def test_run_candidate_channel_only(candidate_sim, candidate_channel_only_sim,
output_dir, ncores, capsys):
if candidate_sim.model.model_config.lower().find('nwm') < 0:
pytest.skip('Channel-only test only applicable to nwm_ana config')
with capsys.disabled():
print("\nQuestion: The candidate channel-only mode runs successfully?",
end='')
# Dont recompile the model, just use the candidate's model.
candidate_channel_only_sim.model = candidate_sim.model
# Set the forcing directory
# Set run directory and go for execution.
run_dir = output_dir / 'run_candidate_channel_only'
run_dir.mkdir(parents=True)
os.chdir(str(run_dir))
# Job
exe_command = ('mpirun -np {0} ./wrf_hydro.exe').format(str(ncores))
job = wrfhydropy.Job(job_id='run_candidate_channel_only',
exe_cmd=exe_command)
candidate_channel_only_sim.add(job)
candidate_channel_only_sim.jobs[0]._hrldas_namelist['noahlsm_offline']['indir'] = \
str(output_dir / 'run_candidate')
# Run
with warnings.catch_warnings():
warnings.simplefilter("ignore")
candidate_channel_only_sim.compose()
with capsys.disabled():
print('\nwaiting for job to complete...', end='')
candidate_channel_only_sim.run()
# Wait to collect until job has finished. All test runs are performed on a single job with
# job_id='test_job'
wait_job(candidate_channel_only_sim)
candidate_channel_only_sim.collect()
candidate_channel_only_sim.pickle(
run_dir.joinpath('WrfHydroSim_collected.pkl'))
# Check job run statuses
for job in candidate_channel_only_sim.jobs:
assert job.exit_status == 0, \
"Candidate channel-only code run exited with non-zero status"
def test_perfrestart_candidate_channel_only(output_dir):
candidate_channel_only_sim_file = \
output_dir / 'run_candidate_channel_only' / 'WrfHydroSim.pkl'
candidate_channel_only_collected_file = \
output_dir / 'run_candidate_channel_only' / 'WrfHydroSim_collected.pkl'
if candidate_channel_only_collected_file.is_file() is False:
pytest.skip(
'candidate_channel_only run object not found, skipping test')
print(
"\nQuestion: The candidate_channel_only outputs from a restart run match the outputs "
"from standard run?\n",
end='')
print('\n')
# Load initial run model object and copy
candidate_channel_only_sim = \
pickle.load(candidate_channel_only_sim_file.open(mode="rb"))
candidate_channel_only_sim_expected = \
pickle.load(candidate_channel_only_collected_file.open(mode="rb"))
candidate_channel_only_sim_restart = copy.deepcopy(
candidate_channel_only_sim)
# Set run directory
run_dir = output_dir / 'restart_candidate_channel_only'
run_dir.mkdir(parents=True)
os.chdir(str(run_dir))
# Get a new start time 1 hour later
restart_job = candidate_channel_only_sim_restart.jobs[0]
restart_job.model_start_time = restart_job.model_start_time + dt.timedelta(
hours=2)
# Get restart files from previous run and symlink into restart sim dir
# Hydro: Use actual time listed in meta data not filename or positional list index
for restart_file in candidate_channel_only_sim_expected.output.restart_hydro:
restart_time = restart_file.open().Restart_Time
restart_time = pd.to_datetime(restart_time, format='%Y-%m-%d_%H:%M:%S')
if restart_time == restart_job.model_start_time:
candidate_hydro_restart_file = pathlib.Path(restart_file.name)
candidate_hydro_restart_file.symlink_to(restart_file)
# Nudging: Use actual time listed in meta data not filename or positional list index
for restart_file in candidate_channel_only_sim_expected.output.restart_nudging:
restart_time = restart_file.open().modelTimeAtOutput
restart_time = pd.to_datetime(restart_time, format='%Y-%m-%d_%H:%M:%S')
if restart_time == restart_job.model_start_time:
candidate_nudging_restart_file = pathlib.Path(restart_file.name)
candidate_nudging_restart_file.symlink_to(restart_file)
# Compose and run
# Catch warnings related to missing start and end job times
with warnings.catch_warnings():
