Parsing floats in URDF yields wrong value depending on locale

[dbasa]

urdfdom_headers uses std::stod to parse floating point values like position (xyz) and rotation (rpy) from URDF files. The result of std::stod is locale dependant. So with e.g. a German locale (de_DE.UTF-8) "1.57079632679" is parsed to 1.
This is the case in urdfdom_headers/urdf_model/include/urdf_model/pose.h Line 70 (not necessarily exclusive)

In my opinion the issue described here can be linked to this error, as can be the issue described here.

This link states that "[...] using a stringstream to convert it worked perfectly fine."

If desired I can fix this issue using stringstream and create a pull request, but I have to admit I have no background knowledge in ros package building and testing procedures.

Cheers

[traversaro]

Related urdfdom issue: ros/urdfdom#98 .

[clalancette]

So I ended up looking into this a bit. @traversaro did some good investigating of the issue in robotology/idyntree#288 (comment), and digging down into the URDF schema, this is represented as an xs:double. Looking at the XSD document here, it says that the mantissa for a double is represented as a "decimal" number. And looking at the XSD document here, a decimal number is a "finite-length sequence of decimal digits (#x30-#x39) separated by a period as a decimal indicator". Thus, the locale should have no effect on how the document is parsed, and using str::stod is incorrect. I've prepared a patch for urdfdom_headers which switches to using the stream operator; I'll open that presently.

[greimela-si]

We also ran into this issue over at ros2/rviz when migrating the RobotModel display.