[OptApp / SIApp] Updating the Expressions to TensorAdaptors by sunethwarna · Pull Request #14214 · KratosMultiphysics/Kratos

sunethwarna · 2026-02-17T13:03:36Z

📝 Description
This PR removes all the use cases of Expression in OptimizationApplication and SystemIdentificationApplication, and replaces them with new TensorAdaptors.

This requires following PRs to be merged first.

🆕 Changelog

Replaces the uses of Expression with TensorAdaptor

…_adaptor_update

talhah-ansari

Very impressive. @sunethwarna Thanks

applications/OptimizationApplication/python_scripts/controls/material/simp_control.py

...cations/OptimizationApplication/python_scripts/controls/thickness/shell_thickness_control.py

.../OptimizationApplication/python_scripts/processes/optimization_problem_vtu_output_process.py

...alysis_based_tests/algorithm_relaxed_gradient_projection/test_relaxed_gradient_projection.py

talhah-ansari · 2026-03-05T08:45:17Z

applications/OptimizationApplication/tests/control/material/test_simp_control.py

        for i in range(20):
-            control_field *= 1.2
-            simp_control.Update(control_field)
+            control_field.data += 1.2


is this * -> + intentional? Also line 257

yeah, this is because control_field was zeros before.

talhah-ansari · 2026-03-05T08:46:14Z

...ationApplication/tests/algorithm_tests/scipy_tests/slsqp_shell_thickness_opt/summary_ref.csv

      2,       2.6811e+03,                1.5885e-03
      3,       2.6677e+03,                1.5436e-03
-      4,       2.6191e+03,                1.5046e-03
+      4,       2.6841e+03,                1.5475e-03


Is this value change intentional?

talhah-ansari

Thanks very much @sunethwarna

Igarizza · 2026-03-06T12:41:42Z

applications/OptimizationApplication/python_scripts/algorithms/algorithm_steepest_descent.py

+    def ComputeControlUpdate(self, alpha: Kratos.TensorAdaptors.DoubleCombinedTensorAdaptor) -> None:
+        update: Kratos.TensorAdaptors.DoubleCombinedTensorAdaptor = self.algorithm_data.GetBufferedData()["search_direction"].Clone()
+        update.data[:] *= alpha.data[:]
+        Kratos.TensorAdaptors.DoubleCombinedTensorAdaptor(update, perform_store_data_recursively=False, copy=False).StoreData()


As I understood the documentation, we can simply say update.StoreData() or ??

You can, but if the CombinedTensorAdaptor is created with perform_store_data_recursively=true, then it will do the Store data recursively which will write the values to model part also.

So the way which I did is safer, it does not copy the data, it just perform the store data on existing array, without recursive calling.

What is dangerous to update the MP var, as we are working with OptApp var here only ?

Igarizza

Please answer to my question that I can better understand TA functions

Igarizza · 2026-03-06T12:57:54Z

...s/OptimizationApplication/python_scripts/algorithms/algorithm_relaxed_gradient_projection.py

                # scaling constraints grad
-                norm = KratosOA.ExpressionUtils.NormInf(constr_grad[i])
+                norm = numpy.linalg.norm(constr_grad[i].data.ravel(), ord=numpy.inf)
+                ta = Kratos.TensorAdaptors.DoubleCombinedTensorAdaptor(constr_grad[i], perform_collect_data_recursively=False, perform_store_data_recursively=False)


give please more describtive name to ta. Is it Talhah Ansari?)

hehe.... This stands for TensorAdaptor... ;) [ Talhah can sell it as Talhah Ansari ;)

@talhah-ansari

Igarizza · 2026-03-09T14:47:18Z

applications/OptimizationApplication/custom_python/add_custom_response_utilities_to_python.cpp

    m.def_submodule("LinearStrainEnergyResponseUtils")
        .def("CalculateValue", &LinearStrainEnergyResponseUtils::CalculateValue)
-        .def("CalculateGradient", &LinearStrainEnergyResponseUtils::CalculateGradient, py::arg("list_of_gradient_variables"), py::arg("list_of_gradient_required_model_parts"), py::arg("list_of_gradient_computed_model_parts"), py::arg("list_of_container_expressions"), py::arg("perturbation_size"))
+        .def("CalculateGradient", &LinearStrainEnergyResponseUtils::CalculateGradient, py::arg("physical_variable"), py::arg("value_influencing_model_part"), py::arg("combined_tensor_adaptor"), py::arg("perturbation_size"))


Now this function works with 1 physical variable only?

