def test_stmt_mode_unread():
"Test whether we can only create statements from unread content."
# Prep the inputs.
db = get_db_with_pubmed_content()
tcids = {tcid for tcid, in db.select_all(db.TextContent.id)}
# Test with just sparser for tolerable speeds.
readers = get_readers('SPARSER')
# First create some readings.
some_tcids = random.sample(tcids, len(tcids) // 2)
workers0 = rdb.run_reading(readers, some_tcids, db=db, verbose=True)
pre_stmt_hash_set = {
sd.statement.get_hash(shallow=False)
for sd in workers0[0].statement_outputs
}
# Now only make statements for the content that was not read.
workers = rdb.run_reading(readers,
tcids,
db=db,
verbose=True,
stmt_mode='unread')
stmt_hash_set = {
sd.statement.get_hash(shallow=False)
for sd in workers[0].statement_outputs
}
assert stmt_hash_set.isdisjoint(pre_stmt_hash_set), \
"There were overlapping statements."
def _run_reading(self, db, tcids, reader_name):
ids_per_job = 5000
if len(tcids) > ids_per_job:
raise ReadingUpdateError("Too many id's to run locally. Try "
"running on batch (use_batch).")
logger.info("Producing readings locally for %d new text refs."
% len(tcids))
base_dir = path.join(THIS_DIR, 'read_all_%s' % reader_name)
readers = rdb.construct_readers([reader_name], base_dir=base_dir,
n_proc=self.n_proc)
rdb.run_reading(readers, tcids, db=db, batch_size=ids_per_job,
verbose=self.verbose)
return
def test_multi_batch_run():
"Test that reading works properly with multiple batches run."
db = get_db_with_pubmed_content()
readers = get_readers()
tcids = {tcid for tcid, in db.select_all(db.TextContent.id)}
rdb.run_reading(readers, tcids, batch_size=len(tcids)//2, db=db,
stmt_mode='none')
# This should catch any repeated readings.
num_readings = db.filter_query(db.Reading).count()
# NOTE: This might need some special consideration given the new TRIPS
# reader, which only reads titles
num_expected = len(readers)*len(tcids)
assert num_readings == num_expected, \
"Expected %d readings, only found %d." % (num_expected, num_readings)
def test_multiproc_statements():
"Test the multiprocessing creation of statements."
db = get_db_with_pubmed_content()
readers = get_readers()
tcids = {tcid for tcid, in db.select_all(db.TextContent.id)}
workers = rdb.run_reading(readers, tcids, db=db)
assert not any(worker.extant_readings for worker in workers)
outputs = [rd for worker in workers for rd in worker.new_readings]
stmts = rdb.make_statements(outputs, 2)
assert len(stmts)
def test_produce_readings():
"Comprehensive test of the high level production of readings."
# Prep the inputs.
db = get_db_with_pubmed_content()
tcids = {tcid for tcid, in db.select_all(db.TextContent.id)}
# Test with just sparser for tollerable speeds.
readers = get_readers('SPARSER')
# Test the reading_mode='none' option (should yield nothing, because there
# aren't any readings yet.)
workers = rdb.run_reading(readers, tcids, verbose=True, db=db,
reading_mode='none', stmt_mode='none')
assert all(len(worker.new_readings) == 0 for worker in workers)
assert all(len(worker.extant_readings) == 0 for worker in workers)
# Test just getting a pickle file (Nothing should be posted to the db.).
pkl_file = 'test_db_res.pkl'
workers = rdb.run_reading(readers, tcids, verbose=True, db=db,
upload_readings=False, reading_pickle=pkl_file)
N_new = len(workers[0].new_readings)
N_old = len(workers[0].extant_readings)
N_exp = len(readers)*len(tcids)
assert N_new == N_exp, "Expected %d readings, got %d." % (N_exp, N_new)
assert N_old == 0, "Found old readings, when there should be none."
assert path.exists(pkl_file), "Pickle file not created."
with open(pkl_file, 'rb') as f:
N_pkl = len(pickle.load(f))
assert N_pkl == N_exp, \
"Expected %d readings in pickle, got %d." % (N_exp, N_pkl)
N_readings = db.filter_query(db.Reading).count()
assert N_readings == 0, \
"There shouldn't be any readings yet, but found %d." % N_readings
# Test reading and insert to the database.
rdb.run_reading(readers, tcids, verbose=True, db=db)
N_db = db.filter_query(db.Reading).count()
assert N_db == N_exp, "Expected %d readings, got %d." % (N_exp, N_db)
# Test reading again, without read_mode='all', ('unread' by default)
workers = rdb.run_reading(readers, tcids, verbose=True, db=db)
N_old = len(workers[0].extant_readings)
N_new = len(workers[0].new_readings)
assert N_old == N_exp, \
"Got %d old readings, expected %d." % (N_old, N_exp)
assert N_new == 0, \
"Got %d new readings, when none should have been read." % N_new
assert all([rd.reading_id is not None
for rd in workers[0].extant_readings])
# Test with read_mode='none' again.
workers = rdb.run_reading(readers, tcids, verbose=True, db=db,
reading_mode='none')
N_old = len(workers[0].extant_readings)
assert N_old == N_exp
assert all([rd.reading_id is not None
for rd in workers[0].extant_readings])
# Test the read_mode='all'.
workers = rdb.run_reading(readers, tcids, verbose=True, db=db,
reading_mode='all')
old = workers[0].extant_readings
new = workers[0].new_readings
assert len(new) == N_exp
assert len(old) == 0
assert all([rd.reading_id is not None for rd in new])
def main():
arg_parser = get_parser()
args = arg_parser.parse_args()
s3 = boto3.client('s3')
s3_log_prefix = get_s3_job_log_prefix(args.s3_base, args.job_name)
logger.info("Using log prefix \"%s\"" % s3_log_prefix)
id_list_key = args.s3_base + 'id_list'
logger.info("Looking for id list on s3 at \"%s\"" % id_list_key)
try:
id_list_obj = s3.get_object(Bucket=bucket_name, Key=id_list_key)
except botocore.exceptions.ClientError as e:
# Handle a missing object gracefully
if e.response['Error']['Code'] == 'NoSuchKey':
logger.info('Could not find PMID list file at %s, exiting' %
id_list_key)
sys.exit(1)
# If there was some other kind of problem, re-raise the exception
else:
raise e
# Get the content from the object
id_list_str = id_list_obj['Body'].read().decode('utf8').strip()
id_str_list = id_list_str.splitlines()[args.start_index:args.end_index]
random.shuffle(id_str_list)
tcids = [int(line.strip()) for line in id_str_list]
# Get the reader objects
if not os.path.exists(args.out_dir):
os.makedirs(args.out_dir)
kwargs = {'base_dir': args.out_dir, 'n_proc': args.num_cores}
readers = construct_readers(args.readers, **kwargs)
# Record the reader versions used in this run.
reader_versions = {}
for reader in readers:
reader_versions[reader.name] = reader.get_version()
s3.put_object(Bucket=bucket_name,
Key=get_s3_reader_version_loc(args.s3_base, args.job_name),
Body=json.dumps(reader_versions))
# Some combinations of options don't make sense:
forbidden_combos = [('all', 'unread'), ('none', 'unread'),
('none', 'none')]
assert (args.read_mode, args.rslt_mode) not in forbidden_combos, \
("The combination of reading mode %s and statement mode %s is not "
"allowed." % (args.reading_mode, args.rslt_mode))
# Get a handle for the database
if args.test:
from indra_db.tests.util import get_temp_db
db = get_temp_db(clear=True)
else:
db = None
# Read everything ========================================
if args.batch is None:
run_reading(readers,
tcids,
verbose=True,
db=db,
reading_mode=args.read_mode,
rslt_mode=args.rslt_mode)
else:
for tcid_batch in batch_iter(tcids, args.batch):
run_reading(readers,
tcid_batch,
verbose=True,
db=db,
reading_mode=args.read_mode,
rslt_mode=args.rslt_mode)
# Preserve the sparser logs
contents = os.listdir('.')
logger.info("Checking for any log files to cache:\n" + '\n'.join(contents))
sparser_logs = []
trips_logs = []
for fname in contents:
# Check if this file is a sparser log
if fname.startswith('sparser') and fname.endswith('log'):
sparser_logs.append(fname)
elif is_trips_datestring(fname):
for sub_fname in os.listdir(fname):
if sub_fname.endswith('.log') or sub_fname.endswith('.err'):
trips_logs.append(os.path.join(fname, sub_fname))
_dump_logs_to_s3(s3, s3_log_prefix, 'sparser', sparser_logs)
_dump_logs_to_s3(s3, s3_log_prefix, 'trips', trips_logs)
return
