def _pers_loop(session, fx, obs, index, data_start, data_end, issue_times):
load_data = _preload_load_data(session, obs, data_start, data_end)
out = defaultdict(list)
logger.info('Making persistence forecast for %s:%s from %s to %s', fx.name,
fx.forecast_id, issue_times[0], issue_times[-1])
for issue_time in issue_times:
run_time = issue_time
try:
fx_out = run_persistence(session,
obs,
fx,
run_time,
issue_time,
index=index,
load_data=load_data)
except ValueError as e:
logger.error('Unable to generate persistence forecast: %s', e)
else:
if hasattr(fx, 'constant_values'):
cv_ids = [f.forecast_id for f in fx.constant_values]
for id_, fx_ser in zip(cv_ids, fx_out):
out[id_].append(fx_ser)
else:
out[fx.forecast_id].append(fx_out)
for id_, serlist in out.items():
if len(serlist) > 0:
ser = pd.concat(serlist)
for cser in generate_continuous_chunks(ser, fx.interval_length):
if type(fx) == datamodel.Forecast:
session.post_forecast_values(id_, cser)
else:
session.post_probabilistic_forecast_constant_value_values(
id_, cser)
def test_generate_continuous_chunks(data, freq, expected):
lg = list(utils.generate_continuous_chunks(data, freq))
assert len(lg) == len(expected)
for i, g in enumerate(lg):
if isinstance(expected[i], pd.DataFrame):
pdt.assert_frame_equal(expected[i], g, check_column_type=False)
else:
pdt.assert_series_equal(expected[i], g)
def make_latest_probabilistic_persistence_forecasts(
token, max_run_time, base_url=None):
"""Make all reference probabilistic persistence forecasts that need to
be made up to *max_run_time*.
Parameters
----------
token : str
Access token for the API
max_run_time : pandas.Timestamp
Last possible run time of the forecast generation
base_url : str or None, default None
Alternate base_url of the API
"""
session = api.APISession(token, base_url=base_url)
forecasts = session.list_probabilistic_forecasts()
observations = session.list_observations()
params = generate_reference_persistence_forecast_parameters(
session, forecasts, observations, max_run_time)
for fx, obs, index, data_start, issue_times in params:
load_data = _preload_load_data(session, obs, data_start, max_run_time)
out = defaultdict(list)
logger.info('Making persistence forecast for %s:%s from %s to %s',
fx.name, fx.forecast_id, issue_times[0], issue_times[-1])
for issue_time in issue_times:
run_time = issue_time
try:
fx_list = run_persistence(
session, obs, fx, run_time, issue_time,
index=index, load_data=load_data)
except ValueError as e:
logger.error('Unable to generate persistence forecast: %s', e)
else:
# api requires a post per constant value
cv_ids = [f.forecast_id for f in fx.constant_values]
for id_, fx_ser in zip(cv_ids, fx_list):
out[id_].append(fx_ser)
for id_, serlist in out.items():
if len(serlist) > 0:
ser = pd.concat(serlist)
for cser in generate_continuous_chunks(
ser, fx.interval_length):
session.post_probabilistic_forecast_constant_value_values(
id_, cser)
def test_generate_continuous_chunks_errs(data, freq, err):
with pytest.raises(err):
list(utils.generate_continuous_chunks(data, freq))