def run(
departement,
prediction_beginning_date=None,
return_df_etab_if_successful=False,
):
"""
Returns True if computation successful and False otherwise.
"""
if prediction_beginning_date is None:
prediction_beginning_date = compute_prediction_beginning_date()
check_prediction_beginning_date(prediction_beginning_date)
# get df_etab with monthly hirings for each hiring_type
# hirings are not yet grouped by period (i.e. 6 months for DPAE / 6 months for Alternance)
df_etab = get_df_etab_with_hiring_monthly_aggregates(
departement, prediction_beginning_date)
logger.debug(
"df_etab_with_hiring_monthly_aggregates loaded for departement %s",
departement)
if df_etab is None:
logger.warning("no etab/hiring data found for departement %s",
departement)
return False # failed computation (e.g. no data)
# LBB / DPAE.
train(
df_etab,
departement,
prediction_beginning_date=prediction_beginning_date,
last_historical_data_date=go_back_last_day_of_the_month(
DpaeStatistics.get_last_historical_data_date(DpaeStatistics.DPAE)),
months_per_period=6,
training_periods=7,
prefix_for_fields='dpae',
score_field_name='score',
)
logger.debug("DPAE/LBB training done for departement %s", departement)
# LBA / Alternance.
train(df_etab,
departement,
prediction_beginning_date=prediction_beginning_date,
last_historical_data_date=go_back_last_day_of_the_month(
DpaeStatistics.get_last_historical_data_date(
DpaeStatistics.APR)),
months_per_period=6,
training_periods=7,
prefix_for_fields='alt',
score_field_name='score_alternance',
is_lbb=False)
logger.debug("Alternance/LBA training done for departement %s",
departement)
# Final data export.
export_df_etab_to_db(df_etab, departement)
if return_df_etab_if_successful:
return df_etab # only used in test_compute_score.py for inspection
return True # successful computation
def run(self):
"""
Tricky parallelization.
In some cases the jobs take a long time to run, so there's COMPUTE_SCORE_TIMEOUT for that.
If a job was not finished before the timeout, it will be added to a list to be processed sequentially.
"""
results = []
pool = multiprocessing.Pool(processes=8, maxtasksperchild=1)
async_results = {}
most_recent_data_date = DpaeStatistics.get_most_recent_data_date()
if '75' in settings.DEPARTEMENTS:
departements = deque(settings.DEPARTEMENTS)
departements.remove('75')
departements.appendleft('75')
else:
departements = list(settings.DEPARTEMENTS)
if COMPUTE_SCORES_DEBUG_MODE:
if len(COMPUTE_SCORES_DEBUG_DEPARTEMENTS) >= 1:
departements = COMPUTE_SCORES_DEBUG_DEPARTEMENTS
for departement in departements:
abortable_func = partial(abortable_worker,
compute,
timeout=COMPUTE_SCORE_TIMEOUT)
async_result = pool.apply_async(
abortable_func,
args=(settings.OFFICE_TABLE, settings.DPAE_TABLE, departement,
most_recent_data_date))
async_results[departement] = async_result
logger.info("going to close process pool...")
pool.close()
logger.info("going to join pool")
pool.join()
logger.info("all compute_score done, analyzing results...")
