Fix copy-view issue in epochs

doc/changes/devel.rst

@@ -56,6 +56,7 @@ Bugs
 - Fix bug where ``encoding`` argument was ignored when reading annotations from an EDF file (:gh:`11958` by :newcontrib:`Andrew Gilbert`)
 - Mark tests ``test_adjacency_matches_ft`` and ``test_fetch_uncompressed_file`` as network tests (:gh:`12041` by :newcontrib:`Maksym Balatsko`)
 - Fix bug with :func:`mne.channels.read_ch_adjacency` (:gh:`11608` by :newcontrib:`Ivan Zubarev`)
+- Fix bug where ``epochs.get_data(..., scalings=...)`` would errantly modify the preloaded data (:gh:`12121` by :newcontrib:`Pablo Mainar` and `Eric Larson`_)
 - Fix bugs with saving splits for :class:`~mne.Epochs` (:gh:`11876` by `Dmitrii Altukhov`_)
 - Fix bug with multi-plot 3D rendering where only one plot was updated (:gh:`11896` by `Eric Larson`_)
 - Fix bug where ``verbose`` level was not respected inside parallel jobs (:gh:`12154` by `Eric Larson`_)
@@ -92,6 +93,7 @@ Bugs
 
 API changes
 ~~~~~~~~~~~
+- The default for :meth:`mne.Epochs.get_data` of ``copy=False`` will change to ``copy=True`` in 1.7. Set it explicitly to avoid a warning (:gh:`12121` by :newcontrib:`Pablo Mainar` and `Eric Larson`_)
 - ``mne.preprocessing.apply_maxfilter`` and ``mne maxfilter`` have been deprecated and will be removed in 1.7. Use :func:`mne.preprocessing.maxwell_filter` (see :ref:`this tutorial <tut-artifact-sss>`) in Python or the command-line utility from MEGIN ``maxfilter`` and :func:`mne.bem.fit_sphere_to_headshape` instead (:gh:`11938` by `Eric Larson`_)
 - :func:`mne.io.kit.read_mrk` reading pickled files is deprecated using something like ``np.savetxt(fid, pts, delimiter="\t", newline="\n")`` to save your points instead (:gh:`11937` by `Eric Larson`_)
 - Replace legacy ``inst.pick_channels`` and ``inst.pick_types`` with ``inst.pick`` (where ``inst`` is an instance of :class:`~mne.io.Raw`, :class:`~mne.Epochs`, or :class:`~mne.Evoked`) wherever possible (:gh:`11907` by `Clemens Brunner`_)

doc/changes/names.inc

@@ -406,6 +406,8 @@
 
 .. _Pablo-Arias: https://github.com/Pablo-Arias
 
+.. _Pablo Mainar: https://github.com/pablomainar
+
 .. _Padma Sundaram: https://www.nmr.mgh.harvard.edu/user/8071
 
 .. _Paul Pasler: https://github.com/ppasler

examples/datasets/kernel_phantom.py

@@ -49,6 +49,7 @@
 
 # %%
 # The data covariance has an interesting structure because of densely packed sensors:
+
 cov = mne.compute_covariance(epochs, tmax=-0.01)
 mne.viz.plot_cov(cov, raw.info)
 

examples/datasets/limo_data.py

@@ -37,7 +37,6 @@
 # License: BSD-3-Clause
 
 # %%
-
 import matplotlib.pyplot as plt
 import numpy as np
 

examples/decoding/decoding_csp_eeg.py

@@ -78,8 +78,8 @@
 
 # Define a monte-carlo cross-validation generator (reduce variance):
 scores = []
-epochs_data = epochs.get_data()
-epochs_data_train = epochs_train.get_data()
+epochs_data = epochs.get_data(copy=False)
+epochs_data_train = epochs_train.get_data(copy=False)
 cv = ShuffleSplit(10, test_size=0.2, random_state=42)
 cv_split = cv.split(epochs_data_train)
 

examples/decoding/decoding_csp_timefreq.py

@@ -105,7 +105,7 @@
     epochs.drop_bad()
     y = le.fit_transform(epochs.events[:, 2])
 
-    X = epochs.get_data()
+    X = epochs.get_data(copy=False)
 
     # Save mean scores over folds for each frequency and time window
     freq_scores[freq] = np.mean(
@@ -165,7 +165,7 @@
         w_tmax = w_time + w_size / 2.0
 
         # Crop data into time-window of interest
-        X = epochs.copy().crop(w_tmin, w_tmax).get_data()
+        X = epochs.get_data(tmin=w_tmin, tmax=w_tmax, copy=False)
 
         # Save mean scores over folds for each frequency and time window
         tf_scores[freq, t] = np.mean(

examples/decoding/decoding_time_generalization_conditions.py

@@ -77,12 +77,12 @@
 
