Clumps, contours, n_contours
Hi, all--
I'm seeing a glitch in the matplotlib countor function, as employed by annotate_clumps. As is, I'm seeing some clumps that are not plotted by the countour plotter unless I bump the number of contours-- I get all the plots if I change the contour line to:
self.rv = plot._axes.contour(buff, len(self.clumps)+5, extent=extent,**self.plot_args)
instead of
self.rv = plot._axes.contour(buff, len(self.clumps)+1, extent=extent,**self.plot_args)
Is this a glitch that anyone is aware of? Is there a more general fix? The clump contour is not the only time I've run into this glitch. It seems to happen on large contours, for what that's worth.
Thanks! d.
Hi Dave,
I haven't seen this. Could this be due to the colorbar? I believe it picks the contour colors based on a colorbar. If the colors it picks are invisible on your plot, you won't see the contours.
It would help to figure out what's going on if you could come up with a sample script that illustrates the issue. Bonus points if you use one of the datasets on yt-project.org/data.
-Nathan
On Mon, May 20, 2013 at 3:17 PM, david collins dcollins4096@gmail.comwrote:
Hi, all--
I'm seeing a glitch in the matplotlib countor function, as employed by annotate_clumps. As is, I'm seeing some clumps that are not plotted by the countour plotter unless I bump the number of contours-- I get all the plots if I change the contour line to:
self.rv = plot._axes.contour(buff, len(self.clumps)+5, extent=extent,**self.plot_args)
instead of
self.rv = plot._axes.contour(buff, len(self.clumps)+1, extent=extent,**self.plot_args)
Is this a glitch that anyone is aware of? Is there a more general fix? The clump contour is not the only time I've run into this glitch. It seems to happen on large contours, for what that's worth.
Thanks! d.
yt-users mailing list yt-users@lists.spacepope.org http://lists.spacepope.org/listinfo.cgi/yt-users-spacepope.org
Hi, Nathan--
I thought about the colorbar thing, so I made the image grayscale and it still doesnt' show up. I also set the zlim so that the data was either all white or black, and it shows up in neither.
I've attached two representative images. By construction all of these contours are disjoint. The missing clump is pretty obvious, as it takes up most of the image where it exists.
The two attached plots are two tweaks to plot_modifications.py/ClumpContourCallback, at line ~700 the "plus_3" plot uses self.rv = plot._axes.contour(buff, len(self.clumps)+3, extent=extent,**self.plot_args) the "plus_5" plot uses self.rv = plot._axes.contour(buff, len(self.clumps)+5, extent=extent,**self.plot_args)
The code to reproduce the plot is pretty benign, and is below for generality. the list clumps_to_use comes from a set of pickled clumps for this dataset. I'll see if I can reproduce it with something more universally accessible. I'm quite sure that it doesn't have anything to do with the clump itself, as I've seen this behavior with other calls to 'contour'. I'm working on a simpler thing that demonstrates this-- I've seen it with code that uses no piece of yt, so I'm 97.2% sure that it's contour and not yt. (Might be something else dumb I'm doing, that's pretty common, but I haven't been able to sort out what it is.)
<code> field='Density' center = [0.5]*3 proj = pf.h.proj(ax,field,center=center) plot = proj.to_pw(origin='domain') plot.set_cmap('Density','gray') subset = 'black_map_fixed' plot.set_zlim('Density',1e6,1e7) if clumps_to_use is not None: plot.annotate_clumps(clumps_to_use) print plot.save('%s_n%04d_%s'%(name,frame,subset)) </code>
On Mon, May 20, 2013 at 4:21 PM, Nathan Goldbaum nathan12343@gmail.comwrote:
Hi Dave,
I haven't seen this. Could this be due to the colorbar? I believe it picks the contour colors based on a colorbar. If the colors it picks are invisible on your plot, you won't see the contours.
It would help to figure out what's going on if you could come up with a sample script that illustrates the issue. Bonus points if you use one of the datasets on yt-project.org/data.
-Nathan
On Mon, May 20, 2013 at 3:17 PM, david collins dcollins4096@gmail.comwrote:
Hi, all--
I'm seeing a glitch in the matplotlib countor function, as employed by annotate_clumps. As is, I'm seeing some clumps that are not plotted by the countour plotter unless I bump the number of contours-- I get all the plots if I change the contour line to:
self.rv = plot._axes.contour(buff, len(self.clumps)+5, extent=extent,**self.plot_args)
instead of
self.rv = plot._axes.contour(buff, len(self.clumps)+1, extent=extent,**self.plot_args)
Is this a glitch that anyone is aware of? Is there a more general fix? The clump contour is not the only time I've run into this glitch. It seems to happen on large contours, for what that's worth.
Thanks! d.
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
Hi Dave,
I think the way contours works is that it sets N contours. What might work better, since we're contouring images, would be to explicitly specify the contours. Can you test if something like range(N+1) works better, rather than len(contours)?
