[MOAB-dev] Writing a tag

[Vijay S. Mahadevan]

> "Primary fields" is a relative term and maybe not even meaningful in certain
> circumstances.

True. The intent of my statement was that currently DMMoabOutput only
writes out tags that have associated fields in the discrete operator.
I didn't generalize this to write out arbitrary PETSc Vecs defined on
MOAB tags, although as I mentioned, this should be easy to do once an
API is defined to collect the tag names or Vec handles to be
serialized.

Vijay

On Tue, Aug 19, 2014 at 4:16 PM, Gerd Heber <gheber at hdfgroup.org> wrote:
> Yes, having the ability to write arbitrary MOAB tag-backed PETSc Vecs would be great.
> "Primary fields" is a relative term and maybe not even meaningful in certain
> circumstances.
>
> Thanks, G.
>
> -----Original Message-----
> From: Vijay S. Mahadevan [mailto:vijay.m at gmail.com]
> Sent: Tuesday, August 19, 2014 2:38 PM
> To: Gerd Heber
> Cc: MOAB dev
> Subject: Re: [MOAB-dev] Writing a tag
>
> You can load a file with few tags already defined, create a new tag on a range of entities and write it back out with write_file to see the additional tag also. Use mbsize -t option to see the tags.
>
> If this is still related to DMMoab, are you looking for a way to define an auxillary (derived) variable, that is not part of the original linear system (containing only primary variables) and the dump out all the data with DMMoabOutput ? Currently, DMMoabOutput explicitly only loops over the primary fields and writes out the tags in the file. If there is a need to define auxillary variables, I can certain add that capability since I can see several different use cases that would benefit.
>
> Vijay
>
> On Tue, Aug 19, 2014 at 2:08 PM, Gerd Heber <gheber at hdfgroup.org> wrote:
>> How can I append a tag to an existing MOAB file (which has a few other tags already)?
>> Maybe this is so obvious that the documentation doesn't talk about it,
>> but it's unclear to me. I don't see, for example, how to sell this to
>> write_file. Maybe that's not the right call anyway...
>>
>> Thanks for your help.
>> G.
>>
>>