 # Fit classifiers on the epochs where the stimulus was presented to the left.
 # Note that the experimental condition y indicates auditory or visual
-time_gen.fit(X=epochs["Left"].get_data(), y=epochs["Left"].events[:, 2] > 2)
+time_gen.fit(X=epochs["Left"].get_data(copy=False), y=epochs["Left"].events[:, 2] > 2)
 
 # %%
 # Score on the epochs where the stimulus was presented to the right.
 scores = time_gen.score(
-    X=epochs["Right"].get_data(), y=epochs["Right"].events[:, 2] > 2
+    X=epochs["Right"].get_data(copy=False), y=epochs["Right"].events[:, 2] > 2
 )
 
 # %%

examples/decoding/decoding_unsupervised_spatial_filter.py

@@ -58,7 +58,7 @@
     verbose=False,
 )
 
-X = epochs.get_data()
+X = epochs.get_data(copy=False)
 
 ##############################################################################
 # Transform data with PCA computed on the average ie evoked response

examples/decoding/ems_filtering.py

@@ -64,7 +64,7 @@
 epochs.pick("grad")
 
 # Setup the data to use it a scikit-learn way:
-X = epochs.get_data()  # The MEG data
+X = epochs.get_data(copy=False)  # The MEG data
 y = epochs.events[:, 2]  # The conditions indices
 n_epochs, n_channels, n_times = X.shape
 

examples/decoding/linear_model_patterns.py

@@ -60,7 +60,7 @@
 
 # get MEG data
 meg_epochs = epochs.copy().pick(picks="meg", exclude="bads")
-meg_data = meg_epochs.get_data().reshape(len(labels), -1)
+meg_data = meg_epochs.get_data(copy=False).reshape(len(labels), -1)
 
 # %%
 # Decoding in sensor space using a LogisticRegression classifier

examples/decoding/ssd_spatial_filters.py

@@ -146,7 +146,7 @@
         h_trans_bandwidth=1,
     ),
 )
-ssd_epochs.fit(X=epochs.get_data())
+ssd_epochs.fit(X=epochs.get_data(copy=False))
 
 # Plot topographies.
 pattern_epochs = mne.EvokedArray(data=ssd_epochs.patterns_[:4].T, info=ssd_epochs.info)

examples/preprocessing/otp.py

@@ -13,7 +13,6 @@
 # License: BSD-3-Clause
 
 # %%
-
 import numpy as np
 
 import mne
@@ -70,7 +69,7 @@ def compute_bias(raw):
     sphere = mne.make_sphere_model(r0=(0.0, 0.0, 0.0), head_radius=None, verbose=False)
     cov = mne.compute_covariance(epochs, tmax=0, method="oas", rank=None, verbose=False)
     idx = epochs.time_as_index(0.036)[0]
-    data = epochs.get_data()[:, :, idx].T
+    data = epochs.get_data(copy=False)[:, :, idx].T
     evoked = mne.EvokedArray(data, epochs.info, tmin=0.0)
     dip = fit_dipole(evoked, cov, sphere, n_jobs=None, verbose=False)[0]
     actual_pos = mne.dipole.get_phantom_dipoles()[0][dipole_number - 1]

examples/stats/sensor_regression.py

@@ -18,10 +18,6 @@
 of the words for which we have EEG activity.
 
 For the general methodology, see e.g. :footcite:`HaukEtAl2006`.
