# Based on ais_normalize.py from https://github.com/schwehr/noaadata

import sys
import traceback
import warnings

import ais
from ais.stream import checksum

warnings.warn(
  "The stream module is deprecated and will be removed in 1.0",
  FutureWarning,
  stacklevel=2
)


def ErrorPrinter(e,
                 stats,
                 verbose=False,
                 max_errors=None, # In % of total number of input lines
                 **kw):
  if max_errors != None and float(stats["error_num_total"]) / float(stats["line_num"]) * 100.0 > max_errors:
    raise TooManyErrorsError(**stats)
  if verbose:
    sys.stderr.write('%s\n' % e)


# def ErrorRaiser(e, stats, **kw):
#   raise e


class StreamError(Exception):

  description = 'Stream error'

  def __init__(self, **kw):
    self.kw = kw
    self.kw['description'] = self.description

  def __str__(self):
    return '%(description)s: %(line_num)s: %(line)s' % self.kw


class InvalidChecksumError(StreamError):
  description = 'Invalid checksum'


class InvalidChecksumInConstructedError(StreamError):
  description = 'Invalid checksum in constructed one-liner'


class NoStationFoundError(StreamError):
  description = 'No station found'


class TooFewFieldsError(StreamError):
  description = 'Too few fields'

  def __str__(self):
    return '%(description)s, got %(fields)s but needed 6: %(line_num)s: %(line)s' % self.kw


class MissingTimestampsError(StreamError):
  description = 'Timestamps missing'

  def __str__(self):
    return '%(description)s: %(line_num)s: %(line)s, parts: %(parts)s' % self.kw

class DifferingTimestampsError(StreamError):
  description = 'Timestamps not all the same'

  def __str__(self):
    return '%(description)s for %(timestamp)s: %(line_num)s: %(line)s, parts: %(parts)s' % self.kw

class OnlyMessageEndError(StreamError):
  description = 'Do not have the preceeding packets for'

  def __str__(self):
    return '%(description)s for %(bufferSlot)s:\n%(line)s\n' % self.kw

class UnfinishedMessagesError(StreamError):
  description = 'Unfinished messages at end of file'

  def __str__(self):
    return '%(description)s:\n%(buffers)s\n' % self.kw


class TooManyErrorsError(StreamError):
  description = 'Too many errors'

  def __str__(self):
    res = dict(self.kw)
    res['error_lines'] = ""
    if 'error_num' in res:
      res['error_lines'] = "\n" + "\n".join("  %s: %s" % (error, num) for error, num in res['error_num'].iteritems())
    return '%(description)s: %(error_num_total)s errors in %(line_num)s lines:%(error_lines)s' % res


def parseTagBlock(line):
  if not line.startswith("\\"):
    return {}, line
  tagblock, line = line[1:].split("\\", 1)
  tagblock, checksum = tagblock.rsplit("*", 1)

  tags = {}
  for field in tagblock.split(","):
    key, value = field.split(":")

    if key == 'c':
      key = 'timestamp'
      value = int(value)
      # This can be either seconds or milliseconds...
      if value > 40000000000:
        value = value / 1000.0
    elif key == 'n':
      key = 'line_count'
      value = int(value)
    elif key == 'r':
      key = 'relative_time'
      value = int(value)
    elif key == 'd':
      key = 'destination'
    elif key == 's':
      key = 'station'
    elif key == 't':
      key = 'text'
    elif key == 'g':
      key = 'group'
      value = dict(zip(["sentence", "groupsize", "id"],
                       [int(part) for part in value.split("-")]))

    tags["tagblock_" + key] = value
  return tags, line


def add_error_to_stats(e, stats):
  if "error_num_total" not in stats:
    stats["error_num_total"] = 0
  stats["error_num_total"] += 1
  if "error_num" not in stats:
     stats["error_num"] = {}
  name = getattr(e, "description", getattr(e, "message", str(type(e))))
  if name not in stats["error_num"]:
     stats["error_num"][name] = 0
  stats["error_num"][name] += 1


def normalize(nmea=sys.stdin,
              uscg=True,
              validate_checksum=True,
              allow_unknown=False,
              window=2,
              ignore_tagblock_station=False,
              treat_ab_equal=False,
              pass_invalid_checksums=False,
              allow_missing_timestamps=False,
              errorcb=ErrorPrinter,
              stats=None,
              **kw):
  """Assemble multi-line messages

  Args:
    nmea: iterator of lines to read
    window: number of seconds to allow the later parts of a multiline message to span
  """

  buffers = {} # Put partial messages in a queue by station so that they can be reassembled
  if stats is None: stats={}
  stats['line_num'] = stats.pop('line_num', 0)
  stats['error_num_total'] = stats.pop('error_num_total', 0)
  stats['invalid_checksums'] = stats.pop('invalid_checksums', 0)

  def report_error(e):
    add_error_to_stats(e, stats)
    errorcb(e, stats, **kw)

  if not uscg:
    report_error('Need to make a faster version that does not worry about the extra args and stations dict')
    assert False

  for idx, origline in enumerate(nmea):
    try:
      tagblock, line = parseTagBlock(origline)

      line = line.strip() + '\n'  # Get rid of DOS issues.
      stats["line_num"] += 1
      if len(line) < 7 or line[3:6] not in ('VDM', 'VDO'):
        yield tagblock, line, origline
        continue

      if validate_checksum and not checksum.isChecksumValid(line):
        stats["invalid_checksums"] += 1
        report_error(InvalidChecksumError(line_num=stats["line_num"], line=line.strip()))
        if not pass_invalid_checksums:
          continue

      fields = line.split(',')

      if len(fields) < 6:
        report_error(TooFewFieldsError(line_num=stats["line_num"],
                                  line=line.strip(),
                                  fields=len(fields)))
        continue

