def test_skip_delete_new_blocks(self):
now = util.utcnow()
today_0000 = now.replace(hour=0, minute=0, second=0, tzinfo=pytz.UTC)
yesterday_0000 = today_0000 - timedelta(days=1)
yesterday_2359 = today_0000 - timedelta(seconds=1)
old = now - timedelta(days=5)
session = self.db_master_session
for i in range(100, 150, 10):
block = ObservationBlock()
block.measure_type = ObservationType.cell
block.start_id = i
block.end_id = i + 10
block.s3_key = 'fake_key'
block.archive_sha = 'fake_sha'
block.archive_date = None
session.add(block)
observations = []
for i in range(100, 110):
observations.append(CellObservation(id=i, created=old))
for i in range(110, 120):
observations.append(CellObservation(id=i, created=yesterday_0000))
for i in range(120, 130):
observations.append(CellObservation(id=i, created=yesterday_2359))
for i in range(130, 140):
observations.append(CellObservation(id=i, created=today_0000))
for i in range(140, 150):
observations.append(CellObservation(id=i, created=now))
session.add_all(observations)
session.commit()
def _archived_blocks():
blocks = session.query(ObservationBlock).all()
return len([b for b in blocks if b.archive_date is not None])
def _delete(days=7):
with patch.object(S3Backend,
'check_archive',
lambda x, y, z: True):
delete_cellmeasure_records.delay(days_old=days).get()
session.commit()
_delete(days=7)
self.assertEquals(session.query(CellObservation).count(), 50)
self.assertEqual(_archived_blocks(), 0)
_delete(days=2)
self.assertEquals(session.query(CellObservation).count(), 40)
self.assertEqual(_archived_blocks(), 1)
_delete(days=1)
self.assertEquals(session.query(CellObservation).count(), 20)
self.assertEqual(_archived_blocks(), 3)
_delete(days=0)
self.assertEquals(session.query(CellObservation).count(), 0)
self.assertEqual(_archived_blocks(), 5)
def test_mcc_latlon(self):
sample = dict(radio=Radio.gsm,
mnc=6,
lac=1,
cid=2,
lat=GB_LAT,
lon=GB_LON)
assert CellObservation.create(mcc=GB_MCC, **sample) is not None
assert CellObservation.create(mcc=262, **sample) is None
def add_line_of_cells_and_scan_lac(self):
session = self.session
big = 1.0
small = big / 10
keys = dict(radio=Radio.cdma, mcc=1, mnc=1, lac=1)
observations = [
CellObservation(lat=ctr + xd, lon=ctr + yd, cid=cell, **keys)
for cell in range(10)
for ctr in [cell * big]
for (xd, yd) in [(small, small),
(small, -small),
(-small, small),
(-small, -small)]
]
session.add_all(observations)
cells = [
Cell(lat=ctr, lon=ctr, cid=cell,
new_measures=4, total_measures=1, **keys)
for cell in range(10)
for ctr in [cell * big]
]
session.add_all(cells)
session.commit()
result = location_update_cell.delay(min_new=0,
max_new=9999,
batch=len(observations))
self.assertEqual(result.get(), (len(cells), 0))
scan_areas.delay()
def test_unthrottle_cell_observations(self):
session = self.db_master_session
block = ObservationBlock()
block.measure_type = ObservationType.cell
block.start_id = 120
block.end_id = 140
block.s3_key = 'fake_key'
block.archive_sha = 'fake_sha'
block.archive_date = None
session.add(block)
gsm = RADIO_TYPE['gsm']
k = dict(radio=gsm, mcc=1, mnc=2, lac=4, lat=1.0, lon=1.0)
for i in range(100, 150):
session.add(
CellObservation(id=i, cid=i, created=self.really_old, **k))
session.add(Cell(total_measures=11000, cid=i, **k))
session.commit()
with patch.object(S3Backend, 'check_archive', lambda x, y, z: True):
delete_cellmeasure_records.delay(batch=3).get()
cell_unthrottle_measures.delay(10000, 1000).get()
cells = session.query(Cell).all()
self.assertEquals(len(cells), 50)
for cell in cells:
if 120 <= cell.cid and cell.cid < 140:
self.assertEquals(cell.total_measures, 0)
else:
self.assertEquals(cell.total_measures, 1)
self.check_stats(counter=['items.cell_unthrottled'])
def test_customjson(self):
now = util.utcnow()
report_id = uuid.uuid1()
obs = CellObservation.create(
radio=Radio.gsm, mcc=GB_MCC, mnc=5, lac=12345, cid=23456,
report_id=report_id, lat=GB_LAT, lon=GB_LON, created=now)
