def update_addon_average_daily_users(chunk_size=250):
"""Update add-ons ADU totals."""
if not waffle.switch_is_active('local-statistics-processing'):
return False
counts = dict(
# In order to reset the `average_daily_users` values of add-ons that
# don't exist in BigQuery, we prepare a set of `(guid, 0)` for most
# add-ons.
Addon.objects.filter(type__in=amo.ADDON_TYPES_WITH_STATS).exclude(
guid__isnull=True
).exclude(guid__exact=''
).exclude(average_daily_users=0
).annotate(count=Value(0, IntegerField())
).values_list('guid', 'count')
# Just to make order predictable in tests, we order by id. This
# matches the GROUP BY being generated so it should be safe.
.order_by('id'))
# Update the `counts` with values from BigQuery.
counts.update(get_addons_and_average_daily_users_from_bigquery())
counts = list(counts.items())
log.info('Preparing update of `average_daily_users` for %s add-ons.',
len(counts))
create_chunked_tasks_signatures(_update_addon_average_daily_users, counts,
chunk_size).apply_async()
def gather_index_stats_tasks(index, addons=None, dates=None):
"""
Return the list of task groups to execute to index statistics for the given
index/dates/addons.
"""
queries = [
(UpdateCount.objects, index_update_counts,
{'date': 'date'}),
(DownloadCount.objects, index_download_counts,
{'date': 'date'}),
]
jobs = []
for qs, task, fields in queries:
date_field = fields['date']
if dates or addons:
qs = qs.order_by('-%s' % date_field)
qs = qs.values_list('id', flat=True)
if addons:
pks = [int(a.strip()) for a in addons.split(',')]
qs = qs.filter(addon__in=pks)
if dates:
if ':' in dates:
qs = qs.filter(**{'%s__range' % date_field:
dates.split(':')})
else:
qs = qs.filter(**{date_field: dates})
if not (dates or addons):
# We're loading the whole world. Do it in stages so we get most
# recent stats first and don't do huge queries.
limits = (qs.model.objects.filter(**{'%s__isnull' %
date_field: False})
.extra(where=['%s <> "0000-00-00"' % date_field])
.aggregate(min=Min(date_field), max=Max(date_field)))
# If there isn't any data at all, skip over.
if not (limits['max'] or limits['min']):
continue
num_days = (limits['max'] - limits['min']).days
for start in range(0, num_days, STEP):
stop = start + STEP - 1
date_range = (limits['max'] - timedelta(days=stop),
limits['max'] - timedelta(days=start))
data = list(qs.filter(**{
'%s__range' % date_field: date_range
}))
if data:
jobs.append(create_chunked_tasks_signatures(
task, data, CHUNK_SIZE, task_args=(index,)))
else:
jobs.append(create_chunked_tasks_signatures(
task, list(qs), CHUNK_SIZE, task_args=(index,)))
return jobs
def run_yara_query_rule(query_rule_pk):
"""
Run a specific ScannerQueryRule on multiple Versions.
"""
# We're not forcing this task to happen on primary db to let the replicas
# handle the Version query below, but we want to fetch the rule using the
# primary db in all cases.
rule = ScannerQueryRule.objects.using('default').get(pk=query_rule_pk)
# Build a huge list of all pks we're going to run the tasks on.
pks = Version.unfiltered.all().filter(
addon__type=amo.ADDON_EXTENSION,
files__is_webextension=True,
).exclude(addon__status=amo.STATUS_DISABLED, ).filter(
Q(channel=amo.RELEASE_CHANNEL_UNLISTED)
| Q(channel=amo.RELEASE_CHANNEL_LISTED,
pk=F('addon___current_version'))).values_list(
'id', flat=True).order_by('pk')
rule.update(state=RUNNING)
# Build the workflow using a group of tasks dealing with 250 files at a
# time, chained to a task that marks the query as completed.
chunk_size = 250
workflow = (create_chunked_tasks_signatures(
run_yara_query_rule_on_versions_chunk,
list(pks),
chunk_size,
task_args=(query_rule_pk, ))
| mark_yara_query_rule_as_completed.si(query_rule_pk))
# Fire it up.
workflow.apply_async()
def handle(self, *args, **options):
ids = AddonGUID.objects.filter(hashed_guid=None).values_list('id',
flat=True)
chunked_tasks = create_chunked_tasks_signatures(backfill_hashed_guids,
items=list(ids),
chunk_size=100)
chunked_tasks.apply_async()
def get_indexing_tasks_for_qs(qs):
index_data_tasks = create_chunked_tasks_signatures(
cls.get_indexing_task(), qs, cls.CHUNK_SIZE,
task_args=(index_name,))
# Unwrap the tasks from the group create_chunked_tasks_signatures()
# returned, we'll create our own flat group with all the tasks,
# no need to create unnecessary nesting.
return index_data_tasks.tasks
def reindex_tasks_group(index_name):
"""
Return the group of tasks to execute for a full reindex of addons on the
index called `index_name` (which is not an alias but the real index name).
