[yt-dev] New code frontend

[Matthew Turk]

Hi Nicola,

Today, Jill and I sat down and worked out how to get a voronoi
tesselation plotted in yt, using only existing machinery.  (We're
going to try out getting a new set of machinery very soon...)
Quantitative analysis should work just fine without any chances, but
for visualizing things like slices and projections, we came up with
this:

def nearest_neighbor(field, data):
    pos = data["PartType0", "particle_position"].in_units("code_length")
    dens = data["PartType0", "Density"].in_units("g/cm**3")
    rv = data.smooth(pos, [dens], method="nearest", nneighbors=1)
    # Now some quick unit conversions.
    rv = data.apply_units(rv, "g/cm**3")
    return rv
yt.add_field(("deposit", "PartType0_nn"), function = nearest_neighbor,
            validators = [yt.ValidateSpatial(0)],
            units = "g/cm**3")

This sets up a derived field that deposits the density of the nearest
particle into each mesh point for the image.  It own't look good
unless you also increase the resolution; we did this, although keep in
mind that our simulation is pretty small:

ds = yt.load("snap_029.hdf5", over_refine_factor=4, n_ref=64)

the over_refine_factor=4 means to use 2**4 == 16 zones on a side for
the octree mesh, where a new level is added every time n_ref (64)
particles are in an Oct.

-Matt

On Mon, Apr 27, 2015 at 10:08 AM, Jill Naiman <jnaiman at gmail.com> wrote:
> Hi Nicola,
>
> I am also interested in supporting Voroni-Delaunay tessellations so feel
> free to keep me in the loop!  I was going to try to start writing a front
> end for Arepo at some point (which uses them), but I didn't have a specific
> time schedule in mind.
>
> Cheers,
> -Jill
>
> On Mon, Apr 27, 2015 at 6:16 AM, Matthew Turk <matthewturk at gmail.com> wrote:
>>
>> Hi Nicola,
>>
>> Thanks for writing!  There's work being done on unstructured meshes
>> right now, and the timeline is that we're aiming to have support for
>> tetrahedra, hexahedra, and wedges by the end of May.  Jill Naiman and
>> I are sprinting on this the next couple weeks.  That doesn't *quite*
>> get you to what you're looking for, but it's a slightly different
>> situation for voronoi tesselations anyway, since they are both more
>> complex and simpler to define.
>>
>> What would be possible, right now, would be to load in the data as
>> particles, visualize it (by over-refining the mesh by a good amount
>> and using nearest-neighbor mesh deposition), and if you have the
>> volumes stored in advance, do quantitative analysis.  Computing the
>> volumes on the fly isn't yet possible (but could be implemented) and
>> hopefully in the next couple weeks it'll be a lot easier to make
>> visualizations without over-refining the mesh.
>>
>> Please feel free to stop by IRC sometime to chat more about this, and
>> if you are able to give a try to loading in as particles, let us know
>> how that goes.
>>
>> -Matt
>>
>> On Sun, Apr 26, 2015 at 11:57 AM, Nicola Clementel <clementel at saao.ac.za>
>> wrote:
>> > Dear yt developer,
>> >
>> > I would like to use yt to visualise the output from the SimpleX
>> > radiative transfer code.
>> > Due to the "Voroni-Delaunay" unstructured grid use in the code the
>> > output ‘cells’ are tetrahedra.
>> >
>> > From a firs look into the code and in the documentation, I was not able
>> > to find if this kind of unstructured grid are already supported in yt or if
>> > something similar is under development.
>> > If one of the other code have something similar already implemented that
>> > I can use as guideline it would be incredibly helpful!
>> >
>> > Thanks,
>> >
>> > Nicola Clementel
>> > _______________________________________________
>> > yt-dev mailing list
>> > yt-dev at lists.spacepope.org
>> > http://lists.spacepope.org/listinfo.cgi/yt-dev-spacepope.org
>> _______________________________________________
>> yt-dev mailing list
>> yt-dev at lists.spacepope.org
>> http://lists.spacepope.org/listinfo.cgi/yt-dev-spacepope.org
>
>
>
>
> --
> Jill P. Naiman, Ph.D.
> Harvard-Smithsonian Center for Astrophysics
> jill.naiman at cfa.harvard.edu
> www.astroblend.com
> www.avriot.com
>
> _______________________________________________
> yt-dev mailing list
> yt-dev at lists.spacepope.org
> http://lists.spacepope.org/listinfo.cgi/yt-dev-spacepope.org
>
_______________________________________________
yt-dev mailing list
yt-dev at lists.spacepope.org
http://lists.spacepope.org/listinfo.cgi/yt-dev-spacepope.org