Function to write metadata tsv files in Genfi-to-BIDS could be split and tested

[AliceJoubert]

In Genfi-to-BIDS, writing the different metadata types of files is done in only one function :

clinica/clinica/converters/genfi_to_bids/_utils.py

Lines 783 to 873 in 4247631

	def write_bids(
	to: Path,
	participants: DataFrame,
	sessions: DataFrame,
	scans: DataFrame,
	) -> None:
	"""This function writes the BIDS
	Parameters
	----------
	to: Path
	The path where the BIDS should be written
	participants: DataFrame
	DataFrame containing the data for the participants.tsv
	sessions: DataFrame
	DataFrame containing the data for the sessions.tsv
	scans: DataFrame
	DataFrame containing the data for the scans.tsv
	"""
	import os
	from fsspec.implementations.local import LocalFileSystem
	from clinica.converters._utils import run_dcm2niix, write_to_tsv
	from clinica.dataset import BIDSDatasetDescription
	from clinica.utils.stream import cprint
	cprint("Starting to write the BIDS.", lvl="info")
	fs = LocalFileSystem(auto_mkdir=True)
	# Ensure BIDS hierarchy is written first.
	participants = (
	participants.reset_index()
	.drop(["session_id", "modality", "run_num", "bids_filename", "source"], axis=1)
	.drop_duplicates()
	.set_index("participant_id")
	)
	with fs.transaction:
	with fs.open(to / "dataset_description.json", "w") as dataset_description_file:
	BIDSDatasetDescription(name="GENFI").write(to=dataset_description_file)
	with fs.open(to / "participants.tsv", "w") as participant_file:
	write_to_tsv(participants, participant_file)
	for participant_id, data_frame in sessions.groupby("participant_id"):
	sessions = data_frame.droplevel(
	["participant_id", "modality", "bids_filename", "run_num"]
	).drop_duplicates()
	sessions_filepath = to / str(participant_id) / f"{participant_id}_sessions.tsv"
	with fs.open(sessions_filepath, "w") as sessions_file:
	write_to_tsv(sessions, sessions_file)
	scans = scans.reset_index().set_index(["bids_full_path"], verify_integrity=True)
	for bids_full_path, metadata in scans.iterrows():
	bids_full_path = Path(bids_full_path)
	try:
	os.makedirs(to / bids_full_path.parent)
	except OSError:
	pass
	dcm2niix_success = run_dcm2niix(
	Path(metadata["source_path"]).parent,
	to / Path(bids_full_path).parent,
	metadata["bids_filename"],
	True,
	)
	if dcm2niix_success:
	scans_filepath = (
	to
	/ str(metadata.participant_id)
	/ str(metadata.session_id)
	/ f"{metadata.participant_id}_{metadata.session_id}_scan.tsv"
	)
	row_to_write = _serialize_row(
	metadata.drop(["participant_id", "session_id"]),
	write_column_names=not scans_filepath.exists(),
	)
	with open(scans_filepath, "a") as scans_file:
	scans_file.write(f"{row_to_write}\n")
	if (
	"dwi" in metadata["bids_filename"]
	and "Philips" in metadata.manufacturer
	):
	_merge_philips_diffusion(
	to / bids_full_path.with_suffix(".json"),
	metadata.number_of_parts,
	metadata.run_num,
	)
	_correct_fieldmaps_name(to)
	_delete_real_and_imaginary_files(to)

Rather than having one big function which is not tested, it would be better to use this one as an interface to three separate functions specific to one type of file to test them more easily.