[ENH] Adds initial workflow class

neuromaps/datasets/annotations.py

@@ -207,6 +207,8 @@ def fetch_annotation(*, source=None, desc=None, space=None, den=None, res=None,
     token = _get_token(token=token)
     return_restricted = False if (token is None or not token) else True
     data_dir = get_data_dir(data_dir=data_dir)
+    if space == 'MNI152' and hemi == ('L', 'R'):
+        hemi = None
     info = _match_annot(get_dataset_info('annotations', return_restricted),
                         source=source, desc=desc, space=space, den=den,
                         res=res, hemi=hemi, tags=tags, format=format)
@@ -223,8 +225,9 @@ def fetch_annotation(*, source=None, desc=None, space=None, den=None, res=None,
     for dset in info:
         fn = Path(data_dir) / 'annotations' / dset['rel_path'] / dset['fname']
         if not fn.exists():
-            dl_file = _fetch_file(dset['url'], str(fn.parent), verbose=verbose,
-                                  md5sum=dset['checksum'], session=session)
+            dl_file = _fetch_file(dset['url'], str(fn.parent),
+                                  verbose=verbose, md5sum=dset['checksum'],
+                                  session=session)
             shutil.move(dl_file, fn)
         data.append(str(fn))
 

neuromaps/datasets/tests/test_utils.py

@@ -33,6 +33,8 @@ def test__get_token():
     assert utils._get_token('test') == 'test'
     if orig is not None:  # reset env variable
         os.environ['NEUROMAPS_OSF_TOKEN'] = orig
+    else:
+        os.environ.pop('NEUROMAPS_OSF_TOKEN', None)
 
 
 @pytest.mark.xfail

neuromaps/images.py

@@ -7,6 +7,7 @@
 import os
 from pathlib import Path
 from typing import Iterable
+import warnings
 
 import nibabel as nib
 from nibabel.filebasedimages import ImageFileError
@@ -133,7 +134,9 @@ def load_nifti(img):
     """
 
     try:
-        img = nib.load(img)
+        with warnings.catch_warnings():
+            warnings.filterwarnings('ignore', category=DeprecationWarning)
+            img = nib.load(img)
     except (TypeError) as err:
         msg = ('stat: path should be string, bytes, os.PathLike or integer, '
                'not Nifti1Image')
@@ -161,7 +164,9 @@ def load_gifti(img):
     """
 
     try:
-        img = nib.load(img)
+        with warnings.catch_warnings():
+            warnings.filterwarnings('ignore', category=DeprecationWarning)
+            img = nib.load(img)
     except (ImageFileError, TypeError) as err:
         # it's gzipped, so read the gzip and pipe it in
         if isinstance(err, ImageFileError) and str(err).endswith('.gii.gz"'):
@@ -199,8 +204,9 @@ def load_data(data):
         out = np.hstack([load_gifti(img).agg_data() for img in data])
     except (AttributeError, TypeError):
         out = np.stack([load_nifti(img).get_fdata() for img in data], axis=3)
-    except ValueError as err:
-        if str(err) == 'stat: embedded null character in path':
+    except (ValueError, FileNotFoundError) as err:
+        if (str(err) == 'stat: embedded null character in path'
+                or str(err).startswith('No such file or no access:')):
             out = np.stack(data, axis=-1)
 
     return np.squeeze(out)

neuromaps/nulls/nulls.py

@@ -408,7 +408,7 @@ def _surf_surrogates(data, atlas, density, parcellation, distmat, n_proc):
         yield hdata[mask], dist[np.ix_(mask, mask)], None, idx[mask]
 
 
-def _vol_surrogates(data, atlas, density, parcellation, distmat, **kwargs):
+def _vol_surrogates(data, atlas, density, parcellation, distmat, *args):
     if atlas != 'MNI152':
         raise ValueError('Cannot compute volumetric surrogates if atlas is '
                          f'"MNI152". Received: {atlas}')
@@ -418,18 +418,24 @@ def _vol_surrogates(data, atlas, density, parcellation, distmat, **kwargs):
 
     # get data + coordinates of valid datapoints
     data = load_data(data)
-    if mni152.shape != data.shape:
-        raise ValueError('Provided `data` array must have same affine as '
-                         'specified MNI152 atlas')
-    data *= load_data(atlas['brainmask'])
-    mask = np.logical_not(np.logical_or(np.isclose(data, 0), np.isnan(data)))
-    xyz = nib.affines.apply_affine(affine, np.column_stack(np.where(mask)))
-
-    if parcellation is not None:
+    if parcellation is None:
+        if mni152.shape != data.shape:
+            raise ValueError('Provided `data` array must have same affine as '
+                             'specified MNI152 atlas')
+        data *= load_data(atlas['brainmask'])
+        mask = ~np.logical_or(np.isclose(data, 0), np.isnan(data))
+        data = data[mask]
+    else:
         parcellation = load_data(
             mni152_to_mni152(parcellation, density, 'nearest')
-        )[mask]
+        )
         labels = np.trim_zeros(np.unique(parcellation))
+        mask = ~np.logical_or(
+            np.isclose(parcellation, 0), np.isnan(parcellation)
+        )
+        parcellation = parcellation[mask]
+
+    xyz = nib.affines.apply_affine(affine, np.column_stack(np.where(mask)))
 
     # calculate distance matrix
     index = None
@@ -458,7 +464,13 @@ def _vol_surrogates(data, atlas, density, parcellation, distmat, **kwargs):
         if parcellation is None:
             index = np.argsort(dist, axis=-1)
 
-    yield data[mask], dist, index, slice(-1)
+    if parcellation is not None:
+        dist = np.row_stack([
+            dist[parcellation == lab].mean(axis=0) for lab in labels
+        ])
+        dist[np.diag_indices_from(dist)] = 0
+
+    yield data, dist, index, slice(None)
 
 
 def _make_surrogates(data, method, atlas='fsaverage', density='10k',

neuromaps/parcellate.py

@@ -10,7 +10,8 @@
 from neuromaps.datasets import ALIAS, DENSITIES
 from neuromaps.images import construct_shape_gii, load_gifti
 from neuromaps.resampling import resample_images
-from neuromaps.transforms import _check_hemi
+from neuromaps.transforms import (_check_hemi, _estimate_density,
+                                  _estimate_resolution)
 from neuromaps.nulls.spins import vertices_to_parcels, parcels_to_vertices
 
 
@@ -60,6 +61,12 @@ def __init__(self, parcellation, space, resampling_target='data',
         self.hemi = hemi
         self._volumetric = self.space == 'MNI152'
 
+        if not self._volumetric:
+            self._density, = _estimate_density((self.parcellation,), self.hemi)
+        else:
+            self._density, = _estimate_resolution((self.parcellation,))
+            self.hemi = ('L', 'R') if hemi is None else hemi
+
         if self.resampling_target == 'parcellation':
             self._resampling = 'transform_to_trg'
         else: