[petsc-dev] PetscObjectViewFromOptions() pull request 1005 1006

[Jed Brown]

"Kong, Fande" <fande.kong at inl.gov> writes:

> On Thu, Jun 28, 2018 at 3:03 PM, Jed Brown <jed at jedbrown.org> wrote:
>
>> "Kong, Fande" <fande.kong at inl.gov> writes:
>>
>> >     There is also a question about making -help useful.  We would like
>> these
>> >
>> >> sort of diagnostic options to be provided in -help exactly once in on
>> >> organized way.  The normal way to do that is to put it in the
>> >> *SetFromOptions.
>> >>
>> >
>> > Like this one. It is even better if we have an option to dump all petsc
>> > options.
>>
>> What do you mean?  Like before execution?  That dump would not be
>> finite.
>>
>
> Yes, before execution. For example, if I say "-help snes", and then all
> snes related options should be printed out. Of course, we do not need to
> add prefix to these options.
>
> It may be impossible with the current PETSc design.

The number of such options is not finite because we can -npc_snes_type
and spawn a new level of SNES options.  Would we print a grammar?  Would
-help pc print all the (top-level only, even for composite and
multilevel?) options for all the PC types?  I don't know how to make the
user interface work outside of an interactive configuration tool that
had the ability to keep re-configuring a (trivial sized) solver.