+n]

def hadd( files, n, prefix="tmp_" ) :

	chunky = list( chunks( files, n ) )

	n_to_make = len(chunky)
	print "Length :", n_to_make

	index = 0
	new_files = []
	for c in chunky :
		
		if n_to_make > 1 :
			new_name = prefix + str(index) + ".root"
		else :
			new_name = "merged.root"

		new_files.append( new_name )
		cmd = "hadd " + new_name + " " + " ".join(c)
		print cmd
		os.system( cmd )
		index = index+1
	print " created :", " ".join( new_files )

	if len( new_files ) > 1 :
		hadd( new_files, n, prefix + "tmp_" )
		

parser = argparse.ArgumentParser( description="Creates the configs for Tasks" );
parser.add_argument( "needle", default="*.root", help="selects all of the files that match the glob" )
parser.add_argument( "N", default=50, type=int, help="split the hadd into N files per batch" )


args = parser.parse_args()
files = glob.glob( args.needle )

hadd( files, args.N )