Fix deprecation of setting shape on a numpy array in 2.5+

doc/changes/dev/13585.other.rst

@@ -0,0 +1 @@
+Fix deprecation of setting a shape on an array directly in ``numpy`` 2.5+, by `Mathieu Scheltienne`_.

mne/_fiff/_digitization.py

@@ -7,6 +7,7 @@
 
 import numpy as np
 
+from ..fixes import _reshape_view
 from ..utils import Bunch, _check_fname, _validate_type, logger, verbose, warn
 from .constants import FIFF, _coord_frame_named
 from .tag import read_tag
@@ -335,7 +336,7 @@ def _get_data_as_dict_from_dig(dig, exclude_ref_channel=True):
             f"Only single coordinate frame in dig is supported, got {dig_coord_frames}"
         )
     dig_ch_pos_location = np.array(dig_ch_pos_location)
-    dig_ch_pos_location.shape = (-1, 3)  # empty will be (0, 3)
+    dig_ch_pos_location = _reshape_view(dig_ch_pos_location, (-1, 3))
     return Bunch(
         nasion=fids.get("nasion", None),
         lpa=fids.get("lpa", None),

mne/_fiff/tag.py

@@ -12,6 +12,7 @@
 import numpy as np
 from scipy.sparse import csc_array, csr_array
 
+from ..fixes import _reshape_view
 from ..utils import _check_option, warn
 from ..utils.numerics import _julian_to_date
 from .constants import (
@@ -177,7 +178,7 @@ def _read_matrix(fid, tag, shape, rlims):
             data = data.view(">c8")
         elif matrix_type == FIFF.FIFFT_COMPLEX_DOUBLE:
             data = data.view(">c16")
-        data.shape = dims
+        data = _reshape_view(data, dims)
     else:
         # Find dimensions and return to the beginning of tag data
         ndim = int(np.frombuffer(fid.read(4), dtype=">i4").item())

mne/beamformer/_rap_music.py

@@ -8,7 +8,7 @@
 from scipy import linalg
 
 from .._fiff.pick import pick_channels_forward, pick_info
-from ..fixes import _safe_svd
+from ..fixes import _reshape_view, _safe_svd
 from ..forward import convert_forward_solution, is_fixed_orient
 from ..inverse_sparse.mxne_inverse import _make_dipoles_sparse
 from ..minimum_norm.inverse import _log_exp_var
@@ -68,9 +68,9 @@ def _apply_rap_music(
     phi_sig = eig_vectors[:, -n_dipoles:]
 
     n_orient = 3 if is_free_ori else 1
-    G.shape = (G.shape[0], -1, n_orient)
+    G = _reshape_view(G, (G.shape[0], -1, n_orient))
     gain = forward["sol"]["data"].copy()
-    gain.shape = G.shape
+    gain = _reshape_view(gain, G.shape)
     n_channels = G.shape[0]
     A = np.empty((n_channels, n_dipoles))
     gain_dip = np.empty((n_channels, n_dipoles))
@@ -122,7 +122,7 @@ def _apply_rap_music(
     sol = linalg.lstsq(A, M)[0]
     if n_orient == 3:
         X = sol[:, np.newaxis] * oris[:, :, np.newaxis]
-        X.shape = (-1, len(times))
+        X = _reshape_view(X, (-1, len(times)))
     else:
         X = sol
 

mne/beamformer/tests/test_dics.py

@@ -25,6 +25,7 @@
 from mne.beamformer._dics import _prepare_noise_csd
 from mne.beamformer.tests.test_lcmv import _assert_weight_norm
 from mne.datasets import testing
+from mne.fixes import _reshape_view
 from mne.io import read_info
 from mne.proj import compute_proj_evoked, make_projector
 from mne.surface import _compute_nearest
@@ -269,7 +270,7 @@ def test_make_dics(tmp_path, _load_forward, idx, whiten):
         exp=None,
         noise_cov=noise_cov,
     )
-    G.shape = (n_channels, n_verts, n_orient)
+    G = _reshape_view(G, (n_channels, n_verts, n_orient))
     G = G.transpose(1, 2, 0).conj()  # verts, orient, ch
     _assert_weight_norm(filters, G)
 