json_data = kombu_dumps(obs)
self.assertTrue('accuracy' not in json_data)
result = kombu_loads(json_data)
self.assertTrue(type(result), CellObservation)
self.assertTrue(result.accuracy is None)
self.assertEqual(type(result.report_id), uuid.UUID)
self.assertEqual(result.report_id, report_id)
self.assertEqual(type(result.radio), Radio)
self.assertEqual(result.radio, Radio.gsm)
self.assertEqual(result.mcc, GB_MCC)
self.assertEqual(result.mnc, 5)
self.assertEqual(result.lac, 12345)
self.assertEqual(result.cid, 23456)
self.assertEqual(result.lat, GB_LAT)
self.assertEqual(result.lon, GB_LON)
self.assertEqual(type(result.created), datetime.datetime)
self.assertEqual(result.created, now)
def test_schedule_cell_observations(self):
session = self.db_master_session
blocks = schedule_cellmeasure_archival.delay(batch=1).get()
self.assertEquals(len(blocks), 0)
observations = []
for i in range(20):
observations.append(CellObservation(created=self.really_old))
session.add_all(observations)
session.flush()
start_id = observations[0].id
blocks = schedule_cellmeasure_archival.delay(batch=15).get()
self.assertEquals(len(blocks), 1)
block = blocks[0]
self.assertEquals(block, (start_id, start_id + 15))
blocks = schedule_cellmeasure_archival.delay(batch=6).get()
self.assertEquals(len(blocks), 0)
blocks = schedule_cellmeasure_archival.delay(batch=5).get()
self.assertEquals(len(blocks), 1)
block = blocks[0]
self.assertEquals(block, (start_id + 15, start_id + 20))
blocks = schedule_cellmeasure_archival.delay(batch=1).get()
self.assertEquals(len(blocks), 0)
def test_fields(self):
obs = CellObservation.create(radio=Radio.gsm,
mcc=GB_MCC,
mnc=5,
lac=12345,
cid=23456,
lat=GB_LAT,
lon=GB_LON,
pressure=1010.2,
source='gnss',
timestamp=1405602028568,
asu=26,
signal=-61,
ta=10)
assert obs.lat == GB_LAT
assert obs.lon == GB_LON
assert obs.pressure == 1010.2
assert obs.source == ReportSource.gnss
assert obs.timestamp == 1405602028568
assert obs.radio == Radio.gsm
assert obs.mcc == GB_MCC
assert obs.mnc == 5
assert obs.lac == 12345
assert obs.cid == 23456
assert obs.asu == 26
assert obs.signal == -61
assert obs.ta == 10
assert obs.shard_id == 'gsm'
def test_location_update_cell(self):
now = util.utcnow()
before = now - timedelta(days=1)
schema = ValidCellKeySchema()
session = self.db_master_session
k1 = dict(radio=1, mcc=1, mnc=2, lac=3, cid=4)
k2 = dict(radio=1, mcc=1, mnc=2, lac=6, cid=8)
k3 = dict(radio=1,
mcc=1,
mnc=2,
lac=schema.fields['lac'].missing,
cid=schema.fields['cid'].missing)
data = [
Cell(new_measures=3, total_measures=5, **k1),
CellObservation(lat=1.0, lon=1.0, created=now, **k1),
CellObservation(lat=1.002, lon=1.003, created=now, **k1),
CellObservation(lat=1.004, lon=1.006, created=now, **k1),
# The lac, cid are invalid and should be skipped
CellObservation(lat=1.5, lon=1.5, created=now, **k3),
CellObservation(lat=1.502, lon=1.503, created=now, **k3),
Cell(lat=2.0, lon=2.0, new_measures=2, total_measures=4, **k2),
# the lat/lon is bogus and mismatches the line above on purpose
# to make sure old observations are skipped
CellObservation(lat=-1.0, lon=-1.0, created=before, **k2),
CellObservation(lat=-1.0, lon=-1.0, created=before, **k2),
CellObservation(lat=2.002, lon=2.004, created=now, **k2),
CellObservation(lat=2.002, lon=2.004, created=now, **k2),
]
session.add_all(data)
session.commit()
result = location_update_cell.delay(min_new=1)
self.assertEqual(result.get(), (2, 0))
self.check_stats(
total=2,
timer=['task.data.location_update_cell'],
gauge=['task.data.location_update_cell.new_measures_1_100'],
)
cells = session.query(Cell).all()
self.assertEqual(len(cells), 2)
self.assertEqual([c.new_measures for c in cells], [0, 0])
for cell in cells:
if cell.cid == 4:
self.assertEqual(cell.lat, 1.002)
self.assertEqual(cell.lon, 1.003)
elif cell.cid == 8:
self.assertEqual(cell.lat, 2.001)
self.assertEqual(cell.lon, 2.002)
def process_report(self, data):
def add_missing_dict_entries(dst, src):
