def test_remove_short_tracks_min_length3(self):
"""Ensures correct output from remove_short_tracks.
In this case, minimum track length is 3 storm objects.
"""
this_storm_object_table = best_tracks.remove_short_tracks(
MAIN_STORM_OBJECT_TABLE, min_objects_in_track=3)
self.assertTrue(
numpy.array_equal(
this_storm_object_table[tracking_io.STORM_ID_COLUMN].values,
STORM_OBJECT_TABLE_TRACK_LENGTH_GEQ3[
tracking_io.STORM_ID_COLUMN].values))
self.assertTrue(
numpy.array_equal(
this_storm_object_table[tracking_io.TIME_COLUMN].values,
STORM_OBJECT_TABLE_TRACK_LENGTH_GEQ3[
tracking_io.TIME_COLUMN].values))
def run_best_track(
smart_file_dict=None,
max_extrap_time_for_breakup_sec=best_tracks.
DEFAULT_MAX_EXTRAP_TIME_SEC,
max_prediction_error_for_breakup_metres=best_tracks.
DEFAULT_MAX_PREDICTION_ERROR_METRES,
max_join_time_sec=best_tracks.DEFAULT_MAX_JOIN_TIME_SEC,
max_join_distance_metres=best_tracks.DEFAULT_MAX_JOIN_DISTANCE_METRES,
max_mean_join_error_metres=best_tracks.
DEFAULT_MAX_MEAN_JOIN_ERROR_METRES,
max_velocity_diff_for_join_m_s01=None,
num_main_iters=best_tracks.DEFAULT_NUM_MAIN_ITERS,
num_breakup_iters=best_tracks.DEFAULT_NUM_BREAKUP_ITERS,
min_objects_in_track=best_tracks.DEFAULT_MIN_OBJECTS_IN_TRACK):
"""Runs the full w2besttrack algorithm with smart IO.
:param smart_file_dict: Dictionary created by find_files_for_smart_io.
:param max_extrap_time_for_breakup_sec: See documentation for
`best_tracks.run_best_track`.
:param max_prediction_error_for_breakup_metres: See doc for
`best_tracks.run_best_track`.
:param max_join_time_sec: See doc for `best_tracks.run_best_track`.
:param max_join_distance_metres: See doc for `best_tracks.run_best_track`.
:param max_mean_join_error_metres: See doc for `best_tracks.run_best_track`.
:param max_velocity_diff_for_join_m_s01: See doc for `best_tracks.run_best_track`.
:param num_main_iters: See doc for `best_tracks.run_best_track`.
:param num_breakup_iters: See doc for `best_tracks.run_best_track`.
:param min_objects_in_track: See doc for
`best_tracks.run_best_track`.
"""
best_tracks.check_best_track_params(
max_extrap_time_for_breakup_sec=max_extrap_time_for_breakup_sec,
max_prediction_error_for_breakup_metres=
max_prediction_error_for_breakup_metres,
max_join_time_sec=max_join_time_sec,
max_join_distance_metres=max_join_distance_metres,
max_mean_join_error_metres=max_mean_join_error_metres,
max_velocity_diff_for_join_m_s01=max_velocity_diff_for_join_m_s01,
num_main_iters=num_main_iters,
num_breakup_iters=num_breakup_iters,
min_objects_in_track=min_objects_in_track)
spc_dates_unix_sec = smart_file_dict[SPC_DATES_KEY]
num_spc_dates = len(spc_dates_unix_sec)
storm_object_table = None
for i in range(num_main_iters):
print('Starting main iteration ' + str(i + 1) + '/' +
str(num_main_iters) + '...\n\n')
for j in range(num_breakup_iters):
print('Starting break-up iteration ' + str(j + 1) + '/' +
str(num_breakup_iters) + '...\n\n')
for k in range(num_spc_dates):
storm_object_table = _shuffle_data_with_smart_io(
storm_object_table=storm_object_table,
file_dict=smart_file_dict,
working_spc_date_unix_sec=spc_dates_unix_sec[k],
read_from_intermediate=i > 0 or j > 0)
if k == 0:
best_track_start_time_unix_sec = numpy.min(
storm_object_table[tracking_io.TIME_COLUMN].values)
if k == num_spc_dates - 1:
best_track_end_time_unix_sec = numpy.max(
