#!/usr/bin/python2.7

"""report: make a table summarizing output from one or more runs of `sample`."""

import collections
import csv
import datetime
import os
import re
import sys

import ifstats
import iperf
import isostream
import options
import wifipacket

optspec = """
report [options...] <journal>
--
r,report_dir=  path to a single report directory to be parsed
"""

NFILE = 'n-datarates.tsv'
nrates = {}

CHANNELFILE = 'channels.tsv'
channels = {}


def _Resource(name):
  return os.path.join(os.path.dirname(os.path.abspath(__file__)), name)


def LoadNRates():
  """Loads 802.11n coding and data rates into a global variable."""
  if nrates: return

  raw = []

  with open(_Resource(NFILE), 'rb') as csvfile:
    reader = csv.reader(csvfile, delimiter='\t')
    next(reader)  # skip header row when reading by machine
    for mcs, width, gi, rate in reader:
      raw.append([int(mcs), int(width), int(gi), float(rate)])

  # Load global table, computing MCS 8-31 statistics from MCS 0-7.
  for mcs, width, gi, rate in raw:
    for i in range(4):
      nrates[(8*i + mcs, width, gi)] = rate * (i + 1)


def LoadChannels():
  """Load 802.11n channels and frequencies into a global variable."""
  if channels: return

  with open(_Resource(CHANNELFILE), 'rb') as csvfile:
    reader = csv.reader(csvfile, delimiter='\t')
    next(reader)

    for channel, freq in reader:
      channels[int(channel)] = int(freq)


def ParseMCSFile(outfile, width=20):
  """Extract MCS and PHY rate statistics from an MCS report file."""
  LoadNRates()

  # assume long guard interval
  guard = 800

  counter = collections.Counter()
  for line in outfile:
    for tok in line.split():
      if tok == '.': continue

      mcs = int(tok)
      counter[mcs] += 1

  phy = 0.0
  alltimes = 0
  for mcs, times in counter.iteritems():
    phy += nrates[(mcs, width, guard)] * times
    alltimes += times

  return counter.most_common()[0][0], phy / alltimes


def ParsePcap(stream, known_ssids):
  """ParsePcap computes PHY performance for a packet capture in `stream`.

  Args:
    stream: an open `file` object pointing to a file with pcap data.
    known_ssids: a dict mapping known BSSIDs to SSIDs.

  Returns:
    a dict containing PHY performance information. keys are descriptive strings,
    values are strings or numbers in Mbit/s.
  """

  rates = collections.defaultdict(list)
  beacon_powers = collections.defaultdict(list)
  times_seen = collections.Counter()
  start_secs = None

  for opt, frame in wifipacket.Packetize(stream):
    if start_secs is None:
      start_secs = opt.pcap_secs

    # Fill in known SSIDs array if our scan didn't see anything.
    ssid = wifipacket.IdentifySSID(opt, frame)
    if ssid is not None:
      known_ssids[opt.ta] = ssid

    for sta, direction in [('ra', 'up'), ('ta', 'down'), ('xa', 'across')]:
      bssid = opt.get(sta)
      ssid = known_ssids.get(bssid)
      if ssid:
        if opt.type == 0x08:  # Beacon
          rssi = opt.get('dbm_antsignal')
          if rssi is not None: beacon_powers[ssid].append(rssi)
        else:
          rates[ssid].append((opt.pcap_secs - start_secs,
                              direction,
                              opt.get('dbm_antsignal'),
                              opt.rate,
                              len(frame)))
        times_seen[ssid] += 1
        break

  if not times_seen:
    return {}

  modal_ssid, _ = times_seen.most_common(1)[0]
  summary = {}
  for _, direction, rssi, rate, size in rates[modal_ssid]:
    size_weighted_rate = rate * float(size)
    if direction not in summary:
      summary[direction] = [size_weighted_rate, size, 0, 0]
    else:
      summary[direction][0] += size_weighted_rate
      summary[direction][1] += size
    if rssi is not None:
      summary[direction][2] += rssi
      summary[direction][3] += 1

