In BIDSPath.update(), don't infer the datatype from the suffix

I needed to set datatype to None to make `find_emptyroom()`
deal with the inheritance principle correctly, but turns out it
never took effect because `BIDSPath.update()` would silently
replace my `datatype=None` with an inferred `datatype='meg'`.

doc/whats_new.rst

@@ -48,6 +48,8 @@ Detailed list of changes
 🧐 API and behavior changes
 ^^^^^^^^^^^^^^^^^^^^^^^^^^^
 
+- In many places, we used to infer the `datatype` of a :class:`~mne_bids.BIDSPath` from the `suffix`, if not explicitly provided. However, this has lead to trouble in certain edge cases. In an effort to reduce the amount of implicit behavior in MNE-BIDS, we now require users to explicitly specify a `datatype` whenever the invoked functions or methods expect one, by `Richard Höchenberger`_ (:gh:`1030`)
+
 - :func:`mne_bids.make_dataset_description` now accepts keyword arguments only, and can now also write the following metadata: ``HEDVersion``, ``EthicsApprovals``, ``GeneratedBy``, and ``SourceDatasets``, by `Stefan Appelhoff`_ (:gh:`406`)
 
 - The deprecated function ``mne_bids.mark_bad_channels`` has been removed in favor of :func:`mne_bids.mark_channels`, by `Richard Höchenberger`_ (:gh:`1009`)

mne_bids/path.py

@@ -22,7 +22,7 @@
 from mne_bids.config import (
     ALLOWED_PATH_ENTITIES, ALLOWED_FILENAME_EXTENSIONS,
     ALLOWED_FILENAME_SUFFIX, ALLOWED_PATH_ENTITIES_SHORT,
-    ALLOWED_DATATYPES, SUFFIX_TO_DATATYPE, ALLOWED_DATATYPE_EXTENSIONS,
+    ALLOWED_DATATYPES, ALLOWED_DATATYPE_EXTENSIONS,
     ALLOWED_SPACES,
     reader, ENTITY_VALUE_TYPE)
 from mne_bids.utils import (_check_key_val, _check_empty_room_basename,
@@ -51,10 +51,6 @@ def _find_matched_empty_room(bids_path):
                          'date set. Cannot get matching empty-room file.')
 
     ref_date = raw.info['meas_date']
-    if not isinstance(ref_date, datetime):  # pragma: no cover
-        # for MNE < v0.20
-        ref_date = datetime.fromtimestamp(raw.info['meas_date'][0])
-
     emptyroom_dir = BIDSPath(root=bids_root, subject='emptyroom').directory
 
     if not emptyroom_dir.exists():
@@ -95,6 +91,7 @@ def _find_matched_empty_room(bids_path):
         er_bids_path = get_bids_path_from_fname(er_fname, check=False)
         er_bids_path.subject = 'emptyroom'  # er subject entity is different
         er_bids_path.root = bids_root
+        er_bids_path.datatype = 'meg'
         er_meas_date = None
 
         # Try to extract date from filename.
@@ -229,7 +226,7 @@ class BIDSPath(object):
     Generate a BIDSPath object and inspect it
 
     >>> bids_path = BIDSPath(subject='test', session='two', task='mytask',
-    ...                      suffix='ieeg', extension='.edf')
+    ...                      suffix='ieeg', extension='.edf', datatype='ieeg')
     >>> print(bids_path.basename)
     sub-test_ses-two_task-mytask_ieeg.edf
     >>> bids_path
@@ -733,11 +730,6 @@ def update(self, *, check=None, **kwargs):
                 getattr(self, f'{key}') if hasattr(self, f'_{key}') else None
             setattr(self, f'_{key}', val)
 
-        # infer datatype if suffix is uniquely the datatype
-        if self.datatype is None and \
-                self.suffix in SUFFIX_TO_DATATYPE:
-            self._datatype = SUFFIX_TO_DATATYPE[self.suffix]
-
         # Perform a check of the entities and revert changes if check fails
         try:
             self._check()
@@ -899,7 +891,11 @@ def find_empty_room(self, use_sidecar_only=False, verbose=None):
                              'Please use `bids_path.update(root="<root>")` '
                              'to set the root of the BIDS folder to read.')
 
-        sidecar_fname = _find_matching_sidecar(self, extension='.json')
+        sidecar_fname = _find_matching_sidecar(
+            # needed to deal with inheritance principle
+            self.copy().update(datatype=None),
+            extension='.json'
+        )
         with open(sidecar_fname, 'r', encoding='utf-8') as f:
             sidecar_json = json.load(f)
 
@@ -1237,14 +1233,15 @@ def _parse_ext(raw_fname):
     return fname, ext
 
 
-def _infer_datatype_from_path(fname):
+def _infer_datatype_from_path(fname: Path):
     # get the parent
-    datatype = Path(fname).parent.name
-
-    if any([datatype.startswith(entity) for entity in ['sub', 'ses']]):
-        datatype = None
-
-    if not datatype:
+    if fname.exists():
+        datatype = fname.parent.name
+        if any([datatype.startswith(entity) for entity in ['sub', 'ses']]):
+            datatype = None
+    elif fname.stem.split('_')[-1] in ('meg', 'eeg', 'ieeg'):
+        datatype = fname.stem.split('_')[-1]
+    else:
         datatype = None
 
     return datatype

mne_bids/sidecar_updates.py

@@ -85,7 +85,8 @@ def update_sidecar_json(bids_path, entries, verbose=None):
     >>> from pathlib import Path
     >>> root = Path('./mne_bids/tests/data/tiny_bids').absolute()
     >>> bids_path = BIDSPath(subject='01', task='rest', session='eeg',
-    ...                      suffix='eeg', extension='.json', root=root)
+    ...                      suffix='eeg', extension='.json', datatype='eeg',
+    ...                      root=root)
     >>> entries = {'PowerLineFrequency': 60}
     >>> update_sidecar_json(bids_path, entries, verbose=False)
 

