analyze_groundtruth

#!/usr/bin/env python2.7
import argparse
import glob
import json
import csv
import sys
import os

EXTRA_PATH = "/data/bff-results/bff/results/crashers/%s/%s"

"""
Something this code needs to do:
  The ground truth data will include potentially multiple seps per bug. What 
  we should do is keep a map from the sets of fixes to (potentailly) fresh
  integers, and "flatten" the data out into a first order map from one crasher
  to one identifier. Then our comparison algorithms will work normally. 

  We might have some different knobs to turn to make those flattening decisions 
  differently in the future. 
"""

def ed_data(args):
  truths = {}
  truths["name"] = args.name
  truths["labels"] = []
  labels = {}
  fresh_label = 0
  reader = csv.reader(open(args.groundtruth, 'r'))
  for i in reader:
    status = i[1]
    path = i[3]
    vfixes = i[8]
    if status == "NORMAL":
      fixes = [ i.split(":")[0] for i in vfixes.split(" ")]
      fixes_s = frozenset(fixes)
      fixes_label = None
      if args.merge:
        found = False
        for j in labels:
          pass

        if found == False:
          labels[fixes_s] = fresh_label
          fresh_label = fresh_label + 1

        for k in labels:
          pass

        fixes_label = 0
      else:
        if labels.has_key(fixes_s):
          fixes_label = labels[fixes_s]
        else:
          labels[fixes_s] = fresh_label
          fixes_label = fresh_label
          fresh_label = fresh_label + 1
        realpath = args.prefixdir + path.replace(EXTRA_PATH % (args.name, args.name), "")
        truths["labels"].append((realpath, fixes_label))

  output = open(args.outputfile, 'w')
  json.dump(truths, output)
  output.close()
  return 0

# In this case, the data formatting is a little different. 
# We need to take a base dir for where the crashers are, and a 
# base dir for where the ground truth data is. 
def benji_data(args):
  # Base dir where the ground truth .txt files are. 
  groundtruth_base_dir = args.groundtruth
  # Base dir where all the crashing files are. 
  crashers_dir = args.prefixdir 

  truths = {}
  truths["name"] = args.name
  truths["labels"] = []
  cluster_id = 0

  for c in glob.glob(groundtruth_base_dir+"/*.txt"):
    # Ignore the crash_list entry.
    if os.path.basename(c) == "crash_list.txt":
      continue
  
    buf = open(c, 'r').read()
    crashers = [ i for i in buf.split("\n") if len(i) > 0 ]
    for f in crashers:
      realpath = crashers_dir + f
      truths["labels"].append((realpath, cluster_id))
    cluster_id = cluster_id + 1
  
  output = open(args.outputfile, 'w')
  json.dump(truths, output)
  output.close()
  return 0

def andrew_data(args):
  q = open(args.groundtruth, 'r')
  qr = csv.reader(q)
  qr.next()

  truths = {}
  truths["name"] = args.name
  truths["labels"] = []
  cluster_map = {}
  fresh_cluster_id = 0
  skip = set(["2015-11-17 06:37:14", "unfixed"])
 
  for (inputfile,classname) in qr:
    if classname in skip:
      continue
    if classname not in cluster_map.keys():
      cluster_map[classname] = fresh_cluster_id
      fresh_cluster_id = fresh_cluster_id + 1
    truths["labels"].append((inputfile, cluster_map[classname]))
  
  output = open(args.outputfile, 'w')
  json.dump(truths, output)
  output.close()
  return 0


def main(args):
  if args.data_source == 'ed':
    return ed_data(args)
  elif args.data_source == 'benji':
    return benji_data(args)
  elif args.data_source == 'andrew':
    return andrew_data(args)
  return 1


if __name__ == '__main__':
  parser = argparse.ArgumentParser("csvtool")
  parser.add_argument("name")
  parser.add_argument("groundtruth")
  parser.add_argument("outputfile")
  parser.add_argument("prefixdir")
  parser.add_argument("--merge", default=False, action="store_true")
  parser.add_argument('--data-source', type=str, default='ed', choices=['ed', 'benji', 'andrew'])
  args = parser.parse_args()
  sys.exit(main(args))