On Sun, Aug 17, 2014 at 6:58 PM, Cameron Hummels <chummels@gmail.com> wrote:
Hey everyone,

Now that the righthandedness has gone into the PlotWindow objects, making slices and projections of isolated disk galaxy simulations yields the y-projection having a vertically oriented disk as seen here in an image generated by Mike Butler:


There was a change in the way slices and projections through the y axis are oriented to make the plot axis a right-handed coordinate system.  This happened in the 3.0 branch sometime in the spring.

It might be the case that the script was written before we made these changes and needs to be updated (probably by calling transpose on the raw image array for the y projections).



I suppose we could modify the ProjectionPlot framework to allow a keyword to specify the x-axis and y-axis vectors, but I was just trying to make equivalent projections using the OffAxisProjectionPlot object since that allows for one to explicitly specify a north_vector.  Unfortunately, I'm finding some inconsistencies between the ProjectionPlot (on-axis) images and the OffAxisProjectionPlot images (specifying vectors to be on-axis).

This simple script demonstrates what I'm talking about:


Running this script yields 4 images:

off_axis_proj.png - http://i.imgur.com/CKbq5iO.png

It looks like something is going wrong with the raw image values in this plot.  This seems like a pretty clear bug. 

Those values make sense for a volume density, perhaps the projection conversion is not being correctly applied?
 
on_axis_proj.png - http://i.imgur.com/2iBucuv.png
off_axis_weighted_proj.png - http://i.imgur.com/RQV2REP.png
on_axis_weighted_proj.png - http://i.imgur.com/CNWHyQS.png

I'm not terribly surprised that the image extrema aren't exactly equal here since the algorithms used to calculate the projections are very different. I think you might also need to set the no_ghost parameter to False to get a more accurate off-axis projection.

I'm not sure whether the off axis projection machinery will ever perfectly agree with the QuadTree projection, someone who is more knowledgeable about the volume renderer would need to comment on this.
 

As you can see, the off_axis_proj is using significantly different limits than the on_axis_proj, which makes me think it isn't doing a true line integral, but I haven't investigated this closely.  While the weighted projections are better, they still have slight differences in their limits.  Is this expected and I'm just messing something up here, or is this a true bug?

Cameron

--
Cameron Hummels
Postdoctoral Researcher
Steward Observatory
University of Arizona

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