-Matt
On Mon, May 20, 2013 at 7:04 PM, David Collins dcollins4096@gmail.com wrote:
Hi, Nathan--
I thought about the colorbar thing, so I made the image grayscale and it still doesnt' show up. I also set the zlim so that the data was either all white or black, and it shows up in neither.
I've attached two representative images. By construction all of these contours are disjoint. The missing clump is pretty obvious, as it takes up most of the image where it exists.
The two attached plots are two tweaks to plot_modifications.py/ClumpContourCallback, at line ~700 the "plus_3" plot uses self.rv = plot._axes.contour(buff, len(self.clumps)+3, extent=extent,**self.plot_args) the "plus_5" plot uses
self.rv = plot._axes.contour(buff, len(self.clumps)+5, extent=extent,**self.plot_args)
The code to reproduce the plot is pretty benign, and is below for generality. the list clumps_to_use comes from a set of pickled clumps for this dataset. I'll see if I can reproduce it with something more universally accessible. I'm quite sure that it doesn't have anything to do with the clump itself, as I've seen this behavior with other calls to 'contour'. I'm working on a simpler thing that demonstrates this-- I've seen it with code that uses no piece of yt, so I'm 97.2% sure that it's contour and not yt. (Might be something else dumb I'm doing, that's pretty common, but I haven't been able to sort out what it is.)
<code> field='Density' center = [0.5]*3 proj = pf.h.proj(ax,field,center=center) plot = proj.to_pw(origin='domain') plot.set_cmap('Density','gray') subset = 'black_map_fixed' plot.set_zlim('Density',1e6,1e7) if clumps_to_use is not None: plot.annotate_clumps(clumps_to_use) print plot.save('%s_n%04d_%s'%(name,frame,subset)) </code>
On Mon, May 20, 2013 at 4:21 PM, Nathan Goldbaum nathan12343@gmail.com wrote:
Hi Dave,
I haven't seen this. Could this be due to the colorbar? I believe it picks the contour colors based on a colorbar. If the colors it picks are invisible on your plot, you won't see the contours.
It would help to figure out what's going on if you could come up with a sample script that illustrates the issue. Bonus points if you use one of the datasets on yt-project.org/data.
-Nathan
On Mon, May 20, 2013 at 3:17 PM, david collins dcollins4096@gmail.com wrote:
Hi, all--
I'm seeing a glitch in the matplotlib countor function, as employed by annotate_clumps. As is, I'm seeing some clumps that are not plotted by the countour plotter unless I bump the number of contours-- I get all the plots if I change the contour line to:
self.rv = plot._axes.contour(buff, len(self.clumps)+5, extent=extent,**self.plot_args)
instead of
self.rv = plot._axes.contour(buff, len(self.clumps)+1, extent=extent,**self.plot_args)
Is this a glitch that anyone is aware of? Is there a more general fix? The clump contour is not the only time I've run into this glitch. It seems to happen on large contours, for what that's worth.
Thanks! d.
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
Hi, Matt-
That's a good idea. I tried np.unique(buff) and it fixed the problem. Should I make a PR with this change?
d.
On Tue, May 21, 2013 at 7:16 AM, Matthew Turk matthewturk@gmail.com wrote:
Hi Dave,
I think the way contours works is that it sets N contours. What might work better, since we're contouring images, would be to explicitly specify the contours. Can you test if something like range(N+1) works better, rather than len(contours)?
-Matt
On Mon, May 20, 2013 at 7:04 PM, David Collins dcollins4096@gmail.com wrote:
Hi, Nathan--
I thought about the colorbar thing, so I made the image grayscale and it still doesnt' show up. I also set the zlim so that the data was either
all
white or black, and it shows up in neither.
I've attached two representative images. By construction all of these contours are disjoint. The missing clump is pretty obvious, as it takes
up
most of the image where it exists.
The two attached plots are two tweaks to plot_modifications.py/ClumpContourCallback, at line ~700 the "plus_3" plot uses self.rv = plot._axes.contour(buff, len(self.clumps)+3, extent=extent,**self.plot_args) the "plus_5" plot uses
self.rv = plot._axes.contour(buff, len(self.clumps)+5, extent=extent,**self.plot_args)
The code to reproduce the plot is pretty benign, and is below for generality. the list clumps_to_use comes from a set of pickled clumps
for
this dataset. I'll see if I can reproduce it with something more universally accessible. I'm quite sure that it doesn't have anything to
do
with the clump itself, as I've seen this behavior with other calls to 'contour'. I'm working on a simpler thing that demonstrates this-- I've seen it with code that uses no piece of yt, so I'm 97.2% sure that it's contour and not yt. (Might be something else dumb I'm doing, that's
pretty
common, but I haven't been able to sort out what it is.)
<code> field='Density' center = [0.5]*3 proj = pf.h.proj(ax,field,center=center) plot = proj.to_pw(origin='domain') plot.set_cmap('Density','gray') subset = 'black_map_fixed' plot.set_zlim('Density',1e6,1e7) if clumps_to_use is not None: plot.annotate_clumps(clumps_to_use) print plot.save('%s_n%04d_%s'%(name,frame,subset)) </code>
On Mon, May 20, 2013 at 4:21 PM, Nathan Goldbaum nathan12343@gmail.com wrote:
Hi Dave,
I haven't seen this. Could this be due to the colorbar? I believe it picks the contour colors based on a colorbar. If the colors it picks
are
invisible on your plot, you won't see the contours.