# x.update(y) overwrites entries in x with those in y;
# We want to only add those not already present.
# We also only want to copy the top-level base report data
# and not any nested values like cell or wifi.
for (key, value) in src.items():
if key != 'radio' and key not in dst \
and not isinstance(value, (tuple, list, dict)):
dst[key] = value
report_data = Report.validate(data)
if report_data is None:
return ([], [])
cell_observations = {}
wifi_observations = {}
if data.get('cell'):
# flatten report / cell data into a single dict
for cell in data['cell']:
# only validate the additional fields
cell = CellReport.validate(cell)
if cell is None:
continue
add_missing_dict_entries(cell, report_data)
cell_key = CellObservation.to_hashkey(cell)
if cell_key in cell_observations:
existing = cell_observations[cell_key]
if existing['ta'] > cell['ta'] or \
(existing['signal'] != 0 and
existing['signal'] < cell['signal']) or \
existing['asu'] < cell['asu']:
cell_observations[cell_key] = cell
else:
cell_observations[cell_key] = cell
cell_observations = cell_observations.values()
# flatten report / wifi data into a single dict
if data.get('wifi'):
for wifi in data['wifi']:
# only validate the additional fields
wifi = WifiReport.validate(wifi)
if wifi is None:
continue
add_missing_dict_entries(wifi, report_data)
wifi_key = WifiObservation.to_hashkey(wifi)
if wifi_key in wifi_observations:
existing = wifi_observations[wifi_key]
if existing['signal'] != 0 and \
existing['signal'] < wifi['signal']:
wifi_observations[wifi_key] = wifi
else:
wifi_observations[wifi_key] = wifi
wifi_observations = wifi_observations.values()
return (cell_observations, wifi_observations)
def process_report(self, data):
def add_missing_dict_entries(dst, src):
# x.update(y) overwrites entries in x with those in y;
# We want to only add those not already present.
# We also only want to copy the top-level base report data
# and not any nested values like cell or wifi.
for (key, value) in src.items():
if key != 'radio' and key not in dst \
and not isinstance(value, (tuple, list, dict)):
dst[key] = value
report_data = Report.validate(data)
if report_data is None:
return ([], [])
cell_observations = {}
wifi_observations = {}
if data.get('cell'):
# flatten report / cell data into a single dict
for cell in data['cell']:
# only validate the additional fields
cell = CellReport.validate(cell)
if cell is None:
continue
add_missing_dict_entries(cell, report_data)
cell_key = CellObservation.to_hashkey(cell)
if cell_key in cell_observations:
existing = cell_observations[cell_key]
if existing['ta'] > cell['ta'] or \
(existing['signal'] != 0 and
existing['signal'] < cell['signal']) or \
existing['asu'] < cell['asu']:
cell_observations[cell_key] = cell
else:
cell_observations[cell_key] = cell
cell_observations = cell_observations.values()
# flatten report / wifi data into a single dict
if data.get('wifi'):
for wifi in data['wifi']:
# only validate the additional fields
wifi = WifiReport.validate(wifi)
if wifi is None:
continue
add_missing_dict_entries(wifi, report_data)
wifi_key = WifiObservation.to_hashkey(wifi)
if wifi_key in wifi_observations:
existing = wifi_observations[wifi_key]
if existing['signal'] != 0 and \
existing['signal'] < wifi['signal']:
wifi_observations[wifi_key] = wifi
else:
wifi_observations[wifi_key] = wifi
wifi_observations = wifi_observations.values()
return (cell_observations, wifi_observations)
def test_max_min_range_update(self):
session = self.session
k1 = dict(radio=Radio.cdma, mcc=1, mnc=2, lac=3, cid=4)
data = [
Cell(lat=1.001,
lon=-1.001,
max_lat=1.002,
min_lat=1.0,
max_lon=-1.0,
min_lon=-1.002,
new_measures=2,
total_measures=4,
**k1),
CellObservation(lat=1.001, lon=-1.003, **k1),
CellObservation(lat=1.005, lon=-1.007, **k1),
]
session.add_all(data)
session.commit()
result = location_update_cell.delay(min_new=1)
self.assertEqual(result.get(), (1, 0))
cells = session.query(Cell).all()
self.assertEqual(len(cells), 1)
cell = cells[0]
self.assertEqual(cell.lat, 1.002)
self.assertEqual(cell.max_lat, 1.005)
self.assertEqual(cell.min_lat, 1.0)
self.assertEqual(cell.lon, -1.003)
self.assertEqual(cell.max_lon, -1.0)
self.assertEqual(cell.min_lon, -1.007)
# independent calculation: the cell bounding box is
# (1.000, -1.007) to (1.005, -1.000), with centroid
# at (1.002, -1.003); worst distance from centroid
# to any corner is 556m
self.assertEqual(cell.range, 556)
def test_json(self):
obs = CellObservationFactory.build(accuracy=None)
result = CellObservation.from_json(simplejson.loads(
simplejson.dumps(obs.to_json())))
self.assertTrue(type(result), CellObservation)
self.assertTrue(result.accuracy is None)
self.assertEqual(type(result.radio), Radio)
self.assertEqual(result.radio, obs.radio)
self.assertEqual(result.mcc, obs.mcc)
self.assertEqual(result.mnc, obs.mnc)
self.assertEqual(result.lac, obs.lac)
self.assertEqual(result.cid, obs.cid)
self.assertEqual(result.lat, obs.lat)
