add curryreader as full dep (#1475)

* fixups from #1430

* fix wrong orig_unit sometimes written to channels.tsv

* fix doctest (bad syntax; add verbose; avoid np types)

* add curryreader to full and test deps

* importorskip

Author: drammock

doc/install.rst

@@ -25,6 +25,7 @@ Optional:
 * ``matplotlib`` (>=3.6, for using the interactive data inspector)
 * ``pandas`` (>=1.3.2, for generating event statistics)
 * ``edfio`` (>=0.4.10, for writing EDF data)
+* ``curryreader`` (>=0.1.2, for reading Curry data)
 * ``defusedxml`` (for writing reading EGI MFF data and BrainVision montages)
 * ``filelock`` (for atomic file writing, and parallel processing support)
 

doc/whats_new.rst

@@ -27,6 +27,7 @@ The following authors contributed for the first time. Thank you so much! 🤩
 The following authors had contributed before. Thank you for sticking around! 🤘
 
 * `Stefan Appelhoff`_
+* `Daniel McCloy`_
 
 
 Detailed list of changes
@@ -44,8 +45,9 @@ Detailed list of changes
 🧐 API and behavior changes
 ^^^^^^^^^^^^^^^^^^^^^^^^^^^
 
--  `tracksys` accepted as argument in :class:`mne_bids.BIDSPath()` by `Julius Welzel`_ (:gh:`1430`)
+-  ``tracksys`` accepted as argument in :class:`mne_bids.BIDSPath()` by `Julius Welzel`_ (:gh:`1430`)
 - :func:`mne_bids.read_raw_bids()` has a new parameter ``on_ch_mismatch`` that controls behaviour when there is a mismatch between channel names in ``channels.tsv`` and the raw data; accepted values are ``'raise'`` (default), ``'reorder'``, and ``'rename'``, by `Kalle Mäkelä`_.
+- Column names read from ``.tsv`` are now converted from :class:`numpy.str_` class to built-in :class:`str` class, by `Daniel McCloy`_ (:gh:`1475`).
 
 🛠 Requirements
 ^^^^^^^^^^^^^^^
@@ -60,12 +62,14 @@ Detailed list of changes
 - Updated MEG/iEEG writers to satisfy the stricter checks in the latest BIDS validator releases: BTi/4D run folders now retain their ``.pdf`` suffix (falling back to the legacy naming when an older validator is detected), KIT marker files encode the run via the ``acq`` entity instead of ``run``, datasets lacking iEEG montages receive placeholder ``electrodes.tsv``/``coordsystem.json`` files, and the ``AssociatedEmptyRoom`` entry stores dataset-relative paths  by `Bruno Aristimunha`_ (:gh:`1449`)
 - Made the lock helpers skip reference counting when the optional ``filelock`` dependency is missing, preventing spurious ``AttributeError`` crashes during reads, by `Bruno Aristimunha`_ (:gh:`1469`)
 - Fixed a bug in :func:`mne_bids.read_raw_bids` that caused it to fail when reading BIDS datasets where the acquisition time was specified in local time rather than UTC only in Windows, by `Bruno Aristimunha`_ (:gh:`1452`)
+- Fixed bug in :func:`~mne_bids.write_raw_bids` where incorrect unit was sometimes written into ``channels.tsv`` file when converting data to BrainVision, EDF, or EEGLAB formats, by `Daniel McCloy`_ (:gh:`1475`)
 
 ⚕️ Code health
 ^^^^^^^^^^^^^^
 
 - Made :func:`mne_bids.copyfiles.copyfile_brainvision` output more meaningful error messages when encountering problematic files, by `Stefan Appelhoff`_ (:gh:`1444`)
 - Raised the minimum ``edfio`` requirement to ``0.4.10``,  eeglabio to ``0.1.0``  by `Bruno Aristimunha`_ (:gh:`1449`)
 - Relaxed EDF padding warnings in the test suite to accommodate upstream changes by `Bruno Aristimunha`_ (:gh:`1449`)
+- Adapt to upstream switch to new reader package for Neuroscan-Curry-format files, by `Daniel McCloy`_ (:gh:`1475`)
 