  line = {
      'PHY ssid': modal_ssid,
      'Beacon RSSI': (sum(beacon_powers[modal_ssid]) /
                      float(len(beacon_powers[modal_ssid]))
                      if beacon_powers.get(modal_ssid) else 0.0)
  }

  for direction, accum in summary.items():
    size_weighted_rate, size = accum[:2]
    line['PHY {}'.format(direction)] = ((size_weighted_rate / size) if size
                                        else 0.0)

    rssi, count = accum[2:]
    line['PHY RSSI {}'.format(direction)] = ((rssi / float(count)) if count
                                             else 0.0)

  return line


def Channel(text_channel):
  """Given a text channel spec like 149,+1 return the central freq and width."""
  LoadChannels()

  if ',' in text_channel:
    control, second = text_channel.split(',')
    control = int(control)
    second = int(second)

    if second == 80:
      five_ghz_channels = sorted(ch for ch in channels.keys() if ch >= 36)
      i = five_ghz_channels.index(control)
      base = five_ghz_channels[i - i % 4]
      return control, channels[base] + 30, 80
    elif second in [-1, 1]:
      freq = channels[control]
      offset = second
      return control, freq + offset * 10, 40
    else:
      raise AssertionError('text channel "{}" does not match any known format'
                           .format(text_channel))
  else:
    control = int(text_channel)
    return control, channels[control], 20


def Overlap(c1, w1, c2, w2):
  """Return True if two WiFi channels overlap, or False otherwise."""
  # TODO(willangley): replace with code from Waveguide
  b1 = c1 - w1 / 2
  t1 = c1 + w1 / 2
  b2 = c2 - w2 / 2
  t2 = c2 + w2 / 2

  return ((b1 <= b2 <= t1) or (b2 <= b1 <= t2)
          or (b1 <= t2 <= t1) or (b2 <= t1 <= t2))


def ReportLine(report_dir, series=None):
  """Condense the output of a sample.py run into a one-line summary report."""
  line = collections.OrderedDict()
  if series:
    line['Series'] = series

  _, stamp, steps = os.path.basename(report_dir).split('-')
  line['Time'] = datetime.datetime.fromtimestamp(float(stamp))
  line['Steps'] = int(steps)

  system, cache = ifstats.Restore(report_dir)

  # known_ssids is a map from BSSID(string) => SSID(string)
  known_ssids = {}

  if 'Darwin' in system:
    result = ifstats.Parse('Darwin', cache)
    airport = result.get('link')
    control, freq, width = Channel(airport['channel'])
    shared = 0
    overlap = 0

    scan = result.get('scan')
    if len(scan) > 1:
      for row in scan:
        oc, of, ow = Channel(row['CHANNEL'])
        if control == oc:
          shared += 1
        elif Overlap(freq, width, of, ow):
          overlap += 1

        known_ssids[row['BSSID']] = row['SSID']

    line.update({
        'Channel': freq,
        'Width': width,
        'RSSI': airport['agrCtlRSSI'],
        'Noise': airport['agrCtlNoise'],
        'Shared': shared,
        'Interfering': overlap
    })

  if 'Linux' in system:
    result = ifstats.Parse('Linux', cache)
    iwlink = result.get('link')
    signal = int(iwlink.get('signal', '0 dBm').split()[0])
    channel = int(iwlink.get('freq', '0'))
    width = 20
    m = re.search(r'(\d+)MHz', iwlink.get('tx bitrate', ''), flags=re.I)
    if m:
      width = int(m.group(1))

    # Noise and contention not yet gathered in samples run on Linux systems.
    line.update({
        'Channel': channel,
        'Width': width,
        'RSSI': signal,
    })

  # TODO(willangley): integrate skid statistics with the rest of the benchmark
  #   framework system detection.
  try:
    with open(os.path.join(report_dir, 'status_wireless')) as status_wireless:
      result = ifstats.skids.ParseStatusWireless(status_wireless.read())
      width = result['Bandwidth'].split()[0]
      line.update({
          'Channel': result['Channel'],
          'Width': width,
          'RSSI': result['RSSI'],
      })
  except IOError:
    pass

  try:
    ppath = os.path.join(report_dir, 'testnetwork.pcap')
    with open(ppath) as stream:
      line.update(ParsePcap(stream, known_ssids))
  except IOError:
    try:
      mpath = os.path.join(report_dir, 'mcs')
      with open(mpath) as mf:
        mcs, phy = ParseMCSFile(mf, width)
        line['MCS'] = mcs
        line['PHY'] = phy
    except IOError:
      pass