storm_object_table[tracking_io.TIME_COLUMN].values)
storm_track_table = best_tracks.storm_objects_to_tracks(
storm_object_table)
storm_track_table = best_tracks.theil_sen_fit_for_each_track(
storm_track_table)
these_working_indices = numpy.where(
storm_object_table[tracking_io.SPC_DATE_COLUMN].values ==
spc_dates_unix_sec[k])[0]
storm_object_table, storm_track_table = (
best_tracks.break_storm_tracks(
storm_object_table=storm_object_table,
storm_track_table=storm_track_table,
working_object_indices=these_working_indices,
max_extrapolation_time_sec=
max_extrap_time_for_breakup_sec,
max_prediction_error_metres=
max_prediction_error_for_breakup_metres))
for k in range(num_spc_dates):
storm_object_table = _shuffle_data_with_smart_io(
storm_object_table=storm_object_table,
file_dict=smart_file_dict,
working_spc_date_unix_sec=spc_dates_unix_sec[k],
read_from_intermediate=True)
storm_track_table = best_tracks.storm_objects_to_tracks(
storm_object_table)
storm_track_table = best_tracks.theil_sen_fit_for_each_track(
storm_track_table)
these_working_indices = _find_tracks_with_spc_date(
storm_object_table,
storm_track_table,
spc_date_unix_sec=spc_dates_unix_sec[k])
storm_object_table, storm_track_table = (
best_tracks.merge_storm_tracks(
storm_object_table=storm_object_table,
storm_track_table=storm_track_table,
working_track_indices=these_working_indices,
max_join_time_sec=max_join_time_sec,
max_join_distance_metres=max_join_distance_metres,
max_mean_prediction_error_metres=max_mean_join_error_metres,
max_velocity_diff_m_s01=max_velocity_diff_for_join_m_s01))
for k in range(num_spc_dates):
storm_object_table = _shuffle_data_with_smart_io(
storm_object_table=storm_object_table,
file_dict=smart_file_dict,
working_spc_date_unix_sec=spc_dates_unix_sec[k],
read_from_intermediate=True)
storm_track_table = best_tracks.storm_objects_to_tracks(
storm_object_table)
storm_track_table = best_tracks.theil_sen_fit_for_each_track(
storm_track_table)
these_working_indices = _find_tracks_with_spc_date(
storm_object_table,
storm_track_table,
spc_date_unix_sec=spc_dates_unix_sec[k])
storm_object_table, storm_track_table = (
best_tracks.break_ties_among_storm_objects(
storm_object_table,
storm_track_table,
working_track_indices=these_working_indices))
for k in range(num_spc_dates):
storm_object_table = _shuffle_data_with_smart_io(
storm_object_table=storm_object_table,
file_dict=smart_file_dict,
working_spc_date_unix_sec=spc_dates_unix_sec[k],
read_from_intermediate=True)
print('Removing storm tracks with < ' + str(min_objects_in_track) +
' objects...')
storm_object_table = best_tracks.remove_short_tracks(
storm_object_table, min_objects_in_track=min_objects_in_track)
print 'Recomputing storm attributes...'
storm_object_table = best_tracks.recompute_attributes(
storm_object_table,
best_track_start_time_unix_sec=best_track_start_time_unix_sec,
best_track_end_time_unix_sec=best_track_end_time_unix_sec)
this_spc_date_indices = numpy.where(storm_object_table[
tracking_io.SPC_DATE_COLUMN].values == spc_dates_unix_sec[k])[0]
best_tracks.write_output_storm_objects(
storm_object_table.iloc[this_spc_date_indices],
input_file_names=smart_file_dict[INPUT_FILE_NAMES_KEY][k],
output_file_names=smart_file_dict[OUTPUT_FILE_NAMES_KEY][k])
for k in range(num_spc_dates):
print 'Deleting temp files...'
os.remove(smart_file_dict[TEMP_FILE_NAMES_KEY][k])