We never supported computing gradients at once for list of variables in this cpp function. It is done by the python level. So, I corrected it in here.

Following is the block, where we go through the list of variables...

Kratos/applications/OptimizationApplication/python_scripts/responses/linear_strain_energy_response_function.py

Lines 71 to 77 in a334d63

for physical_variable, merged_model_part in merged_model_part_map.items():

KratosOA.ResponseUtils.LinearStrainEnergyResponseUtils.CalculateGradient(

physical_variable,

merged_model_part,

intersected_model_part_map[physical_variable],

physical_variable_collective_expressions[physical_variable].GetContainerExpressions(),

self.perturbation_size)

Igarizza · 2026-03-09T18:30:13Z

applications/OptimizationApplication/custom_utilities/filtering/explicit_filter_utils.cpp

 {
 template<class TContainerType>
-Expression::ConstPointer GetNodalDomainSizeExpression(
+void GetNodalDomainSizes(


This is a bit strange design, get function return void. CalculateNodalDomainSize ?

True.... I changed it to CalculateNodalDomainSize. :D

Igarizza · 2026-03-09T18:35:22Z

applications/OptimizationApplication/custom_utilities/filtering/explicit_filter_utils.cpp

+        << "Only scalar values are allowed for the filter radius tensor adaptor."
+        << "Provided tensor adaptor = " << *pTensorAdaptor << ".\n";
+
+    if (std::holds_alternative<ModelPart::NodesContainerType::Pointer>(pTensorAdaptor->GetContainer())) {


Instead of this IF blocks we can use TMeshDependencyType may be ?

Mmm.... I am not sure I am getting your point. TensorAdaptor can hold pointers to different types of containers such as nodes, elements, conditions, master_slave_constraints. This If is checking what type of a pointer the TensorAdaptor is having. TMeshDependencyType is a compile time thing, and this If block is a runtime thing.

Igarizza · 2026-03-09T18:38:14Z

applications/OptimizationApplication/custom_utilities/filtering/explicit_filter_utils.cpp

-        << "Filter radius container expression model part and filter model part mismatch."
-        << "\n\tFilter = " << *this
-        << "\n\tContainerExpression = " << rContainerExpression;
+    if (std::holds_alternative<ModelPart::NodesContainerType::Pointer>(rTensorAdaptor.GetContainer())) {


may be here as well we can simplify with TMeshDependencyType ?

same as the previous comment

Igarizza · 2026-03-09T18:46:21Z

...ptimizationApplication/custom_utilities/tensor_adaptors/properties_variable_tensor_adaptor.h

+ * @see @ref DataValueContainer::GetValue   Variable value retrieval/update method.
+ * @see @ref DataValueContainer::SetValue   Variable value set method.
+ */
+class KRATOS_API(OPTIMIZATION_APPLICATION) PropertiesVariableTensorAdaptor: public TensorAdaptor<double> {


where do you use these propoerties? I can't find it =(

Kratos/applications/OptimizationApplication/custom_utilities/response/linear_strain_energy_response_utils.cpp

Lines 160 to 161 in c207a56

const auto& sensitivity_variable = KratosComponents<Variable<double>>::Get(pVariable->Name() + "_SENSITIVITY");

PropertiesVariableTensorAdaptor(*p_tensor_adaptor, &sensitivity_variable, false).CollectData();

Igarizza · 2026-03-09T19:33:33Z

applications/OptimizationApplication/custom_utilities/optimization_utils.cpp

    KRATOS_CATCH("");
 }

+TensorAdaptor<double>::Pointer OptimizationUtils::MapContainerDataToNodalData(


add please documentation how it is expected to work

It is there in the optimization_utils.h. We usually add documentation only in the header so that doxygen can read it and create appropriate html files. [ I improved it slightly ]

the documentation is

Kratos/applications/OptimizationApplication/custom_utilities/optimization_utils.h

Lines 115 to 149 in a7d96bf

/**

* @brief Maps container data to nodal data

*

* @details This method returns a mapped container data given in @p rInput to nodal given by @p pNodes .

* It distributes the each entity's data to its nodes equally.

*

* @throws std::runtime_error if the container in @p rInput is not of type conditions or elements.

* @throws std::runtime_error if the entities in @p rInput 's container contain nodes which are not found in the @p pNodes.

*

* @param rInput Input containing values in which needs to be mapped to nodes (container should be conditions or elements).

* @param pNodes List of nodes to which the values will be mapped out.

