def get_classifications(self):
logger.info('Loading dual cursors of dump and caesar classifications')
cursor1 = DB().classifications.getClassifications()
cursor2 = DB().caesar.getClassifications()
return DualCursor(cursor1, cursor2)
def run(self, amount=None):
def _amt(stats):
return stats['first_classifications']
if amount is None:
amount = _amt(DB().classifications.get_stats())
amount += _amt(DB().caesar._gen_stats(upload=False))
super().run(amount=amount)
def call(self, args):
"""
Define what to do if this interface's command was passed
"""
if args.upload_dump:
fname = args.upload_dump[0]
DB().classifications.upload_project_dump(fname)
if args.upload_golds:
fname = args.upload_golds[0]
DB().golds.upload_golds_csv(fname)
if args.gen_stats:
DB().classifications._gen_stats()
def get_cursor():
"""
Generate a cursor with classifications
Returns
-------
swap.db.Cursor
Classifications
"""
cursor = DB().classifications.aggregate([{
'$match': {
'gold_label': {
'$ne': -1
}
}
}, {
'$group': {
'_id': '$subject_id',
'gold': {
'$first': "$gold_label"
},
'total': {
'$sum': 1
},
'votes': {
'$sum': "$annotation"
}
}
}])
return cursor
def test_batch_size(self, mock):
DB().classifications.getClassifications(batch_size=50)
args, kwargs = mock.call_args
print(args, kwargs)
assert 'batchSize' in args[1]
assert args[1]['batchSize'] == 50
def _init_subjects(self, golds):
subjects = {}
for id_, gold in golds.items():
stats = SubjectStats.from_static(id_, DB())
subjects[id_] = self.Subject(id_, gold, stats)
return subjects
def upload_data(data):
db = DB()
requests = []
def write():
nonlocal requests
print('writing')
if len(requests) > 0:
db.subjects.bulk_write(requests)
requests = []
i = 0
print(i)
for subject, metadata in data.items():
r = db.subjects.update_metadata(subject, metadata, False)
requests.append(r)
if i % 10000:
sys.stdout.flush()
sys.stdout.write('%d\r' % i)
if len(requests) > 1e5:
write()
requests = []
i += 1
write()
def test_get_cursor_type(self):
DB._instances = {}
db = DB()._db
query = [{'$limit': 5}]
c = Cursor(query, db.classifications)
assert isinstance(c.getCursor(), pymongo.command_cursor.CommandCursor)
def test_length(self):
DB._reset()
db = DB()._db
query = [{'$limit': 5}]
c = Cursor(query, db.classifications)
print(c.next())
assert len(c) == 5
def get_classifications():
"""
Get the cursor containing classifications from db
Returns
-------
swap.db.Cursor
Cursor with classifications
"""
return DB().classifications.getClassifications()
def subjects(self, subject_ids):
"""
Get the gold labels for a set of subjects
Parameters
----------
subject_ids : list
List of subject ids (int)
"""
logger.debug('getting %d subjects', len(subject_ids))
return DB().golds.get_golds(subject_ids)
def random(self, size, gold=None):
"""
Get a random sample of gold labels
Parameters
----------
size : int
Sample size
"""
logger.debug('Size %d gold filter %s', size, gold)
return DB().golds.get_random_golds(size, gold)
def test_get_classifications_1(self, mock):
query = [
{'$sort': OrderedDict([
('seen_before', 1),
('classification_id', 1)])},
{'$match': {'seen_before': False}},
{'$project': {
'user_id': 1, 'subject_id': 1,
'annotation': 1, 'session_id': 1}}]
DB().classifications.getClassifications()
mock.assert_called_with(query, {'batchSize': 100000})
def consensus(self, size):
"""
Get the gold labels for the most consensus subjects
Parameters
----------
size : int
Number of subjects
"""
logger.debug('Size %d', size)
subjects = db_cv().get_consensus(size)
return DB().golds.get_golds(subjects)
def run(self, amount=None):
"""
Process all classifications in DB with SWAP
.. note::
Iterates through the classification collection of the
database and proccesss each classification one at a time
in the order returned by the db.
