Oops, I completely misunderstood the question! The idea is that ProjectionPlot accepts an argument describing how to "center" the domain. It makes more sense sometimes to show relative differences, so it by default sets the origin at the center of the plot. If you call ProjectionPlot with the argument origin="left-domain" it should give the behavior you're looking for, with 0 .. 1 instead of -0.5 to 0.5. -Matt On Mon, Aug 13, 2012 at 3:15 PM, Matthew Turk <matthewturk@gmail.com> wrote:
Hi Suoqing,
yt is trying to anti-alias your data. At cell-boundaries, this will be subject to roundoff error. The colorbar is being c onstricted to around machine precision, as you can see, in an effort to draw out the very small differences that result from the anti-aliasing procedure over a constant field. You can disable anti-aliasing on the plot object by doing:
p1._frb.antialias = False
ixing the ticks like John suggested would, in all likelihood, cause the colorbar to be distantly-enough spaced as to eliminate the cross-hatching, which would give you the constant density image you are seeking.
-Matt
On Mon, Aug 13, 2012 at 3:11 PM, Ji Suoqing <jisuoqing@gmail.com> wrote:
Hi John,
I'm grateful for your explanation.
The tick problem might be a small issue, but it's not a wise choice to modify by hand.
Best wishes, Suoqing
在 2012-8-13，下午3:07，John ZuHone <jzuhone@gmail.com> 写道：
Suoqing,
The weird lines you are seeing for a constant density plot are likely due to the fact that by default yt is looking for the min/max of your dataset to set the colorbar, and because you have a constant density the min/max is whatever differences from that value arise due to machine precision.
Best,
John ZuHone
On Aug 13, 2012, at 3:04 PM, Ji Suoqing wrote:
Hi Nathan,
what you mean by 'set the domain.'
This is defined in FLASH4 as runtime parameter, which means that in FLASH hdf5 file, the domain has already been intrinsically set to [0.0, 1.0, 0.0, 1.0].
So a simple script for this plot is just:
from yt.mods import * pf=load('pringle_hdf5_chk_0000') p1 = ProjectionPlot(pf, "z", "Density") p1.save()
So I just attach two figures generated by YT and VisIt respectively, from which you could see the axis tick difference. In VisIt the axis ticks is correct, while in YT is not.
I suspect in YT, the domain edge should be firstly grasped, based on which the axis ticks could be correctly set.
Thanks for your care.
Best wishes, Suoqing
VisIt plot: http://dl.dropbox.com/u/83265784/visit0000.png YT plot: http://dl.dropbox.com/u/83265784/YT_pringle_hdf5_chk_0000_Projection_z_Densi...
P.S. What is plotted out is a constant density field with AMR, while YT seems have weird plot in such case, just as the figure shown. But I do not care too much about this because I do not treat constant field.
在 2012-8-13，下午2:44，Nathan Goldbaum <nathan12343@gmail.com> 写道：
Hi Suoqing,
I'm unable to reproduce the error you're seeing and I'm not sure what you mean by 'set the domain.' Would you mind forwarding the full script you used to produce the plot?
Cheers,
Nathan
On Aug 13, 2012, at 11:40 AM, Ji Suoqing wrote:
Hi all,
I've found a small problem in ProjectionPlot function: in FLASH4 data, I've already set the domain to be, say:
xmin=0.0 xmax=1.0 ymin=0.0 ymax=1.0
But when I use ProjectionPlot to plot the figure, the axis ticks become -0.5<x<0.5, -0.5<y<0.5; while in VisIt, the axis ticks are right. Also, in YT if I print out:
pf.domain_left_edge & pf.domain_right_edge
the domain is correct: 0<x<1, 0<y<1.
I've found this in 2d cylindrical case, where x represents for r direction, and y for z direction. I'm not sure whether in cartesian coordinate the same problem exists or not.
Best wishes, Suoqing _______________________________________________ yt-users mailing list yt-users@lists.spacepope.org http://lists.spacepope.org/listinfo.cgi/yt-users-spacepope.org
_______________________________________________ yt-users mailing list yt-users@lists.spacepope.org http://lists.spacepope.org/listinfo.cgi/yt-users-spacepope.org
_______________________________________________ yt-users mailing list yt-users@lists.spacepope.org http://lists.spacepope.org/listinfo.cgi/yt-users-spacepope.org
_______________________________________________ yt-users mailing list yt-users@lists.spacepope.org http://lists.spacepope.org/listinfo.cgi/yt-users-spacepope.org
_______________________________________________ yt-users mailing list yt-users@lists.spacepope.org http://lists.spacepope.org/listinfo.cgi/yt-users-spacepope.org