open TELEMAC-MASCARET The mathematically superior suite of solvers

Hi all,

I experience a strange behaviour with the Nearest Neighbour method. In some cases, if the coordinates of a point (source) match exactly the coordinates of the mesh node (z to interpolate), the interpolator doesn't take this point but the next nearest point. This behaviour happens also with the Inverse Distance method. With the linear method all is fine since this method, unlike the two other methods, doesn't have a distance based algorithm.
In my view, in such situations all the methods should give the same results.

Enclosed a mesh file and an interpolator file. See mesh node number 19...
I use v3.3.4, hence, maybe this issue has been solved already.

With best regards,

Clemens

Hi Clemens
The problem remain in the latest beta version 3.10.1
I tested your case...

I think it's a small bug but a bug and probably linked to the fact that the points have exactly the same X coordinate...

On the latest version, there is also another small bug because it's impossible to load your in2 file... I generate a new one and save it to see differences (there are some in the header) but BK cannot reopen the new one also...

Kind regards

Hi Christophe!

Thanks for the confirmation and yes, it seems so.
The inverse distance method needs a limiter (threshold) for a minimum distance > 0 which could explain this behaviour. However, I don't know why the Nearest neighbour method should be affected too.

With best regards,

Clemens

Hi Clemens,

There is a known bug in the interpolator (as Christophe correctly diagnosed). I'll chat with Alan next time I see him about getting it fixed.

Hi Christophe,

It looks like you have exposed a recently introduced bug that is in Alans wheelhouse.

I think Alan is on vacation right now so patience please.

Hope your summer is going well...
Cheers... Martin