Parameters like max_batch_size are hard-coded.
Prints status.
"""
if amount is None:
amount = DB().classifications.get_stats()
amount = amount['first_classifications']
self.init_swap()
# get classifications
cursor = self.get_classifications()
# loop over classification cursor to process
# classifications one at a time
logger.info('Start: SWAP Processing %d classifications', amount)
count = 0
with progressbar.ProgressBar(max_value=amount) as bar:
bar.update(count)
# Loop over all classifications of the query
# Note that the exact size of the query might be lower than
# n_classifications if not all classifications are being queried
for cl in cursor:
# process classification in swap
cl = Classification.generate(cl)
self._delegate(cl)
bar.update(count)
count += 1
if config.control.debug and count > config.control.amount:
break
if config.back_update:
logger.info('back_update active: processing changes')
self.swap.process_changes()
logger.info('done')
def classify(self, raw_cl):
# Add classification from caesar
data = self.parse_raw(raw_cl)
cl = self.gen_cl(data)
logger.debug('Checking if already received classification')
if not self.cl_exists(data):
logger.debug('Uploading classification to caesar db: %s',
str(data))
DB().caesar.insert(data)
logger.debug('Adding classification from network: %s',
str(cl))
self.swap.classify(cl)
subject = self.swap.subjects.get(cl.subject)
return subject
def getClassifications(self):
""" Returns Iterator over all Classifications """
# fields to project
fields = ['user_name', 'subject_id', 'annotation', 'gold_label']
# if meta data is requested
if self.meta_data is not None:
meta_data_field = 'metadata' + "." + self.meta_data
fields.append('metadata')
fields[fields.index('metadata')] = meta_data_field
# Define a query
q = Query()
q.project(fields)
# range query on metadata
if self.meta_lower is not None and self.meta_upper is not None:
q.match_range(meta_data_field, self.meta_lower, self.meta_upper)
# perform query on classification data
classifications = DB().classifications.aggregate(q.build())
return classifications
def __init__(self):
self.getters = []
self._golds = None
self.db = DB().golds
def all(self):
"""
Get all gold labels
"""
return DB().golds.get_golds()
def db_cv():
return DB().controversial
def cl_exists(cl):
def id_(cl):
return cl['classification_id']
return DB().caesar.exists(id_(cl)) or \
DB().classifications.exists(id_(cl))
def db_cl():
return DB().classifications
def get_swap_scores():
return DB().subjects.get_scores()
def call(self, args):
swap = None
scores = None
if args.load:
obj = self.load(args.load[0])
if isinstance(obj, SWAP):
swap = obj
scores = swap.score_export()
elif isinstance(obj, ScoreExport):
scores = obj
if args.scores_from_csv:
fname = args.scores_from_csv[0]
scores = ScoreExport.from_csv(fname)
if args.run:
swap = self.run_swap(args)
scores = swap.score_export()
if swap is not None:
if args.save:
manifest = self.manifest(swap, args)
self.save(swap, self.f(args.save[0]), manifest)
if args.log:
fname = self.f(args.log[0])
write_log(swap, fname)
if args.stats:
s = swap.stats_str()
print(s)
logger.debug(s)
if args.test:
from swap.utils.golds import GoldGetter
gg = GoldGetter()
logger.debug('applying new gold labels')
swap.set_gold_labels(gg.golds)
swap.process_changes()
logger.debug('done')
if args.test_reorder:
self.reorder_classifications(swap)
if args.export_user_scores:
fname = self.f(args.export_user_scores[0])
self.export_user_scores(swap, fname)
if scores is not None:
if args.save_scores:
DB().subjects.save_scores(scores)
if args.scores_to_csv:
self.scores_to_csv(scores, args.scores_to_csv[0])
self.plot(args, swap, scores)
if args.shell:
import code
code.interact(local=locals())
return swap