BUG: Update Ad operator evaluation

Should have been included in previous commits
pmgbergen · Feb 7, 2025 · a4123a6 · a4123a6
1 parent f931949
commit a4123a6
Show file tree

Hide file tree

Showing 2 changed files with 18 additions and 12 deletions.
diff --git a/src/porepy/numerics/nonlinear/line_search.py b/src/porepy/numerics/nonlinear/line_search.py
@@ -341,14 +341,14 @@ def compute_constraint_weights(
         f_1_vals = cast(
             np.ndarray,
             model.equation_system.evaluate(
-                constraint_function, x_0 + solution_update * b
+                constraint_function, state=x_0 + solution_update * b
             ),
         )
         f_1 = f_1_vals[crossing_inds]
 
         def f(x):
             return model.equation_system.evaluate(
-                constraint_function, x_0 + solution_update * x
+                constraint_function, state=x_0 + solution_update * x
             )[crossing_inds]
 
         alpha, a, b = self.recursive_spline_interpolation(
@@ -658,12 +658,12 @@ def constraint_weights(
         f_1 = cast(
             np.ndarray,
             model.equation_system.evaluate(
-                constraint_function, x_0 + max_weight * solution_update
+                constraint_function, state=x_0 + max_weight * solution_update
             ),
         )
         weight = max_weight
         weights = max_weight * np.ones(f_1.shape)
-        f_0 = model.equation_system.evaluate(constraint_function, x_0)
+        f_0 = model.equation_system.evaluate(constraint_function, state=x_0)
         active_inds = np.ones(f_1.shape, dtype=bool)
         for i in range(10):
             # Only consider dofs where the constraint indicator has changed sign.
@@ -711,7 +711,7 @@ def constraint_weights(
             f_1 = cast(
                 np.ndarray,
                 model.equation_system.evaluate(
-                    constraint_function, x_0 + weight * solution_update
+                    constraint_function, state=x_0 + weight * solution_update
                 ),
             )
             active_inds = np.logical_and(

diff --git a/tests/numerics/ad/test_operators.py b/tests/numerics/ad/test_operators.py
@@ -574,15 +574,17 @@ def _compare_ad_objects(a, b):
     inds_var = np.hstack(
         [eq_system.dofs_of(eq_system.get_variables([var], [g])) for g in subdomains]
     )
-    assert np.allclose(true_iterate[inds_var], eq_system.evaluate(var_ad, true_iterate))
+    assert np.allclose(
+        true_iterate[inds_var], eq_system.evaluate(var_ad, state=true_iterate)
+    )
 
     # Check evaluation when no state is passed to the parser, and information must
     # instead be glued together from the MixedDimensionalGrid
     assert np.allclose(true_iterate[inds_var], eq_system.evaluate(var_ad))
 
     # Evaluate the equation using the double iterate
     assert np.allclose(
-        2 * true_iterate[inds_var], eq_system.evaluate(var_ad, double_iterate)
+        2 * true_iterate[inds_var], eq_system.evaluate(var_ad, state=double_iterate)
     )
 
     # Represent the variable on the previous time step. This should be a numpy array
@@ -593,7 +595,9 @@ def _compare_ad_objects(a, b):
 
     # Also check that state values given to the ad parser are ignored for previous
     # values
-    assert np.allclose(prev_evaluated, eq_system.evaluate(prev_var_ad, double_iterate))
+    assert np.allclose(
+        prev_evaluated, eq_system.evaluate(prev_var_ad, state=double_iterate)
+    )
 
     ## Next, test edge variables. This should be much the same as the grid variables,
     # so the testing is less thorough.
@@ -607,7 +611,7 @@ def _compare_ad_objects(a, b):
         [eq_system.dofs_of([var]) for var in variable_interfaces]
     )
     interface_values = np.hstack(
-        [eq_system.evaluate(var, true_iterate) for var in variable_interfaces]
+        [eq_system.evaluate(var, state=true_iterate) for var in variable_interfaces]
     )
     assert np.allclose(
         true_iterate[interface_inds],
@@ -622,11 +626,13 @@ def _compare_ad_objects(a, b):
     ind1 = eq_system.dofs_of(eq_system.get_variables([var], [g]))
     ind2 = eq_system.dofs_of(eq_system.get_variables([var2], [g]))
 
-    assert np.allclose(true_iterate[ind1], eq_system.evaluate(v1, true_iterate))
-    assert np.allclose(true_iterate[ind2], eq_system.evaluate(v2, true_iterate))
+    assert np.allclose(true_iterate[ind1], eq_system.evaluate(v1, state=true_iterate))
+    assert np.allclose(true_iterate[ind2], eq_system.evaluate(v2, state=true_iterate))
 
     v1_prev = v1.previous_timestep()
-    assert np.allclose(true_state[ind1], eq_system.evaluate(v1_prev, true_iterate))
+    assert np.allclose(
+        true_state[ind1], eq_system.evaluate(v1_prev, state=true_iterate)
+    )
 
 
 @pytest.mark.parametrize("prev_time", [True, False])