format_results.py

#!/usr/bin/env python


"""
Helper script to extract all the data from *.dat files and generate one table
"""

import argparse
import csv
import glob
import os
import re


# Values we care about
keys = []
keys.append('num_events')
keys.append('num_edges')
keys.append('num_ops')
keys.append('num_read')
keys.append('num_writes')

keys.append('num_time_edges')

keys.append('num_races')
keys.append('num_harmful')
keys.append('num_commute')
keys.append('num_covered')
keys.append('num_time_filtered_races')

keys.append('num_pkts')
keys.append('num_per_pkt_races')
keys.append('num_per_pkt_inconsistent')
keys.append('num_per_pkt_inconsistent_covered')
keys.append('num_per_pkt_entry_version_race')
keys.append('num_per_pkt_inconsistent_no_repeat')
keys.append('num_versions')
keys.append('num_racing_versions')


table_keys = ['rw_delta', 'ww_delta', 'alt_barr', 'data_deps']
table_keys.extend(keys)

timing_keys = []


timing_keys.append('total_time_sec')
timing_keys.append('load_time_sec')
timing_keys.append('detect_races_time_sec')
timing_keys.append('extract_traces_time_sec')
timing_keys.append('find_reactive_cmds_time_sec')
timing_keys.append('find_proactive_cmds_time_sec')
timing_keys.append('find_covered_races_time')
timing_keys.append('per_packet_inconsistent_time_sec')
timing_keys.append('find_inconsistent_update_time_sec')

timing_table_keys = ['rw_delta', 'ww_delta', 'alt_barr', 'data_deps']
timing_table_keys.extend(timing_keys)


okay_to_ignore = ['key', 'rw_delta', 'ww_delta', 'alt_barr', 'data_deps']


def natural_keys(text):
  def atoi(text):
    return int(text) if text.isdigit() else text
  return [atoi(c) for c in re.split('(\d+)', text)]


def read_dat_files(file_keys, infiles):
  table = {}
  for key in file_keys:
    table[key] = {}
  for fname in infiles:
    print "READING FILE", fname
    with open(fname, 'r') as csvfile:
      data = csv.reader(csvfile, delimiter=',')
      t = None
      ab = None
      for row in data:
        key, value = row[0].strip(), row[1].strip()
        if key in ['alt_barr']:
          ab = value
        if key in ['rw_delta', 'ww_delta']:
          if value != 'inf':
            value = int(value)
          if t is None:
            t = value
          else:
            # TODO(jm): This is a really weird way of checking that rw_delta == ww_delta.
            #           We should make t a tuple (t_rw, t_ww) and print that out.
            #           If this assertion fails you might need to delete any *.dat files
            #           remaining from previous runs where rw_delta != ww_delta
            assert t == value
        if key in table:
          table[key][str(ab) + '-' + str(t)] = value
        elif key not in okay_to_ignore:
          print "Ignored row", row
  return table


def save_to_csv(file_keys, table, outname):
  print "Saving results to", outname
  with open(outname, 'w') as f:
    wr = csv.writer(f, delimiter=',')
    for key, values in table.iteritems():
      row = ['key/t'] + list(sorted(values.keys(), key=natural_keys))
      wr.writerow(row)
      break
    for key in file_keys:
      values = table[key]
      times = sorted(values.keys(), key=natural_keys)
      row = [key] + [values[t] for t in times]
      wr.writerow(row)


def read_dat_files_table(file_keys, infiles):
  table = []
  for fname in infiles:
    print "READING FILE ALT", fname
    loaded = {}
    with open(fname, 'r') as csvfile:
      data = csv.reader(csvfile, delimiter=',')
      t = None
      for row in data:
        key, value = row[0].strip(), row[1].strip()
        if key in ['rw_delta', 'ww_delta']:
          if value != 'inf':
            value = int(value)
          if t is None:
            t = value
          else:
            # TODO(jm): This is a really weird way of checking that rw_delta == ww_delta.
            #           We should make t a tuple (t_rw, t_ww) and print that out.
            #           If this assertion fails you might need to delete any *.dat files
            #           remaining from previous runs where rw_delta != ww_delta
            assert t == value
        loaded[key] = value
    row = []
    for k in file_keys:
      row.append(loaded[k])
    table.append(row)
  table.sort(key=lambda x: x[0])
  table.insert(0, file_keys)
  return table


def save_to_csv_table(table, outname):
  print "Saving results to", outname
  with open(outname, 'w') as f:
    wr = csv.writer(f, delimiter=',')
    for row in table:
      wr.writerow(row)
  return table


def main(result_dir):
  # Reading files
  table = read_dat_files(keys, glob.glob(os.path.join(result_dir, 'results*.dat')))
  timings_table = read_dat_files(timing_keys, glob.glob(os.path.join(result_dir, 'timings*.dat')))

  table_tbl = read_dat_files_table(table_keys, glob.glob(os.path.join(result_dir, 'results*.dat')))
  timings_tbl= read_dat_files_table(timing_table_keys, glob.glob(os.path.join(result_dir, 'timings*.dat')))

  # Saving summary
  save_to_csv(keys, table, os.path.join(result_dir, 'summary.csv'))
  # Saving timing results
  save_to_csv(timing_keys, timings_table,
              os.path.join(result_dir, 'summary_timings.csv'))

  save_to_csv_table(table_tbl, os.path.join(result_dir, 'summary_tbl.csv'))
  save_to_csv_table(timings_tbl, os.path.join(result_dir, 'summary_timings_tbl.csv'))


if __name__ == '__main__':
  parser = argparse.ArgumentParser(
    description='Helper script to extract all the data from *.dat files and'
                ' generates two csv files: summary.csv and summary_timings.csv')
  parser.add_argument('result_dir',
                      help='Directory where [results|timings]*.dat files are located')
  args = parser.parse_args()
  main(args.result_dir)