      # Total NMEA lines that compose this message [1..9].
      totNumSentences = int(fields[1])
      if 1 == totNumSentences:
        # A single line needs no work, so pass it along.
        yield tagblock, line, origline
        continue

      sentenceNum = int(fields[2])  # Message sequence number 1..9 (packetNum)
      payload = fields[5]  # AIS binary data encoded in whacky ways
      timestamp = fields[-1].strip()   # Seconds since Epoch UTC.  Always the last field

      station = None  # USCG Receive Stations        # if None==station:
      for i in range(len(fields)-1, 5, -1):
        if len(fields[i]) and fields[i][0] in ('r', 'b'):
          station = fields[i]
          break  # Found it so ditch the for loop.

      if ignore_tagblock_station:
        tagblock_station = None
      else:
        tagblock_station = tagblock.get('tagblock_station', None)

      if station is None and allow_unknown:
        station = 'UNKNOWN'

      if station is None and tagblock_station is None:
        report_error(NoStationFoundError(line_num=stats["line_num"], line=line.strip()))
        continue

      bufferSlot = (tagblock_station, station, fields[3])  # seqId and Channel make a unique stream

      if not treat_ab_equal:
        bufferSlot += (fields[4],)  # channel id

      newPacket = {"payload": payload,
                   "timestamp": timestamp,
                   "tagblock": tagblock,
                   "origline": origline}
      if sentenceNum == 1:
        buffers[bufferSlot] = [newPacket]  # Overwrite any partials
        continue

      if totNumSentences == sentenceNum:
        # Finished a message
        if bufferSlot not in buffers:
          report_error(OnlyMessageEndError(line=line, bufferSlot=bufferSlot))
          continue
        buffers[bufferSlot].append(newPacket)
        parts = buffers[bufferSlot]  # Now have all the pieces.
        del buffers[bufferSlot]  # Clear out the used packets to save memory.

        # Sanity check
        ok = True
        ts1 = None
        for part in parts:
          try:
            ts1 = float(part['timestamp'])
            ts2 = float(timestamp)
          except ValueError:
            try:
              ts1 = float(part['tagblock']['tagblock_timestamp'])
              ts2 = float(tagblock['tagblock_timestamp'])
            except:
              if allow_missing_timestamps:
                ts1 = 0
                ts2 = 0
              else:
                report_error(MissingTimestampsError(line_num=stats["line_num"],
                                                    line=line.strip(),
                                                    timestamp=timestamp, parts=parts))
                ok = False
                break
          if ts1 > ts2+window or ts1 < ts2-window:
            report_error(DifferingTimestampsError(line_num=stats["line_num"],
                                                  line=line.strip(),
                                                  timestamp=timestamp, parts=parts))
            ok = False
            break
        if not ok:
          continue

        payload = ''.join([p['payload'] for p in parts])
        tagblock = {}
        for p in reversed(parts):
          tagblock.update(p['tagblock'])

        # Try to mirror the packet as much as possible... same seqId and channel.
        checksumed_str = ','.join((fields[0], '1,1', fields[3], fields[4],
                                   payload, fields[6].split('*')[0]+'*'))
        if ts1 == 0:
            # allowed missing timestamp and it is missing
            if len(fields[7:-1]) == 0:
              out_str = checksumed_str + checksum.checksumStr(checksumed_str)
            else:
              out_str = checksumed_str + checksum.checksumStr(checksumed_str) +',' + ','.join(fields[7:-1])
        else:
          out_str = checksumed_str + checksum.checksumStr(checksumed_str) +',' + ','.join(fields[7:])

        if not checksum.isChecksumValid(out_str):
          report_error(InvalidChecksumInConstructedError(line_num=stats["line_num"], line=line.strip()))

        out_str = out_str.strip()+'\n'  # FIX: Why do I have to do this last strip?
        origstr = ''.join([p['origline'] for p in parts])

        yield tagblock, out_str, origstr

        continue

      buffers[bufferSlot].append(newPacket)
    except TooManyErrorsError:
      raise
    except Exception as inst:
      report_error(inst)

  if buffers:
    report_error(UnfinishedMessagesError(buffers=buffers))


def decode(nmea=sys.stdin,
           errorcb=ErrorPrinter,
           keep_nmea=False,
           stats=None,
           **kw):
  """Decodes a stream of AIS messages. Takes the same arguments as normalize."""

  if stats is None: stats={}

  def report_error(e):
    add_error_to_stats(e, stats)
    errorcb(e, stats, **kw)

  for tagblock, line, origline in normalize(nmea=nmea, errorcb=errorcb, stats=stats, **kw):
    try:
      body = ''.join(line.split(',')[5])
      pad = int(line.split('*')[0][-1])
      res = ais.decode(body, pad)
      res.update(tagblock)
      if keep_nmea:
        res['nmea'] = origline
      yield res
    except TooManyErrorsError:
      raise
    except Exception as e:
      report_error(e)