Problem with output Cahn-Hilliard demo due to function names

[Steven Vandekerckhove]

When outputting part of a mixed formulation, e.g., file << (u.split()[0], t) in the cahn-Hilliard demo the split function gives new names to the output every time step.

That way the results can not be properly played in Paraview.

[Patrick Farrell]

Try the following diff:

diff --git site-packages/dolfin/functions/function.py site-packages/dolfin/functions/function.py
index 5897875..a3b8d08 100644
--- site-packages/dolfin/functions/function.py
+++ site-packages/dolfin/functions/function.py
@@ -362,9 +362,9 @@ class Function(ufl.Coefficient, cpp.Function):

         # Create and instantiate the Function
         if deepcopy:
-            return Function(self.function_space().sub(i), cpp.Function.sub(self, i))
+            return Function(self.function_space().sub(i), cpp.Function.sub(self, i), name='%s:%d' % (str(self), i))
         else:
-            return Function(self, i)
+            return Function(self, i, name='%s:%d' % (str(self), i))

     def assign(self, rhs):
         """

Suggestions for naming scheme (e.g. should the first component of a function called 'z' be 'z:0', 'z.0', 'z-0', ..?) welcome.

[Martin Sandve Alnæs]

Is : ok if this name is used in a filename on all platforms? Maybe z-0 is safer.

[Steven Vandekerckhove]

That does the trick, Thank you! I have no naming preference...

[Patrick Farrell]

OK. Can you run the tests, choose 'z-0', and propose a merge?

[Steven Vandekerckhove]

Tests are running, if they check out, I'll post a pull request.

Thanks!

[Prof Garth Wells]

Fixes Issue #335 (fix from Steven Vandekerckhove)

→ <<cset 10e90e138df5>>