self.assertEqual(result.lon, obs.lon)
def test_json(self):
obs = CellObservationFactory.build(accuracy=None, source="fixed")
result = CellObservation.from_json(json.loads(json.dumps(obs.to_json())))
assert type(result) is CellObservation
assert result.accuracy is None
assert type(result.radio), Radio
assert result.radio == obs.radio
assert result.mcc == obs.mcc
assert result.mnc == obs.mnc
assert result.lac == obs.lac
assert result.cid == obs.cid
assert result.lat == obs.lat
assert result.lon == obs.lon
assert result.source is ReportSource.fixed
assert type(result.source) is ReportSource
def test_fields(self):
obs = CellObservation.create(
radio=Radio.gsm, mcc=GB_MCC, mnc=5, lac=12345, cid=23456,
lat=GB_LAT, lon=GB_LON,
asu=26, signal=-61, ta=10)
self.assertEqual(obs.lat, GB_LAT)
self.assertEqual(obs.lon, GB_LON)
self.assertEqual(obs.radio, Radio.gsm)
self.assertEqual(obs.mcc, GB_MCC)
self.assertEqual(obs.mnc, 5)
self.assertEqual(obs.lac, 12345)
self.assertEqual(obs.cid, 23456)
self.assertEqual(obs.asu, 26)
self.assertEqual(obs.signal, -61)
self.assertEqual(obs.ta, 10)
def test_internaljson(self):
obs = CellObservation.create(
radio=Radio.gsm, mcc=GB_MCC, mnc=5, lac=12345, cid=23456,
lat=GB_LAT, lon=GB_LON)
result = internal_loads(internal_dumps(obs))
self.assertTrue(type(result), CellObservation)
self.assertTrue(result.accuracy is None)
self.assertEqual(type(result.radio), Radio)
self.assertEqual(result.radio, Radio.gsm)
self.assertEqual(result.mcc, GB_MCC)
self.assertEqual(result.mnc, 5)
self.assertEqual(result.lac, 12345)
self.assertEqual(result.cid, 23456)
self.assertEqual(result.lat, GB_LAT)
self.assertEqual(result.lon, GB_LON)
def test_backup_cell_to_s3(self):
session = self.db_master_session
batch_size = 10
observations = []
for i in range(batch_size):
observations.append(CellObservation(created=self.really_old))
session.add_all(observations)
session.flush()
start_id = observations[0].id
blocks = schedule_cellmeasure_archival.delay(batch=batch_size).get()
self.assertEquals(len(blocks), 1)
block = blocks[0]
self.assertEquals(block, (start_id, start_id + batch_size))
with mock_s3():
with patch.object(S3Backend,
'backup_archive', lambda x, y, z: True):
write_cellmeasure_s3_backups.delay(cleanup_zip=False).get()
msgs = self.heka_client.stream.msgs
info_msgs = [m for m in msgs if m.type == 'oldstyle']
self.assertEquals(1, len(info_msgs))
info = info_msgs[0]
fname = info.payload.split(":")[-1]
myzip = ZipFile(fname)
try:
contents = set(myzip.namelist())
expected_contents = set(['alembic_revision.txt',
'cell_measure.csv'])
self.assertEquals(expected_contents, contents)
finally:
myzip.close()
blocks = session.query(ObservationBlock).all()
self.assertEquals(len(blocks), 1)
block = blocks[0]
actual_sha = hashlib.sha1()
actual_sha.update(open(fname, 'rb').read())
self.assertEquals(block.archive_sha, actual_sha.digest())
self.assertTrue(block.s3_key is not None)
self.assertTrue('/cell_' in block.s3_key)
self.assertTrue(block.archive_date is None)
def test_json(self):
obs = CellObservationFactory.build(
accuracy=None, source='fixed')
result = CellObservation.from_json(simplejson.loads(
simplejson.dumps(obs.to_json())))
assert type(result) is CellObservation
assert result.accuracy is None
assert type(result.radio), Radio
assert result.radio == obs.radio
assert result.mcc == obs.mcc
assert result.mnc == obs.mnc
assert result.lac == obs.lac
assert result.cid == obs.cid
assert result.lat == obs.lat
assert result.lon == obs.lon
assert result.source is ReportSource.fixed
assert type(result.source) is ReportSource
def test_monitor_measures(self):
session = self.db_master_session
result = monitor_measures.delay().get()
self.check_stats(
gauge=[('table.cell_measure', 1), ('table.wifi_measure', 1)],
)
self.assertEqual(result, {'cell_measure': -1, 'wifi_measure': -1})
# add some observations
session.add_all([CellObservation() for i in range(3)])
session.add_all([WifiObservation() for i in range(5)])
session.flush()
result = monitor_measures.delay().get()
self.check_stats(
gauge=[('table.cell_measure', 2), ('table.wifi_measure', 2)],
)
self.assertEqual(result, {'cell_measure': 3, 'wifi_measure': 5})
def test_scan_lacs_asymmetric(self):
session = self.db_master_session
big = 0.1
small = big / 10
keys = dict(radio=1, mcc=1, mnc=1, lac=1)
observations = [
CellObservation(lat=ctr + xd, lon=ctr + yd, cid=cell, **keys)
for cell in range(6) for ctr in [(2**cell) * big]
for (xd, yd) in [(small, small), (small,
-small), (-small,
small), (-small,
-small)]
]
session.add_all(observations)
cells = [
Cell(lat=ctr,
lon=ctr,
cid=cell,
new_measures=4,
total_measures=1,
**keys) for cell in range(6) for ctr in [(2**cell) * big]
]
session.add_all(cells)
session.commit()
result = location_update_cell.delay(min_new=0,
max_new=9999,
batch=len(observations))
self.assertEqual(result.get(), (len(cells), 0))
scan_lacs.delay()
lac = session.query(CellArea).filter(CellArea.lac == 1).first()
