[L-IE] supporting fracture intersections in mechanics #2235

norihiro-w · 2018-10-08T23:14:57Z

supporting branch/junction intersections of fractures in 2D M problems

norihiro-w · 2018-12-03T00:36:07Z

i can't find out why Jenkins is failing. anyone can help me?

endJunction · 2018-12-03T11:05:08Z

I think it's a temporary internet access error:
Get https://registry-1.docker.io/v2/library/ubuntu/manifests/16.04: received unexpected HTTP status: 503 Service Unavailable
Just restarted appveyor and jenkins.

norihiro-w · 2018-12-03T12:39:36Z

@endJunction can you please restart the Jenkins again?

endJunction · 2018-12-03T14:50:41Z

The jenkins run also fine, but has new warnings; Github cannot display the "unstable" build, so it is displaying a failure. The warnings are:
on windows: for postutils.cpp: 59 and 67: 'return': conversion from '__int64' to 'int', possible loss of data and 'return': conversion from 'const __int64' to 'unsigned int', possible loss of data
Then there are new warning for doxygen in levelsetfunction.h; there the @parm is wrong, should be @param.

The builds automatically starts, if there is a change in commit hash, e.g. by changing the last commit's date.

norihiro-w · 2018-12-04T02:24:04Z

@endJunction thanks!

norihiro-w · 2018-12-04T15:33:22Z

i think now this PR is ready for reviewing

endJunction · 2018-12-04T16:15:02Z

ProcessLib/LIE/Common/BranchProperty.h

+	// unit vector normal to the master fracture in a direction to the slave
+	Eigen::Vector3d normal_vector_branch;
+
+    virtual ~BranchProperty() = default;


Since nothing is derived from this class, I'd drop the dtor and mark the class final.

Reorder the members as {coords, normal, node_id, further ints} s.t. to reduce the holes in the memory because of alignment.

Not sure, but might be necessary to have aligned new operator http://eigen.tuxfamily.org/dox/group__TopicStructHavingEigenMembers.html

endJunction · 2018-12-04T16:18:18Z

ProcessLib/LIE/Common/FractureProperty.h

+	MeshLib::Node const& branchNode,
+	FractureProperty const& master_frac,
+	FractureProperty const& slave_frac,
+	BranchProperty& branch)


Consideration: I'd rather return a new Branch object, or maybe create a constructor to the Branch class.

endJunction · 2018-12-04T16:19:28Z

ProcessLib/LIE/Common/LevelSetFunction.cpp


 namespace
 {
 // Heaviside step function
 inline double Heaviside(double v)
 {
-    return (v < 0.0) ? 0.0 : 1.0;
+	return (v < 0.0) ? 0.0 : 1.0;


It seems, that the clang-format is not using the .clang-format from the source root...

will check it

endJunction · 2018-12-04T16:21:49Z

ProcessLib/LIE/Common/LevelSetFunction.cpp

+}
+
+template <class T_VEC, class T_V>
+inline bool inList(T_VEC const& vec, T_V const& v)


Would suggest to move it to BaseLib/Algorithm.h, and rename to contains or similar, because it is not acting on lists only.

endJunction · 2018-12-04T16:24:45Z

ProcessLib/LIE/Common/LevelSetFunction.cpp

+	std::unordered_map<int,int> const& fracID_to_local,
+	Eigen::Vector3d const& x)
+{
+	auto this_frac_local_index = fracID_to_local.at(this_frac_id);


Not sure if bound checked access is intended; if not use operator[].

update: i remember it now. i had to use at() because fracID_to_local is const. operator[] cannot be used because it is non-const.

endJunction · 2018-12-04T16:27:25Z

ProcessLib/LIE/Common/MeshUtils.cpp

+            frac_nodeID_to_matIDs[node->getID()].push_back(vec_fracture_mat_IDs[i]);
+
+    auto opt_material_ids(
+        mesh.getProperties().getPropertyVector<int>("MaterialIDs"));


There is MeshLib::materialIDs(mesh) function returning a pointer with additional checks of the property.

endJunction · 2018-12-04T16:32:51Z

ProcessLib/LIE/Common/PostUtils.cpp

+    return (itr2 != vec_nodes.end());
+}
+
+std::vector<int> const& getmatids(


longer names are preferred and use camelCase for function names, e.g. getMaterialIdsForNode(...)

endJunction · 2018-12-04T16:39:51Z

ProcessLib/LIE/SmallDeformation/SmallDeformationProcess.cpp

+
+        master_frac.branches_master.emplace_back(branch);
+
+        BranchProperty* branch2 = new BranchProperty();


In the new code new is not welcome (unless you have to as in copying the MeshLib's nodes). I also didn't see the reason to store pointers in the *.branches_master/slave vectors instead of the branch objects.
In any case this is a memory leak.
I was wrong, no memory leak, because the branches_master/slave are vectors of unique_ptr's.
Still the question remains, why store pointers instead of values?