mne/beamformer/tests/test_external.py

@@ -11,6 +11,7 @@
 from mne.beamformer import apply_lcmv, apply_lcmv_cov, make_lcmv
 from mne.beamformer.tests.test_lcmv import _get_data
 from mne.datasets import testing
+from mne.fixes import _reshape_view
 
 data_path = testing.data_path(download=False)
 ft_data_path = data_path / "fieldtrip" / "beamformer"
@@ -98,7 +99,7 @@ def test_lcmv_fieldtrip(_get_bf_data, bf_type, weight_norm, pick_ori, pwr):
     ft_fname = ft_data_path / ("ft_source_" + bf_type + "-vol.mat")
     stc_ft_data = pymatreader.read_mat(ft_fname)["stc"]
     if stc_ft_data.ndim == 1:
-        stc_ft_data.shape = (stc_ft_data.size, 1)
+        stc_ft_data = _reshape_view(stc_ft_data, (stc_ft_data.size, 1))
 
     if stc_mne.data.ndim == 2:
         signs = np.sign((stc_mne.data * stc_ft_data).sum(-1, keepdims=True))

mne/beamformer/tests/test_lcmv.py

@@ -42,6 +42,7 @@
 )
 from mne.beamformer._compute_beamformer import _prepare_beamformer_input
 from mne.datasets import testing
+from mne.fixes import _reshape_view
 from mne.minimum_norm import apply_inverse, make_inverse_operator
 from mne.minimum_norm.tests.test_inverse import _assert_free_ori_match
 from mne.simulation import simulate_evoked
@@ -1185,7 +1186,7 @@ def test_unit_noise_gain_formula(pick_ori, weight_norm, reg, inversion):
     )
     n_channels, n_sources = G.shape
     n_sources //= 3
-    G.shape = (n_channels, n_sources, 3)
+    G = _reshape_view(G, (n_channels, n_sources, 3))
     G = G.transpose(1, 2, 0)  # verts, orient, ch
     _assert_weight_norm(filters, G)
 

mne/channels/montage.py

@@ -28,6 +28,7 @@
 from .._fiff.pick import _picks_to_idx, channel_type, pick_types
 from .._freesurfer import get_mni_fiducials
 from ..defaults import HEAD_SIZE_DEFAULT
+from ..fixes import _reshape_view
 from ..transforms import (
     Transform,
     _ensure_trans,
@@ -973,9 +974,9 @@ def read_dig_hpts(fname, unit="mm"):
         label[ii]: this_xyz for ii, this_xyz in enumerate(xyz) if kind[ii] == "eeg"
     }
     hpi = np.array([this_xyz for ii, this_xyz in enumerate(xyz) if kind[ii] == "hpi"])
-    hpi.shape = (-1, 3)  # in case it's empty
+    hpi = _reshape_view(hpi, (-1, 3))  # in case it's empty
     hsp = np.array([this_xyz for ii, this_xyz in enumerate(xyz) if kind[ii] == "extra"])
-    hsp.shape = (-1, 3)  # in case it's empty
+    hsp = _reshape_view(hsp, (-1, 3))  # in case it's empty
     return make_dig_montage(ch_pos=ch_pos, **fid, hpi=hpi, hsp=hsp)
 