# We produced a sequence of 0.02-degree-on-a-side
# cell bounding boxes centered at
# [0, 0.2, 0.4, 0.8, 1.6, 3.2] degrees.
# So the lower-left corner is at (-0.01, -0.01)
# and the upper-right corner is at (3.21, 3.21)
# we should therefore see a LAC centroid at (1.05, 1.05)
# with a range of 339.540m
self.assertEqual(lac.lat, 1.05)
self.assertEqual(lac.lon, 1.05)
self.assertEqual(lac.range, 339540)
def test_scan_areas_race_with_location_update(self):
session = self.session
# First batch of cell observations for CID 1
keys = dict(radio=Radio.cdma, mcc=1, mnc=1, lac=1, cid=1)
cell = Cell(new_measures=4, total_measures=1, **keys)
observations = [
CellObservation(lat=1.0, lon=1.0, **keys),
CellObservation(lat=1.0, lon=1.0, **keys),
CellObservation(lat=1.0, lon=1.0, **keys),
CellObservation(lat=1.0, lon=1.0, **keys),
]
session.add(cell)
session.add_all(observations)
session.commit()
# Periodic location_update_cell runs and updates CID 1
# to have a location, inserts LAC 1 with new_measures=1
# which will be picked up by the next scan_lac.
result = location_update_cell.delay(min_new=1)
self.assertEqual(result.get(), (1, 0))
# Second batch of cell observations for CID 2
keys['cid'] = 2
cell = Cell(new_measures=4, total_measures=1, **keys)
observations = [
CellObservation(lat=1.0, lon=1.0, **keys),
CellObservation(lat=1.0, lon=1.0, **keys),
CellObservation(lat=1.0, lon=1.0, **keys),
CellObservation(lat=1.0, lon=1.0, **keys),
]
session.add(cell)
session.add_all(observations)
session.commit()
# Periodic LAC scan runs, picking up LAC 1; this could
# accidentally pick up CID 2, but it should not since it
# has not had its location updated yet. If there's no
# exception here, CID 2 is properly ignored.
scan_areas.delay()
def location_update_cell(self, min_new=10, max_new=100, batch=10):
cells = []
redis_client = self.app.redis_client
with self.db_session() as session:
emit_new_observation_metric(self.stats_client, session, self.shortname,
Cell, min_new, max_new)
query = (session.query(Cell).filter(
Cell.new_measures >= min_new).filter(
Cell.new_measures < max_new).limit(batch))
cells = query.all()
if not cells:
return 0
moving_cells = set()
updated_lacs = set()
for cell in cells:
query = session.query(
CellObservation.lat, CellObservation.lon,
CellObservation.id).filter(*CellObservation.joinkey(cell))
# only take the last X new_measures
query = query.order_by(CellObservation.created.desc()).limit(
cell.new_measures)
observations = query.all()
if observations:
moving = calculate_new_position(cell, observations,
CELL_MAX_DIST_KM)
if moving:
moving_cells.add(cell)
updated_lacs.add(CellArea.to_hashkey(cell))
if updated_lacs:
session.on_post_commit(enqueue_lacs, redis_client, updated_lacs,
UPDATE_KEY['cell_lac'])
if moving_cells:
# some cells found to be moving too much
blacklist_and_remove_moving_cells(session, moving_cells)
session.commit()
return (len(cells), len(moving_cells))
def test_fields(self):
report_id = uuid.uuid1()
self.session.add(CellObservation.create(
radio=Radio.gsm, mcc=GB_MCC, mnc=5, lac=12345, cid=23456,
report_id=report_id, lat=GB_LAT, lon=GB_LON,
asu=26, signal=-61, ta=10))
self.session.flush()
result = self.session.query(CellObservation).first()
self.assertEqual(result.report_id, report_id)
self.assertEqual(result.lat, GB_LAT)
self.assertEqual(result.lon, GB_LON)
self.assertEqual(result.radio, Radio.gsm)
self.assertEqual(result.mcc, GB_MCC)
self.assertEqual(result.mnc, 5)
self.assertEqual(result.lac, 12345)
self.assertEqual(result.cid, 23456)
self.assertEqual(result.asu, 26)
self.assertEqual(result.signal, -61)
self.assertEqual(result.ta, 10)
def test_cell_histogram(self):
session = self.db_master_session
today = util.utcnow().date()
yesterday = (today - timedelta(1))
two_days = (today - timedelta(2))
long_ago = (today - timedelta(3))
observations = [
CellObservation(lat=1.0, lon=2.0, created=today),
CellObservation(lat=1.0, lon=2.0, created=today),
CellObservation(lat=1.0, lon=2.0, created=yesterday),
CellObservation(lat=1.0, lon=2.0, created=two_days),
CellObservation(lat=1.0, lon=2.0, created=two_days),
CellObservation(lat=1.0, lon=2.0, created=two_days),
CellObservation(lat=1.0, lon=2.0, created=long_ago),
]
session.add_all(observations)
session.commit()
cell_histogram.delay(ago=3).get()
stats = session.query(Stat).order_by(Stat.time).all()
self.assertEqual(len(stats), 1)
self.assertEqual(stats[0].key, StatKey.cell)
self.assertEqual(stats[0].time, long_ago)
self.assertEqual(stats[0].value, 1)