 :doc:`Find out what was new in previous releases <whats_new_previous_releases>`

mne_bids/config.py

@@ -10,6 +10,7 @@
 
 PYBV_VERSION = "0.7.3"
 EEGLABIO_VERSION = "0.0.2"
+CURRYREADER_VERSION = "0.1.2"
 
 DOI = """https://doi.org/10.21105/joss.01896"""
 

mne_bids/path.py

@@ -1766,7 +1766,8 @@ def get_entities_from_fname(fname, on_error="raise", verbose=None):
 'space': None, \
 'recording': None, \
 'split': None, \
-'description': None}
+'description': None, \
+'tracking_system': None}
     """
     if on_error not in ("warn", "raise", "ignore"):
         raise ValueError(

mne_bids/read.py

@@ -541,14 +541,16 @@ def _handle_info_reading(sidecar_fname, raw):
     return raw
 
 
-def events_file_to_annotation_kwargs(events_fname: str | Path) -> dict:
+@verbose
+def events_file_to_annotation_kwargs(events_fname: str | Path, *, verbose=None) -> dict:
     r"""
     Read the ``events.tsv`` file and extract onset, duration, and description.
 
     Parameters
     ----------
     events_fname : str
         The file path to the ``events.tsv`` file.
+    %(verbose)s
 
     Returns
     -------
@@ -592,8 +594,8 @@ def events_file_to_annotation_kwargs(events_fname: str | Path) -> dict:
     ...     'duration': [0.1, 0.1, 0.1],
     ...     'trial_type': ['event1', 'event2', 'event1'],
     ...     'value': [1, 2, 1],
-    ...     'sample': [10, 20, 30]
-            'foo': ['a', 'b', 'c'],
+    ...     'sample': [10, 20, 30],
+    ...     'foo': ['a', 'b', 'c'],
     ... }
     >>> df = pd.DataFrame(data)
     >>>
@@ -603,15 +605,11 @@ def events_file_to_annotation_kwargs(events_fname: str | Path) -> dict:
     >>> df.to_csv(events_file, sep='\t', index=False)
     >>>
     >>> # Read the events file using the function
-    >>> events_dict = events_file_to_annotation_kwargs(events_file)
+    >>> events_dict = events_file_to_annotation_kwargs(events_file, verbose=False)
     >>> events_dict
-    {'onset': array([0.1, 0.2, 0.3]),
-    'duration': array([0.1, 0.1, 0.1]),
-    'description': array(['event1', 'event2', 'event1'], dtype='<U6'),
-    'event_id': {'event1': 1, 'event2': 2},
-    'extras': [{'foo': 'a'}, {'foo': 'b'}, {'foo': 'c'}]}
+    {'onset': array([0.1, 0.2, 0.3]), 'duration': array([0.1, 0.1, 0.1]), 'description': array(['event1', 'event2', 'event1'], dtype='<U6'), 'event_id': {'event1': 1, 'event2': 2}, 'extras': [{'foo': 'a'}, {'foo': 'b'}, {'foo': 'c'}]}
 
-    """
+    """  # noqa E501
     logger.info(f"Reading events from {events_fname}.")
     events_dict = _from_tsv(events_fname)
 
@@ -677,9 +675,11 @@ def events_file_to_annotation_kwargs(events_fname: str | Path) -> dict:
                     culled_vals = culled_vals.astype(int)
                 except ValueError:  # numeric, but has some non-integer values
                     pass
+            # purge any np.str_, np.int_, or np.float_ types
+            culled_vals = np.asarray(culled_vals).tolist()
             event_id = dict(zip(culled[trial_type_col_name], culled_vals))
         else:
-            event_id = dict(zip(trial_types, np.arange(len(trial_types))))
+            event_id = dict(zip(trial_types, list(range(len(trial_types)))))
         descrs = np.asarray(trial_types, dtype=str)
 