for departement, async_result in async_results.iteritems():
try:
result = async_result.get()
if not bool(result):
logger.info("departement with error : %s", departement)
results.append([departement, bool(result)])
except:
logger.error("traceback for unprocessed_departement: %s",
traceback.format_exc())
logger.info("compute_scores FINISHED")
return results
def run_task(self):
date_insertion = datetime.now()
logger.info("extracting %s ", self.input_filename)
# this pattern matches the first date
# e.g. 'lbb_xdpdpae_delta_201611102200.bz2'
# will match 2018-09-12
date_pattern = r'.*_(\d\d\d\d\d\d\d\d)\d\d\d\d' #We keep only the date in the file name, ex: 20190910 = 10th september 2019
date_match = re.match(date_pattern, self.input_filename)
if date_match:
date_part = date_match.groups()[0]
self.last_historical_data_date_in_file = datetime.strptime(
date_part, "%Y%m%d")
logger.debug("identified last_historical_data_date_in_file=%s",
self.last_historical_data_date_in_file)
else:
raise Exception(
"couldn't find a date pattern in filename. filename should be \
like lbb_xdpdpae_delta_YYYYMMDDHHMM.csv")
count = 0
statements = []
something_new = False
query = """
INSERT into %s(
siret,
hiring_date,
contract_type,
departement,
contract_duration,
iiann,
tranche_age,
handicap_label,
duree_pec,
date_insertion
)
values(%%s, %%s, %%s, %%s, %%s, %%s, %%s, %%s, %%s, %%s)
""" % settings.HIRING_TABLE
imported_dpae = 0
imported_dpae_distribution = {}
not_imported_dpae = 0
last_historical_data_date_in_db = db_session.query(func.max(Hiring.hiring_date)) \
.filter(Hiring.contract_type.in_((Hiring.CONTRACT_TYPE_CDI,
Hiring.CONTRACT_TYPE_CDD,
Hiring.CONTRACT_TYPE_CTT))).first()[0]
if last_historical_data_date_in_db is None:
last_historical_data_date_in_db = DEFAULT_DATETIME_DPAE
logger.info(
"will now extract all dpae with hiring_date between %s and %s",
last_historical_data_date_in_db,
self.last_historical_data_date_in_file)
with import_util.get_reader(self.input_filename) as myfile:
con, cur = import_util.create_cursor()
header_line = myfile.readline().strip(
) # FIXME detect column positions from header
if b"siret" not in header_line:
logger.debug(header_line)
raise Exception("wrong header line")
for line in myfile:
line = line.decode()
count += 1
if not count % 100000:
logger.debug("reading line %i", count)
try:
try:
cur.executemany(query, statements)
except OperationalError: # retry once in case of deadlock error
time.sleep(10)
cur.executemany(query, statements)
statements = []
con.commit()
something_new = True
except:
logger.error(
"error in executing statement into dpae table: %s",
sys.exc_info()[1])
statements = []
raise
try:
siret, hiring_date, _, contract_type, departement, contract_duration, \
iiann, tranche_age, handicap_label, duree_pec = parse_dpae_line(line)
except ValueError:
self.zipcode_errors += 1
continue
except InvalidRowException:
logger.info("invalid_row met at row: %i", count)
self.invalid_row_errors += 1
continue
dpae_should_be_imported = (
hiring_date > last_historical_data_date_in_db
and hiring_date <= self.last_historical_data_date_in_file
# For DPAE contracts we only keep all CDI, only long enough CDD (at least 31 days)
# and we ignore CTT.
and (contract_type == Hiring.CONTRACT_TYPE_CDI or
(contract_type == Hiring.CONTRACT_TYPE_CDD
and contract_duration is not None
and contract_duration > 31)))
if dpae_should_be_imported:
statement = (siret, hiring_date, contract_type,
departement, contract_duration, iiann,
tranche_age, handicap_label, duree_pec,
date_insertion)
statements.append(statement)
imported_dpae += 1
if hiring_date.year not in imported_dpae_distribution:
imported_dpae_distribution[hiring_date.year] = {}
if hiring_date.month not in imported_dpae_distribution[
hiring_date.year]:
imported_dpae_distribution[hiring_date.year][
hiring_date.month] = {}
if hiring_date.day not in imported_dpae_distribution[
hiring_date.year][hiring_date.month]:
imported_dpae_distribution[hiring_date.year][
hiring_date.month][hiring_date.day] = 0
imported_dpae_distribution[hiring_date.year][
hiring_date.month][hiring_date.day] += 1
else:
not_imported_dpae += 1
# run remaining statements
try:
cur.executemany(query, statements)
something_new = True
except:
logger.error("error in executing statement into dpae table: %s",
sys.exc_info()[1])
raise
logger.info("processed %i dpae...", count)
logger.info("imported dpae: %i", imported_dpae)
logger.info("not imported dpae: %i", not_imported_dpae)
logger.info("zipcode errors: %i", self.zipcode_errors)
logger.info("invalid_row errors: %i", self.invalid_row_errors)
if self.zipcode_errors > settings.MAXIMUM_ZIPCODE_ERRORS:
raise IOError('too many zipcode errors')
if self.invalid_row_errors > settings.MAXIMUM_INVALID_ROWS:
raise IOError('too many invalid_row errors')
logger.info("verifying good number of dpae imported.")
