Enable implicit conversion and change KinDynComputations holder in the iDynTree bindings

[GiulioRomualdi]

As per the title, this PR changes the holder for KinDynComputations and enables the implicit conversion. The former allows the interoperability between iDynTree swig bindings and the blf pybind11 bindings. The latter simplifies calling methods that take as inputs vectors and/or matrices. For instance, now it is possible to call

import idyntree.swig as idyn

kindyn = idyn.KinDynComputations()
kindyn.setJointPos([1, 0, 1, 1, 1, 1, 1])

as you can notice it is not necessary to explicitly create an iDynTree::VectorDynSize
This latter modification should work also in Matlab and in theory, can simplify the implementation of the idyntree high level wrappers

[traversaro]

Does:

vec =  iDynTree.VectorDynSize(7)
kindyn.setJointPos(vec)

still work?

[GiulioRomualdi]

vec =  iDynTree.VectorDynSize(7)
kindyn.setJointPos(vec)

yes, just tested

[traversaro]

Per Beyonce's Rule, you may add a test in https://github.com/robotology/idyntree/tree/master/bindings/python/tests if you want to make sure that this continue to work in the past.

This latter modification should work also in Matlab and in theory, can simplify the implementation of the idyntree high level wrappers

fyi @gabrielenava

[traversaro]

How did you find out of the implicitconv option? I could not find any docs on it.

[traversaro]

Things to do (not that you need to do it yourself @GiulioRomualdi, I just marking them down):

• Check if https://github.com/robotology/idyntree/blob/master/examples/python/KinDynComputationsTutorial.py also works after this change
• Update https://github.com/robotology/idyntree/blob/master/examples/python/KinDynComputationsTutorial.py to avoid explicitly creating VectorDynSize .

[traversaro]

Based on #688, probably we can put the shared_ptr stuff in #ifndef SWIGMATLAB, as the problem should be only in our old swig matlab fork.

[GiulioRomualdi]

Based on #688, probably we can put the shared_ptr stuff in #ifndef SWIGMATLAB, as the problem should be only in our old swig matlab fork.

Done in e74239d

[traversaro]

Based on #688, probably we can put the shared_ptr stuff in #ifndef SWIGMATLAB, as the problem should be only in our old swig matlab fork.

Done in e74239d

MATLAB binding generation still fails (see https://github.com/robotology/idyntree/actions/runs/3687003195/jobs/6240008478):

/home/runner/work/idyntree/idyntree/bindings/matlab/autogenerated/iDynTreeMATLAB_wrap.cxx: In function ‘int _wrap_new_Material__SWIG_1(int, mxArray**, int, mxArray**)’:
/home/runner/work/idyntree/idyntree/bindings/matlab/autogenerated/iDynTreeMATLAB_wrap.cxx:50371:7: error: ‘SWIG_CheckImplicit’ was not declared in this scope
50371 |   if (SWIG_CheckImplicit(SWIGTYPE_p_iDynTree__Material)) SWIG_fail;
      |       ^~~~~~~~~~~~~~~~~~
make[2]: *** [bindings/matlab/CMakeFiles/idyntreeOctaveMex.dir/build.make:76: bindings/matlab/CMakeFiles/idyntreeOctaveMex.dir/autogenerated/iDynTreeMATLAB_wrap.cxx.o] Error 1
make[1]: *** [CMakeFiles/Makefile2:907: bindings/matlab/CMakeFiles/idyntreeOctaveMex.dir/all] Error 2
make: *** [Makefile:136: all] Error 2

Based on https://github.com/search?q=repo%3Aswig%2Fswig%20SWIG_CheckImplicit&type=code, probably also implicitconv is also Python-only .

[GiulioRomualdi]

Let's see if 9be0778 fixes the problem

[traversaro]

Testing a similar commit in https://github.com/robotology/idyntree/actions/runs/3687142976 .

[traversaro]

This latter modification should work also in Matlab and in theory, can simplify the implementation of the idyntree high level wrappers

fyi @gabrielenava

Actually we discovered that that option is not supported in MATLAB.

[traversaro]

Now the generation work: https://github.com/robotology/idyntree/actions/runs/3687142976/jobs/6240337040 (the CI fails as nothing has changed, so actually we do not need to run generation).