[yt-users] Projection Question

Matthew Turk matthewturk at gmail.com
Fri Apr 12 15:38:36 PDT 2013


No prob.  As a bit of an explanation, this is somewhat historical but
also has a good reason.  When the first plotting engine was used in yt
(HippoDraw) it accepted dx's that were *half* widths, and
cell-centered positions.  This was so it could avoid the floating
point operations during the iteration over the cells to deposit.  When
I implemented the pixelizer that underlies the FixedResolutionBuffer
(and, actually, has largely remained unchanged since its first
creation to output to matplotlib) I kept this convention, as I worried
about the data source having a large array, which I did not want to
duplicate, and I also worried about the floating point operations.  I
did some tests not that long ago, and it turns out that in worst case
scenario, you can get considerable impact from doign the fp operations
inside the pixelizer.  So I guess it's a good decision to keep, for
now.

On Fri, Apr 12, 2013 at 6:32 PM, David Collins <dcollins4096 at gmail.com> wrote:
>
> Ah, right on.  Thanks!
>
> d.
>
>
> On Fri, Apr 12, 2013 at 4:17 PM, Matthew Turk <matthewturk at gmail.com> wrote:
>>
>> On Fri, Apr 12, 2013 at 6:13 PM, david collins <dcollins4096 at gmail.com>
>> wrote:
>> > Hi, all--
>> >
>> > I have a dumb projection question.
>> >
>> > Why is it that when I do a projection, the minimum pdx is smaller than
>> > the
>> > minimum cell width in my simulation?
>> >
>> >>>> pf.h.get_smallest_dx()
>> > 0.0001220703125
>> >>>> 1./pf.h.get_smallest_dx()
>> > 8192.0
>> >
>> > Additionally, the sum of the areas is not what I expect.  I expect
>> >
>> >>>> proj=pf.h.proj(0,'Density')
>> >>>> (proj['pdx']*proj['pdy']).sum()
>> > 0.25
>> >
>> >
>> > I get 0.25, instead of 1 (consistent with the above, if all the widths
>> > are
>> > down by a factor of 2.)
>> >
>> > I assumed that proj['pdx'] is the width of the zone of the projection,
>> > which
>> > I naievely expect to be the smallest cell width along any line of sight.
>> >
>> > Could one of you good people fill in what I'm misunderstanding?
>>
>> Sure!  pdx is halfwidth, and the px values are cell centers.  I think
>> this is somewhere in the method paper, but there's no clear place to
>> stick it in the documentation.
>>
>> -Matt
>>
>> >
>> > Thanks a ton!
>> > d.
>> >
>> > _______________________________________________
>> > yt-users mailing list
>> > yt-users at lists.spacepope.org
>> > http://lists.spacepope.org/listinfo.cgi/yt-users-spacepope.org
>> >
>> _______________________________________________
>> yt-users mailing list
>> yt-users at lists.spacepope.org
>> http://lists.spacepope.org/listinfo.cgi/yt-users-spacepope.org
>
>
>
> _______________________________________________
> yt-users mailing list
> yt-users at lists.spacepope.org
> http://lists.spacepope.org/listinfo.cgi/yt-users-spacepope.org
>



More information about the yt-users mailing list