     # convert onsets & durations to floats ("n/a" onsets were already dropped)

mne_bids/tests/test_read.py

@@ -635,7 +635,7 @@ def test_handle_events_reading(tmp_path, with_extras):
             match=re.escape(
                 "The version of MNE-Python you are using (<1.10) "
                 "does not support the extras argument in mne.Annotations. "
-                "The extra column(s) [np.str_('foo')] will be ignored."
+                "The extra column(s) ['foo'] will be ignored."
             ),
         )
         if with_extras and not check_version("mne", "1.10")

mne_bids/tests/test_write.py

@@ -47,6 +47,7 @@
 )
 from mne_bids.config import (
     BIDS_COORD_FRAME_DESCRIPTIONS,
+    CURRYREADER_VERSION,
     EEGLABIO_VERSION,
     PYBV_VERSION,
     REFERENCES,
@@ -3373,8 +3374,12 @@ def test_sidecar_encoding(_bids_validate, tmp_path):
 @testing.requires_testing_data
 def test_convert_eeg_formats(dir_name, fmt, fname, reader, tmp_path):
     """Test conversion of EEG/iEEG manufacturer fmt to BrainVision/EDF."""
-    pytest.importorskip("pybv", PYBV_VERSION)
-    pytest.importorskip("eeglabio", EEGLABIO_VERSION)
+    if dir_name == "BrainVision" or fmt == "BrainVision":
+        pytest.importorskip("pybv", PYBV_VERSION)
+    elif dir_name == "EEGLAB" or fmt == "EEGLAB":
+        pytest.importorskip("eeglabio", EEGLABIO_VERSION)
+    elif dir_name == "curry" or fmt == "curry":
+        pytest.importorskip("curryreader", CURRYREADER_VERSION)
     bids_root = tmp_path / fmt
     raw_fname = data_path / dir_name / fname
 
@@ -3431,7 +3436,7 @@ def test_convert_eeg_formats(dir_name, fmt, fname, reader, tmp_path):
     # load channels.tsv; the unit should be Volts
     channels_fname = bids_output_path.copy().update(suffix="channels", extension=".tsv")
     channels_tsv = _from_tsv(channels_fname)
-    assert channels_tsv["units"][0] == "V"
+    assert channels_tsv["units"][0] == "µV"
 
     if fmt == "BrainVision":
         assert Path(raw2.filenames[0]).suffix == ".eeg"

mne_bids/tsv_handler.py

@@ -158,7 +158,7 @@ def _from_tsv(fname, dtypes=None):
     # If data is 1-dimensional (only header), make it 2D
     data = np.atleast_2d(data)
 
-    column_names = data[0, :]
+    column_names = data[0, :].tolist()  # cast to list to avoid `np.str_()` keys in dict
     info = data[1:, :]
     data_dict = OrderedDict()
     if dtypes is None:

mne_bids/write.py

@@ -136,7 +136,7 @@ def _should_use_bti_pdf_suffix() -> bool:
     return use_pdf_suffix
 
 
-def _channels_tsv(raw, fname, overwrite=False):
+def _channels_tsv(raw, fname, *, convert_fmt, overwrite=False):
     """Create a channels.tsv file and save it.
 
