assemble(1 * dx) and similar don't work

[Nico Schlömer]

I wanted to compute the area (volume) of a mesh, but found that it is not quite easy to do. None of the options below works, except for the last involved Expression. I thought this was easier at some point (but I can't remember exactly)

from dolfin import *

mesh = UnitSquareMesh(10, 10)

#area = assemble(1 * dx)
#area = assemble(1.0 * dx)
#area = assemble(Constant(1.0) * dx)
#area = assemble(Expression('1.0', cell=triangle) * dx)
area = assemble(Expression('1.0', cell=triangle) * dx, mesh=mesh)

print area

[Marc-Antonio Bisotti]

I could get it to work with

area = assemble(Constant(1.0) * dx, mesh=mesh)

[Jan Blechta]

The point to say is that 1.0 * dx should be faster than Constant and Expression because the value gets hardcoded into generated code.

[Johan Hake]

There are 2 problems to address to fix assemble(1*dx):

1. What mesh should the assembler use to iterate over?
2. For what spatial dimension or cell type should the form compiler generate code for?

If we associate dx with a domain we have solved both these problems. That is addressed by issue 66.

[Jan Blechta]

1. User definitely needs to supply mesh as an argument to assemble or pass mesh function to measure before (which is checked from commit https://bitbucket.org/fenics-project/dolfin/commits/c9cccc2d5d3bd72d5267b9ada11ccf2580d88028).

2. That's why 1.0*dx does not work still (even if dx contains mesh function).

[Martin Sandve Alnæs]

Duplicate of #66.