def create_run_result_file_list(config: PlotCrossValidationConfig,
folder: str) -> List[RunResultFiles]:
"""
Creates a list of input files for cross validation analysis, from files stored inside of the test data folder.
:param config: The overall cross validation config
:param folder: The folder to read from, inside of test_data/plot_cross_validation.
:return:
"""
full_folder = full_ml_test_data_path("plot_cross_validation") / folder
files: List[RunResultFiles] = []
previous_dataset_file = None
for split in ["0", "1"]:
for mode in config.execution_modes_to_download():
metrics_file = full_folder / split / mode.value / METRICS_FILE_NAME
dataset_file: Optional[
Path] = full_folder / split / DATASET_CSV_FILE_NAME
if dataset_file.exists(): # type: ignore
# Reduce amount of checked-in large files. dataset files can be large, and usually duplicate across
# runs. Store only a copy in split 0, re-use in split 1.
previous_dataset_file = dataset_file
else:
dataset_file = previous_dataset_file
if metrics_file.exists():
file = RunResultFiles(
execution_mode=mode,
metrics_file=metrics_file,
dataset_csv_file=dataset_file,
run_recovery_id=config.run_recovery_id + "_" +
split, # type: ignore
split_index=split)
files.append(file)
return files
def test_metrics_preparation_for_segmentation(test_config_ensemble: PlotCrossValidationConfig) -> None:
"""
Test if metrics dataframes can be loaded and prepared. The files in question are checked in, but
were downloaded from a run, ID given in DEFAULT_ENSEMBLE_RUN_RECOVERY_ID.
"""
files = create_file_list_for_segmentation_recovery_run(test_config_ensemble)
downloaded_metrics = load_dataframes(files, test_config_ensemble)
for mode in test_config_ensemble.execution_modes_to_download():
expected_df = _get_metrics_df(mode)
# Drop the "mode" column, because that was added after creating the test data
metrics = downloaded_metrics[mode]
assert metrics is not None
actual_df = metrics.drop(COL_MODE, axis=1)
actual_df = actual_df.sort_values(list(actual_df.columns), ascending=True).reset_index(drop=True)
pd.testing.assert_frame_equal(expected_df, actual_df, check_like=True, check_dtype=False)
def test_metrics_preparation_for_segmentation(
drop_column: Optional[str], test_config: PlotCrossValidationConfig,
test_output_dirs: OutputFolderForTests) -> None:
"""
Test if metrics dataframes can be loaded and prepared. The files in question are checked in, but
were downloaded from a run, ID given in DEFAULT_ENSEMBLE_RUN_RECOVERY_ID.
Additionally test that CSV_INSTITUTION_HEADER or CSV_SERIES_HEADER can be dropped from the dataset_csv_file.
"""
files = create_file_list_for_segmentation_recovery_run(test_config)
if drop_column:
def drop_csv_column(path: Path) -> None:
"""
Load a csv file, drop a column, and save the csv file.
:param path: Path to csv file.
"""
df = pd.read_csv(path)
dropped_df = df.drop(drop_column, axis=1)
dropped_df.to_csv(path)
files = copy_run_result_files(files, full_ml_test_data_path(),
test_output_dirs.root_dir,
drop_csv_column)
downloaded_metrics = load_dataframes(files, test_config)
assert test_config.run_recovery_id
for mode in test_config.execution_modes_to_download():
expected_df = _get_metrics_df(test_config.run_recovery_id, mode)
if drop_column:
# If dropped a column from dataset_csv_file, remove it from expected dataframe.
expected_df[drop_column] = ''
# Drop the "mode" column, because that was added after creating the test data
metrics = downloaded_metrics[mode]
assert metrics is not None
actual_df = metrics.drop(COL_MODE, axis=1)
actual_df = actual_df.sort_values(
list(actual_df.columns), ascending=True).reset_index(drop=True)
pd.testing.assert_frame_equal(expected_df,
actual_df,
check_like=True,
check_dtype=False)