What’s new?

Version 0.21 (unreleased)

👩🏽‍💻 Authors

The following authors contributed for the first time. Thank you so much! 🤩

• Hamza Abdelhedi

• Shubham Padkonde

The following authors had contributed before. Thank you for sticking around! 🤘

• Bruno Aristimunha

• Eric Larson

Detailed list of changes

🚀 Enhancements

• Expose MRI defacing function to public API as mne_bids.deface_mri() by Erica Peterson (#1544)

• Add an extra_fields parameter to make_dataset_description() to write additional top-level keys to dataset_description.json, by Bruno Aristimunha (#1674)

• Add mne_bids.open_lock(), the cross-process file lock mne-bids uses for shared files such as participants.tsv, so downstream writers can share it, by Bruno Aristimunha (#1675)

• mne_bids.read_epochs_bids() now reads epoched FIF files (RecordingType: "epoched") with mne.read_epochs(), and mne_bids.read_raw_bids() points to it for such files, by Hamza Abdelhedi (#1679)

• mne_bids.read_raw_bids() now accepts path-like inputs (str or pathlib.Path), converting them via mne_bids.get_bids_path_from_fname(), by Bruno Aristimunha (#1695)

• Add fnames parameter to mne_bids.find_matching_paths() to filter pre-collected file paths without walking the filesystem, by Bruno Aristimunha (#1698)

🧐 API and behavior changes

• Sidecar files are now written to a temporary file and moved into place, so reading them no longer takes a file lock, and a sidecar whose content would not change is no longer rewritten, by Hamza Abdelhedi (#1680)

🛠 Requirements

• None yet

🪲 Bug fixes

• mne_bids.read_raw_bids() no longer fails with an AttributeError when an events.json sidecar is not a JSON object (for example []), by Bruno Aristimunha (#1703)

• Fix readers seeing a partially written sidecar while another process was rewriting it, by Hamza Abdelhedi (#1680)

• Fix parallel mne_bids.write_raw_bids() calls for BTi/4D data overwriting each other’s .bidsignore changes, by Hamza Abdelhedi (#1680)

• find_matching_paths() and BIDSPath.match() now follow symbolic links to directories, and no longer return hidden files (or files inside hidden directories), which could happen when a dataset was searched without datatypes and without ignore_nosub, by Hamza Abdelhedi (#1678)

• Fix find_matching_paths() omitting files with a tracksys entity and returning no matches when tracking_systems is specified, by Shubham Padkonde.

• Allow acquisition="calibration" or "crosstalk" together with a task in BIDSPath when datatype is not "meg", since these labels are reserved only for MEG fine-calibration and crosstalk files, by Bruno Aristimunha (#1673)

• Fix find_matching_paths() and BIDSPath.match() raising an error when the dataset contains a file that cannot be represented as a BIDSPath, such as FreeSurfer output in derivatives; such files are now skipped, by Eric Larson (#1699)

• Fix mne_bids.get_entity_vals() with include_match returning no values when the dataset itself is inside a hidden directory, by Eric Larson (#1699)

• Fix mne_bids.get_bids_path_from_fname() raising an UnboundLocalError instead of a clear error for a scans.tsv file name without subject or session entities, by Eric Larson (#1699)

• mne_bids.read_raw_bids() no longer sets fNIRS channels whose channels.tsv type is NIRSCWHBO, NIRSCWHBR or NIRSCWOPTICALDENSITY to misc: they get the MNE types hbo, hbr and fnirs_od, and channels.tsv is written with these BIDS types for such channels instead of raising a KeyError, by Bruno Aristimunha (#1701)

• mne_bids.read_raw_bids() now matches fNIRS channels on the source, detector and wavelength_nominal columns of channels.tsv when their names differ from the names MNE gives SNIRF channels (for example S1-D1-690.0 in OpenNeuro ds004830), instead of raising a channel mismatch error, by Bruno Aristimunha (#1704)

⚕️ Code health

• mne_bids.read_raw_bids() now parses participants.tsv once per version of the file instead of on every call, so reading every recording of a large dataset no longer gets slower as the number of subjects grows, by Hamza Abdelhedi (#1681)

• Dropping rows from a TSV table no longer deep-copies the table first, by Hamza Abdelhedi (#1681)

• Sped up mne_bids.read_raw_bids() and mne_bids.write_raw_bids() on network filesystems by taking fewer file locks, by Hamza Abdelhedi (#1680)

• Improvements to directory searching that speed up BIDSPath.match(), find_matching_paths(), and read_raw_bids(), especially when searching network file systems, by Hamza Abdelhedi (#1678)

• Creating and updating BIDSPath objects is faster, which speeds up BIDSPath.match() and find_matching_paths() on datasets with many files, by Eric Larson (#1699)

Find out what was new in previous releases