 

mne/chpi.py

@@ -43,7 +43,7 @@
 from .cov import compute_whitener, make_ad_hoc_cov
 from .dipole import _make_guesses
 from .event import find_events
-from .fixes import jit
+from .fixes import _reshape_view, jit
 from .forward import _concatenate_coils, _create_meg_coils, _magnetic_dipole_field_vec
 from .io import BaseRaw, RawArray
 from .io.ctf.trans import _make_ctf_coord_trans_set
@@ -117,7 +117,7 @@ def read_head_pos(fname):
     """
     _check_fname(fname, must_exist=True, overwrite="read")
     data = np.loadtxt(fname, skiprows=1)  # first line is header, skip it
-    data.shape = (-1, 10)  # ensure it's the right size even if empty
+    data = _reshape_view(data, (-1, 10))  # ensure it's the right size even if empty
     if np.isnan(data).any():  # make sure we didn't do something dumb
         raise RuntimeError(f"positions could not be read properly from {fname}")
     return data
@@ -1390,7 +1390,7 @@ def compute_chpi_locs(
     )
     fwd = _magnetic_dipole_field_vec(guesses, meg_coils, too_close)
     fwd = np.dot(fwd, whitener.T)
-    fwd.shape = (guesses.shape[0], 3, -1)
+    fwd = _reshape_view(fwd, (guesses.shape[0], 3, -1))
     fwd = np.linalg.svd(fwd, full_matrices=False)[2]
     guesses = dict(rr=guesses, whitened_fwd_svd=fwd)
     del fwd, R

mne/conftest.py

@@ -207,6 +207,9 @@ def pytest_configure(config: pytest.Config):
     ignore:^'.*' deprecated - use '.*'$:DeprecationWarning
     # dipy
     ignore:'where' used without 'out', expect .*:UserWarning
+    # VTK <-> NumPy 2.5 (https://gitlab.kitware.com/vtk/vtk/-/merge_requests/12796)
+    # nitime <-> NumPy 2.5 (https://github.com/nipy/nitime/pull/236)
+    ignore:Setting the shape on a NumPy array has been deprecated.*:DeprecationWarning
     """  # noqa: E501
     for warning_line in warning_lines.split("\n"):
         warning_line = warning_line.strip()

mne/decoding/receptive_field.py

@@ -15,6 +15,7 @@
 from sklearn.exceptions import NotFittedError
 from sklearn.metrics import r2_score
 
+from ..fixes import _reshape_view
 from ..utils import _validate_type, fill_doc, pinv
 from ._fixes import _check_n_features_3d, validate_data
 from .base import _check_estimator, get_coef
@@ -361,7 +362,7 @@ def predict(self, X):
         else:
             extra = 1
         shape = shape[: self._y_dim + extra]
-        y_pred.shape = shape
+        y_pred = _reshape_view(y_pred, shape)
         return y_pred
 
     def score(self, X, y):

mne/decoding/tests/test_receptive_field.py

@@ -23,6 +23,7 @@
     _times_to_delays,
 )
 from mne.decoding.time_delaying_ridge import _compute_corrs, _compute_reg_neighbors
+from mne.fixes import _reshape_view
 