# fill up newer dates
cell_histogram.delay(ago=2).get()
cell_histogram.delay(ago=1).get()
cell_histogram.delay(ago=0).get()
# test duplicate execution
cell_histogram.delay(ago=1).get()
stats = session.query(Stat).order_by(Stat.time).all()
self.assertEqual(len(stats), 4)
self.assertEqual(stats[0].time, long_ago)
self.assertEqual(stats[0].value, 1)
self.assertEqual(stats[1].time, two_days)
self.assertEqual(stats[1].value, 4)
self.assertEqual(stats[2].time, yesterday)
self.assertEqual(stats[2].value, 5)
self.assertEqual(stats[3].time, today)
self.assertEqual(stats[3].value, 7)
def test_internaljson(self):
obs = CellObservation.create(radio=Radio.gsm,
mcc=GB_MCC,
mnc=5,
lac=12345,
cid=23456,
lat=GB_LAT,
lon=GB_LON)
result = internal_loads(internal_dumps(obs))
self.assertTrue(type(result), CellObservation)
self.assertTrue(result.accuracy is None)
self.assertEqual(type(result.radio), Radio)
self.assertEqual(result.radio, Radio.gsm)
self.assertEqual(result.mcc, GB_MCC)
self.assertEqual(result.mnc, 5)
self.assertEqual(result.lac, 12345)
self.assertEqual(result.cid, 23456)
self.assertEqual(result.lat, GB_LAT)
self.assertEqual(result.lon, GB_LON)
def test_delete_cell_observations(self):
session = self.db_master_session
block = ObservationBlock()
block.measure_type = ObservationType.cell
block.start_id = 120
block.end_id = 140
block.s3_key = 'fake_key'
block.archive_sha = 'fake_sha'
block.archive_date = None
session.add(block)
for i in range(100, 150):
session.add(CellObservation(id=i, created=self.really_old))
session.commit()
with patch.object(S3Backend, 'check_archive', lambda x, y, z: True):
delete_cellmeasure_records.delay(batch=3).get()
self.assertEquals(session.query(CellObservation).count(), 30)
self.assertTrue(block.archive_date is not None)
def test_fields(self):
report_id = uuid.uuid1()
session = self.session
session.add(CellObservation.create(
radio=Radio.gsm, mcc=GB_MCC, mnc=5, lac=12345, cid=23456,
report_id=report_id, lat=GB_LAT, lon=GB_LON,
asu=26, signal=-61, ta=10))
session.flush()
result = session.query(CellObservation).first()
self.assertEqual(result.report_id, report_id)
self.assertEqual(result.lat, GB_LAT)
self.assertEqual(result.lon, GB_LON)
self.assertEqual(result.radio, Radio.gsm)
self.assertEqual(result.mcc, GB_MCC)
self.assertEqual(result.mnc, 5)
self.assertEqual(result.lac, 12345)
self.assertEqual(result.cid, 23456)
self.assertEqual(result.asu, 26)
self.assertEqual(result.signal, -61)
self.assertEqual(result.ta, 10)
def test_fields(self):
obs = CellObservation.create(
radio=Radio.gsm, mcc=GB_MCC, mnc=5, lac=12345, cid=23456,
lat=GB_LAT, lon=GB_LON,
pressure=1010.2, source='gnss', timestamp=1405602028568,
asu=26, signal=-61, ta=10)
assert obs.lat == GB_LAT
assert obs.lon == GB_LON
assert obs.pressure == 1010.2
assert obs.source == ReportSource.gnss
assert obs.timestamp == 1405602028568
assert obs.radio == Radio.gsm
assert obs.mcc == GB_MCC
assert obs.mnc == 5
assert obs.lac == 12345
assert obs.cid == 23456
assert obs.asu == 26
assert obs.signal == -61
assert obs.ta == 10
assert obs.shard_id == 'gsm'
def test_fields(self):
obs = CellObservation.create(radio=Radio.gsm,
mcc=GB_MCC,
mnc=5,
lac=12345,
cid=23456,
lat=GB_LAT,
lon=GB_LON,
asu=26,
signal=-61,
ta=10)
self.assertEqual(obs.lat, GB_LAT)
self.assertEqual(obs.lon, GB_LON)
self.assertEqual(obs.radio, Radio.gsm)
self.assertEqual(obs.mcc, GB_MCC)
self.assertEqual(obs.mnc, 5)
self.assertEqual(obs.lac, 12345)
self.assertEqual(obs.cid, 23456)
self.assertEqual(obs.asu, 26)
self.assertEqual(obs.signal, -61)
self.assertEqual(obs.ta, 10)
self.assertEqual(obs.shard_id, 'gsm')
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 test_blacklist_moving_cells(self):
now = util.utcnow()
long_ago = now - timedelta(days=40)
session = self.session
k1 = dict(radio=Radio.cdma, mcc=1, mnc=2, lac=3, cid=4)
k2 = dict(radio=Radio.cdma, mcc=1, mnc=2, lac=6, cid=8)
k3 = dict(radio=Radio.cdma, mcc=1, mnc=2, lac=9, cid=12)
k4 = dict(radio=Radio.cdma, mcc=1, mnc=2, lac=12, cid=16)
k5 = dict(radio=Radio.cdma, mcc=1, mnc=2, lac=15, cid=20)
k6 = dict(radio=Radio.cdma, mcc=1, mnc=2, lac=18, cid=24)
# keys k2, k3 and k4 are expected to be detected as moving
data = [
# a cell with an entry but no prior position
Cell(new_measures=3, total_measures=0, **k1),
CellObservation(lat=1.001, lon=1.001, **k1),
CellObservation(lat=1.002, lon=1.005, **k1),
CellObservation(lat=1.003, lon=1.009, **k1),
# a cell with a prior known position
Cell(lat=2.0, lon=2.0,