It would help to figure out what's going on if you could come up with a sample script that illustrates the issue. Bonus points if you use one
of
the datasets on yt-project.org/data.
-Nathan
On Mon, May 20, 2013 at 3:17 PM, david collins dcollins4096@gmail.com wrote:
Hi, all--
I'm seeing a glitch in the matplotlib countor function, as employed by annotate_clumps. As is, I'm seeing some clumps that are not plotted
by the
countour plotter unless I bump the number of contours-- I get all the
plots
if I change the contour line to:
self.rv = plot._axes.contour(buff, len(self.clumps)+5, extent=extent,**self.plot_args)
instead of
self.rv = plot._axes.contour(buff, len(self.clumps)+1, extent=extent,**self.plot_args)
Is this a glitch that anyone is aware of? Is there a more general fix? The clump contour is not the only time I've run into this glitch. It
seems
to happen on large contours, for what that's worth.
Thanks! d.
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
Hi Dave,
Yes, please -- I think that is the right solution. Thank you!
-MAtt
On Tue, May 21, 2013 at 11:43 AM, David Collins dcollins@physics.ucsd.edu wrote:
Hi, Matt-
That's a good idea. I tried np.unique(buff) and it fixed the problem. Should I make a PR with this change?
d.
On Tue, May 21, 2013 at 7:16 AM, Matthew Turk matthewturk@gmail.com wrote:
Hi Dave,
I think the way contours works is that it sets N contours. What might work better, since we're contouring images, would be to explicitly specify the contours. Can you test if something like range(N+1) works better, rather than len(contours)?
-Matt
On Mon, May 20, 2013 at 7:04 PM, David Collins dcollins4096@gmail.com wrote:
Hi, Nathan--
I thought about the colorbar thing, so I made the image grayscale and it still doesnt' show up. I also set the zlim so that the data was either all white or black, and it shows up in neither.
I've attached two representative images. By construction all of these contours are disjoint. The missing clump is pretty obvious, as it takes up most of the image where it exists.
The two attached plots are two tweaks to plot_modifications.py/ClumpContourCallback, at line ~700 the "plus_3" plot uses self.rv = plot._axes.contour(buff, len(self.clumps)+3, extent=extent,**self.plot_args) the "plus_5" plot uses
self.rv = plot._axes.contour(buff, len(self.clumps)+5, extent=extent,**self.plot_args)
The code to reproduce the plot is pretty benign, and is below for generality. the list clumps_to_use comes from a set of pickled clumps for this dataset. I'll see if I can reproduce it with something more universally accessible. I'm quite sure that it doesn't have anything to do with the clump itself, as I've seen this behavior with other calls to 'contour'. I'm working on a simpler thing that demonstrates this-- I've seen it with code that uses no piece of yt, so I'm 97.2% sure that it's contour and not yt. (Might be something else dumb I'm doing, that's pretty common, but I haven't been able to sort out what it is.)
<code> field='Density' center = [0.5]*3 proj = pf.h.proj(ax,field,center=center) plot = proj.to_pw(origin='domain') plot.set_cmap('Density','gray') subset = 'black_map_fixed' plot.set_zlim('Density',1e6,1e7) if clumps_to_use is not None: plot.annotate_clumps(clumps_to_use) print plot.save('%s_n%04d_%s'%(name,frame,subset)) </code>
On Mon, May 20, 2013 at 4:21 PM, Nathan Goldbaum nathan12343@gmail.com wrote:
Hi Dave,
I haven't seen this. Could this be due to the colorbar? I believe it picks the contour colors based on a colorbar. If the colors it picks are invisible on your plot, you won't see the contours.
It would help to figure out what's going on if you could come up with a sample script that illustrates the issue. Bonus points if you use one of the datasets on yt-project.org/data.
-Nathan
On Mon, May 20, 2013 at 3:17 PM, david collins dcollins4096@gmail.com wrote:
Hi, all--
I'm seeing a glitch in the matplotlib countor function, as employed by annotate_clumps. As is, I'm seeing some clumps that are not plotted by the countour plotter unless I bump the number of contours-- I get all the plots if I change the contour line to:
self.rv = plot._axes.contour(buff, len(self.clumps)+5, extent=extent,**self.plot_args)
instead of
self.rv = plot._axes.contour(buff, len(self.clumps)+1, extent=extent,**self.plot_args)
Is this a glitch that anyone is aware of? Is there a more general fix? The clump contour is not the only time I've run into this glitch. It seems to happen on large contours, for what that's worth.
Thanks! d.
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
-- Sent from my computer.
yt-users mailing list yt-users@lists.spacepope.org http://lists.spacepope.org/listinfo.cgi/yt-users-spacepope.org
participants (5)
-
david collins -
David Collins
-
David Collins -
Matthew Turk -
Nathan Goldbaum