wildmeshing / wildmeshing-toolkit Goto Github PK

https://github.com/Philip-Trettner/probabilistic-quadrics

https://github.com/wangbolun300/fast-envelope/blob/3574917ccbce6e873cba22ce00c20629ce24970c/src/common_algorithms.cpp

https://github.com/wangbolun300/fast-envelope/blob/3574917ccbce6e873cba22ce00c20629ce24970c/src/Morton.cpp

run ./app_unit_tests triangle-insertion-parallel on bunny to reproduce the error. (code in branch optimize-triangle-insertion)

This function indicates the number of tetrahedra, but not accounting for the removed ones, maybe we need a new name.

This is more similar to capacity in std language.

better loggers (careful with thread-safe init):

• https://github.com/adobe/lagrange/blob/2a49d3ee93c1f1e712c93c5c87ea25b9a83c8f40/modules/core/include/lagrange/Logger.h#L30
• https://github.com/adobe/lagrange/blob/2a49d3ee93c1f1e712c93c5c87ea25b9a83c8f40/modules/core/src/Logger.cpp#L29

@jdumas

Try to match paper results.

Also, the armadillo starts from CGAL.

they all should take the mesh as input and eid is wrong

This is strange behavior, some bug may lie inside.

see also, igl::decimate, and maybe also compare with its result.

need to check boundary

in ShortestEdgeCollapse.cpp:51-52

• Do not use vid to query. Use tuple instead.
• Get one ring edges return only the connected ones, that is not enough. All the previous ones from the tet are invalid now, and needs renewal.

The following test would fail. Switching tet changes the edge it points to.

TEST_CASE("switch_face_tet", "[test_tuple]")
{
    TetMesh m;
    m.init(5, {{{0, 1, 2, 3}}, {{0, 1, 4, 2}}, {{0, 1, 3, 4}}});
    auto e = m.tuple_from_edge(0, 3);

    spdlog::critical("edge {} {} {} {}", e.tid(), e.fid(m), e.eid(m), e.vid());
    e = e.switch_face(m);
    spdlog::critical("edge {} ({}) {} {}", e.tid(), e.fid(m), e.eid(m), e.vid());
    auto edge0 = e.eid(m);
    e = e.switch_tetrahedron(m).value();
    spdlog::critical("edge ({}) {} {} {}", e.tid(), e.fid(m), e.eid(m), e.vid());
    auto edge1 = e.eid(m);
    REQUIRE(edge0 == edge1);
}

The output is as follows, note that the last line changes the id of the edge from 3 to 0.

[2021-12-14 17:36:24.441] [critical] edge 0 1 3 0
[2021-12-14 17:36:24.442] [critical] edge 0 (2) 3 0
[2021-12-14 17:36:24.442] [critical] edge (2) 2 0 0

remove:

• cout from test
• insertion_update_surface_tag
• add_tet_centroid, maybe not remove
• remove tringle soup and change into trimesh
• add mapping for everything else but faces as it is already there

Let's try to have only tests that take < 1 second. @yfzhou42 could you please shorten all the tests for the 2D operations?

Additional Test for Collapse:

• edges on the boundary, and internal edges connecting the boundary
• test timestamp and revert operations

https://github.com/wildmeshing/wildmeshing-toolkit/blob/main/.github/workflows/continuous.yml#L126

We need more tests on real meshes, like bunny.

This requires a system like igl_download_tutorial_data to be in place to avoid additional burden on git.

• check during collapse operation
• Vertex/Face/Edge/Tet Invariants

The name for the timestamp/hash is different in trimesh and tetmesh. We need to decide on one and update it accordingly.

• Envelope check in post-conditions.

Currently not tested.

• Bunny isn't looking correct.

in ShortestEdgeCollapse.cpp:51-52

• Do not use vid to query. Use tuple instead.
• Get one ring edges return only the connected ones, that is not enough. All the previous ones from the tet are invalid now, and needs renewal.