def insert(self, entries, userid=None):
all_observations = []
drop_counter = defaultdict(int)
new_stations = 0
# Process entries and group by validated station key
station_observations = defaultdict(list)
for entry in entries:
self.pre_process_entry(entry)
obs = self.observation_model.create(**entry)
if not obs:
drop_counter['malformed'] += 1
continue
station_observations[obs.hashkey()].append(obs)
# Process observations one station at a time
for key, observations in station_observations.items():
first_blacklisted = None
incomplete = False
station = self.station_model.querykey(self.session, key).first()
if station is None:
# Drop observations for blacklisted stations.
blacklisted, first_blacklisted = self.blacklisted_station(key)
if blacklisted:
drop_counter['blacklisted'] += len(observations)
continue
incomplete = self.incomplete_observation(key)
if not incomplete:
# We discovered an actual new complete station.
new_stations += 1
# Accept all observations
all_observations.extend(observations)
num = len(observations)
# Accept incomplete observations, just don't make stations for them
# (station creation is a side effect of count-updating)
if not incomplete and num > 0:
self.create_or_update_station(station, key, num,
first_blacklisted)
# Credit the user with discovering any new stations.
if userid is not None and new_stations > 0:
scorekey = Score.to_hashkey(
userid=userid,
key=ScoreKey['new_' + self.station_type],
time=self.utcnow.date())
Score.incr(self.session, scorekey, new_stations)
added = len(all_observations)
self.emit_stats(added, drop_counter)
self.session.add_all(all_observations)
return added
def insert(self, entries, userid=None):
all_observations = []
drop_counter = defaultdict(int)
new_stations = 0
# Process entries and group by validated station key
station_observations = defaultdict(list)
for entry in entries:
self.pre_process_entry(entry)
obs = self.observation_model.create(**entry)
if not obs:
drop_counter['malformed'] += 1
continue
station_observations[obs.hashkey()].append(obs)
# Process observations one station at a time
for key, observations in station_observations.items():
first_blacklisted = None
incomplete = False
station = self.station_model.getkey(self.session, key)
if station is None:
# Drop observations for blacklisted stations.
blacklisted, first_blacklisted = self.blacklisted_station(key)
if blacklisted:
drop_counter['blacklisted'] += len(observations)
continue
incomplete = self.incomplete_observation(key)
if not incomplete:
# We discovered an actual new complete station.
new_stations += 1
# Accept all observations
all_observations.extend(observations)
num = len(observations)
# Accept incomplete observations, just don't make stations for them
# (station creation is a side effect of count-updating)
if not incomplete and num > 0:
self.create_or_update_station(station, key, num,
first_blacklisted)
# Credit the user with discovering any new stations.
if userid is not None and new_stations > 0:
scorekey = Score.to_hashkey(
userid=userid,
key=ScoreKey['new_' + self.station_type],
time=self.utcnow.date())
Score.incr(self.session, scorekey, new_stations)
added = len(all_observations)
self.emit_stats(added, drop_counter)
self.session.add_all(all_observations)
return added
def process_reports(self, reports, userid=None):
positions = []
cell_observations = []
wifi_observations = []
for i, report in enumerate(reports):
report['report_id'] = uuid.uuid1()
cell, wifi = self.process_report(report)
cell_observations.extend(cell)
wifi_observations.extend(wifi)
if cell or wifi:
positions.append({
'lat': report['lat'],
'lon': report['lon'],
})
if cell_observations:
# group by and create task per cell key
self.stats_client.incr('items.uploaded.cell_observations',
len(cell_observations))
if self.api_key_log:
self.stats_client.incr(
'items.api_log.%s.uploaded.'
'cell_observations' % self.api_key_name,
len(cell_observations))
cells = defaultdict(list)
for obs in cell_observations:
cells[CellObservation.to_hashkey(obs)].append(obs)
# Create a task per group of 5 cell keys at a time.
# Grouping them helps in avoiding per-task overhead.
cells = list(cells.values())
batch_size = 5
countdown = 0
for i in range(0, len(cells), batch_size):
values = []
for observations in cells[i:i + batch_size]:
values.extend([encode_radio_dict(o) for o in observations])
# insert observations, expire the task if it wasn't processed
# after six hours to avoid queue overload, also delay
# each task by one second more, to get a more even workload
# and avoid parallel updates of the same underlying stations
self.insert_cell_task.apply_async(
args=[values],
kwargs={'userid': userid},
expires=21600,
countdown=countdown)
countdown += 1
if wifi_observations:
# group by WiFi key
self.stats_client.incr('items.uploaded.wifi_observations',
len(wifi_observations))
if self.api_key_log:
self.stats_client.incr(
'items.api_log.%s.uploaded.'