query = "select count(*) from hirings h where hiring_date > %s and hiring_date <= %s and h.contract_type in (1,2,3)"
cur.execute(query, [
last_historical_data_date_in_db,
self.last_historical_data_date_in_file
])
res = cur.fetchone()
if res[0] != imported_dpae:
raise DoublonException(
f"Too many DPAE ({res[0]}) in DB compared to DPAE file ({imported_dpae})."
)
logger.info("verifying number of DPAE: OK.")
con.commit()
cur.close()
con.close()
try:
statistics = DpaeStatistics(
last_import=datetime.now(),
most_recent_data_date=self.last_historical_data_date_in_file,
file_type=self.file_type)
db_session.add(statistics)
db_session.commit()
logger.info("First way to insert DPAE statistics in DB : OK")
except OperationalError:
# For an obscure reason, the DpaeStatistics way to insert does not work on the bonaparte server
# So we insert it directly via an SQL query
# This job has been broken for more than a year, only way to fix it :
db_session.rollback()
last_import_date = datetime.now().strftime('%Y-%m-%d %H:%M:%S')
most_recent_date = self.last_historical_data_date_in_file.strftime(
'%Y-%m-%d %H:%M:%S')
query = f"insert into dpae_statistics (last_import, most_recent_data_date, file_type) values ('{last_import_date}','{most_recent_date}','{self.file_type}')"
con, cur = import_util.create_cursor()
cur.execute(query)
con.commit()
cur.close()
con.close()
logger.info("Second way to insert DPAE statistics in DB : OK")
logger.info("finished importing dpae...")
return something_new
def get_dpae_last_historical_data_date():
return DpaeStatistics.get_last_historical_data_date(DpaeStatistics.DPAE)
def run_task(self):
logger.info("extracting %s ", self.input_filename)
# this pattern matches the first date
# e.g. 'LBB_XDPDPAE_2018-09-12_2017-08-01.bz2'
# will match 2018-09-12
date_pattern = r'.*_(\d\d\d\d-\d\d-\d\d)_'
date_match = re.match(date_pattern, self.input_filename)
if date_match:
date_part = date_match.groups()[0]
self.last_historical_data_date_in_file = datetime.strptime(
date_part, "%Y-%m-%d")
logger.debug("identified last_historical_data_date_in_file=%s",
self.last_historical_data_date_in_file)
else:
raise Exception(
"couldn't find a date pattern in filename. filename should be \
like LBB_XDPDPAE_YYYY-MM-DD_YYYY-MM-DD.csv")
count = 0
statements = []
something_new = False
query = """
INSERT into %s(
siret,
hiring_date,
contract_type,
departement,
contract_duration,
iiann,
tranche_age,
handicap_label,
duree_pec
)
values(%%s, %%s, %%s, %%s, %%s, %%s, %%s, %%s, %%s)
""" % settings.HIRING_TABLE
imported_dpae = 0
imported_dpae_distribution = {}
not_imported_dpae = 0
last_historical_data_date_in_db = DpaeStatistics.get_last_historical_data_date(
)
logger.info(
"will now extract all dpae with hiring_date between %s and %s",
last_historical_data_date_in_db,
self.last_historical_data_date_in_file)
with import_util.get_reader(self.input_filename) as myfile:
con, cur = import_util.create_cursor()
header_line = myfile.readline().strip(
) # FIXME detect column positions from header
if b"siret" not in header_line:
logger.debug(header_line)
raise Exception("wrong header line")
for line in myfile:
line = line.decode()
count += 1
if not count % 100000:
logger.debug("reading line %i", count)
try:
try:
cur.executemany(query, statements)
except OperationalError: # retry once in case of deadlock error
time.sleep(10)
cur.executemany(query, statements)
statements = []
con.commit()
something_new = True
except:
logger.error(
"error in executing statement into dpae table: %s",
sys.exc_info()[1])
statements = []
raise
try:
siret, hiring_date, _, contract_type, departement, contract_duration, \
iiann, tranche_age, handicap_label, duree_pec = parse_dpae_line(line)
except ValueError:
self.zipcode_errors += 1
continue
except InvalidRowException:
logger.info("invalid_row met at row: %i", count)