Because the number of branches varies depending on problem, i store them as a vector of its pointers. i am not sure if I understand your question.

wenqing · 2018-12-05T16:37:50Z

ProcessLib/LIE/Common/BranchProperty.h

+{
+struct BranchProperty
+{
+	int node_id;


These members can be set as const by introducing a constructor.

yes it can be const but i don't see a definitive reason for it.

wenqing · 2018-12-05T16:39:26Z

ProcessLib/LIE/Common/FractureProperty.h

    computeNormalVector(e, dim, frac_prop.normal_vector);
    frac_prop.R.resize(dim, dim);
    computeRotationMatrix(e, frac_prop.normal_vector, dim, frac_prop.R);
 }

+
+inline void setBranchProperty(


can be converted to a constructor of BranchProperty.

here i intentionally separate data and functions (to let myself focus on the data structure during coding).

wenqing

continue my previous review

wenqing · 2018-12-07T09:09:50Z

ProcessLib/LIE/Common/FractureProperty.h

    computeNormalVector(e, dim, frac_prop.normal_vector);
    frac_prop.R.resize(dim, dim);
    computeRotationMatrix(e, frac_prop.normal_vector, dim, frac_prop.R);
 }

+inline BranchProperty* createBranchProperty(MeshLib::Node const& branchNode,


I see that this function is called to fill _process_data._vec_junction_property which is a vector of unique_ptr. I think that you can return std::unique_ptr<BranchProperty> in this function. Then use std::move in line 139 of file
ProcessLib/LIE/SmallDeformation/SmallDeformationProcess.cpp.

i don’t like to use unique ptr too much. current implementation is already readable.

It is just related to the operator of new. If a new is used, one may try to find a delete. If you prefer to use it, it is OK. All other stuffs are OK for me too.

benchmarks failed.

wenqing · 2018-12-07T09:09:56Z

ProcessLib/LIE/Common/FractureProperty.h

+                                            FractureProperty const& master_frac,
+                                            FractureProperty const& slave_frac)
+{
+    BranchProperty* branch = new BranchProperty();


new can be replace with make_unique if the return type is change to unique_ptr.

same as above

wenqing · 2018-12-07T09:10:13Z

ProcessLib/LIE/Common/FractureProperty.h

+    return branch;
+}
+
+inline JunctionProperty* createJunctionProperty(


Here the same.

wenqing · 2018-12-07T09:12:00Z

ProcessLib/LIE/Common/LevelSetFunction.cpp

+    for (std::size_t i = 0; i < junction_props.size(); i++)
+    {
+        auto const* junction = junction_props[i];
+        auto fid1 = fracID_to_local.at(junction->fracture_IDs[0]);


It is good to add const to these three definitions.

i don’t see reason for it, though fid1 is already const

wenqing · 2018-12-07T09:12:25Z

ProcessLib/LIE/Common/LevelSetFunction.cpp

+                fracID_to_local.end())
+                continue;
+
+            double singned = boost::math::sign(


not important as the scope is limited

wenqing · 2018-12-07T09:13:34Z

ProcessLib/LIE/Common/LevelSetFunction.cpp

+        if (!BaseLib::contains(junction->fracture_IDs, this_frac.fracture_id))
+            continue;
+
+        auto another_frac_id =


not important as the scope is limited

wenqing · 2018-12-07T09:38:01Z

ProcessLib/LIE/HydroMechanics/LocalAssembler/HydroMechanicsLocalAssemblerMatrixNearFracture.h

@@ -72,6 +76,11 @@ class HydroMechanicsLocalAssemblerMatrixNearFracture

    static const int displacement_jump_index =
        displacement_index + displacement_size;
+
+    std::vector<FractureProperty*> _fracture_props;


Might be good to use std::vector<std::reference_wrapper<const unique_ptr> > because the vector is pushed with process_data.fracture_property.get().

thanks for the suggestion. however that way is too complicated for me

wenqing · 2018-12-07T09:46:14Z

...sLib/LIE/HydroMechanics/LocalAssembler/HydroMechanicsLocalAssemblerMatrixNearFracture-impl.h