  # If the initial ping test fails, we won't collect performance information.
  # deal with this gracefully.
  ips = iperf.Restore(report_dir)
  if 'iperf' in ips:
    # pylint:disable=line-too-long
    for key, perf in [('TCP BW up', iperf.ParseIperfTCP(ips.get('iperf', ''))),
                      ('UDP BW up', iperf.ParseIperfUDP(ips.get('iperfu', '')))]:
      line[key] = perf.get('bandwidth')
      line['{} units'.format(key)] = perf.get('bandwidth_unit')
  elif 'iperf3' in ips:
    for name in (key for key in ips
                 if key.startswith('iperf3')):
      perf = iperf.ParseIperf3(ips[name])
      if not perf or 'error' in perf:
        continue

      test_start = perf['start']['test_start']
      protocol = test_start['protocol']
      direction = 'down' if test_start['reverse'] else 'up'
      key = '{protocol} BW {direction}'.format(protocol=protocol,
                                               direction=direction)

      if protocol == 'TCP':
        line[key] = perf['end']['sum_received']['bits_per_second']
      elif protocol == 'UDP':
        line[key] = (perf['end']['sum']['bits_per_second'] *
                     (100 - float(perf['end']['sum']['lost_percent'])) / 100.0)
      else:
        continue

      line['{} units'.format(key)] = 'bit/s'

  try:
    with open(os.path.join(report_dir, 'isostream')) as istm:
      text = istm.read()
      line['isostream'] = isostream.ParseIsostream(text)
  except IOError:
    pass

  return line


def ReadJournal(jname):
  """Read a journal, returning a series name and its data folders."""
  jname = os.path.realpath(jname)
  series = os.path.basename(jname)
  if series == 'journal':
    series = os.path.basename(os.path.dirname(jname))

  folders = []
  with open(jname) as journal:
    for line in journal:
      line = line.strip()
      if line.startswith('#'):
        continue
      folders.append(os.path.join(os.path.dirname(jname), line))

  return series, folders


def Report(journals):
  """Given the name of a journal file, return a list of ReportLines."""
  report = []
  bad = []

  for jname in journals:
    series, folders = ReadJournal(jname)
    for folder in folders:
      try:
        report += [ReportLine(folder, series=series)]
      except (TypeError, IOError, ValueError) as e:
        bad += [collections.OrderedDict(folder=folder, error=repr(e))]

  return report, bad


def WriteReport(lines):
  """Write a network testing report in .tsv format to stdout."""
  # include every field we can write in the header row
  header = ['Series', 'Time', 'Steps', 'Channel', 'Width', 'RSSI',
            'Beacon RSSI', 'Noise', 'Shared', 'Interfering',
            'MCS', 'PHY', 'PHY ssid', 'PHY up', 'PHY down', 'PHY across',
            'PHY RSSI up', 'PHY RSSI down', 'PHY RSSI across',
            'TCP BW up', 'TCP BW up units', 'UDP BW up', 'UDP BW up units',
            'TCP BW down', 'TCP BW down units', 'UDP BW down',
            'UDP BW down units', 'isostream']

  writer = csv.DictWriter(sys.stdout, header, dialect=csv.excel_tab)
  writer.writeheader()
  writer.writerows(lines)


def main():
  o = options.Options(optspec)
  (opt, _, extra) = o.parse(sys.argv[1:])

  if opt.report_dir:
    report = [ReportLine(opt.report_dir)]
  elif extra:
    report, bad = Report(extra)
    if bad:
      writer = csv.DictWriter(sys.stdout, bad[0].keys(), dialect=csv.excel_tab)
      writer.writeheader()
      writer.writerows(bad)
      print

  if len(report) < 1:
    o.fatal("Didn't find any samples. Did you supply at least one report dir"
            ' or journal?')

  WriteReport(report)


if __name__ == '__main__':
  main()