self.invalid_row_errors += 1
continue
dpae_should_be_imported = (
hiring_date > last_historical_data_date_in_db
and hiring_date <= self.last_historical_data_date_in_file
# For DPAE contracts we only keep all CDI, only long enough CDD (at least 31 days)
# and we ignore CTT.
and (contract_type == Hiring.CONTRACT_TYPE_CDI or
(contract_type == Hiring.CONTRACT_TYPE_CDD
and contract_duration is not None
and contract_duration > 31)))
if dpae_should_be_imported:
statement = (
siret,
hiring_date,
contract_type,
departement,
contract_duration,
iiann,
tranche_age,
handicap_label,
duree_pec,
)
statements.append(statement)
imported_dpae += 1
if hiring_date.year not in imported_dpae_distribution:
imported_dpae_distribution[hiring_date.year] = {}
if hiring_date.month not in imported_dpae_distribution[
hiring_date.year]:
imported_dpae_distribution[hiring_date.year][
hiring_date.month] = {}
if hiring_date.day not in imported_dpae_distribution[
hiring_date.year][hiring_date.month]:
imported_dpae_distribution[hiring_date.year][
hiring_date.month][hiring_date.day] = 0
imported_dpae_distribution[hiring_date.year][
hiring_date.month][hiring_date.day] += 1
else:
not_imported_dpae += 1
# run remaining statements
try:
cur.executemany(query, statements)
something_new = True
except:
logger.error("error in executing statement into dpae table: %s",
sys.exc_info()[1])
raise
logger.info("processed %i dpae...", count)
logger.info("imported dpae: %i", imported_dpae)
logger.info("not imported dpae: %i", not_imported_dpae)
logger.info("zipcode errors: %i", self.zipcode_errors)
logger.info("invalid_row errors: %i", self.invalid_row_errors)
if self.zipcode_errors > settings.MAXIMUM_ZIPCODE_ERRORS:
raise IOError('too many zipcode errors')
if self.invalid_row_errors > settings.MAXIMUM_INVALID_ROWS:
raise IOError('too many invalid_row errors')
con.commit()
cur.close()
con.close()
statistics = DpaeStatistics(
last_import=datetime.now(),
most_recent_data_date=self.last_historical_data_date_in_file,
)
statistics.save()
logger.info("finished importing dpae...")
return something_new
def run_task(self):
date_insertion = datetime.now()
logger.info("extracting %s ", self.input_filename)
# this pattern matches the first date
# e.g. '20200803ExtractApp'
# will match 20200803
date_string = self.input_filename.split('/')[-1][0:8]
try:
self.last_historical_data_date_in_file = datetime.strptime(date_string, "%Y%m%d")
except ValueError:
raise Exception("couldn't find a date pattern in filename. filename should be \
like 20200803ExtractApp.csv")
count = 0
statements = []
something_new = False
query = """
INSERT into %s(
siret,
hiring_date,
contract_type,
departement,
contract_duration,
iiann,
tranche_age,
handicap_label,
duree_pec,
date_insertion
)
values(%%s, %%s, %%s, %%s, %%s, %%s, %%s, %%s, %%s, %%s)
""" % settings.HIRING_TABLE
imported_alternance_contracts = 0
imported_alternance_contracts_distribution = {}
not_imported_alternance_contracts = 0
last_historical_data_date_in_db = db_session.query(func.max(Hiring.hiring_date))\
.filter(Hiring.contract_type == self.contract_type).first()[0]
logger.info("will now extract all alternance contracts with hiring_date between %s and %s",
last_historical_data_date_in_db, self.last_historical_data_date_in_file)
with import_util.get_reader(self.input_filename) as myfile:
con, cur = import_util.create_cursor()
header_line = myfile.readline().strip() # FIXME detect column positions from header
if b"SIRET" not in header_line:
logger.debug(header_line)
raise Exception("wrong header line")
for line in myfile:
line = line.decode()
count += 1
if not count % 10000:
logger.debug("reading line %i", count)
try:
try:
cur.executemany(query, statements)
except OperationalError: # retry once in case of deadlock error
time.sleep(10)
cur.executemany(query, statements)
