Feature/add urdf parser

[Levi-Armstrong]

This add additional functionality over what urdfdom provides related to additional shapes and support of quaternions for the origin tag.

Additional Shapes

• Cone
• Capsule
• Mesh
• Convex Mesh with ability to convert to Convex mesh
• SDF Mesh
• Octomap with sub shape types (Box, Sphere Inside, Sphere Outside) and an option to prune. This has a custom prune operation to limit the number of sub shapes created. Also provides support for supplying point cloud to define the octomap with resolution.
• Origin with quaternion support for rotation. You are able to define both to not break comparability with urdfdom but this solver default to quaternion if both are available.

[traversaro]

In the proposed parser std::stod is used for converting from string to double, but unfortunately std::stod depends on the currently set global locale, so this would make this parser subject to subtle locale-related bugs such as the one described in ros/urdfdom_headers#41, ros/urdfdom#98 and robotology/idyntree#288 .

[Levi-Armstrong]

@traversaro Thank you for pointing out the issue with using std::stod. I have made changes based on one of the links you provided to avoid using std::stod.

[Levi-Armstrong]

@mpowelson This should be ready to review. I currently working on adding documentation on the additional feature available in this parser over urdfdom.

[mpowelson]

@Levi-Armstrong Is there a minimum version of tinyxml? On 16.04 (kinetic) I get

error: no matching function for call to 'tinyxml2::XMLElement::DoubleAttribute(const char [17], int) const'
   s->soft_lower_limit = xml_element->DoubleAttribute("soft_lower_limit", 0);
                                                                           ^
In file included from /home/mpowelson/workspaces/temp2/src/tesseract-1/tesseract/tesseract_urdf/include/tesseract_urdf/urdf_parser.h:33:0,
                 from /home/mpowelson/workspaces/temp2/src/tesseract-1/tesseract/tesseract_urdf/src/urdf_parser.cpp:27:
/usr/include/tinyxml2.h:1205:14: note: candidate: double tinyxml2::XMLElement::DoubleAttribute(const char*) const
     double   DoubleAttribute( const char* name ) const {
              ^'
/usr/include/tinyxml2.h:1205:14: note:   candidate expects 1 argument, 2 provided

[mpowelson]

That review was not done. For some reason when I commented, it posted the whole review. If you want, I'll have to look at the rest next week.

[mpowelson]

Sorry in advance for the sprawling comments.

[mpowelson]

@Levi-Armstrong Is there a minimum version of tinyxml? On 16.04 (kinetic) I get

error: no matching function for call to 'tinyxml2::XMLElement::DoubleAttribute(const char [17], int) const'
   s->soft_lower_limit = xml_element->DoubleAttribute("soft_lower_limit", 0);
                                                                           ^
In file included from /home/mpowelson/workspaces/temp2/src/tesseract-1/tesseract/tesseract_urdf/include/tesseract_urdf/urdf_parser.h:33:0,
                 from /home/mpowelson/workspaces/temp2/src/tesseract-1/tesseract/tesseract_urdf/src/urdf_parser.cpp:27:
/usr/include/tinyxml2.h:1205:14: note: candidate: double tinyxml2::XMLElement::DoubleAttribute(const char*) const
     double   DoubleAttribute( const char* name ) const {
              ^'
/usr/include/tinyxml2.h:1205:14: note:   candidate expects 1 argument, 2 provided

@Levi-Armstrong Ok I was mistaken. I still get this error, and it looks like CI does too.

[mpowelson]

@Levi-Armstrong I accidentally pushed it to your branch again. I can submit a PR if you want from my fork with the same name.

[Levi-Armstrong]

@mpowelson It is ready for a second look.

[mpowelson]

@Levi-Armstrong It looks good to me. Because the tests changed files and also were modified in the same commit, I did not look at them again. I assume you fixed the things I pointed out. Everything else looks good once we get the fixes for 16.04 in there.

[Levi-Armstrong]

@mpowelson Can you check the latest and make sure it build on kinetic?

[mpowelson]

@Levi-Armstrong It needs Levi-Armstrong#23

[mpowelson]

@Levi-Armstrong I didn't realize how bad the Eigen alignment issue was. Basically all of my examples now no longer build with this PR. Until we fix the those I'm not sure we should merge this. I'm not sure how common this issue will be.

[Levi-Armstrong]

@mpowelson I have fixed the parsing of available materials and added unit tests to verify that it works and fails when expected.

[mpowelson]

@Levi-Armstrong Does this change suggest that urdf_dom had a default name, but this one doesn't?

[Levi-Armstrong]

I will test it with urdfdom and see if I get the same error I was getting from tesseract_urdf. It looks like tesseract_motion_planners tests are only failing on kinetic. Could you run the tests on your machine and see why they are failing?

[gavanderhoorn]

Anonymous material definitions are definitely supported by urdfdom, so the name should not be necessary -- if you want to maintain some level of compatibility with urdfdom that is.

[Levi-Armstrong]

@gavanderhoorn Thank you. I have added support for anonymous material names.

[Levi-Armstrong]

@mpowelson It looks like the issue with boost and pcl on xenial is related to pcl not supporting c++11 on xenial. They recommend if this becomes an issue to build pcl from source. To avoid doing this I added a compiler definition to disable the ability to parse point cloud files defined in the urdf for xenial builds. I could not find any workarounds.

[mpowelson]

@Levi-Armstrong The tests run on my PC as well now.