groupFilesInBlocks.py

Go to the documentation of this file.

#! /usr/bin/env python

import re,os,sys,shutil,math
import optparse

copyargs = sys.argv[:]
for i in range(len(copyargs)):
    if copyargs[i] == "":
        copyargs[i] = "\"\""
    if copyargs[i].find(" ") != -1:
        copyargs[i] = "\"%s\"" % copyargs[i]
commandline = " ".join(copyargs)

prog = sys.argv[0]

usage='./%(prog)s NBLOCKS INFILE OUTFILE [options]\n'+\
  'takes list of files produced by findQualityFiles.py as INFILE,\n'+\
  'groups them into maximum NBLOCKS blocks with approximately similar #events.'


######################################################
# To parse commandline args


parser=optparse.OptionParser(usage)

parser.add_option("-v", "--verbose",
  help="debug verbosity level",
  type="int",
  default=0,
  dest="debug")

options,args=parser.parse_args()

if len(sys.argv) < 4:
    raise SystemError("Too few arguments.\n\n"+parser.format_help())

NBLOCKS = int(sys.argv[1])
INFILE = sys.argv[2]
OUTFILE = sys.argv[3]



def makeJobBlock(mylist, evtn):
  n = mylist[0][0]
  block = [mylist[0]]
  choosen = [0]
  while n<evtn:
    #print "n,evtn=",n,evtn
    # find the biggest unused #evt that would give n<evtn
    for i in range(len(mylist)):
      # get last not choosen i
      last_i=len(mylist)-1
      while last_i in choosen: last_i += -1
      if i==last_i:
        #print i,"last element reached"
        n += mylist[i][0]
        #print "   new last append: ",i, mylist[i][0], n
        block.append(mylist[i])
        choosen.append(i)
        break
      if i in choosen:
        #print i,"  in choosen, continue..."
        continue
      if n+mylist[i][0]<evtn:
        n += mylist[i][0]
        #print "   new append: ",i, mylist[i][0], n
    block.append(mylist[i])
    choosen.append(i)
    break
    if len(choosen)==len(mylist):
      #print " got everything"
      break
  # pick up unused elements
  newlist = []
  for i in range(len(mylist)):
    if not i in choosen:
      newlist.append(mylist[i])
  print "done makeJobBlock n =",n," len =",len(block)
  return block, newlist, n



comment1RE = re.compile (r'^#.+$')
fileLineRE = re.compile (r'^.*\'(.*)\'.+# (\d*).*$')
#fileLineRE = re.compile (r'^.*\'(.*)\'.+# (\d*),(\d*).*$')

if not os.access(INFILE, os.F_OK): 
  print "Cannot find input file ", INFILE
  sys.exit()

fin = open(INFILE, "r")
lines = fin.readlines()
fin.close()


eventsFiles = []
ntotal = 0
commentLines=[]

for line in lines:
  #line = comment1RE.sub ('', line)
  #line = line.strip()
  #if not line: continue
  match = comment1RE.match(line)
  if match:
    commentLines.append(line)
  
  match = fileLineRE.match(line)
  if match:
    #print int(match.group(3)), str(match.group(1))
    #eventsFiles.append((int(match.group(3)), str(match.group(1)), str(match.group(2))))
    eventsFiles.append((int(match.group(2)), str(match.group(1))))
    ntotal += int(match.group(2))
  #else: print line,

if len(eventsFiles)==0:
  print "no file description strings found"
  sys.exit()

#print "len=", len(eventsFiles), ntotal
#tmp = set(eventsFiles)
#eventsFiles = list(tmp)
#ntotal = 0
#for ff in eventsFiles:  ntotal += ff[0]
#print "len=", len(eventsFiles), ntotal
#sys.exit()

eventsFiles.sort(reverse=True)
#print eventsFiles

evtPerJob = int(math.ceil(float(ntotal)/NBLOCKS))
print "Total = ",ntotal, "  per block =", evtPerJob,"(would give total of ", evtPerJob*NBLOCKS, ")", "  list length =",len(eventsFiles)
if eventsFiles[0][0] > evtPerJob:
  print "the biggest #evt is larger then #evt/block:",eventsFiles[0][0],">",evtPerJob
  print "consider lowering NBLOCKS"


jobsBlocks=[]
temp = eventsFiles

tt = 0
for j in range(NBLOCKS):
  print j
  if len(temp)==0:
    print "done!"
    break
  block, temp, nn = makeJobBlock(temp,evtPerJob)
  tt+=nn
  if len(block)>0:
    jobsBlocks.append((block,nn))
    print block
  else:
    print "empty block!"
  
print tt
print commandline


fout = open(OUTFILE, mode="w")

fout.write("### job-split file list produced by:\n")
fout.write("### "+commandline+"\n")
fout.write("### Total #evt= "+str(ntotal)+"  #files ="+str(len(eventsFiles))+"  per job #evt="
           +str(evtPerJob)+" (would give total of"+str(evtPerJob*NBLOCKS)+")\n###\n")
fout.write("### previously produced by:\n")
fout.write("".join(commentLines))
fout.write("\nfileNamesBlocks = [\n")

commax = ","
for b in range(len(jobsBlocks)):
  fout.write('  [ # job '+str(b)+' with nevt='+str(jobsBlocks[b][1])+'\n')
  comma = ","
  for i in range(len(jobsBlocks[b][0])):
    if i==len(jobsBlocks[b][0])-1:
        comma=""
    #fout.write("    '"+ jobsBlocks[b][0][i][1] +"'"+comma+" # "+ str(jobsBlocks[b][0][i][2]) +','+ str(jobsBlocks[b][0][i][0]) + "\n")
    fout.write("    '"+ jobsBlocks[b][0][i][1] +"'"+comma+" # "+ str(jobsBlocks[b][0][i][0]) + "\n")
  if b==len(jobsBlocks)-1:
    commax=""
  fout.write('  ]'+commax+'\n')
fout.write(']\n')
fout.close()