use get_data instead of ._data (#128)

Author: sappelhoff

pyprep/find_noisy_channels.py

@@ -90,7 +90,7 @@ def __init__(self, raw, do_detrend=True, random_state=None, matlab_strict=False)
         ch_names = np.asarray(self.raw_mne.info["ch_names"])
         self.ch_names_original = ch_names
         self.n_chans_original = len(ch_names)
-        self.n_samples = raw._data.shape[1]
+        self.n_samples = raw.get_data().shape[1]
 
         # Before anything else, flag bad-by-NaNs and bad-by-flats
         self.find_bad_by_nan_flat()

pyprep/utils.py

@@ -367,7 +367,7 @@ def _eeglab_interpolate_bads(raw):
     _normalize_vectors(pos_bad)
 
     # Interpolate bad channels
-    interp = _eeglab_interpolate(raw._data[good_idx, :], pos_good, pos_bad)
+    interp = _eeglab_interpolate(raw.get_data()[good_idx, :], pos_good, pos_bad)
     raw._data[bad_idx, :] = interp
 
     # Clear all bad EEG channels

tests/test_find_noisy_channels.py

@@ -58,7 +58,7 @@ def _generate_signal(fmin, fmax, timepoints, fcount=1):
 def test_bad_by_nan(raw_tmp):
     """Test the detection of channels containing any NaN values."""
     # Insert a NaN value into a random channel
-    n_chans = raw_tmp._data.shape[0]
+    n_chans = raw_tmp.get_data().shape[0]
     nan_idx = int(RNG.randint(0, n_chans, 1))
     raw_tmp._data[nan_idx, 3] = np.nan
 
@@ -74,9 +74,9 @@ def test_bad_by_nan(raw_tmp):
 def test_bad_by_flat(raw_tmp):
     """Test the detection of channels with flat or very weak signals."""
     # Make the signal for a random channel extremely weak
-    n_chans = raw_tmp._data.shape[0]
+    n_chans = raw_tmp.get_data().shape[0]
     flat_idx = int(RNG.randint(0, n_chans, 1))
-    raw_tmp._data[flat_idx, :] = raw_tmp._data[flat_idx, :] * 1e-12
+    raw_tmp._data[flat_idx, :] = raw_tmp.get_data()[flat_idx, :] * 1e-12
 
     # Test automatic detection of flat channels on NoisyChannels init
     nd = NoisyChannels(raw_tmp, do_detrend=False)
@@ -99,7 +99,7 @@ def test_bad_by_deviation(raw_tmp):
     high_dev_factor = 4.0
 
     # Make the signal for a random channel have a very high amplitude
-    n_chans = raw_tmp._data.shape[0]
+    n_chans = raw_tmp.get_data().shape[0]
     high_dev_idx = int(RNG.randint(0, n_chans, 1))
     raw_tmp._data[high_dev_idx, :] *= high_dev_factor
 
@@ -125,7 +125,7 @@ def test_bad_by_deviation(raw_tmp):
 def test_bad_by_hf_noise(raw_tmp):
     """Test detection of channels with high-frequency noise."""
     # Add some noise between 70 & 80 Hz to the signal of a random channel
-    n_chans = raw_tmp._data.shape[0]
+    n_chans = raw_tmp.get_data().shape[0]
     hf_noise_idx = int(RNG.randint(0, n_chans, 1))
     hf_noise = _generate_signal(70, 80, raw_tmp.times, 5) * 10
     raw_tmp._data[hf_noise_idx, :] += hf_noise
@@ -147,7 +147,7 @@ def test_bad_by_hf_noise(raw_tmp):
 def test_bad_by_dropout(raw_tmp):
     """Test detection of channels with excessive portions of flat signal."""
     # Add large dropout portions to the signal of a random channel
-    n_chans, n_samples = raw_tmp._data.shape
+    n_chans, n_samples = raw_tmp.get_data().shape
     dropout_idx = int(RNG.randint(0, n_chans, 1))
     x1, x2 = (int(n_samples / 10), int(2 * n_samples / 10))
     raw_tmp._data[dropout_idx, x1:x2] = 0  # flatten 10% of signal
@@ -161,7 +161,7 @@ def test_bad_by_dropout(raw_tmp):
 def test_bad_by_correlation(raw_tmp):
     """Test detection of channels that correlate poorly with others."""
     # Replace a random channel's signal with uncorrelated values
-    n_chans, n_samples = raw_tmp._data.shape
+    n_chans, n_samples = raw_tmp.get_data().shape
     low_corr_idx = int(RNG.randint(0, n_chans, 1))
     raw_tmp._data[low_corr_idx, :] = _generate_signal(10, 30, raw_tmp.times, 5)
 
