On Sep 27, 2010, at 12:09 PM, Jed Brown wrote: > On Mon, Sep 27, 2010 at 19:07, Barry Smith <bsmith at mcs.anl.gov> wrote: >> This could be done by putting into the C code the usual "case" handling for >> the new viewer and then have >> it dispatch back to the python code. > > Right, but then it sounds like you're distributing the viewer with > PETSc (instead of petsc4py of some third-party plugin).
No, the viewer is distributed with petsc4py or some other package, but yes the PETSc source code is augmented also. In fact one could organize it so one extra "dispatcher" could support many different python viewers; essentially a shell Viewer :-) > We had a long > thread a while ago about making multiple dispatch runtime-extensible > in both arguments. I was just pointing out that something "quick and dirty" can be done now without the multiple dispatch system. I am not opposed to a multiple dispatch system to handle this; but no one has proposed specifics for such a system that pass the "good enough for PETSc" test. Barry > > Jed