warnings.simplefilter("ignore")
candidate_channel_only_sim_restart.compose(force=True)
print('\nwaiting for job to complete...', end='')
candidate_channel_only_sim_restart.run()
# Wait to collect until job has finished. All test runs are performed on a single job with
wait_job(candidate_channel_only_sim_restart)
candidate_channel_only_sim_restart.collect()
candidate_channel_only_sim_restart.pickle(
run_dir.joinpath('WrfHydroSim_collected.pkl'))
# Check outputs
with warnings.catch_warnings():
warnings.simplefilter("ignore")
diffs = wrfhydropy.outputdiffs.OutputDataDiffs(
candidate_channel_only_sim_restart.output,
candidate_channel_only_sim_expected.output)
# Assert all diff values are 0 and print diff stats if not
has_diffs = any(value != 0 for value in diffs.diff_counts.values())
if has_diffs:
eprint(diffs.diff_counts)
for key, value in diffs.diff_counts.items():
if value != 0:
diffs = getattr(diffs, key)
eprint('\n' + key + '\n')
for diff in diffs:
eprint(diff)
assert has_diffs is False, \
'Outputs for candidate run do not match outputs from candidate restart run'
def test_ncores_candidate_channel_only(output_dir):
candidate_channel_only_sim_file = \
output_dir / 'run_candidate_channel_only' / 'WrfHydroSim.pkl'
candidate_channel_only_collected_file = \
output_dir / 'run_candidate_channel_only' / 'WrfHydroSim_collected.pkl'
if candidate_channel_only_collected_file.is_file() is False:
pytest.skip(
'candidate_channel_only collected run object not found, skipping test.'
)
print(
"\nQuestion: The candidate_channel-only output files from an ncores runmatch those "
"from an ncores-1 run?\n",
end='')
print('\n')
candidate_channel_only_sim = \
pickle.load(candidate_channel_only_sim_file.open("rb"))
candidate_channel_only_sim_expected = \
pickle.load(candidate_channel_only_collected_file.open("rb"))
candidate_channel_only_sim_ncores = copy.deepcopy(
candidate_channel_only_sim)
run_dir = output_dir / 'ncores_candidate_channel_only'
run_dir.mkdir(parents=True)
os.chdir(str(run_dir))
old_job = candidate_channel_only_sim.jobs[0]
new_job = wrfhydropy.Job(job_id='ncores_candidate',
exe_cmd=old_job._exe_cmd)
# Remove old job and add new job
candidate_channel_only_sim_ncores.jobs.pop(0)
candidate_channel_only_sim_ncores.add(new_job)
# Edit the sim object number of cores
if candidate_channel_only_sim_ncores.scheduler is not None:
candidate_channel_only_sim_ncores.scheduler.nproc = \
candidate_channel_only_sim_ncores.scheduler.nproc - 1
else:
orig_exe_cmd = candidate_channel_only_sim_ncores.jobs[0]._exe_cmd
orig_exe_cmd = orig_exe_cmd.replace('-np 2', '-np 1')
# Recompose into new directory and run
# catch warnings related to missing start and end job times
with warnings.catch_warnings():
warnings.simplefilter("ignore")
candidate_channel_only_sim_ncores.compose(force=True)
print('\nwaiting for job to complete...', end='')
candidate_channel_only_sim_ncores.run()
# Wait to collect until job has finished. All test runs are performed on a single job with
# job_id='test_job'
wait_job(candidate_channel_only_sim_ncores)
candidate_channel_only_sim_ncores.collect()
candidate_channel_only_sim_ncores.pickle(
run_dir.joinpath('WrfHydroSim_collected.pkl'))
# Check outputs
with warnings.catch_warnings():
warnings.simplefilter("ignore")
diffs = wrfhydropy.outputdiffs.OutputDataDiffs(
candidate_channel_only_sim_ncores.output,
candidate_channel_only_sim_expected.output)
# Assert all diff values are 0 and print diff stats if not
has_diffs = any(value != 0 for value in diffs.diff_counts.values())
if has_diffs:
eprint(diffs.diff_counts)
for key, value in diffs.diff_counts.items():
if value != 0:
diffs = getattr(diffs, key)
eprint('\n' + key + '\n')
for diff in diffs:
eprint(diff)
assert has_diffs is False, \
'Outputs for candidate_channel_only run with ncores do not match outputs with ncores-1'