Closed jwallwork23 closed 10 months ago
Oh and the Cofunction
changes also made me realise a problem with the error estimation code... The flux terms were summarised by a Function
which came from a linear solve, whereas the cell terms came from assembling a 1-form, which is a Cofunction
. After changing the latter to reflect the former, I found a load of bugs in the error estimation testing. Hopefully, it gives better results now :)
I'm gonna go ahead and get this merged as I'm confident it meets the requests. Thanks again for the review @stephankramer!
Closes #61. Take 2 of #62.
Okay there ended up being quite a few ways we had got out of date with Firedrake... apologies for the mega PR.
Summary:
Cofunction
andFunctionSpace.dual()
.norm
anderrornorm
to acceptCofunction
s.project
to acceptCofunction
s.function2cofunction
andcofunction2function
utilities.y = Function(FunctionSpace(mesh, "R", 0)).assign(x)
rather thany = Constant(x)
.