 data_dir = Path(__file__).parents[2] / "io" / "tests" / "data"
 raw_fname = data_dir / "test_raw.fif"
@@ -271,7 +272,7 @@ def test_time_delaying_fast_calc(n_jobs):
     smin, smax = 1, 2
     X_del = _delay_time_series(X, smin, smax, 1.0)
     # (n_times, n_features, n_delays) -> (n_times, n_features * n_delays)
-    X_del.shape = (X.shape[0], -1)
+    X_del = _reshape_view(X_del, (X.shape[0], -1))
     expected = np.array([[0, 1, 2], [0, 0, 1], [0, 5, 7], [0, 0, 5]]).T
     assert_allclose(X_del, expected)
     Xt_X = np.dot(X_del.T, X_del)
@@ -282,7 +283,7 @@ def test_time_delaying_fast_calc(n_jobs):
     # all positive
     smin, smax = -2, -1
     X_del = _delay_time_series(X, smin, smax, 1.0)
-    X_del.shape = (X.shape[0], -1)
+    X_del = _reshape_view(X_del, (X.shape[0], -1))
     expected = np.array([[3, 0, 0], [2, 3, 0], [11, 0, 0], [7, 11, 0]]).T
     assert_allclose(X_del, expected)
     Xt_X = np.dot(X_del.T, X_del)
@@ -293,7 +294,7 @@ def test_time_delaying_fast_calc(n_jobs):
     # both sides
     smin, smax = -1, 1
     X_del = _delay_time_series(X, smin, smax, 1.0)
-    X_del.shape = (X.shape[0], -1)
+    X_del = _reshape_view(X_del, (X.shape[0], -1))
     expected = np.array(
         [[2, 3, 0], [1, 2, 3], [0, 1, 2], [7, 11, 0], [5, 7, 11], [0, 5, 7]]
     ).T
@@ -315,7 +316,7 @@ def test_time_delaying_fast_calc(n_jobs):
     X = np.array([[1, 2, 3, 5]]).T
     smin, smax = 0, 3
     X_del = _delay_time_series(X, smin, smax, 1.0)
-    X_del.shape = (X.shape[0], -1)
+    X_del = _reshape_view(X_del, (X.shape[0], -1))
     expected = np.array([[1, 2, 3, 5], [0, 1, 2, 3], [0, 0, 1, 2], [0, 0, 0, 1]]).T
     assert_allclose(X_del, expected)
     Xt_X = np.dot(X_del.T, X_del)
@@ -328,7 +329,7 @@ def test_time_delaying_fast_calc(n_jobs):
     X = np.array([[1, 2, 3], [5, 7, 11]]).T
     smin, smax = 0, 2
     X_del = _delay_time_series(X, smin, smax, 1.0)
-    X_del.shape = (X.shape[0], -1)
+    X_del = _reshape_view(X_del, (X.shape[0], -1))
     expected = np.array(
         [[1, 2, 3], [0, 1, 2], [0, 0, 1], [5, 7, 11], [0, 5, 7], [0, 0, 5]]
     ).T
@@ -366,7 +367,7 @@ def test_time_delaying_fast_calc(n_jobs):
             x_yt_true = einsum("tfd,to->ofd", X_del, y)
             x_yt_true = np.reshape(x_yt_true, (x_yt_true.shape[0], -1)).T
             assert_allclose(x_yt, x_yt_true, atol=1e-7, err_msg=(smin, smax))
-            X_del.shape = (X.shape[0], -1)
+            X_del = _reshape_view(X_del, (X.shape[0], -1))
             x_xt_true = np.dot(X_del.T, X_del).T
             assert_allclose(x_xt, x_xt_true, atol=1e-7, err_msg=(smin, smax))
 
@@ -388,7 +389,7 @@ def test_receptive_field_1d(n_jobs):
             y[delay:] = x[:-delay, 0]
             slims += [(1, 2)]
         for ndim in (1, 2):
-            y.shape = (y.shape[0],) + (1,) * (ndim - 1)
+            y = _reshape_view(y, (y.shape[0],) + (1,) * (ndim - 1))
             for slim in slims:
                 smin, smax = slim
                 lap = TimeDelayingRidge(

mne/decoding/transformer.py

@@ -17,6 +17,7 @@
 from ..cov import _check_scalings_user
 from ..epochs import BaseEpochs
 from ..filter import filter_data
+from ..fixes import _reshape_view
 from ..time_frequency import psd_array_multitaper
 from ..utils import _check_option, _validate_type, check_version, fill_doc
 from ._fixes import validate_data  # TODO VERSION remove with sklearn 1.4+
@@ -118,7 +119,7 @@ def _sklearn_reshape_apply(func, return_result, X, *args, **kwargs):
     X = np.reshape(X.transpose(0, 2, 1), (-1, orig_shape[1]))
     X = func(X, *args, **kwargs)
     if return_result:
-        X.shape = (orig_shape[0], orig_shape[2], orig_shape[1])
+        X = _reshape_view(X, (orig_shape[0], orig_shape[2], orig_shape[1]))
         X = X.transpose(0, 2, 1)
         return X
 