@@ -186,7 +186,7 @@ def test_bad_by_correlation(raw_tmp):
 def test_bad_by_SNR(raw_tmp):
     """Test detection of channels that have low signal-to-noise ratios."""
     # Replace a random channel's signal with uncorrelated values
-    n_chans = raw_tmp._data.shape[0]
+    n_chans = raw_tmp.get_data().shape[0]
     low_snr_idx = int(RNG.randint(0, n_chans, 1))
     raw_tmp._data[low_snr_idx, :] = _generate_signal(10, 30, raw_tmp.times, 5)
 
@@ -275,7 +275,7 @@ def test_find_bad_by_ransac_err(raw_tmp):
         nd.find_bad_by_ransac(n_samples=n_samples)
 
     # Test IOError when too few good channels for RANSAC sample size
-    n_chans = raw_tmp._data.shape[0]
+    n_chans = raw_tmp.get_data().shape[0]
     nd = NoisyChannels(raw_tmp, do_detrend=False)
     nd.bad_by_deviation = raw_tmp.info["ch_names"][0 : int(n_chans * 0.8)]
     with pytest.raises(IOError):

tests/test_matprep_compare.py

@@ -184,9 +184,9 @@ def test_compare_removeTrend(matprep_artifacts):
     sample_rate = matprep_raw.info["sfreq"]
 
     # Apply removeTrend to raw artifact to get expected and actual signals
-    expected = matprep_detrended._data
+    expected = matprep_detrended.get_data()
     actual = removeTrend(
-        matprep_raw._data, sample_rate, detrendType="high pass", matlab_strict=True
+        matprep_raw.get_data(), sample_rate, detrendType="high pass", matlab_strict=True
     )
 
     # Check MATLAB equivalence at start of recording
@@ -386,11 +386,11 @@ def test_full_signal(self, pyprep_reference, matprep_reference):
         win_size = 500  # window of samples to check
 
         # Compare signals at start of recording
-        pyprep_start = pyprep_reference.raw._data[:, win_size]
-        matprep_start = matprep_reference._data[:, win_size]
+        pyprep_start = pyprep_reference.raw.get_data()[:, win_size]
+        matprep_start = matprep_reference.get_data()[:, win_size]
         assert np.allclose(pyprep_start, matprep_start)
 
         # Compare signals at end of recording
-        pyprep_end = pyprep_reference.raw._data[:, -win_size:]
-        matprep_end = matprep_reference._data[:, -win_size:]
+        pyprep_end = pyprep_reference.raw.get_data()[:, -win_size:]
+        matprep_end = matprep_reference.get_data()[:, -win_size:]
         assert np.allclose(pyprep_end, matprep_end)

tests/test_prep_pipeline.py

@@ -215,7 +215,7 @@ def test_prep_pipeline_non_eeg(raw, montage):
 
     # make arbitrary non eeg channels from the register
     sfreq = raw_copy.info["sfreq"]  # Sampling frequency
-    times = np.array(list(range(raw_copy._data.shape[1])))
+    times = np.array(list(range(raw_copy.get_data().shape[1])))
     ch_names_non_eeg = ["misc" + str(i) for i in range(4)]
     ch_types_non_eeg = ["misc" for i in range(4)]
     raw_non_eeg, _, _ = make_random_mne_object(
@@ -244,11 +244,11 @@ def test_prep_pipeline_non_eeg(raw, montage):
     # names of raw (only eeg)  same as full names - non eeg names
     assert set(prep.raw_eeg.ch_names) == set(raw_copy.ch_names) - set(ch_names_non_eeg)
     # quantity of raw (only eeg) same as quantity of all - non eeg lists
-    assert prep.raw_eeg._data.shape[0] == len(raw_copy.ch_names) - len(
+    assert prep.raw_eeg.get_data().shape[0] == len(raw_copy.ch_names) - len(
         prep.ch_names_non_eeg
     )
     # quantity of channels in is the same as qty of full raw
-    assert raw_copy._data.shape[0] == prep.raw._data.shape[0]
+    assert raw_copy.get_data().shape[0] == prep.raw.get_data().shape[0]
 
 
 @pytest.mark.usefixtures("raw", "montage")