new_measures=2, total_measures=1, **k2),
CellObservation(lat=2.0, lon=2.0, **k2),
CellObservation(lat=4.0, lon=2.0, **k2),
# a cell with a very different prior position
Cell(lat=1.0, lon=1.0,
new_measures=2, total_measures=1, **k3),
CellObservation(lat=3.0, lon=3.0, **k3),
CellObservation(lat=-3.0, lon=3.0, **k3),
# another cell with a prior known position (and negative lat)
Cell(lat=-4.0, lon=4.0,
new_measures=2, total_measures=1, **k4),
CellObservation(lat=-4.0, lon=4.0, **k4),
CellObservation(lat=-6.0, lon=4.0, **k4),
# an already blacklisted cell
CellBlacklist(time=now, count=1, **k5),
CellObservation(lat=5.0, lon=5.0, **k5),
CellObservation(lat=8.0, lon=5.0, **k5),
# a cell with an old different record we ignore, position
# estimate has been updated since
Cell(lat=6.0, lon=6.0,
new_measures=2, total_measures=1, **k6),
CellObservation(lat=6.9, lon=6.9, time=long_ago, **k6),
CellObservation(lat=6.0, lon=6.0, **k6),
CellObservation(lat=6.001, lon=6, **k6),
]
session.add_all(data)
session.commit()
result = location_update_cell.delay(min_new=1)
self.assertEqual(result.get(), (5, 3))
moving = [k2, k3, k4, k5]
black = session.query(CellBlacklist).all()
self.assertEqual(set([b.hashkey() for b in black]),
set([CellBlacklist.to_hashkey(k) for k in moving]))
keys = [k1, k2, k3, k4, k5, k6]
observations = session.query(CellObservation).all()
self.assertEqual(len(observations), 14)
self.assertEqual(set([obs.hashkey() for obs in observations]),
set([CellObservation.to_hashkey(k) for k in keys]))
# test duplicate call
result = location_update_cell.delay(min_new=1)
self.assertEqual(result.get(), (0, 0))
self.check_stats(
total=6,
timer=[
# We made duplicate calls
('task.data.location_update_cell', 2),
# One of those would've scheduled a remove_cell task
('task.data.remove_cell', 1)
],
gauge=[
('task.data.location_update_cell.new_measures_1_100', 2),
])
def test_blacklist_moving_cells(self):
now = util.utcnow()
long_ago = now - timedelta(days=40)
session = self.session
k1 = dict(radio=Radio.cdma, mcc=1, mnc=2, lac=3, cid=4)
k2 = dict(radio=Radio.cdma, mcc=1, mnc=2, lac=6, cid=8)
k3 = dict(radio=Radio.cdma, mcc=1, mnc=2, lac=9, cid=12)
k4 = dict(radio=Radio.cdma, mcc=1, mnc=2, lac=12, cid=16)
k5 = dict(radio=Radio.cdma, mcc=1, mnc=2, lac=15, cid=20)
k6 = dict(radio=Radio.cdma, mcc=1, mnc=2, lac=18, cid=24)
# keys k2, k3 and k4 are expected to be detected as moving
data = [
# a cell with an entry but no prior position
Cell(new_measures=3, total_measures=0, **k1),
CellObservation(lat=1.001, lon=1.001, **k1),
CellObservation(lat=1.002, lon=1.005, **k1),
CellObservation(lat=1.003, lon=1.009, **k1),
# a cell with a prior known position
Cell(lat=2.0, lon=2.0, new_measures=2, total_measures=1, **k2),
CellObservation(lat=2.0, lon=2.0, **k2),
CellObservation(lat=4.0, lon=2.0, **k2),
# a cell with a very different prior position
Cell(lat=1.0, lon=1.0, new_measures=2, total_measures=1, **k3),
CellObservation(lat=3.0, lon=3.0, **k3),
CellObservation(lat=-3.0, lon=3.0, **k3),
# another cell with a prior known position (and negative lat)
Cell(lat=-4.0, lon=4.0, new_measures=2, total_measures=1, **k4),
CellObservation(lat=-4.0, lon=4.0, **k4),
CellObservation(lat=-6.0, lon=4.0, **k4),
# an already blacklisted cell
CellBlacklist(time=now, count=1, **k5),
CellObservation(lat=5.0, lon=5.0, **k5),
CellObservation(lat=8.0, lon=5.0, **k5),
# a cell with an old different record we ignore, position
# estimate has been updated since
Cell(lat=6.0, lon=6.0, new_measures=2, total_measures=1, **k6),
CellObservation(lat=6.9, lon=6.9, time=long_ago, **k6),
CellObservation(lat=6.0, lon=6.0, **k6),
CellObservation(lat=6.001, lon=6, **k6),
]
session.add_all(data)
session.commit()
result = location_update_cell.delay(min_new=1)
self.assertEqual(result.get(), (5, 3))
moving = [k2, k3, k4, k5]
black = session.query(CellBlacklist).all()
self.assertEqual(set([b.hashkey() for b in black]),
set([CellBlacklist.to_hashkey(k) for k in moving]))
keys = [k1, k2, k3, k4, k5, k6]
observations = session.query(CellObservation).all()
self.assertEqual(len(observations), 14)
self.assertEqual(set([obs.hashkey() for obs in observations]),
set([CellObservation.to_hashkey(k) for k in keys]))
# test duplicate call
result = location_update_cell.delay(min_new=1)
self.assertEqual(result.get(), (0, 0))
self.check_stats(
total=6,
timer=[
# We made duplicate calls
('task.data.location_update_cell', 2),
# One of those would've scheduled a remove_cell task
('task.data.remove_cell', 1)
