def test_deduplicate_by_key(self):
collection_of_dicts = [
{'index': 'one', 'other': '1'},
{'index': 'two', 'other': '2'},
{'index': 'two', 'other': '3'},
]
no_duplicates = deduplicate_by_key(collection_of_dicts, lambda r: r['index'])
expected = collection_of_dicts[0:2]
assert_equal(sorted(no_duplicates), expected)
def test_deduplicate_by_key(self):
collection_of_dicts = [
{
'index': 'one',
'other': '1'
},
{
'index': 'two',
'other': '2'
},
{
'index': 'two',
'other': '3'
},
]
no_duplicates = deduplicate_by_key(collection_of_dicts,
lambda r: r['index'])
expected = collection_of_dicts[0:2]
assert_equal(sorted(no_duplicates), expected)
def validate_records(self, session, cohort):
"""
Fetches the wiki_user(s) already added for self.cohort_id and validates
their mediawiki_username against their stated project as either a user_id
or user_name. Once done, sets the valid state and deletes any duplicates.
Then, it finishes filling in the data model by inserting corresponding
records into the cohort_wiki_users table.
This is meant to execute asynchronously on celery
Parameters
session : an active wikimetrics db session to use
cohort : the cohort to validate; must belong to session
"""
# reset the cohort validation status so it can't be used for reports
cohort.validated = False
session.execute(
WikiUser.__table__.update().values(valid=None).where(
WikiUser.validating_cohort == cohort.id
)
)
session.execute(CohortWikiUser.__table__.delete().where(
CohortWikiUser.cohort_id == cohort.id
))
session.commit()
wikiusers = session.query(WikiUser) \
.filter(WikiUser.validating_cohort == cohort.id) \
.all()
deduplicated = deduplicate_by_key(
wikiusers,
lambda r: (r.mediawiki_username, r.project)
)
wikiusers_by_project = {}
for wu in deduplicated:
try:
normalized_project = normalize_project(wu.project)
if normalized_project is None:
wu.reason_invalid = 'invalid project: {0}'.format(wu.project)
wu.valid = False
continue
wu.project = normalized_project
if wu.project not in wikiusers_by_project:
wikiusers_by_project[wu.project] = []
wikiusers_by_project[wu.project].append(wu)
# validate bunches of records to update the UI but not kill performance
if len(wikiusers_by_project[wu.project]) > 999:
validate_users(
wikiusers_by_project[wu.project],
wu.project,
self.validate_as_user_ids
)
session.commit()
wikiusers_by_project[wu.project] = []
except:
continue
# validate anything that wasn't big enough for a batch
for project, wikiusers in wikiusers_by_project.iteritems():
if len(wikiusers) > 0:
validate_users(wikiusers, project, self.validate_as_user_ids)
session.commit()
unique_and_validated = deduplicate_by_key(
deduplicated,
lambda r: (r.mediawiki_username, r.project)
)
session.execute(
CohortWikiUser.__table__.insert(), [
{
'cohort_id' : cohort.id,
'wiki_user_id' : wu.id,
} for wu in unique_and_validated
]
)
# clean up any duplicate wiki_user records
session.execute(WikiUser.__table__.delete().where(and_(
WikiUser.validating_cohort == cohort.id,
WikiUser.id.notin_([wu.id for wu in unique_and_validated])
)))
cohort.validated = True
session.commit()
def validate_records(self, session, cohort):
"""
Fetches the wiki_user(s) already added for self.cohort_id and validates
their raw_id_or_name field against their stated project as either a user_id
or user_name. Once done, sets the valid state and deletes any duplicates.
Then, it finishes filling in the data model by inserting corresponding
records into the cohort_wiki_users table.
This is meant to execute asynchronously on celery
Parameters
session : an active wikimetrics db session to use
cohort : the cohort to validate; must belong to session
"""
# reset the cohort validation status so it can't be used for reports
cohort.validated = False
session.execute(WikiUserStore.__table__.update().values(
valid=None).where(WikiUserStore.validating_cohort == cohort.id))
session.execute(CohortWikiUserStore.__table__.delete().where(
CohortWikiUserStore.cohort_id == cohort.id))
session.commit()
wikiusers = session.query(WikiUserStore) \
.filter(WikiUserStore.validating_cohort == cohort.id) \
.all()
deduplicated = deduplicate_by_key(
wikiusers, lambda r:
(r.raw_id_or_name, normalize_project(r.project) or r.project))
wikiusers_by_project = {}
for wu in deduplicated:
normalized_project = normalize_project(wu.project)
if normalized_project is None:
wu.reason_invalid = 'invalid project: {0}'.format(wu.project)
wu.valid = False
continue
wu.project = normalized_project
if wu.project not in wikiusers_by_project:
wikiusers_by_project[wu.project] = []
wikiusers_by_project[wu.project].append(wu)
# validate bunches of records to update the UI but not kill performance
if len(wikiusers_by_project[wu.project]) > 999:
validate_users(wikiusers_by_project[wu.project], wu.project,
self.validate_as_user_ids)
session.commit()
wikiusers_by_project[wu.project] = []
# validate anything that wasn't big enough for a batch
for project, wikiusers in wikiusers_by_project.iteritems():
if len(wikiusers) > 0:
validate_users(wikiusers, project, self.validate_as_user_ids)
session.commit()
session.execute(CohortWikiUserStore.__table__.insert(),
[{
'cohort_id': cohort.id,
'wiki_user_id': wu.id,
} for wu in deduplicated])
# clean up any duplicate wiki_user records
session.execute(WikiUserStore.__table__.delete().where(
and_(WikiUserStore.validating_cohort == cohort.id,
WikiUserStore.id.notin_([wu.id for wu in deduplicated]))))
cohort.validated = True
session.commit()
