def test_happy_path(self):
settings.SCORE_COEFFICIENT_OF_VARIATION_MAX = 1.0
settings.HIGH_SCORE_COMPANIES_COUNT_MIN = 0
make_office()
make_dpae()
dpae_date = datetime.now()
departement = "57"
compute_score.run(settings.OFFICE_TABLE, settings.DPAE_TABLE,
departement, dpae_date)
def test_happy_path(self):
make_offices()
make_hirings()
departement = "57"
prediction_beginning_date = get_prediction_beginning_date()
result = compute_score.run(departement, prediction_beginning_date)
self.assertEqual(result, True) # successful computation
def test_unhappy_path(self):
make_offices()
make_hirings()
departement = "58"
prediction_beginning_date = get_prediction_beginning_date()
result = compute_score.run(departement, prediction_beginning_date)
self.assertEqual(
result, False) # failed computation (no data for this departement)
def compute(etab, dpae, departement, dpae_date):
try:
result = compute_score.run(etab, dpae, departement, dpae_date)
logger.info("finished compute_score.run (%s)", departement)
except:
logger.error("error in departement %s : %s", departement,
sys.exc_info()[1])
logger.error("compute_score traceback: %s", traceback.format_exc())
result = None
return result
def compute(departement):
result = compute_score.run(departement)
logger.info("finished compute_score.run (%s)", departement)
return result
def test_happy_path_investigation(self):
make_offices()
make_hirings()
departement = "57"
prediction_beginning_date = get_prediction_beginning_date()
df_etab = compute_score.run(departement,
prediction_beginning_date,
return_df_etab_if_successful=True)
columns = df_etab.columns.values
self.assertEqual(len(df_etab), OFFICES_HAVING_HIRINGS)
# The realistic (past) situation we simulate here is the following:
# Today is 2012 Jan 10th (as we usually run the importer each 10 of the month)
# thus the beggining of prediction is 2012 Jan 1st (1st of current month since we are
# in the first half of the current month),
# note that DPAE ends at 2011 Dec 31th,
# and that last alternance data is at 2011 Aug 31th.
# There is gap of several months for the alternance data, and this is what happens
# in real life as of now :/
self.assertEqual(get_dpae_last_historical_data_date(),
datetime(2011, 12, 31))
#self.assertEqual(importer_settings.ALTERNANCE_LAST_HISTORICAL_DATA_DATE, datetime(2011, 8, 31))
self.assertEqual(prediction_beginning_date,
datetime(2012, 1, 1)) # for both DPAE and Alternance
# --- DPAE/LBB checks
# we should have exactly 5 years of hirings including the last month (2011-12)
self.assertNotIn('dpae-2006-12', columns)
self.assertIn('dpae-2007-1', columns)
self.assertIn('dpae-2011-12', columns)
self.assertNotIn('dpae-2012-1', columns)
# Reminder: for DPAE 1 period = 6 months.
# We expect 7+2+2=11 (past) periods to be computed, here is why:
# LIVE set is based on 7 last periods (i.e. number of features fed to the model).
# TEST set is like LIVE set slided 12 months earlier,
# and thus ignores last 2 periods and is based on 7 periods before that.
# TRAIN set is like LIVE set slided 24 months earlier,
# and thus ignores last 4 periods and is based on 7 periods before that.
self.assertNotIn('dpae-period-0', columns)
self.assertIn('dpae-period-1', columns)
self.assertIn('dpae-period-11', columns)
self.assertNotIn('dpae-period-12', columns)
# final score columns
self.assertIn('score', columns)
self.assertIn('score_regr', columns)
# --- Alternance/LBA checks
# we should have exactly 5 years of hirings including the last month (2011-12)
self.assertNotIn('alt-2006-12', columns)
self.assertIn('alt-2007-1', columns)
self.assertIn('alt-2011-12', columns)
self.assertNotIn('alt-2012-1', columns)
# Reminder: for Alternance 1 period = 6 months.
# We expect 7+2+2+1=12 (past) periods to be computed, here is why:
# There is a gap of 4 months between the last Alternance data and today: 2011-9,10,11,12
# rounded up to a 1 period data gap.
# LIVE set is based on the 7 last periods (i.e. number of features fed to the model)
# before the 1 period of the data gap.
# TEST set is like LIVE set slided 12 months earlier,
# and thus ignores 2 more periods and is based on 7 periods before that.
# TRAIN set is like LIVE set slided 24 months earlier,
# and thus ignores 4 more periods and is based on 7 periods before that.
self.assertNotIn('alt-period-0', columns)
self.assertIn('alt-period-1', columns)
self.assertIn('alt-period-11', columns)
self.assertNotIn('alt-period-12', columns)
# final score columns
self.assertIn('score_alternance', columns)
self.assertIn('score_alternance_regr', columns)
