def test_h5py_dataset_typecheck():
ds = datasets['uni2small']
fd, fpath = tempfile.mkstemp('mvpa', 'test')
os.close(fd)
fd, fpath2 = tempfile.mkstemp('mvpa', 'test')
os.close(fd)
h5save(fpath2, [[1, 2, 3]])
assert_raises(ValueError, AttrDataset.from_hdf5, fpath2)
# this one just catches if there is such a group
assert_raises(ValueError, AttrDataset.from_hdf5, fpath2, name='bogus')
hdf = h5py.File(fpath, 'w')
ds = AttrDataset([1, 2, 3])
obj2hdf(hdf, ds, name='non-bogus')
obj2hdf(hdf, [1, 2, 3], name='bogus')
hdf.close()
assert_raises(ValueError, AttrDataset.from_hdf5, fpath, name='bogus')
ds_loaded = AttrDataset.from_hdf5(fpath, name='non-bogus')
assert_array_equal(ds, ds_loaded) # just to do smth useful with ds ;)
# cleanup and ignore stupidity
os.remove(fpath)
os.remove(fpath2)
def test_dataset_without_chunks(fname):
# ValueError: All chunk dimensions must be positive (Invalid arguments to routine: Out of range)
# MH: This is not about Dataset chunks, but about an empty samples array
ds = AttrDataset([8], a=dict(custom=1))
save(ds, fname, compression='gzip')
ds_loaded = h5load(fname)
ok_(ds_loaded.a.custom == ds.a.custom)
def test_h5py_dataset_typecheck():
ds = datasets["uni2small"]
fd, fpath = tempfile.mkstemp("mvpa", "test")
os.close(fd)
fd, fpath2 = tempfile.mkstemp("mvpa", "test")
os.close(fd)
h5save(fpath2, [[1, 2, 3]])
assert_raises(ValueError, AttrDataset.from_hdf5, fpath2)
# this one just catches if there is such a group
assert_raises(ValueError, AttrDataset.from_hdf5, fpath2, name="bogus")
hdf = h5py.File(fpath, "w")
ds = AttrDataset([1, 2, 3])
obj2hdf(hdf, ds, name="non-bogus")
obj2hdf(hdf, [1, 2, 3], name="bogus")
hdf.close()
assert_raises(ValueError, AttrDataset.from_hdf5, fpath, name="bogus")
ds_loaded = AttrDataset.from_hdf5(fpath, name="non-bogus")
assert_array_equal(ds, ds_loaded) # just to do smth useful with ds ;)
# cleanup and ignore stupidity
os.remove(fpath)
os.remove(fpath2)
def query_byid(self, vertexid):
"""Given a vertex ID give us indices of dataset features (voxels)
Parameters
----------
vertexid: int
Index of searchlight center vertex on the surface.
This value should be an element in self.ids
Returns
-------
voxel_ids: list of int or AttrDataset
The linear indices of voxels near the vertex with index vertexid.
If the instance was constructed with add_fa=None, then voxel_ids
is a list; otherwise it is a AttrDataset with additional feature
attributes stored in voxel_ids.fa.
"""
if self._map_voxel_coord is None:
raise ValueError("No voxel mapping - did you train?")
voxel_unmasked_ids = self.voxsel.get(vertexid)
# map into dataset
voxel_dataset_ids = [
self._map_voxel_coord[i] for i in voxel_unmasked_ids
]
voxel_dataset_ids_flat = sum(voxel_dataset_ids, [])
if self._add_fa is not None:
# optionally add additional information from voxsel
ds = AttrDataset(np.asarray(voxel_dataset_ids_flat)[np.newaxis])
for n in self._add_fa:
fa_values = self.voxsel.get_aux(vertexid, n)
assert (len(fa_values) == len(voxel_dataset_ids))
ds.fa[n] = sum(
[[x] * len(ids)
for x, ids in zip(fa_values, voxel_dataset_ids)], [])
return ds
return voxel_dataset_ids_flat
def query_byid(self, vertexid):
"""Given a vertex ID give us indices of dataset features (voxels)
Parameters
----------
vertexid: int
Index of searchlight center vertex on the surface.
This value should be an element in self.ids
Returns
-------
voxel_ids: list of int or AttrDataset
The linear indices of voxels near the vertex with index vertexid.
If the instance was constructed with add_fa=None, then voxel_ids
is a list; otherwise it is a AttrDataset with additional feature
attributes stored in voxel_ids.fa.
"""
if self._map_voxel_coord is None:
raise ValueError("No voxel mapping - did you train?")
voxel_unmasked_ids = self.voxsel.get(vertexid)
# map into dataset
voxel_dataset_ids = [self._map_voxel_coord[i]
for i in voxel_unmasked_ids]
voxel_dataset_ids_flat = sum(voxel_dataset_ids, [])
if self._add_fa is not None:
# optionally add additional information from voxsel
ds = AttrDataset(np.asarray(voxel_dataset_ids_flat)[np.newaxis])
for n in self._add_fa:
fa_values = self.voxsel.get_aux(vertexid, n)
assert(len(fa_values) == len(voxel_dataset_ids))
ds.fa[n] = sum([[x] * len(ids)
for x, ids in zip(fa_values,
voxel_dataset_ids)], [])
return ds
return voxel_dataset_ids_flat
def test_flat_surface_plotting_exception_wrong_size(self):
s = surf.generate_plane((0, 0, 0), (0, 0, 1), (0, 1, 0), 6, 6)
for offset in (-1, 0, 1):
nfeatures = s.nvertices + offset
ds = AttrDataset(samples=np.random.normal(size=(1, nfeatures)))
# simple 1d -- would have worked before as well
np.array([{
'd': np.empty(shape=(2, 3))
}], dtype=object),
# 2d -- before fix would be reconstructed incorrectly
np.array([[{
'd': np.empty(shape=(2, 3))
}]], dtype=object),
# a bit more elaborate storage
np.array([[{
'd': np.empty(shape=(2, 3)),
'k': 33
}] * 2] * 3, dtype=object),
# Swaroop's use-case
AttrDataset(np.array([{
'd': np.empty(shape=(2, 3))
}], dtype=object)),
# as it would be reconstructed before the fix -- obj array of obj arrays
np.array([np.array([{
'd': np.empty(shape=(2, 3))
}], dtype=object)],
dtype=object),
np.array([], dtype='int64'),
))
def test_save_load_object_dtype_ds(obj=None):
"""Test saving of custom object ndarray (GH #84)
"""
aobjf = np.asanyarray(obj).flatten()
if not aobjf.size and externals.versions['hdf5'] < '1.8.7':
raise SkipTest(