statements = []
con.commit()
something_new = True
except:
logger.error("error in executing statement into hirings table: %s", sys.exc_info()[1])
statements = []
raise
try:
siret, hiring_date, departement = parse_alternance_line(line)
except InvalidRowException:
logger.info("invalid_row met at row: %i", count)
self.invalid_row_errors += 1
continue
except InvalidSiretException:
error_message = traceback.format_exc()
logger.info("invalid siret met at row: %i", count)
logger.info(error_message)
self.invalid_siret_errors += 1
continue
except InvalidZipCodeException:
logger.info("invalid zip code met at row: %i", count)
self.invalid_zipcode_errors += 1
continue
# This part of code is useless :
# The data used has a lot of late contracts inputs
# So we have to insert ALL the contracts from different dates
# alternance_contract_should_be_imported = (
# hiring_date > last_historical_data_date_in_db
# and hiring_date <= self.last_historical_data_date_in_file
#)
if hiring_date <= self.last_historical_data_date_in_file:
statement = (
siret,
hiring_date,
self.contract_type,
departement,
None, #contract_duration
None, #iiann
None, #tranche_age
None, #handicap_label
None, #duree_pec
date_insertion
)
statements.append(statement)
imported_alternance_contracts += 1
if hiring_date.year not in imported_alternance_contracts_distribution:
imported_alternance_contracts_distribution[hiring_date.year] = {}
if hiring_date.month not in imported_alternance_contracts_distribution[hiring_date.year]:
imported_alternance_contracts_distribution[hiring_date.year][hiring_date.month] = {}
if hiring_date.day not in imported_alternance_contracts_distribution[hiring_date.year][hiring_date.month]:
imported_alternance_contracts_distribution[hiring_date.year][hiring_date.month][hiring_date.day] = 0
imported_alternance_contracts_distribution[hiring_date.year][hiring_date.month][hiring_date.day] += 1
# run remaining statements
try:
cur.executemany(query, statements)
something_new = True
except:
logger.error("error in executing statement into hirings table: %s", sys.exc_info()[1])
raise
logger.info(f"Types de contrats à importer : {self.contract_name}")
logger.info(f"processed {count} lba_contracts...")
logger.info(f"imported lba_contracts: {imported_alternance_contracts}")
logger.info(f"not imported lba_contracts: {not_imported_alternance_contracts}")
logger.info(f"zipcode errors: {self.invalid_zipcode_errors}")
logger.info(f"invalid_row errors: {self.invalid_row_errors}")
logger.info(f"invalid siret errors: {self.invalid_siret_errors}")
# if self.zipcode_errors > settings.MAXIMUM_ZIPCODE_ERRORS:
# raise IOError('too many zipcode errors')
# if self.invalid_row_errors > settings.MAXIMUM_INVALID_ROWS:
# raise IOError('too many invalid_row errors')
con.commit()
cur.close()
con.close()
try:
statistics = DpaeStatistics(
last_import=datetime.now(),
most_recent_data_date=self.last_historical_data_date_in_file,
file_type=self.file_type
)
db_session.add(statistics)
db_session.commit()
logger.info("First way to insert DPAE statistics in DB : OK")
except OperationalError:
# For an obscure reason, the DpaeStatistics way to insert does not work on the bonaparte server
# So we insert it directly via an SQL query
# This job has been broken for more than a year, only way to fix it :
db_session.rollback()
last_import_date = datetime.now().strftime('%Y-%m-%d %H:%M:%S')
most_recent_date = self.last_historical_data_date_in_file.strftime('%Y-%m-%d %H:%M:%S')
query = f"insert into dpae_statistics (last_import, most_recent_data_date, file_type) values ('{last_import_date}','{most_recent_date}','{self.file_type}')"
con, cur = import_util.create_cursor()
cur.execute(query)
con.commit()
cur.close()
con.close()
logger.info("Second way to insert DPAE statistics in DB : OK")
logger.info("finished importing dpae...")