"""
from olympia.addons.models import Addon
from olympia.addons.tasks import index_addons
ids = Addon.unfiltered.values_list('id', flat=True).order_by('id')
chunk_size = 150
return create_chunked_tasks_signatures(index_addons, list(ids), chunk_size)
def reindex_tasks_group(index_name):
"""
Return the group of tasks to execute for a full reindex of addons on the
index called `index_name` (which is not an alias but the real index name).
"""
from olympia.addons.models import Addon
from olympia.addons.tasks import index_addons
ids = Addon.unfiltered.values_list('id', flat=True).order_by('id')
chunk_size = 150
return create_chunked_tasks_signatures(index_addons, list(ids), chunk_size)
def run_yara_query_rule(query_rule_pk):
"""
Run a specific ScannerQueryRule on multiple Versions.
Needs the rule to be a the SCHEDULED state, otherwise does nothing.
"""
# We're not forcing this task to happen on primary db to let the replicas
# handle the Version query below, but we want to fetch the rule using the
# primary db in all cases.
rule = ScannerQueryRule.objects.using('default').get(pk=query_rule_pk)
try:
rule.change_state_to(RUNNING)
except ImproperScannerQueryRuleStateError:
log.error(
'Not proceeding with run_yara_query_rule on rule %s because '
'its state is %s',
rule.pk,
rule.get_state_display(),
)
return
log.info('Fetching versions for run_yara_query_rule on rule %s', rule.pk)
# Build a huge list of all pks we're going to run the tasks on.
qs = Version.unfiltered.filter(
addon__type=amo.ADDON_EXTENSION,
files__is_webextension=True,
)
if not rule.run_on_disabled_addons:
qs = qs.exclude(addon__status=amo.STATUS_DISABLED)
qs = qs.values_list('id', flat=True).order_by('pk')
# Build the workflow using a group of tasks dealing with 250 files at a
# time, chained to a task that marks the query as completed.
chunk_size = 250
chunked_tasks = create_chunked_tasks_signatures(
run_yara_query_rule_on_versions_chunk,
list(qs),
chunk_size,
task_args=(query_rule_pk, ),
)
# Force the group id to be generated for those tasks, and store it in the
# result backend.
group_result = chunked_tasks.freeze()
group_result.save()
rule.update(task_count=len(chunked_tasks),
celery_group_result_id=uuid.UUID(group_result.id))
workflow = chunked_tasks | mark_yara_query_rule_as_completed_or_aborted.si(
query_rule_pk)
log.info(
'Running workflow of %s tasks for run_yara_query_rule on rule %s',
len(chunked_tasks),
rule.pk,
)
# Fire it up.
workflow.apply_async()
def handle(self, *args, **options):
log = olympia.core.logger.getLogger('z.files')
files = File.objects.filter(
is_webextension=True,
version__addon__type=amo.ADDON_EXTENSION).order_by('pk')
pks = files.values_list('pk', flat=True)
log.info('Using %s file pks to extract permissions' % pks.count())
if pks:
chunked_tasks = create_chunked_tasks_signatures(
extract_optional_permissions, list(pks), chunk_size=100)
chunked_tasks.apply_async()
def deliver_hotness(chunk_size=300):
"""
Calculate hotness of all add-ons.
a = avg(users this week)
b = avg(users three weeks before this week)
threshold = 250 if addon type is theme, else 1000
hotness = (a-b) / b if a > threshold and b > 1 else 0
"""
frozen_guids = list(set(fa.addon.guid for fa in FrozenAddon.objects.all()))
averages = get_averages_by_addon_from_bigquery(today=date.today(),
exclude=frozen_guids)
create_chunked_tasks_signatures(update_addon_hotness, averages.items(),
chunk_size).apply_async()
# Reset add-ons that won't be returned by BigQuery.
addon_ids = (Addon.objects.filter(status__in=amo.REVIEWED_STATUSES).filter(
hotness__gt=0).exclude(guid__in=averages.keys()).values_list(
'id', flat=True))
create_chunked_tasks_signatures(reset_addon_hotness, addon_ids,
chunk_size).apply_async()
def update_addon_hotness(chunk_size=300):
"""
Calculate hotness of all add-ons.
a = avg(users this week)
b = avg(users three weeks before this week)
threshold = 250 if addon type is theme, else 1000
hotness = (a-b) / b if a > threshold and b > 1 else 0
"""
frozen_guids = list(
{fa.addon.guid for fa in FrozenAddon.objects.all() if fa.addon.guid}
)
log.info('Found %s frozen add-on GUIDs.', len(frozen_guids))
amo_guids = (
Addon.objects.exclude(guid__in=frozen_guids)
.exclude(guid__isnull=True)
.exclude(guid__exact='')
.exclude(hotness=0)
.values_list('guid', flat=True)
)
averages = {
guid: {'avg_this_week': 1, 'avg_three_weeks_before': 1} for guid in amo_guids
}
log.info('Found %s add-on GUIDs in AMO DB.', len(averages))
bq_averages = get_averages_by_addon_from_bigquery(
today=date.today(), exclude=frozen_guids
)
log.info('Found %s add-on GUIDs with averages in BigQuery.', len(bq_averages))
averages.update(bq_averages)
log.info('Preparing update of `hotness` for %s add-ons.', len(averages))
create_chunked_tasks_signatures(
_update_addon_hotness, averages.items(), chunk_size
).apply_async()
def update_addon_weekly_downloads(chunk_size=250):
"""
Update 7-day add-on download counts.