     Parameters
@@ -145,6 +145,10 @@ def _channels_tsv(raw, fname, overwrite=False):
         The data as MNE-Python Raw object.
     fname : str | mne_bids.BIDSPath
         Filename to save the channels.tsv to.
+    convert_fmt : str | None
+        Which format the data are being converted to (determines what gets written in
+        the "unit" column). If ``None`` then we assume no conversion is happening (the
+        raw data files are being copied from the source tree into the BIDS folder tree).
     overwrite : bool
         Whether to overwrite the existing file.
         Defaults to False.
@@ -193,11 +197,32 @@ def _channels_tsv(raw, fname, overwrite=False):
         ch_type.append(map_chs[_channel_type])
         description.append(map_desc[_channel_type])
     low_cutoff, high_cutoff = (raw.info["highpass"], raw.info["lowpass"])
-    if raw._orig_units:
+    # If data are being *converted*, unit is determined by destination format:
+    #   - `eeglabio.raw.export_set` always converts V to µV, cf:
+    # https://github.com/jackz314/eeglabio/blob/3961bb29daf082767ea44e7c7d9da2df10971c37/eeglabio/raw.py#L57
+    #
+    #   - `mne.export._edf_bdf._export_raw_edf_bdf` always converts V to µV, cf:
+    # https://github.com/mne-tools/mne-python/blob/1b921f4af5154bad40202d87428a2583ef896a00/mne/export/_edf_bdf.py#L61-L63
+    #
+    #   - `pybv.write_brainvision` converts V to µV by default (and we don't alter that)
+    # https://github.com/bids-standard/pybv/blob/2832c80ee00d12990a8c79f12c843c0d4ddc825b/pybv/io.py#L40
+    # https://github.com/mne-tools/mne-bids/blob/1e0a96e132fc904ba856d42beaa9ddddb985f1ed/mne_bids/write.py#L1279-L1280
+    if convert_fmt:
+        volt_like = "µV" if convert_fmt in ("BrainVision", "EDF", "EEGLAB") else "V"
+        units = [
+            volt_like
+            if ch_i["unit"] == FIFF.FIFF_UNIT_V
+            else _unit2human.get(ch_i["unit"], "n/a")
+            for ch_i in raw.info["chs"]
+        ]
+    # if raw data is merely copied, check `raw._orig_units`
+    elif raw._orig_units:
         units = [raw._orig_units.get(ch, "n/a") for ch in raw.ch_names]
+    # If `raw._orig_units` is missing, assume SI units
     else:
         units = [_unit2human.get(ch_i["unit"], "n/a") for ch_i in raw.info["chs"]]
-        units = [u if u not in ["NA"] else "n/a" for u in units]
+    # fixup "NA" (from `_unit2human`) → "n/a"
+    units = [u if u not in ["NA"] else "n/a" for u in units]
 
     # Translate from MNE to BIDS unit naming
     for idx, mne_unit in enumerate(units):
@@ -2229,7 +2254,6 @@ def write_raw_bids(
         emptyroom_fname=associated_er_path,
         overwrite=overwrite,
     )
-    _channels_tsv(raw, channels_path.fpath, overwrite)
 
     # create parent directories if needed
     _mkdir_p(os.path.dirname(data_path))
@@ -2288,6 +2312,14 @@ def write_raw_bids(
                 f"for {datatype} datatype."
             )
 
+    # this can't happen until after value of `convert` has been determined
+    _channels_tsv(
+        raw,
+        channels_path.fpath,
+        convert_fmt=format if convert else None,
+        overwrite=overwrite,
+    )
+
     # raise error when trying to copy files (copyfile_*) into same location
     # (src == dest, see https://github.com/mne-tools/mne-bids/issues/867)
     if (

pyproject.toml

@@ -62,6 +62,7 @@ doc = [
 ]
 # Dependencies for using all mne_bids features
 full = [
+  "curryreader >= 0.1.2",
   "defusedxml",  # For reading EGI MFF data and BrainVision montages
   "edfio >= 0.4.10",
   "eeglabio >= 0.1.0",
@@ -74,7 +75,7 @@ full = [
   "pymatreader",
 ]
 # Dependencies for running the test infrastructure
-test = ["mne_bids[full]", "pytest", "pytest-cov", "pytest-rerunfailures", "pytest-sugar", "ruff"]
+test = ["curryreader", "mne_bids[full]", "pytest", "pytest-cov", "pytest-rerunfailures", "pytest-sugar", "ruff"]
 
 [project.urls]
 "Bug Tracker" = "https://github.com/mne-tools/mne-bids/issues/"