Streamlining function evaluation for simple cases

[RemDelaporteMathurin]

While I understand the method presented here provides enough flexibility and is very robust for parallel runs, it is a bit cumbersome and lengthy and I wondered if we could provide a convenience function for simple cases.

In legacy-fenics, we could evaluate a function simply by u.eval((0, 0, 0)). This was a bit fragile in parallel of course.

Could we imagine a convenience function to add this?

[finsberg]

I agree this would be nice to have. Maybe we could have

def evaluate_function(u: dolfinx.fem.Function, points: npt.NDArray[np.float64]): ...

I think it would also be nice to have a convenient way to evaluate ufl expressions, provided the user supplies an appropriate function space, i.e

def evaluate_expression(u: ufl.core.expr.Expr, V: dolfinx.fem.FunctionSpace, points: npt.NDArray[np.float64]): ...

To handle issues with parallelism we could also add a argument broadcast=True that would ensure that values are present on all processors, i.e

def evaluate_function(u: dolfinx.fem.Function, points: npt.NDArray[np.float64], broadcast: bool = True): ...

[RemDelaporteMathurin]

@finsberg that would be lovely. I really like the idea of evaluation ufl expressions!!!

Not sure I understand the broadcast argument here. If True, it would run an internal check and return an error if the point is not on all procs?

[finsberg]

@RemDelaporteMathurin This has now been resolved for function evaluation here: #74