The order of raw1.info["bads"] should not matter when concatenating with raw0.info["bads"] mne-tools#11501 (mne-tools#11502)

Author: moritz-gerster

doc/changes/latest.inc

@@ -57,6 +57,8 @@ Bugs
 - Fix bug in :func:`mne.preprocessing.compute_maxwell_basis` where using ``int_order=0`` would raise an error (:gh:`11562` by `Eric Larson`_)
 - Fix :func:`mne.io.read_raw` for file names containing multiple dots (:gh:`11521` by `Clemens Brunner`_)
 - Fix bug in :func:`mne.export.export_raw` when exporting to EDF with a physical range set smaller than the data range (:gh:`11569` by `Mathieu Scheltienne`_)
+- Fix bug in :func:`mne.concatenate_raws` where two raws could not be merged if the order of the bad channel lists did not match (:gh:`11502` by `Moritz Gerster`_)
+
 
 API changes
 ~~~~~~~~~~~

mne/io/base.py

@@ -2503,22 +2503,27 @@ def _check_raw_compatibility(raw):
     for ri in range(1, len(raw)):
         if not isinstance(raw[ri], type(raw[0])):
             raise ValueError(f'raw[{ri}] type must match')
-        for key in ('nchan', 'bads', 'sfreq'):
+        for key in ('nchan', 'sfreq'):
             a, b = raw[ri].info[key], raw[0].info[key]
             if a != b:
                 raise ValueError(
                     f'raw[{ri}].info[{key}] must match:\n'
                     f'{repr(a)} != {repr(b)}')
-        if not set(raw[ri].info['ch_names']) == set(raw[0].info['ch_names']):
-            raise ValueError('raw[%d][\'info\'][\'ch_names\'] must match' % ri)
-        if not all(raw[ri]._cals == raw[0]._cals):
+        for kind in ('bads', 'ch_names'):
+            set1 = set(raw[0].info[kind])
+            set2 = set(raw[ri].info[kind])
+            mismatch = set1.symmetric_difference(set2)
+            if mismatch:
+                raise ValueError(f'raw[{ri}][\'info\'][{kind}] do not match: '
+                                 f'{sorted(mismatch)}')
+        if any(raw[ri]._cals != raw[0]._cals):
             raise ValueError('raw[%d]._cals must match' % ri)
         if len(raw[0].info['projs']) != len(raw[ri].info['projs']):
             raise ValueError('SSP projectors in raw files must be the same')
         if not all(_proj_equal(p1, p2) for p1, p2 in
                    zip(raw[0].info['projs'], raw[ri].info['projs'])):
             raise ValueError('SSP projectors in raw files must be the same')
-    if not all(r.orig_format == raw[0].orig_format for r in raw):
+    if any(r.orig_format != raw[0].orig_format for r in raw):
         warn('raw files do not all have the same data format, could result in '
              'precision mismatch. Setting raw.orig_format="unknown"')
         raw[0].orig_format = 'unknown'

mne/io/fiff/tests/test_raw_fiff.py

@@ -28,7 +28,7 @@
 from mne.io.tests.test_raw import _test_concat, _test_raw_reader
 from mne import (concatenate_events, find_events, equalize_channels,
                  compute_proj_raw, pick_types, pick_channels, create_info,
-                 pick_info)
+                 pick_info, make_fixed_length_epochs)
 from mne.utils import (requires_pandas, assert_object_equal, _dt_to_stamp,
                        requires_mne, run_subprocess, _record_warnings,
                        assert_and_remove_boundary_annot)
@@ -382,6 +382,54 @@ def test_concatenate_raws(on_mismatch):
             concatenate_raws(**kws)
 
 
+def _create_toy_data(n_channels=3, sfreq=250, seed=None):
+    rng = np.random.default_rng(seed)
+    data = rng.standard_normal(size=(n_channels, 50 * sfreq)) * 5e-6
+    info = create_info(n_channels, sfreq, "eeg")
+    return RawArray(data, info)
+
+
+def test_concatenate_raws_bads_order():
+    """Test concatenation of raw instances."""
+    raw0 = _create_toy_data()
+    raw1 = _create_toy_data()
+
+    # Test bad channel order
+    raw0.info["bads"] = ["0", "1"]
+    raw1.info["bads"] = ["1", "0"]
+
+    # raw0 is modified in-place and therefore copied
+    raw_concat = concatenate_raws([raw0.copy(), raw1])
+
+    # Check data are equal
+    data_concat = np.concatenate([raw0.get_data(), raw1.get_data()], 1)
+    assert np.all(raw_concat.get_data() == data_concat)
+
+    # Check bad channels
+    assert set(raw_concat.info["bads"]) == {"0", "1"}
+
+    # Bad channel mismatch raises
+    raw2 = raw1.copy()
+    raw2.info["bads"] = ["0", "2"]
+    with pytest.raises(ValueError):
+        concatenate_raws([raw0, raw2])
+
+    # Type mismatch raises
+    epochs1 = make_fixed_length_epochs(raw1)
+    with pytest.raises(ValueError):
+        concatenate_raws([raw0, epochs1])
+
+    # Sample rate mismatch
+    raw3 = _create_toy_data(sfreq=500)
+    with pytest.raises(ValueError):
+        concatenate_raws([raw0, raw3])
+
+    # Number of channels mismatch
+    raw4 = _create_toy_data(n_channels=4)
+    with pytest.raises(ValueError):
+        concatenate_raws([raw0, raw4])
+
+
 @testing.requires_testing_data
 @pytest.mark.parametrize('mod', (
     'meg',