Fix get_vector_length incorrectly returning for shared variable without static shape (#1295)

Author: Abhinav-Khot

pytensor/tensor/sharedvar.py

@@ -3,7 +3,6 @@
 import numpy as np
 
 from pytensor.compile import SharedVariable, shared_constructor
-from pytensor.tensor import _get_vector_length
 from pytensor.tensor.type import TensorType
 from pytensor.tensor.variable import TensorVariable
 
@@ -51,11 +50,6 @@ def zero(self, borrow: bool = False):
             self.container.value = 0 * self.container.value
 
 
-@_get_vector_length.register(TensorSharedVariable)
-def _get_vector_length_TensorSharedVariable(var_inst, var):
-    return len(var.get_value(borrow=True))
-
-
 @shared_constructor.register(np.ndarray)
 def tensor_constructor(
     value,

tests/tensor/test_extra_ops.py

@@ -965,8 +965,10 @@ def fn(i, d):
             f_array_array = fn(indices, shape_array)
             np.testing.assert_equal(ref, f_array_array())
 
-            # shape given as an PyTensor variable
-            shape_symb = pytensor.shared(shape_array)
+            # shape given as a shared PyTensor variable with static shape
+            shape_symb = pytensor.shared(
+                shape_array, shape=shape_array.shape, strict=True
+            )
             f_array_symb = fn(indices, shape_symb)
             np.testing.assert_equal(ref, f_array_symb())
 

tests/tensor/test_sharedvar.py

@@ -605,6 +605,7 @@ def test_specify_shape_inplace(self):
         def test_values_eq(self):
             # Test the type.values_eq[_approx] function
             dtype = self.dtype
+
             if dtype is None:
                 dtype = pytensor.config.floatX
 
@@ -691,9 +692,13 @@ def test_scalar_shared_deprecated():
 
 
 def test_get_vector_length():
-    x = pytensor.shared(np.array((2, 3, 4, 5)))
+    arr = np.array((2, 3, 4, 5))
+    x = pytensor.shared(arr, shape=arr.shape, strict=True)
     assert get_vector_length(x) == 4
 
+    with pytest.raises(ValueError):
+        get_vector_length(pytensor.shared(arr))
+
 
 def test_shared_masked_array_not_implemented():
     x = np.ma.masked_greater(np.array([1, 2, 3, 4]), 3)