-
-References
-----------
-.. footbibliography::
 """
 # Authors: Tal Linzen <[email protected]>
 #          Denis A. Engemann <[email protected]>
@@ -43,7 +39,7 @@
 epochs = mne.read_epochs(path)
 print(epochs.metadata.head())
 
-##############################################################################
+# %%
 # Psycholinguistically relevant word characteristics are continuous. I.e.,
 # concreteness or imaginability is a graded property. In the metadata,
 # we have concreteness ratings on a 5-point scale. We can show the dependence
@@ -59,7 +55,7 @@
     evokeds, colors=colors, split_legend=True, cmap=(name + " Percentile", "viridis")
 )
 
-##############################################################################
+# %%
 # We observe that there appears to be a monotonic dependence of EEG on
 # concreteness. We can also conduct a continuous analysis: single-trial level
 # regression with concreteness as a continuous (although here, binned)
@@ -72,7 +68,7 @@
         title=cond, ts_args=dict(time_unit="s"), topomap_args=dict(time_unit="s")
     )
 
-##############################################################################
+# %%
 # Because the :func:`~mne.stats.linear_regression` function also estimates
 # p values, we can --
 # after applying FDR correction for multiple comparisons -- also visualise the
@@ -85,3 +81,8 @@
 reject_H0, fdr_pvals = fdr_correction(res["Concreteness"].p_val.data)
 evoked = res["Concreteness"].beta
 evoked.plot_image(mask=reject_H0, time_unit="s")
+
+# %%
+# References
+# ----------
+# .. footbibliography::

mne/_fiff/tests/test_reference.py

@@ -620,12 +620,10 @@ def test_add_reference():
     assert_equal(epochs_ref._data.shape[1], epochs._data.shape[1] + 1)
     _check_channel_names(epochs_ref, "Ref")
     ref_idx = epochs_ref.ch_names.index("Ref")
-    ref_data = epochs_ref.get_data()[:, ref_idx, :]
+    ref_data = epochs_ref.get_data(picks=[ref_idx])[:, 0]
     assert_array_equal(ref_data, 0)
     picks_eeg = pick_types(epochs.info, meg=False, eeg=True)
-    assert_array_equal(
-        epochs.get_data()[:, picks_eeg, :], epochs_ref.get_data()[:, picks_eeg, :]
-    )
+    assert_array_equal(epochs.get_data(picks_eeg), epochs_ref.get_data(picks_eeg))
 
     # add two reference channels to epochs
     raw = read_raw_fif(fif_fname, preload=True)
@@ -650,12 +648,10 @@ def test_add_reference():
     ref_idy = epochs_ref.ch_names.index("M2")
     assert_equal(epochs_ref.info["chs"][ref_idx]["ch_name"], "M1")
     assert_equal(epochs_ref.info["chs"][ref_idy]["ch_name"], "M2")
-    ref_data = epochs_ref.get_data()[:, [ref_idx, ref_idy], :]
+    ref_data = epochs_ref.get_data([ref_idx, ref_idy])
     assert_array_equal(ref_data, 0)
     picks_eeg = pick_types(epochs.info, meg=False, eeg=True)
-    assert_array_equal(
-        epochs.get_data()[:, picks_eeg, :], epochs_ref.get_data()[:, picks_eeg, :]
-    )
+    assert_array_equal(epochs.get_data(picks_eeg), epochs_ref.get_data(picks_eeg))
 
     # add reference channel to evoked
     raw = read_raw_fif(fif_fname, preload=True)
@@ -725,7 +721,7 @@ def test_add_reference():
     data = data.get_data()
     epochs = make_fixed_length_epochs(raw).load_data()
     data_2 = epochs.copy().add_reference_channels(["REF"]).pick(picks="eeg")
-    data_2 = data_2.get_data()[0]
+    data_2 = data_2.get_data(copy=False)[0]
     assert_allclose(data, data_2)
     evoked = epochs.average()
     data_3 = evoked.copy().add_reference_channels(["REF"]).pick(picks="eeg")

mne/beamformer/_dics.py

@@ -493,7 +493,7 @@ def apply_dics_epochs(epochs, filters, return_generator=False, verbose=None):
     tmin = epochs.times[0]
 
     sel = _check_channels_spatial_filter(epochs.ch_names, filters)
-    data = epochs.get_data()[:, sel, :]
+    data = epochs.get_data(sel)
 
     stcs = _apply_dics(data=data, filters=filters, info=info, tmin=tmin)
 

mne/beamformer/_lcmv.py

@@ -402,7 +402,7 @@ def apply_lcmv_epochs(epochs, filters, *, return_generator=False, verbose=None):
     tmin = epochs.times[0]
 
     sel = _check_channels_spatial_filter(epochs.ch_names, filters)
-    data = epochs.get_data()[:, sel, :]
+    data = epochs.get_data(sel)
     stcs = _apply_lcmv(data=data, filters=filters, info=info, tmin=tmin)
 
     if not return_generator:

mne/channels/channels.py

@@ -1898,7 +1898,10 @@ def combine_channels(
     ch_idx = list(range(inst.info["nchan"]))
     ch_names = inst.info["ch_names"]
     ch_types = inst.get_channel_types()
-    inst_data = inst.data if isinstance(inst, Evoked) else inst.get_data()
+    kwargs = dict()
+    if isinstance(inst, BaseEpochs):
