Wrapper unrestricted to restricted

[ordinary-slim]

As far as I understand the wrappers (e.g. VecSubVectorWrapper) take a restricted vector and give you an unrestricted one. Is there a way to go in the opposite direction? In particular, I want to take this product:

$\textrm{restricted matrix} \cdot \textrm{unrestricted vector} = \textrm{restricted vector}$

and I am looking for a clean and efficient way to go about this.

EDIT: So far the solution I found looks like this

indices = np.array(tuple(restriction.restricted_to_unrestricted.values())[:restriction.index_map.size_local],dtype=np.int32)
indices = restriction.dofmap.index_map.local_to_global(indices).astype(np.int32)
iset = PETSc.IS().createGeneral(indices, MPI.COMM_WORLD)
restricted_f = f.x.petsc_vec.getSubVector(iset)                                           
...
f.x.petsc_vec.restoreSubVector(iset, restricted_f ) 

The costliest line here is the first one although not by a big margin. Let me know if you find something better.

[francesco-ballarin]

Hi,
creating a PETSc index set and then getting the sub vector/sub matrix (although not necessarily with getSubVector) is what VecSubVectorWrapper and the other wrappers do, so it seems to me that your solution is close enough to what the library does.