Thanks again. I'm now up and running in FiPy trunk. Sorry for all the stupid questions. Time to feel sorry for the guys at Sandia, as I'm now trying to get PyTrilinos to work...
Best Regards, Angus Hendrick On Thu, Oct 9, 2008 at 8:25 PM, Jonathan Guyer <[EMAIL PROTECTED]> wrote: > > > On Oct 9, 2008, at 6:40 PM, Angus Hendrick wrote: > > Is there anything you can point me at for a primer on what's changed >> between fipy 1.2.1 and the trunk version? I'm trying to update my old code >> to work with the newer version so I can use PyTrilinos. >> > > The complete changes to the examples since we released 1.2 are: > > > http://www.matforge.org/fipy/changeset?new=trunk%2Fexamples%402565&old=trunk%2Fexamples%402034 > > > > > In short, the x, y, z component of MeshVariables is now first, not last, so > you you can do > > x, y = mesh.getCellCenters() > > or > > x = mesh.getCellCenters()[0] > y = mesh.getCellCenters()[1] > > instead of > > x = mesh.getCellCenters()[...,0] > y = mesh.getCellCenters()[...,1] > > > ImplicitSourceTerms now automatically determine what goes on the matrix > diagonal and what goes on the b-vector, so some of the previous machinations > are no longer necessary (but should still work). > > ConvectionTerms calculate their Péclet number automatically, so no need to > explicitly pass a diffusionTerm to them. > > > This isn't a change in FiPy, but in NumPy, but you need to be careful about > initializing, e.g., CellVariables with floats and not integers: > > var = CellVariable(mesh=mesh, value=0.) > > vs. > > var = CellVariable(mesh=mesh, value=0) > > > > You've already discovered the differences in specifying boundary > conditions. > > > I imagine there's more, but that's what's jumping out at me right now. > > > >