@@ -75,9 +79,9 @@ void HydroMechanicsLocalAssemblerMatrixNearFracture<

    // levelset value of the element
    // remark: this assumes the levelset function is uniform within an element
-    auto const& fracture_props = *_process_data.fracture_property;


fracture_props was set as member of the class as _fracture_props. However I could not found where _fracture_props is initialized. May be I am wrong.

fracture_property is initialized in CreateSmallDeformationProcess.cpp and moved to the process data. see https://github.com/ufz/ogs/blob/master/ProcessLib/LIE/SmallDeformation/CreateSmallDeformationProcess.cpp#L143

I see that is initialized after creating its owner.

ufz#2278

ogsbot · 2020-06-19T12:55:15Z

OpenGeoSys development has been moved to GitLab.

See this pull request on GitLab.

bilke force-pushed the master branch from 7e4d94d to 26445e2 Compare November 16, 2018 13:23

norihiro-w force-pushed the lie-intersects-pr branch 3 times, most recently from 8b582ba to bb1a621 Compare December 3, 2018 00:02

endJunction added the please review label Dec 4, 2018

endJunction reviewed Dec 4, 2018

View reviewed changes

wenqing reviewed Dec 5, 2018

View reviewed changes

norihiro-w force-pushed the lie-intersects-pr branch 2 times, most recently from 5e27400 to f8cf098 Compare December 7, 2018 00:19

wenqing reviewed Dec 7, 2018

View reviewed changes

norihiro-w closed this Dec 8, 2018

norihiro-w reopened this Dec 8, 2018

norihiro-w added 5 commits December 10, 2018 15:50

[LIE] extend MeshUtils to find intersections

e525556

[LIE] add Branch/JunctionProperty

dbfa9b2

[LIE] extend levelset for interesections

778dd29

[LIE] primitive support of intersections in SmallDeformation process

659be2a

[LIE] update postLie for intersections

aa3e1e7

norihiro-w and others added 15 commits December 10, 2018 15:50

[LIE] fix compiling HM

c13ba33

[LIE] add two benchmarks for intersected cracks in M

ef01b43

[LIE] fix post

907d0cd

[LIE] clang-format

5ad68b5

[LIE] bugfix in levelset

7ec0407

[Tests] add sfrac to LIE/M

659c08f

[LIE] add sfrac test

d2443b1

[Base] add contains() to Algorithm

4675690

[LIE] use contains() in BaseLib

b9084fa

[LIE] use MeshLib::materialIDs

a5e2711

[LIE] rename getmatids to getMaterialIdsForNode

0e4a457

[LIE] corrected branch and junction properties

d72513c

[LIE] fix compiling

9b3e127

[LIE] clang-format.

e287c11

[T] LIE; Update output according to PR ufz#2278.

e3d6fa9

ufz#2278

endJunction force-pushed the lie-intersects-pr branch from 7e27f44 to e3d6fa9 Compare December 10, 2018 15:49

[PL] LIE; Fix doxygen errors in LevelSetFunction.

46cc9d2

endJunction force-pushed the lie-intersects-pr branch from 4904a1b to 46cc9d2 Compare December 10, 2018 16:24

endJunction merged commit bfa0fdf into ufz:master Dec 10, 2018

endJunction removed the please review label Dec 10, 2018

endJunction mentioned this pull request Dec 10, 2018

LIE intersect (continued). #2293

Merged

norihiro-w deleted the lie-intersects-pr branch September 26, 2022 15:32


		master_frac.branches_master.emplace_back(branch);

		BranchProperty* branch2 = new BranchProperty();

[L-IE] supporting fracture intersections in mechanics #2235

[L-IE] supporting fracture intersections in mechanics #2235

Conversation

norihiro-w commented Oct 8, 2018 • edited Loading

norihiro-w commented Dec 3, 2018

endJunction commented Dec 3, 2018

norihiro-w commented Dec 3, 2018

endJunction commented Dec 3, 2018 • edited Loading

norihiro-w commented Dec 4, 2018

norihiro-w commented Dec 4, 2018

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

endJunction Dec 4, 2018 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

wenqing left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

ogsbot commented Jun 19, 2020

norihiro-w commented Oct 8, 2018 •

edited

Loading

endJunction commented Dec 3, 2018 •

edited

Loading

endJunction Dec 4, 2018 •

edited

Loading