[yt-users] Odd array size issue

[Matthew Turk]

Hi Elizabeth,

There are two ways that the covering grid (not the smoothed covering
grid) functions.  One is that if the data exists on disk, it fills
each zone from that.  If it's a derived field (as, for instance,
'x','y','z' are) it will calculate it locally.  *However*, this can be
overridden by the validator NeedsOriginalGrid, which is good for
things like "GridIndices" and whatnot.

Okay, so all of that out of the way, the smoothed covering grid does
*not* operate like this.  All fields, even if htye are derived fields
created in grids, are interpolated downward as needed.  I remember
making this decision, but for the life of me I cannot think why I
would; your field is the most obvious, but it seems to me that
CellMassMsun would also be incorrect if CellVolume varies between
cells.

Normally, I would consider this to be a bug and would fix it right
out; however, I think this deserves a bit more thought before I just
make this change.  The change is between:

1) Always interpolating all fields from lower resolution grids
2) Always generating (unless mandated not!) fields that can be
generated *from* the interpolated fields.

I'm in agreement with point #2.  Hearing no objections, I'll make this change.

For what it's worth, this should affect a small minority of users,
potentially only those already involved in this thread.  :)

-Matt

On Wed, Jun 13, 2012 at 3:36 AM, Elizabeth Tasker
<tasker at astro1.sci.hokudai.ac.jp> wrote:
> Hi Matt,
>
> Sorry to re-visit this issue but while your fix totally solved the
> previous problem, I have unearthed a second.
>
> If I create a scg, the "z" position is correct up to the first & last
> handful of cells. At this point, is mysteriously starts to decrease
> along with associate changes in density and acceleration fields.
>
> e.g.
>
> scg details:  AMRSmoothedCoveringGrid
> (GR_Enzo2_128amr5_zeus_turb_psupp_rr_0001): level=3, left_edge=[  0.
> 31.   0.], right_edge=[ 64.  33.  64.], ActiveDimensions=[2048   64
> 2048]
>
>
> z position for z-index 2035-2047:  [ 63.609375  63.640625  63.671875
> 63.703125  63.734375  63.765625
>  62.796875  60.828125  57.859375  53.890625  48.921875  42.953125]
>
>
> z position for z-index 0-20:  [ 28.015625  21.046875  15.078125
> 10.109375   6.140625   3.171875
>   1.203125   0.234375   0.265625   0.296875   0.328125   0.359375
>   0.390625   0.421875   0.453125   0.484375   0.515625   0.546875
>   0.578125   0.609375]
>
>
> The same is true in the x direction but the (smaller) y direction is fine:
>
> y position for y-index 40-63:  [ 32.265625  32.296875  32.328125
> 32.359375  32.390625  32.421875
>  32.453125  32.484375  32.515625  32.546875  32.578125  32.609375
>  32.640625  32.671875  32.703125  32.734375  32.765625  32.796875
>  32.828125  32.859375  32.890625  32.921875  32.953125]
> y position for y-index 0-20:  [ 31.015625  31.046875  31.078125
> 31.109375  31.140625  31.171875
>  31.203125  31.234375  31.265625  31.296875  31.328125  31.359375
>  31.390625  31.421875  31.453125  31.484375  31.515625  31.546875
>  31.578125  31.609375]
>
>
> The script that generates that output is here:
>
>
> http://paste.yt-project.org/show/2454/
>
> I actually end up chopping out the middle of my data and using it
> someplace else, so it's maybe not a problem but it is a bit worrying.
>
> Elizabeth
>
>
> On 5 June 2012 23:37, Matthew Turk <matthewturk at gmail.com> wrote:
>> Hi Elizabeth,
>>
>> Sorry for the delay.  I have identified the problem.
>>
>> Your covering grid was only selecting level 0 grids.  This resulted in
>> the smoothed covering grid making an early exit.  I've added
>> additional steps to produce grid-free refinement/interpolation to
>> higher-levels in changeset 2cd5fa6d84ae.
>>
>> -Matt
>>
>> On Fri, Jun 1, 2012 at 9:40 PM, Elizabeth Tasker
>> <tasker at astro1.sci.hokudai.ac.jp> wrote:
>>> Hi Matt,
>>>
>>>> It's a possiblity; John reverted some changes several weeks ago, but I
>>>> thought that restored it to the previous operations.  None of the
>>>> answer tests have thrown anything, and they do test this
>>>> functionality.
>>>>
>>>> Can you provide a script (pastebinned) that I can run?  And, what's
>>>> the size of your domain?  32, 32, 32 for extent, and dims of ...
>>>> 128^3?
>>>
>>> Actually, double my usual:
>>>
>>> 64 x 64 x 64 kpc boxsize
>>> 256 root
>>> 3 AMR (static grids)
>>>
>>> This script produces nothing good:
>>>
>>> http://paste.yt-project.org/show/2414/
>>>
>>>
>>> Elizabeth
>>>
>>> _______________________________________________
>>> 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