mne_bids/tests/test_path.py

@@ -260,10 +260,8 @@ def test_parse_ext():
 @pytest.mark.parametrize('fname', [
     'sub-01_ses-02_task-test_run-3_split-01_meg.fif',
     'sub-01_ses-02_task-test_run-3_split-01',
-    ('/bids_root/sub-01/ses-02/meg/' +
-     'sub-01_ses-02_task-test_run-3_split-01_meg.fif'),
-    ('sub-01/ses-02/meg/' +
-     'sub-01_ses-02_task-test_run-3_split-01_meg.fif')
+    '/bids_root/sub-01/ses-02/meg/sub-01_ses-02_task-test_run-3_split-01_meg.fif',  # noqa: E501
+    'sub-01/ses-02/meg/sub-01_ses-02_task-test_run-3_split-01_meg.fif'
 ])
 def test_get_bids_path_from_fname(fname):
     bids_path = get_bids_path_from_fname(fname)
@@ -598,7 +596,7 @@ def test_bids_path(return_bids_test_dir):
     # ... but raises an error with check=True
     match = r'space \(foo\) is not valid for datatype \(eeg\)'
     with pytest.raises(ValueError, match=match):
-        BIDSPath(subject=subject_id, space='foo', suffix='eeg')
+        BIDSPath(subject=subject_id, space='foo', suffix='eeg', datatype='eeg')
 
     # error check on space for datatypes that do not support space
     match = 'space entity is not valid for datatype anat'
@@ -612,7 +610,8 @@ def test_bids_path(return_bids_test_dir):
         bids_path_tmpcopy.update(space='CapTrak', check=True)
 
     # making a valid space update works
-    bids_path_tmpcopy.update(suffix='eeg', space="CapTrak", check=True)
+    bids_path_tmpcopy.update(suffix='eeg', datatype='eeg',
+                             space="CapTrak", check=True)
 
     # suffix won't be error checks if initial check was false
     bids_path.update(suffix=suffix)
@@ -628,7 +627,7 @@ def test_bids_path(return_bids_test_dir):
 
     # test repr
     bids_path = BIDSPath(subject='01', session='02',
-                         task='03', suffix='ieeg',
+                         task='03', suffix='ieeg', datatype='ieeg',
                          extension='.edf')
     assert repr(bids_path) == ('BIDSPath(\n'
                                'root: None\n'
@@ -680,7 +679,8 @@ def test_make_filenames():
     # All keys work
     prefix_data = dict(subject='one', session='two', task='three',
                        acquisition='four', run=1, processing='six',
-                       recording='seven', suffix='ieeg', extension='.json')
+                       recording='seven', suffix='ieeg', extension='.json',
+                       datatype='ieeg')
     expected_str = ('sub-one_ses-two_task-three_acq-four_run-01_proc-six_'
                     'rec-seven_ieeg.json')
     assert BIDSPath(**prefix_data).basename == expected_str
@@ -974,7 +974,7 @@ def test_find_emptyroom_ties(tmp_path):
                         'sample_audvis_trunc_raw.fif')
 
     bids_root = str(tmp_path)
-    bids_path = _bids_path.copy().update(root=bids_root)
+    bids_path = _bids_path.copy().update(root=bids_root, datatype='meg')
     session = '20010101'
     er_dir_path = BIDSPath(subject='emptyroom', session=session,
                            datatype='meg', root=bids_root)

mne_bids/tests/test_update.py

@@ -107,7 +107,7 @@ def test_update_sidecar_jsons(_get_bids_test_dir, _bids_validate,
                        ('SEEGChannelCount', None, 0)]
 
     # get the sidecar json
-    sidecar_path = bids_path.copy().update(extension='.json')
+    sidecar_path = bids_path.copy().update(extension='.json', datatype='meg')
     sidecar_fpath = sidecar_path.fpath
     with open(sidecar_fpath, 'r', encoding='utf-8') as fin:
         sidecar_json = json.load(fin)
@@ -212,11 +212,6 @@ def test_update_anat_landmarks(tmp_path):
     update_anat_landmarks(bids_path=bids_path_mri_no_ext,
                           landmarks=landmarks_new)
 
-    # Check without datatytpe provided
-    bids_path_mri_no_datatype = bids_path_mri.copy().update(datatype=None)
-    update_anat_landmarks(bids_path=bids_path_mri_no_datatype,
-                          landmarks=landmarks)
-
     # Check handling of invalid input
     bids_path_invalid = bids_path_mri.copy().update(datatype='meg')
     with pytest.raises(ValueError, match='Can only operate on "anat"'):

mne_bids/tests/test_write.py

@@ -446,7 +446,7 @@ def test_line_freq(line_freq, _bids_validate, tmp_path):
     _bids_validate(bids_root)
 
     eeg_json_fpath = (bids_path.copy()
-                      .update(suffix='eeg', extension='.json')
+                      .update(suffix='eeg', datatype='eeg', extension='.json')
                       .fpath)
     with open(eeg_json_fpath, 'r', encoding='utf-8') as fin:
         eeg_json = json.load(fin)
@@ -916,7 +916,7 @@ def test_kit(_bids_validate, tmp_path):
     # ensure the marker file is produced in the right place
     marker_fname = BIDSPath(
         subject=subject_id, session=session_id, task=task, run=run,
-        suffix='markers', extension='.sqd',
+        suffix='markers', extension='.sqd', datatype='meg',
         root=bids_root)
     assert op.exists(marker_fname)