Add support for nested models in TPE

[iche033]

Depends on gazebosim/sdformat#316

This PR adds support for nested models in TPE plugin (dartsim plugin does not support nested models yet).

Example sdf description of a nested model:

  <model>
      <link>...  </link>
      <model>
      ...
      </model>
  <model>

The main changes are:

• Added functions to allow Models to construct child models.
• Added support for getting canonical links from a nested model tree
• Updated SetWorldPose in FreeGroup to work with nested models

[scpeters]

@azeey so now we want to declare that ignition::physics::sdf::ConstructSdfModel does not support nested models? If we require SDFormat 9.3, then we can use the new DOM API's to check if an sdf::Model passed to this API has any nested models. Should we add this as a check in our shim function (the one usually in a detail folder)?

• https://github.com/ignitionrobotics/ign-physics/blob/ignition-physics2_2.2.0/sdf/include/ignition/physics/sdf/ConstructModel.hh#L48-L55

At a minimum, we should document the new restriction on the ignition::physics::sdf::ConstructSdfModel feature. We can do this in a separate PR, as long as this doesn't get released before the follow-up is merged

[iche033]

At a minimum, we should document the new restriction on the ignition::physics::sdf::ConstructSdfModel feature. We can do this in a separate PR, as long as this doesn't get released before the follow-up is merged

done. added comment in ef6503e

[codecov]

Codecov Report

Merging #86 into ign-physics2 will increase coverage by 0.15%.
The diff coverage is 86.23%.

@@               Coverage Diff                @@
##           ign-physics2      #86      +/-   ##
================================================
+ Coverage         83.28%   83.43%   +0.15%     
================================================
  Files               106      106              
  Lines              3804     3900      +96     
================================================
+ Hits               3168     3254      +86     
- Misses              636      646      +10     

Impacted Files	Coverage Δ
sdf/include/ignition/physics/sdf/ConstructModel.hh	100.00% <ø> (ø)
dartsim/src/SDFFeatures.cc	66.57% <25.00%> (-0.49%)	⬇️
tpe/plugin/src/FreeGroupFeatures.cc	80.00% <76.00%> (+2.22%)	⬆️
tpe/plugin/src/SDFFeatures.cc	83.08% <86.48%> (+3.08%)	⬆️
tpe/lib/src/Model.cc	98.33% <96.96%> (-1.67%)	⬇️
...clude/ignition/physics/sdf/ConstructNestedModel.hh	100.00% <100.00%> (ø)
tpe/plugin/src/Base.hh	98.00% <100.00%> (ø)

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update c5ad325...359ca9a. Read the comment docs.

[scpeters]

@iche033 I've made a prerelease of sdformat 9.3.0; to use prereleases in CI for this PR, we need the following: gazebo-tooling/gzdev#16

[scpeters]

@osrf-jenkins run tests please

[scpeters]

@osrf-jenkins run tests please

[scpeters]

CI looks good with the prerelease, except for the slight decrease in code coverage

[scpeters]

will a std::set maintain the same order of nested models?

[iche033]

I don't think so. Similarly, all the children (links and nested mdoels) in a TPE Model are stored in an std::map which also does not guarantee order. Will that be an issue?

[scpeters]

the spec calls for the first link to be the canonical link if the attribute //model/@canonical_link is unspecified, so if this ignores the xml order then the choice of canonical link may not be consistent

it's a bit pedantic I guess. we can make an issue for it ("TPE canonical link choice may be inconsistent") instead of spending more time on it

[iche033]

ok I added some logic to mark the first link added to the model as the canonical link, and also added a todo note about handling (or preventing) removal of this link. The order in which the links are added should be consistent with the xml order so the canonical link should also be consistent now. d2a9481

[scpeters]

the handling of links looks good now. for a model with no links and multiple nested models, I'm not sure that it will always pick the same model since it is storing them in a std::set, but I'm fine with adding a todo comment and opening an issue instead of iterating further

[iche033]

sounds good. Added note in 359ca9a and ticketed issue #100

[scpeters]

I think we can ignore the abichecker failure, since it was not configured to use prereleases. It should be fixed by gazebo-tooling/release-tools#273, but we don't need to wait for that.

[scpeters]

looks ok to me, though I haven't tested it

[iche033]

@osrf-jenkins run tests please