Typed physiological archives (PR 3)

[MaximeBICMTL]

Builds on #1389 (diff)

Description

Move the "create physiological file archive" and "create physiological event archive" from the untyped Eeg class to a new lib.import_bids_dataset.archive typed module.

Bug fix

The old code contained this function, which returns a tuple with the relative and full path of the archive:

def get_archive_paths(self, archive_rel_name):
        package_path = get_eeg_pre_package_download_dir_path_config(self.env)
        if package_path:
            raw_package_dir = os.path.join(package_path, 'raw')
            os.makedirs(raw_package_dir, exist_ok=True)
            archive_rel_name = os.path.basename(archive_rel_name)
            archive_full_path = os.path.join(raw_package_dir, archive_rel_name)
        else:
            archive_full_path = os.path.join(self.data_dir, archive_rel_name)

        return (archive_rel_name, archive_full_path)

As can be seen, the relative path is relative to the ${package_path}/raw directory if the package_raw configuration is defined, and relative to the LORIS data directory otherwise. This is a bug, it means that the path of an archive cannot be interpreted without reading the LORIS configuration (although I am rather confident that only one of these two paths works in practice), and that changing setting/unsetting that configuration breaks all archive paths in the database.

This PR makes all paths relative to the LORIS data directory, which fixes this issue and is a common practice in the LORIS database but may be a breaking change for some projects.

Details

• Shorten DbPhysioFileArchive.file_path and DbPhysioEventArchive.file_path to DbPhysioFileArchive.path and DbPhysioEventArchive.path.
• Add a new typed loris_utils.archive module.
• Add a new lib.import_bids_dataset.physio_archive module.
• Adopt the aforementioned module in the BIDS EEG import pipeline.
• Fix the aforementioned bug.
• Improve the BIDS EEG import integration test with archive checking.
• Remove or deprecate the old code.
• Raise an exception if the archive file already exists instead of ignoring it.
• Remove a check to know if the file already has an archive. The file sent argument has just been created by the import pipeline, it cannot already have an archive.