+        kwargs["copy"] = False
+    inst_data = inst.get_data(**kwargs)
     groups = OrderedDict(deepcopy(groups))
 
     # Convert string values of ``method`` into callables

mne/channels/tests/test_channels.py

@@ -615,7 +615,7 @@ def test_equalize_channels():
     assert raw2.ch_names == ["CH1", "CH2"]
     assert_array_equal(raw2.get_data(), [[1.0], [2.0]])
     assert epochs2.ch_names == ["CH1", "CH2"]
-    assert_array_equal(epochs2.get_data(), [[[3.0], [2.0]]])
+    assert_array_equal(epochs2.get_data(copy=False), [[[3.0], [2.0]]])
     assert cov2.ch_names == ["CH1", "CH2"]
     assert cov2["bads"] == cov["bads"]
     assert ave2.ch_names == ave.ch_names

mne/channels/tests/test_interpolation.py

@@ -232,10 +232,10 @@ def test_interpolation_meg():
     assert len(raw_meg.info["bads"]) == len(raw_meg.info["bads"])
 
     # MEG -- epochs
-    data1 = epochs_meg.get_data()[:, pick, :].ravel()
+    data1 = epochs_meg.get_data(pick).ravel()
     epochs_meg.info.normalize_proj()
     epochs_meg.interpolate_bads(mode="fast")
-    data2 = epochs_meg.get_data()[:, pick, :].ravel()
+    data2 = epochs_meg.get_data(pick).ravel()
     assert np.corrcoef(data1, data2)[0, 1] > thresh
     assert len(epochs_meg.info["bads"]) == 0
 

mne/decoding/tests/test_base.py

@@ -317,7 +317,7 @@ def test_get_coef_multiclass_full(n_classes, n_channels, n_times):
     )
     scorer = "roc_auc_ovr_weighted"
     time_gen = GeneralizingEstimator(clf, scorer, verbose=True)
-    X = epochs.get_data()
+    X = epochs.get_data(copy=False)
     y = epochs.events[:, 2]
     n_splits = 3
     cv = StratifiedKFold(n_splits=n_splits)

mne/decoding/tests/test_csp.py

@@ -123,7 +123,7 @@ def test_csp():
         preload=True,
         proj=False,
     )
-    epochs_data = epochs.get_data()
+    epochs_data = epochs.get_data(copy=False)
     n_channels = epochs_data.shape[1]
     y = epochs.events[:, -1]
 
@@ -182,7 +182,7 @@ def test_csp():
         proj=False,
         preload=True,
     )
-    epochs_data = epochs.get_data()
+    epochs_data = epochs.get_data(copy=False)
     n_channels = epochs_data.shape[1]
 
     n_channels = epochs_data.shape[1]
@@ -256,7 +256,7 @@ def test_regularized_csp():
     epochs = Epochs(
         raw, events, event_id, tmin, tmax, picks=picks, baseline=(None, 0), preload=True
     )
-    epochs_data = epochs.get_data()
+    epochs_data = epochs.get_data(copy=False)
     n_channels = epochs_data.shape[1]
 
     n_components = 3

mne/decoding/tests/test_ems.py

@@ -76,7 +76,7 @@ def test_ems():
     raw.close()
 
     # EMS transformer, check that identical to compute_ems
-    X = epochs.get_data()
+    X = epochs.get_data(copy=False)
     y = epochs.events[:, 2]
     X = X / np.std(X)  # X scaled outside cv in compute_ems
     Xt, coefs = list(), list()

mne/decoding/tests/test_transformer.py

@@ -55,7 +55,7 @@ def test_scaler(info, method):
     epochs = Epochs(
         raw, events, event_id, tmin, tmax, picks=picks, baseline=(None, 0), preload=True
     )
-    epochs_data = epochs.get_data()
+    epochs_data = epochs.get_data(copy=False)
     y = epochs.events[:, -1]
 
     epochs_data_t = epochs_data.transpose([1, 0, 2])
@@ -115,7 +115,7 @@ def test_scaler(info, method):
         picks=np.arange(len(raw.ch_names)),
     )  # non-data chs
     scaler = Scaler(epochs_bad.info, None)
-    pytest.raises(ValueError, scaler.fit, epochs_bad.get_data(), y)
+    pytest.raises(ValueError, scaler.fit, epochs_bad.get_data(copy=False), y)
 
 
 def test_filterestimator():
@@ -129,7 +129,7 @@ def test_filterestimator():
     epochs = Epochs(
         raw, events, event_id, tmin, tmax, picks=picks, baseline=(None, 0), preload=True
     )
-    epochs_data = epochs.get_data()
+    epochs_data = epochs.get_data(copy=False)
 
     # Add tests for different combinations of l_freq and h_freq
     filt = FilterEstimator(epochs.info, l_freq=40, h_freq=80)
@@ -180,7 +180,7 @@ def test_psdestimator():
     epochs = Epochs(
         raw, events, event_id, tmin, tmax, picks=picks, baseline=(None, 0), preload=True
     )
-    epochs_data = epochs.get_data()
+    epochs_data = epochs.get_data(copy=False)
     psd = PSDEstimator(2 * np.pi, 0, np.inf)
     y = epochs.events[:, -1]
     X = psd.fit_transform(epochs_data, y)
@@ -244,7 +244,7 @@ def test_unsupervised_spatial_filter():
     pytest.raises(ValueError, UnsupervisedSpatialFilter, KernelRidge(2))
 
     # Test fit
-    X = epochs.get_data()
+    X = epochs.get_data(copy=False)
     n_components = 4
     usf = UnsupervisedSpatialFilter(PCA(n_components))
     usf.fit(X)