ENH: sparse: Generalize coo_array to support 1d shapes

scipy/sparse/_base.py

@@ -67,7 +67,15 @@ class _spbase:
 
     __array_priority__ = 10.1
     _format = 'und'  # undefined
-    ndim = 2
+
+    @property
+    def ndim(self) -> int:
+        return len(self._shape)
+
+    @property
+    def _shape_as_2d(self):
+        s = self._shape
+        return (1, s[-1]) if len(s) == 1 else s
 
     @property
     def _bsr_container(self):
@@ -355,9 +363,11 @@ def imag(self):
     def __repr__(self):
         _, format_name = _formats[self.format]
         sparse_cls = 'array' if isinstance(self, sparray) else 'matrix'
-        return f"<%dx%d sparse {sparse_cls} of type '%s'\n" \
-               "\twith %d stored elements in %s format>" % \
-               (self.shape + (self.dtype.type, self.nnz, format_name))
+        shape_str = 'x'.join(str(x) for x in self.shape)
+        return (
+            f"<{shape_str} sparse {sparse_cls} of type '{self.dtype.type}'\n"
+            f"\twith {self.nnz} stored elements in {format_name} format>"
+        )
 
     def __str__(self):
         maxprint = self._getmaxprint()
@@ -568,7 +578,11 @@ def _mul_dispatch(self, other):
         # This method has to be different from `__matmul__` because it is also
         # called by sparse matrix classes.
 
-        M, N = self.shape
+        # Currently matrix multiplication is only supported
+        # for 2D arrays. Hence we unpacked and use only the
+        # two last axes' lengths.
+        N = self.shape[-1]
+        M = self.shape[-2] if self.ndim > 1 else 1
 
         if other.__class__ is np.ndarray:
             # Fast path for the most common case

scipy/sparse/_construct.py

@@ -931,7 +931,7 @@ def block_diag(mats, format=None, dtype=None):
     c_idx = 0
     for a in mats:
         if isinstance(a, (list, numbers.Number)):
-            a = coo_array(a)
+            a = coo_array(np.atleast_2d(a))
         nrows, ncols = a.shape
         if issparse(a):
             a = a.tocoo()