],
gauge=[
('task.data.location_update_cell.new_measures_1_100', 2),
])
def test_mcc_latlon(self):
sample = dict(radio=Radio.gsm, mnc=6, lac=1, cid=2,
lat=GB_LAT, lon=GB_LON)
assert CellObservation.create(mcc=GB_MCC, **sample) is not None
assert CellObservation.create(mcc=262, **sample) is None
def process_observations(observations,
session,
userid=None,
api_key_log=False,
api_key_name=None):
stats_client = get_stats_client()
positions = []
cell_observations = []
wifi_observations = []
for i, obs in enumerate(observations):
obs['report_id'] = uuid.uuid1()
cell, wifi = process_observation(obs, session)
cell_observations.extend(cell)
wifi_observations.extend(wifi)
if cell or wifi:
positions.append({
'lat': obs['lat'],
'lon': obs['lon'],
})
if cell_observations:
# group by and create task per cell key
stats_client.incr('items.uploaded.cell_observations',
len(cell_observations))
if api_key_log:
stats_client.incr(
'items.api_log.%s.uploaded.cell_observations' % 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(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
insert_measures_cell.apply_async(args=[values],
kwargs={'userid': userid},
expires=21600,
countdown=countdown)
countdown += 1
if wifi_observations:
# group by WiFi key
stats_client.incr('items.uploaded.wifi_observations',
len(wifi_observations))
if api_key_log:
stats_client.incr(
'items.api_log.%s.uploaded.wifi_observations' % 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
insert_measures_wifi.apply_async(args=[values],
kwargs={'userid': userid},
expires=21600,
countdown=countdown)
countdown += 1
if userid is not None:
process_score(userid, len(positions), session)
if positions:
process_mapstat(session, positions)
def process_report(self, data):
def add_missing_dict_entries(dst, src):
# x.update(y) overwrites entries in x with those in y;
# We want to only add those not already present.
# We also only want to copy the top-level base report data
# and not any nested values like cell or wifi.
for (key, value) in src.items():
if key != 'radio' and key not in dst \
and not isinstance(value, (tuple, list, dict)):
dst[key] = value
def better_cell_obs(new, old):
comparators = [
('ta', operator.lt),
('signal', operator.gt),
('asu', operator.gt),
]
for field, better in comparators:
if (None not in (old[field], new[field]) and
better(new[field], old[field])):
return True
return False
def better_wifi_obs(new, old):
if (None not in (old['signal'], new['signal']) and
new['signal'] > old['signal']):
return True
return False
report_data = Report.validate(data)
if report_data is None:
return ([], [])
cell_observations = {}
wifi_observations = {}
if data.get('cell'):
# flatten report / cell data into a single dict
for cell in data['cell']:
# only validate the additional fields
cell = CellReport.validate(cell)
if cell is None:
continue
add_missing_dict_entries(cell, report_data)
cell_key = CellObservation.to_hashkey(cell)
if cell_key in cell_observations:
existing = cell_observations[cell_key]
if better_cell_obs(cell, existing):
cell_observations[cell_key] = cell
else:
cell_observations[cell_key] = cell
cell_observations = cell_observations.values()
# flatten report / wifi data into a single dict
if data.get('wifi'):
for wifi in data['wifi']:
# only validate the additional fields
wifi = WifiReport.validate(wifi)
if wifi is None:
continue
add_missing_dict_entries(wifi, report_data)
wifi_key = WifiObservation.to_hashkey(wifi)
if wifi_key in wifi_observations:
existing = wifi_observations[wifi_key]
if better_wifi_obs(wifi, existing):
wifi_observations[wifi_key] = wifi
else:
wifi_observations[wifi_key] = wifi
wifi_observations = wifi_observations.values()
return (cell_observations, wifi_observations)
def process_reports(self, reports, userid=None):
positions = set()
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) and report.get('lat') and report.get('lon'):
positions.add((report['lat'], 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 and 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 100 cell keys at a time.
# Grouping them helps in avoiding per-task overhead.
cells = list(cells.values())
batch_size = 100
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 and 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 100 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 = 100
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
self.process_mapstat(positions)
self.process_score(userid, positions)