return something_new
def run_task(self):
logger.info("extracting %s ", self.input_filename)
date_pattern = ".*_(\d\d\d\d\d\d\d\d)"
date_match = re.match(date_pattern, self.input_filename)
if date_match:
date_part = date_match.groups()[-1]
self.most_recent_data_date = datetime.strptime(date_part, "%Y%m%d")
logger.debug("identified most_recent_data_date=%s" %
self.most_recent_data_date)
else:
raise Exception(
"couldn't find a date pattern in filename. filename should be dpae_XYZ_20xxxxxx.tar.gz"
)
count = 0
statements = []
something_new = False
query = "INSERT into %s(siret, hiring_date, zipcode, contract_type, departement, contract_duration, iiann, tranche_age, handicap_label) values(%%s, %%s, %%s, %%s, %%s, %%s, %%s, %%s, %%s)" % settings.DPAE_TABLE
imported_dpae = 0
imported_dpae_distribution = {}
not_imported_dpae = 0
initial_most_recent_data_date = DpaeStatistics.get_most_recent_data_date(
)
logger.info(
"will now extract all dpae with hiring_date between %s and %s" %
(initial_most_recent_data_date, self.most_recent_data_date))
with import_util.get_reader(self.input_filename) as myfile:
con, cur = import_util.create_cursor()
header_line = myfile.readline().strip()
if "siret" not in header_line:
logger.debug(header_line)
raise Exception("wrong header line")
for line in myfile:
count += 1
if not count % 100000:
logger.debug("reading line %i", count)
try:
try:
cur.executemany(query, statements)
except OperationalError: # retry once in case of deadlock error
time.sleep(10)
cur.executemany(query, statements)
statements = []
con.commit()
something_new = True
except:
logger.error(
"error in executing statement into dpae table: %s",
sys.exc_info()[1])
statements = []
raise
try:
siret, hiring_date, zipcode, contract_type, departement, contract_duration, iiann, tranche_age, handicap_label = parse_dpae_line(
line)
except ValueError:
self.zipcode_errors += 1
continue
except DepartementException:
self.zipcode_errors += 1
continue
except TooFewFieldsException:
logger.info("invalid_row met at row: %i", count)
self.invalid_row_errors += 1
continue
if hiring_date > initial_most_recent_data_date and hiring_date <= self.most_recent_data_date:
statement = (siret, hiring_date, zipcode, contract_type,
departement, contract_duration, iiann,
tranche_age, handicap_label)
statements.append(statement)
imported_dpae += 1
if hiring_date.year not in imported_dpae_distribution:
imported_dpae_distribution[hiring_date.year] = {}
if hiring_date.month not in imported_dpae_distribution[
hiring_date.year]:
imported_dpae_distribution[hiring_date.year][
hiring_date.month] = {}
if hiring_date.day not in imported_dpae_distribution[
hiring_date.year][hiring_date.month]:
imported_dpae_distribution[hiring_date.year][
hiring_date.month][hiring_date.day] = 0
imported_dpae_distribution[hiring_date.year][
hiring_date.month][hiring_date.day] += 1
else:
not_imported_dpae += 1
# run remaining statements
try:
cur.executemany(query, statements)
something_new = True
except:
logger.error("error in executing statement into dpae table: %s",
sys.exc_info()[1])
raise
logger.info("processed %i dpae...", count)
logger.info("imported dpae: %i", imported_dpae)
logger.info("not imported dpae: %i", not_imported_dpae)
logger.info("zipcode errors: %i", self.zipcode_errors)
logger.info("invalid_row errors: %i", self.invalid_row_errors)
if self.zipcode_errors >= 100:
raise Exception('too many zipcode errors')
if self.invalid_row_errors >= 100:
raise Exception('too many invalid_row errors')
statistics = DpaeStatistics(
last_import=datetime.now(),
most_recent_data_date=self.most_recent_data_date)
statistics.save()
con.commit()
logger.info("finished importing dpae...")
return something_new