Re: [Libmesh-users] how to dynamically add vectors using System::add_vector()

[Yujie <rec...@gm...>]

Dear Roy:

I think you are busy and can't reply me. it is fine. I have tested the
dynamic vector addition in sequential mode. It works for me. I think it
should be work for parallel.

Regarding outputing dynamic vector, I have writing some codes for it.
However I didn't finish it. Because the output final vector containing all
the solutions is variable-major. if there are multiple variables in one
System, large change should be done in output function. In addition, the FE
type of dynamic vectors is another problem. I am very happy to continue
finishing this patch. Could you give me some advice about the above worry.
thanks a lot.

Regards,

Yujie

On Wed, Dec 10, 2008 at 10:37 AM, Roy Stogner <roy...@ic...>wrote:

>
> Sorry, did I not get back to you on this yet?
>
> On Mon, 24 Nov 2008, Yujie wrote:
>
>  I am checking the codes about the output file. If using the method you
>> provide, one can't simultaneously output the variable and vector in a
>> file,
>> right?
>>
>
> Right.  You'd have to have a separate gmv (or whatever) file for each
> output format.
>
>  Because the vectors are the copies of the solutions in my application, I
>> am
>> considering to refer to build_variable_names() and build_solution_vector()
>> to write two functions (build_vector_names() and build_vector_vector()) to
>> output the vector values. I feel it should be simple. could you give me
>> some
>> comments? thanks a lot.
>>
>
> Yes.  And after that refactoring, you could rewrite our output code to
> cause all vectors to be output rather than just the solution.  A
> variable named "concentration" in the vector "old_solution" would be
> written as "old_solution_concentration" in the output file.
>
> For backwards compatibility's sake (and so as to not annoy the many
> users whose only extra vector is an old_solution they don't care to
> output) it would be good if this was optional and not default
> behavior.  But with that quibble, we'd certainly be happy to accept
> such a patch.
>
>
> By the way (and this is what got me looking through old emails from
> you in the first place), did you apply and test that "dynamic vector
> addition" patch yet?  I'd like to commit it to SVN if it works
> correctly.
>
> Thanks,
> ---
> Roy
>