"""
counts = dict(
# In order to reset the `weekly_downloads` values of add-ons that
# don't exist in BigQuery, we prepare a set of `(hashed_guid, 0)`
# for most add-ons.
Addon.objects.filter(type__in=amo.ADDON_TYPES_WITH_STATS).exclude(
guid__isnull=True
).exclude(guid__exact=''
).exclude(weekly_downloads=0
).annotate(count=Value(0, IntegerField())
).values_list('addonguid__hashed_guid',
'count'))
# Update the `counts` with values from BigQuery.
counts.update(get_addons_and_weekly_downloads_from_bigquery())
counts = list(counts.items())
log.info('Preparing update of `weekly_downloads` for %s add-ons.',
len(counts))
create_chunked_tasks_signatures(_update_addon_weekly_downloads, counts,
chunk_size).apply_async()
def update_addon_weekly_downloads(chunk_size=250):
"""
Update 7-day add-on download counts.
"""
if waffle.switch_is_active('use-bigquery-for-download-stats-cron'):
counts = dict(
# In order to reset the `weekly_downloads` values of add-ons that
# don't exist in BigQuery, we prepare a set of `(guid, 0)` for most
# add-ons.
Addon.objects.filter(type__in=amo.ADDON_TYPES_WITH_STATS).exclude(
guid__isnull=True
).exclude(guid__exact=''
).exclude(weekly_downloads=0
).annotate(count=Value(0, IntegerField())
).values_list('guid', 'count'))
# Update the `counts` with values from BigQuery.
counts.update(get_addons_and_weekly_downloads_from_bigquery())
counts = list(counts.items())
log.info('Preparing update of `weekly_downloads` for %s add-ons.',
len(counts))
create_chunked_tasks_signatures(_update_addon_weekly_downloads, counts,
chunk_size).apply_async()
else:
raise_if_reindex_in_progress('amo')
with connection.cursor() as cursor:
cursor.execute("""
SELECT addon_id, SUM(count) AS weekly_count
FROM download_counts
WHERE `date` >= DATE_SUB(CURDATE(), INTERVAL 7 DAY)
GROUP BY addon_id
ORDER BY addon_id""")
counts = cursor.fetchall()
addon_ids = [r[0] for r in counts]
if not addon_ids:
return
with connection.cursor() as cursor:
cursor.execute(
"""
SELECT id, 0
FROM addons
WHERE id NOT IN %s""", (addon_ids, ))
counts += cursor.fetchall()
cursor.execute("""
CREATE TEMPORARY TABLE tmp_wd
(addon_id INT PRIMARY KEY, count INT)""")
cursor.execute(
'INSERT INTO tmp_wd VALUES %s' %
','.join(['(%s,%s)'] * len(counts)),
list(itertools.chain(*counts)))
cursor.execute("""
UPDATE addons INNER JOIN tmp_wd
ON addons.id = tmp_wd.addon_id
SET weeklydownloads = tmp_wd.count""")
cursor.execute("DROP TABLE IF EXISTS tmp_wd")
def gather_index_stats_tasks(index, addons=None, dates=None):
"""
Return the list of task groups to execute to index statistics for the given
index/dates/addons.
"""
queries = [
(UpdateCount.objects, index_update_counts,
{'date': 'date'}),
(DownloadCount.objects, index_download_counts,
{'date': 'date'}),
(ThemeUserCount.objects, index_theme_user_counts,
{'date': 'date'})
]
jobs = []
for qs, task, fields in queries:
date_field = fields['date']
if dates or addons:
qs = qs.order_by('-%s' % date_field)
qs = qs.values_list('id', flat=True)
if addons:
pks = [int(a.strip()) for a in addons.split(',')]
qs = qs.filter(addon__in=pks)
if dates:
if ':' in dates:
qs = qs.filter(**{'%s__range' % date_field:
dates.split(':')})
else:
qs = qs.filter(**{date_field: dates})
if not (dates or addons):
# We're loading the whole world. Do it in stages so we get most
# recent stats first and don't do huge queries.
limits = (qs.model.objects.filter(**{'%s__isnull' %
date_field: False})
.extra(where=['%s <> "0000-00-00"' % date_field])
.aggregate(min=Min(date_field), max=Max(date_field)))
# If there isn't any data at all, skip over.
if not (limits['max'] or limits['min']):
continue
num_days = (limits['max'] - limits['min']).days
for start in range(0, num_days, STEP):
stop = start + STEP - 1
date_range = (limits['max'] - timedelta(days=stop),
limits['max'] - timedelta(days=start))
data = list(qs.filter(**{
'%s__range' % date_field: date_range
}))
if data:
jobs.append(create_chunked_tasks_signatures(
task, data, CHUNK_SIZE, task_args=(index,)))
else:
jobs.append(create_chunked_tasks_signatures(
task, list(qs), CHUNK_SIZE, task_args=(index,)))
return jobs
