[Getdp] some warning message

[Christophe Geuzaine]

Ji, Lin wrote:
> Dear Christophe,
>       Thanks for the information. By enforcing Dimension 3, the false
> warning dissapear. But there still remain two minor issues.
>       First, having zero value at some surface point is not correct in some
> case. How can we output the true solution value on the surface into the same
> file as the solution in the domain? Because eventually, I want to import the
> data into a matrix on regular grid points where there is no distinction
> between surface and interior.

OK... The problem is that the function space is usually *not* defined
on the surface elements, so that we cannot interpolate on the surfaces.
And if it is defined, you will only actually be able to interpolate the
trace of the field...

I guess the best way to avoid this situation is to add a small offset to
your postprocessing grid, so that you don't end up with evaluation
points on the surfaces. Would that be possible in your case? If not,
could you send an example?

>      Second, in the case of a vector solution, I get only one zero value
> instead of 3 zero values on some surface element. This caused a little
> trouble when I tried to import the data (TimeTable format) into a matlab
> matrix because my little importing program counts on the three components of
> a vector. Although it can be worked around, it would be better to keep the
> format uniform for every point. The following is part of the output to show
> what I mean by only one zero value for some point:
> 
>  58 0.05900000000000005  -6 -4.2 1.2  -35.57680985887268
> 0.8073098794431409 -10.60569408924594
> 59 0.06000000000000005  -6 -4.2
> 1.2  -19.60545899750656 -0.1722236169587627 -43.32303149432153
> 0 0.001  -6 -4.2 1.35  0
> 1 0.002  -6 -4.2 1.35  0
> 2 0.003  -6 -4.2 1.35  0
> 3 0.004  -6 -4.2 1.35  0

Yeah, on an undefined evaluation point the code just returns the default
type (scalar) with a zero value.

> 
> By the way, does the new version of getdp have the matrix format for data on
> regular grid points that can be readily used in matlab? Thanks a lot.

Not really, but it should be pretty easy to add--tell me if you're
interested in implementing this. What happens if you simply load a file
containing the "raw" values (using e.g. the SpaceTable or Timetable
formats)?

Christophe

-- 
Christophe A. Geuzaine
Applied and Computational Mathematics, Caltech
geuzaine at acm.caltech.edu - http://geuz.org