* @returns Returns a tensor adaptor having the container @p pNodes with the mapped values.

*/

static TensorAdaptor<double>::Pointer MapContainerDataToNodalData(

const TensorAdaptor<double>& rInput,

ModelPart::NodesContainerType::Pointer pNodes);

/**

* @brief Maps nodal data to a specific container data

*

* @details This method returns a mapped nodal data given in @p rInput to container given by @p pEntities .

* It first divide each node's value by number of neighour entities and adds that value to each neighbouring

* entity.

*

* @throws std::runtime_error if the container in @p rInput is not a nodal container.

* @throws std::runtime_error if the nodal container in @p rInput is not same as nodal container in @p rNeighbourCount .

* @throws std::runtime_error if the entities in @p rInput 's container contain nodes which are not found in the nodal container of @p rInput .

*

* @param rInput Input containing values in which needs to be mapped to conditions / elements (container should be nodal).

* @param pEntities List of entities to which the values will be mapped out.

* @param rNeighbourCount Number of neighbour entities around each node.

* @returns Returns a tensor adaptor having the container @p pEntities with the mapped values.

*/

template<class TContainerPointerType>

static TensorAdaptor<double>::Pointer MapNodalDataToContainerData(

Igarizza · 2026-03-09T19:37:34Z

applications/OptimizationApplication/custom_utilities/optimization_utils.cpp


+TensorAdaptor<double>::Pointer OptimizationUtils::MapContainerDataToNodalData(
+    const TensorAdaptor<double>& rInput,
+    ModelPart::NodesContainerType::Pointer pNodes)


can we pass ModelPart and get nodes and conditions out of it??

Adding this function ... how does it relate to TA ?

can we pass ModelPart and get nodes and conditions out of it??

Sure, you can get nodes / conditions / elements from a model part easily by using MP.Nodes , MP.Conditions, and MP.Elements. Then you can convert elemental / condition data to nodal data using this mehtod.

Adding this function ... how does it relate to TA ?

Earlier this method is there with Expression which is used heavily by the ImplcitFilter. Now i updated it to use TensorAdaptors, hence it in this PR.

Igarizza · 2026-03-09T19:43:08Z

...tions/OptimizationApplication/python_scripts/controls/shape/vertex_morphing_shape_control.py

            # model part
            return self.model_part.GetRootModelPart()

+    def __ExtractTensorData(self, variable, tensor_adaptor: Kratos.TensorAdaptors.DoubleTensorAdaptor, nodes_container: Kratos.NodesArray) -> Kratos.TensorAdaptors.DoubleTensorAdaptor:


Why do you need this function ? And what it does? I can't follow. Add please some doc

Added comments. Could you please check if it is descriptive enough?

sunethwarna · 2026-03-09T21:07:32Z

@Igarizza I addressed all of your comments. could you check?

sunethwarna added 30 commits December 20, 2025 00:30

skip some tests if MeshMovingApplication is not found.