mne/epochs.py

@@ -66,7 +66,7 @@
 from .event import _read_events_fif, make_fixed_length_events, match_event_names
 from .evoked import EvokedArray
 from .filter import FilterMixin, _check_fun, detrend
-from .fixes import rng_uniform
+from .fixes import _reshape_view, rng_uniform
 from .html_templates import _get_html_template
 from .parallel import parallel_func
 from .time_frequency.spectrum import EpochsSpectrum, SpectrumMixin, _validate_method
@@ -4479,7 +4479,7 @@ def _get_epoch_from_raw(self, idx, verbose=None):
         else:
             data = data.astype(np.float64)
 
-        data.shape = raw.epoch_shape
+        data = _reshape_view(data, raw.epoch_shape)
         data *= raw.cals
         return data
 

mne/event.py

@@ -15,6 +15,7 @@
 from ._fiff.tag import read_tag
 from ._fiff.tree import dir_tree_find
 from ._fiff.write import end_block, start_and_end_file, start_block, write_int
+from .fixes import _reshape_view
 from .utils import (
     _check_fname,
     _check_integer_or_list,
@@ -181,7 +182,7 @@ def _read_events_fif(fid, tree):
     if event_list is None:
         raise ValueError("Could not find any events")
     else:
-        event_list.shape = (-1, 3)
+        event_list = _reshape_view(event_list, (-1, 3))
     for d in events["directory"]:
         kind = d.kind
         pos = d.pos

mne/filter.py

@@ -22,7 +22,7 @@
     _setup_cuda_fft_resample,
     _smart_pad,
 )
-from .fixes import minimum_phase
+from .fixes import _reshape_view, minimum_phase
 from .parallel import parallel_func
 from .utils import (
     _check_option,
@@ -349,7 +349,7 @@ def _overlap_add_filter(
         for pp, p in enumerate(picks):
             x[p] = data_new[pp]
 
-    x.shape = orig_shape
+    x = _reshape_view(x, orig_shape)
     return x
 
 
@@ -404,7 +404,7 @@ def _prep_for_filtering(x, copy, picks=None):
     orig_shape = x.shape
     x = np.atleast_2d(x)
     picks = _picks_to_idx(x.shape[-2], picks)
-    x.shape = (np.prod(x.shape[:-1]), x.shape[-1])
+    x = _reshape_view(x, (np.prod(x.shape[:-1]), x.shape[-1]))
     if len(orig_shape) == 3:
         n_epochs, n_channels, n_times = orig_shape
         offset = np.repeat(np.arange(0, n_channels * n_epochs, n_channels), len(picks))
@@ -577,7 +577,7 @@ def _iir_filter(x, iir_params, picks, n_jobs, copy, phase="zero"):
         data_new = parallel(p_fun(x=x[p]) for p in picks)
         for pp, p in enumerate(picks):
             x[p] = data_new[pp]
-    x.shape = orig_shape
+    x = _reshape_view(x, orig_shape)
     return x
 
 
@@ -1657,7 +1657,7 @@ def _mt_spectrum_proc(
     )
     logger.info(f"{kind} notch frequencies (Hz):\n{found_freqs}")
 
-    x.shape = orig_shape
+    x = _reshape_view(x, orig_shape)
     return x
 
 
@@ -2952,5 +2952,5 @@ def _iir_pad_apply_unpad(x, *, func, padlen, padtype, **kwargs):
             x_ext = _smart_pad(x_ext, (padlen, padlen), padtype)
         x_ext = func(x=x_ext, axis=-1, padlen=0, **kwargs)
         this_x[:] = x_ext[padlen : len(x_ext) - padlen]
-    x_out.shape = x.shape
+    x_out = _reshape_view(x_out, x.shape)
     return x_out