interpolation of restricted function broken

[Martin Eigel]

Relating to this thread, the interpolation of a function on a restricted space onto the full space is broken.

As a side remark, it would be nice if FunctionAssigner did support restricted functions. Moreover, getting access to a restricted_to_global dof mapping would be useful. Currently, the information one can obtain about the restricted space is pretty incomplete (no mesh, dof, mapping...).

[Johan Hake]

This looks pretty tricky to fix. FunctionSpace::interpolate would need to know that the GenericFunction v (which is interpolated) is a restricted Function. At the moment one can only check that for a proper Function (via its FunctionSpace).

[Martin Sandve Alnæs]

The problem is in Function.restrict, where w[] is not filled with zeros. I have a fix coming up.

[Martin Sandve Alnæs]

Although interpolate is not fully defined in this case, semi-arbitrary things will happen on the boundary of the restricted space where there is a discontinuity. But the same problem shows up in a projection.

[Martin Sandve Alnæs]

Fix issue 165, making Function.restrict set dofs to zero for spaces on a Restriction.

→ <<cset 1feed0391120>>

[Martin Sandve Alnæs]

https://bitbucket.org/fenics-project/dolfin/pull-request/105/fix-issue-165-making-functionrestrict-set/diff

[Martin Sandve Alnæs]

Apparently bitbucket marks issues as resolved before they are merged into master?

[Martin Sandve Alnæs]

Not merged yet.

[Johan Hake]

Bitbucket marks it as fixed because you wrote "Fix issue 165" in the commit message and pushed it to a dolfin branch.

[Martin Sandve Alnæs]

I got that, just wish it wouldn't be marked before it reached master.

[Martin Sandve Alnæs]

Fixed.

[Prof Garth Wells]

Removing milestone: 1.4 (automated comment)