Re: [yt-dev] field line tracing problems

Hi Jeff, I tried to catch you on IRC but I keep failing. I can't seem to reproduce this with load_uniform_grid, but haven't yet tried on an AMR simulation. Just to be clear, after you integrate the streamlines, you call the .clean_streamlines() before calling path(stream_number)? Any chance you could share the data with me to help debug? Or can we find a time to hack on this together? Sorry for all the trouble. Sam On Thu, Mar 13, 2014 at 1:08 PM, j s oishi wrote:

Hi Sam,

Thanks for the tips. Unfortunately, I sat down this afternoon to implement this fix, but it doesn't seem to help at all. Here's what I tried, in diff format:

http://paste.yt-project.org/show/4377/

I also moved the launching plane 2 root grid cells away from the left edge of the grid. I get exactly the same results. I wasn't quite sure from reading the code how many ghost cells the code uses to construct the vertex data; I assumed it only uses one. Thus, I would think moving the launch point 2 cells away from the edge would alleviate the problem if it were a vertex centering issue.

Do you have any further suggestions?

thanks,

Jeff

On Tue, Mar 11, 2014 at 1:14 AM, Sam Skillman wrote:

...

Hi Jeff,

(Offline we went back and forth a bit, I think what we came to was probably right. I just want to follow up here.)

The issue seems to be that during the construction of vertex-centered data, the periodicity of the fields are assumed, and the streamlines go nuts. The ideal fix is to get non-periodic vertex-centered data working, but that might take a bit. The quick fix/hack is to modify either the get_vertex_centered_data function in yt/data_objects/grid_patch.py or the logic in _integrate_through_brick in yt/visualization/streamlines.py. You could cut off the integration when it gets close to dx away from the left/right domain edges.

Sorry this was buried, we should probably raise some sort of warning if the simulation data is not periodic.

Sam

On Fri, Mar 7, 2014 at 3:19 PM, j s oishi wrote:

...

HI Sam,

Nope, that gives identical results. Any other thoughts?

j

On Fri, Mar 7, 2014 at 5:15 PM, Sam Skillman wrote:

...

Hi Jeff,

After integrate_trhough_volume, can you call: streamlines.clean_streamlines()

In reading this function, I think it will fail since with it complaining about not having a .magnitudes, so perhaps first edit that function to take out the references to self.magnitudes?

The reason I suggest this is that I don't see where in your script it goes through and cleans up the ones that left the domain.

If that doesn't help, let me know and I'll dive further in to this.

Sorry for the hassle, Sam

On Fri, Mar 7, 2014 at 1:34 PM, j s oishi wrote:

...

Hi All (especially Sam),

I'm trying to understand some behavior in the streamlines that I don't completely understand. I've documented the problem in a notebook here:

https://hub.yt-project.org/nb/2tmwor

The notebook has a full statement of the problem. In essense, I'm not sure I understand why streams that I believe should terminate instantly upon launch (because they should walk directly out of the box) have thousands of elements in them (instead of one). I assume there are one or more misconceptions in my understanding, and I'd really welcome any suggestions.

thanks,

Jeff

_______________________________________________ yt-dev mailing list yt-dev@lists.spacepope.org http://lists.spacepope.org/listinfo.cgi/yt-dev-spacepope.org

_______________________________________________ yt-dev mailing list yt-dev@lists.spacepope.org http://lists.spacepope.org/listinfo.cgi/yt-dev-spacepope.org

_______________________________________________ yt-dev mailing list yt-dev@lists.spacepope.org http://lists.spacepope.org/listinfo.cgi/yt-dev-spacepope.org

_______________________________________________ yt-dev mailing list yt-dev@lists.spacepope.org http://lists.spacepope.org/listinfo.cgi/yt-dev-spacepope.org

_______________________________________________ yt-dev mailing list yt-dev@lists.spacepope.org http://lists.spacepope.org/listinfo.cgi/yt-dev-spacepope.org