e0b72ec

add properties variable TA

b9f841f

expose to python

b6727b2

minor comment change in PropertiesVarTA

e38e3d5

add unit test to PropertiesVarTA

1c8e7ef

update mass resp utils to TA

fc6c076

update linear strain energy resp utils to TA

f238976

fix resp utils python exposure

262d261

add pGetContainer method for ContainerExps

39b448b

update python linear strain energy resp to TA

517a801

update mass response to TA

29cad0c

response utils minor update

c6bfa80

update sigmoidal utils to TA

a307c6e

update sigmoidal python exp to TA

e3314b2

update sigmoidal proj utils tests to TA

885f3bd

update opt_projection.py to TA

e46673a

Merge remote-tracking branch 'origin/master' into optapp_siapp_tensor…

dca09bc

…_adaptor_update

Merge remote-tracking branch 'origin/master' into optapp_siapp_tensor…

1a6cfc2

…_adaptor_update

change explicit damping

48582af

explicit filter utils update

9fe8fe9

integrated nearest entity update

6272421

undate nearest entity update

0e07e2a

Merge remote-tracking branch 'origin/master' into optapp_siapp_tensor…

88f320a

…_adaptor_update

bugfix

3dee71b

update explicit filter python to TA

4195300

update filter utils to TA

6f44f3f

update texplicit filter test to TA

48b2333

minor

34862e4

move TA to submodule

ad5339c

adding nodal neighour count TA

e4e9e89

sunethwarna mentioned this pull request Feb 20, 2026

[Core / OptApp] Port Kratos::Future::VtuOuput to Kratos #14221

Draft

sunethwarna added 5 commits February 21, 2026 08:26

Merge remote-tracking branch 'origin/master' into optapp_siapp_tensor…

1c64625

…_adaptor_update

minor

d9090ef

minor

325b870

Merge remote-tracking branch 'origin/master' into optapp_siapp_tensor…

2b1593a

…_adaptor_update

revert core changes

b3c2df8

sunethwarna marked this pull request as ready for review March 4, 2026 06:23

sunethwarna requested review from Igarizza and talhah-ansari March 4, 2026 06:23

talhah-ansari reviewed Mar 5, 2026

View reviewed changes

remove duplicated lines

7c99c3e

sunethwarna enabled auto-merge March 5, 2026 12:02

sunethwarna requested a review from talhah-ansari March 5, 2026 12:03

talhah-ansari previously approved these changes Mar 5, 2026

View reviewed changes

matekelemen disabled auto-merge March 5, 2026 13:41

Merge branch 'master' into optapp_siapp_tensor_adaptor_update

6915dd4

sunethwarna dismissed talhah-ansari’s stale review via 6915dd4 March 6, 2026 10:21

sunethwarna requested a review from talhah-ansari March 6, 2026 10:28

minor

c207a56

Igarizza reviewed Mar 6, 2026

View reviewed changes

talhah-ansari previously approved these changes Mar 6, 2026

View reviewed changes

Igarizza reviewed Mar 9, 2026

View reviewed changes

sunethwarna added 2 commits March 9, 2026 21:48

change to CalculateNodalDomainSizes

b2fdb62

more descriptive docs

a7d96bf

sunethwarna dismissed talhah-ansari’s stale review via a7d96bf March 9, 2026 20:48

added comments

346935a

sunethwarna requested a review from Igarizza March 9, 2026 21:07

	for physical_variable, merged_model_part in merged_model_part_map.items():
	KratosOA.ResponseUtils.LinearStrainEnergyResponseUtils.CalculateGradient(
	physical_variable,
	merged_model_part,
	intersected_model_part_map[physical_variable],
	physical_variable_collective_expressions[physical_variable].GetContainerExpressions(),
	self.perturbation_size)

	const auto& sensitivity_variable = KratosComponents<Variable<double>>::Get(pVariable->Name() + "_SENSITIVITY");
	PropertiesVariableTensorAdaptor(*p_tensor_adaptor, &sensitivity_variable, false).CollectData();

	/**
	* @brief Maps container data to nodal data
	*
	* @details This method returns a mapped container data given in @p rInput to nodal given by @p pNodes .
	* It distributes the each entity's data to its nodes equally.
	*
	* @throws std::runtime_error if the container in @p rInput is not of type conditions or elements.
	* @throws std::runtime_error if the entities in @p rInput 's container contain nodes which are not found in the @p pNodes.
	*
	* @param rInput Input containing values in which needs to be mapped to nodes (container should be conditions or elements).
	* @param pNodes List of nodes to which the values will be mapped out.
	* @returns Returns a tensor adaptor having the container @p pNodes with the mapped values.
	*/
	static TensorAdaptor<double>::Pointer MapContainerDataToNodalData(
	const TensorAdaptor<double>& rInput,
	ModelPart::NodesContainerType::Pointer pNodes);

	/**
	* @brief Maps nodal data to a specific container data
	*
	* @details This method returns a mapped nodal data given in @p rInput to container given by @p pEntities .
	* It first divide each node's value by number of neighour entities and adds that value to each neighbouring
	* entity.
	*
	* @throws std::runtime_error if the container in @p rInput is not a nodal container.
	* @throws std::runtime_error if the nodal container in @p rInput is not same as nodal container in @p rNeighbourCount .
	* @throws std::runtime_error if the entities in @p rInput 's container contain nodes which are not found in the nodal container of @p rInput .
	*
	* @param rInput Input containing values in which needs to be mapped to conditions / elements (container should be nodal).
	* @param pEntities List of entities to which the values will be mapped out.
	* @param rNeighbourCount Number of neighbour entities around each node.
	* @returns Returns a tensor adaptor having the container @p pEntities with the mapped values.
	*/
	template<class TContainerPointerType>
	static TensorAdaptor<double>::Pointer MapNodalDataToContainerData(

Conversation

sunethwarna commented Feb 17, 2026

Uh oh!

talhah-ansari left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

talhah-ansari left a comment

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Igarizza left a comment

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

sunethwarna commented Mar 9, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!