Add boost serialization for Environment commands and all underlying types

[mpowelson]

This adds boost serialization for types in tesseract_srdf, tesseract_common, tesseract_geometry, and tessseract_environment. This should be enough to fully serialize planning server requests. This also adds == operators for all of these types that is used to test the serialization. Every type should have a serialization unit test.

There are a few places where additional work can be done. For example we may want to add the ability to change how a mesh geometry is serialized and only serialize the path in some cases. I'll leave that for future work for now.

[codecov]

Codecov Report

Merging #726 (54211fd) into master (bfa7976) will decrease coverage by 0.20%.
The diff coverage is 88.71%.

❗ Current head 54211fd differs from pull request most recent head fba8e37. Consider uploading reports for the commit fba8e37 to get more accurate results

@@            Coverage Diff             @@
##           master     #726      +/-   ##
==========================================
- Coverage   90.82%   90.61%   -0.21%     
==========================================
  Files         189      220      +31     
  Lines       12658    13483     +825     
==========================================
+ Hits        11496    12218     +722     
- Misses       1162     1265     +103     

Impacted Files	Coverage Δ
tesseract_common/src/any.cpp	66.66% <ø> (ø)
tesseract_common/src/eigen_serialization.cpp	56.81% <0.00%> (ø)
tesseract_common/src/joint_state.cpp	100.00% <ø> (ø)
tesseract_common/src/kinematic_limits.cpp	100.00% <ø> (ø)
tesseract_common/src/manipulator_info.cpp	97.43% <ø> (ø)
tesseract_common/src/utils.cpp	94.50% <ø> (ø)
...ry/include/tesseract_geometry/impl/mesh_material.h	100.00% <ø> (ø)
tesseract_common/src/allowed_collision_matrix.cpp	76.47% <76.47%> (ø)
...mands/add_contact_managers_plugin_info_command.cpp	80.00% <80.00%> (ø)
...rc/commands/add_kinematics_information_command.cpp	80.00% <80.00%> (ø)
... and 82 more

[Levi-Armstrong]

This seems odd. Did some logic change somewhere?

[mpowelson]

The isIdentical function did change, but it should be doing the same thing. Looking at the test, it seems like this should be false since acm is nullptr.

[Levi-Armstrong]

I think this should still be returning true. It is comparing if the two vectors are equal and order does not matter which they should be equal. Am I missing something?

[mpowelson]

The problem is that pairs is empty. getCollisionObjectPairs only pushes back pairs if acm !=nullptr. At line 28 we aren't passing it in, so it is empty. When we call it at 48, we pass in the acm, so it is the same as check_pairs.

I'm not sure why this was passing before, but the vectors are not the same. One has the pairs above. The other is empty.

[mpowelson]

Actually, I see the reason. If the first vector is empty, nothing was being checked. This is how isIdentical was checking equality

Since vec1 is empty, it just returns true.

    return std::equal(vec1.begin(), vec1.end(), vec2.begin());

However, I added this up above that. So now it fails

  if (vec1.size() != vec2.size())
    return false;

[Levi-Armstrong]

It looks like getCollisionObjectPairs is broken. If acm is nullptr then it should add the pairs. The fix is below which needs to be updated two places in the function. Can you added it to this PR?

Current:

if (acm != nullptr && !acm(l1, l2))
        clp.push_back(tesseract_collision::getObjectPairKey(l1, l2));

Fixed:

if (acm == nullptr || acm != nullptr && !acm(l1, l2))
        clp.push_back(tesseract_collision::getObjectPairKey(l1, l2));

[mpowelson]

Done