'wifi_observations' % self.api_key_name,
len(wifi_observations))
wifis = defaultdict(list)
for obs in wifi_observations:
wifis[WifiObservation.to_hashkey(obs)].append(obs)
# Create a task per group of 20 WiFi keys at a time.
# We tend to get a huge number of unique WiFi networks per
# batch upload, with one to very few observations per WiFi.
# Grouping them helps in avoiding per-task overhead.
wifis = list(wifis.values())
batch_size = 20
countdown = 0
for i in range(0, len(wifis), batch_size):
values = []
for observations in wifis[i:i + batch_size]:
values.extend(observations)
# insert observations, expire the task if it wasn't processed
# after six hours to avoid queue overload, also delay
# each task by one second more, to get a more even workload
# and avoid parallel updates of the same underlying stations
self.insert_wifi_task.apply_async(
args=[values],
kwargs={'userid': userid},
expires=21600,
countdown=countdown)
countdown += 1
if userid is not None:
scorekey = Score.to_hashkey(
userid=userid,
key=ScoreKey.location,
time=util.utcnow().date())
Score.incr(self.session, scorekey, len(positions))
if positions:
self.process_mapstat(positions)
def process_reports(self, reports, userid=None):
positions = []
cell_observations = []
wifi_observations = []
for i, report in enumerate(reports):
report['report_id'] = uuid.uuid1()
cell, wifi = self.process_report(report)
cell_observations.extend(cell)
wifi_observations.extend(wifi)
if cell or wifi:
positions.append({
'lat': report['lat'],
'lon': report['lon'],
})
if cell_observations:
# group by and create task per cell key
self.stats_client.incr('items.uploaded.cell_observations',
len(cell_observations))
if self.api_key_log:
self.stats_client.incr(
'items.api_log.%s.uploaded.'
'cell_observations' % self.api_key_name,
len(cell_observations))
cells = defaultdict(list)
for obs in cell_observations:
cells[CellObservation.to_hashkey(obs)].append(obs)
# Create a task per group of 5 cell keys at a time.
# Grouping them helps in avoiding per-task overhead.
cells = list(cells.values())
batch_size = 5
countdown = 0
for i in range(0, len(cells), batch_size):
values = []
for observations in cells[i:i + batch_size]:
values.extend([encode_radio_dict(o) for o in observations])
# insert observations, expire the task if it wasn't processed
# after six hours to avoid queue overload, also delay
# each task by one second more, to get a more even workload
# and avoid parallel updates of the same underlying stations
self.insert_cell_task.apply_async(
args=[values],
kwargs={'userid': userid},
expires=21600,
countdown=countdown)
countdown += 1
if wifi_observations:
# group by WiFi key
self.stats_client.incr('items.uploaded.wifi_observations',
len(wifi_observations))
if self.api_key_log:
self.stats_client.incr(
'items.api_log.%s.uploaded.'
'wifi_observations' % self.api_key_name,
len(wifi_observations))
wifis = defaultdict(list)
for obs in wifi_observations:
wifis[WifiObservation.to_hashkey(obs)].append(obs)
# Create a task per group of 20 WiFi keys at a time.
# We tend to get a huge number of unique WiFi networks per
# batch upload, with one to very few observations per WiFi.
# Grouping them helps in avoiding per-task overhead.
wifis = list(wifis.values())
batch_size = 20
countdown = 0
for i in range(0, len(wifis), batch_size):
values = []
for observations in wifis[i:i + batch_size]:
values.extend(observations)
# insert observations, expire the task if it wasn't processed
# after six hours to avoid queue overload, also delay
# each task by one second more, to get a more even workload
# and avoid parallel updates of the same underlying stations
self.insert_wifi_task.apply_async(
args=[values],
kwargs={'userid': userid},
expires=21600,
countdown=countdown)
countdown += 1
if userid is not None:
scorekey = Score.to_hashkey(
userid=userid,
key=ScoreKey.location,
time=util.utcnow().date())
Score.incr(self.session, scorekey, len(positions))
if positions:
self.process_mapstat(positions)
