def tocsc(self, copy=False):
"""Converts the matrix to Compressed Sparse Column format.
Args:
copy (bool): If ``False``, it shares data arrays as much as
possible. Actually this option is ignored because all
arrays in a matrix cannot be shared in dia to csc conversion.
Returns:
cupyx.scipy.sparse.csc_matrix: Converted matrix.
"""
if self.data.size == 0:
return csc.csc_matrix(self.shape, dtype=self.dtype)
num_rows, num_cols = self.shape
num_offsets, offset_len = self.data.shape
row, mask = core.ElementwiseKernel(
'int32 offset_len, int32 offsets, int32 num_rows, '
'int32 num_cols, T data', 'int32 row, bool mask', '''
int offset_inds = i % offset_len;
row = offset_inds - offsets;
mask = (row >= 0 && row < num_rows && offset_inds < num_cols
&& data != T(0));
''', 'dia_tocsc')(offset_len, self.offsets[:, None], num_rows,
num_cols, self.data)
indptr = cupy.zeros(num_cols + 1, dtype='i')
indptr[1:offset_len + 1] = cupy.cumsum(mask.sum(axis=0))
indptr[offset_len + 1:] = indptr[offset_len]
indices = row.T[mask.T].astype('i', copy=False)
data = self.data.T[mask.T]
return csc.csc_matrix((data, indices, indptr),
shape=self.shape,
dtype=self.dtype)
def _compressed_sparse_stack(blocks, axis):
"""Fast path for stacking CSR/CSC matrices
(i) vstack for CSR, (ii) hstack for CSC.
"""
other_axis = 1 if axis == 0 else 0
data = cupy.concatenate([b.data for b in blocks])
constant_dim = blocks[0].shape[other_axis]
idx_dtype = sputils.get_index_dtype(arrays=[b.indptr for b in blocks],
maxval=max(data.size, constant_dim))
indices = cupy.empty(data.size, dtype=idx_dtype)
indptr = cupy.empty(sum(b.shape[axis]
for b in blocks) + 1, dtype=idx_dtype)
last_indptr = idx_dtype(0)
sum_dim = 0
sum_indices = 0
for b in blocks:
if b.shape[other_axis] != constant_dim:
raise ValueError(
'incompatible dimensions for axis %d' % other_axis)
indices[sum_indices:sum_indices+b.indices.size] = b.indices
sum_indices += b.indices.size
idxs = slice(sum_dim, sum_dim + b.shape[axis])
indptr[idxs] = b.indptr[:-1]
indptr[idxs] += last_indptr
sum_dim += b.shape[axis]
last_indptr += b.indptr[-1]
indptr[-1] = last_indptr
if axis == 0:
return csr.csr_matrix((data, indices, indptr),
shape=(sum_dim, constant_dim))
else:
return csc.csc_matrix((data, indices, indptr),
shape=(constant_dim, sum_dim))
def transpose(self, axes=None, copy=False):
"""Returns a transpose matrix.
Args:
axes: This option is not supported.
copy (bool): If ``True``, a returned matrix shares no data.
Otherwise, it shared data arrays as much as possible.
Returns:
cupyx.scipy.sparse.spmatrix: Transpose matrix.
"""
if axes is not None:
raise ValueError(
'Sparse matrices do not support an \'axes\' parameter because '
'swapping dimensions is the only logical permutation.')
shape = self.shape[1], self.shape[0]
return csc.csc_matrix(
(self.data, self.indices, self.indptr), shape=shape, copy=copy)
def tocsc(self, copy=False):
"""Converts the matrix to Compressed Sparse Column format.
Args:
copy (bool): If ``False``, it shares data arrays as much as
possible. Actually this option is ignored because all
arrays in a matrix cannot be shared in coo to csc conversion.
Returns:
cupyx.scipy.sparse.csc_matrix: Converted matrix.
"""
if self.nnz == 0:
return csc.csc_matrix(self.shape, dtype=self.dtype)
# copy is silently ignored (in line with SciPy) because both
# sum_duplicates and coosort change the underlying data
x = self.copy()
x.sum_duplicates()
cusparse.coosort(x, 'c')
x = cusparse.coo2csc(x)
x.has_canonical_format = True
return x
