Hello IDVers.
My opinion on PR naming needs a point of clarification about
what it is exactly.
Is it ONLY an option active during the Create Display moment?
And ONLY if Match Display Region is selected
(not Use Default Region) in the Field Selector?
[cid:8A308110-42A1-4797-B3EC-D0F3C9035454@gateway.2wire.net]
I thought so, and that seems safe and useful, a modest addition.
Maybe that would be called "display-matching subsetting" or something.
The display at a given resolution, once created, would remain. You can
Create another if you want more detail in some sub-area, or a broader
subsampled view over a wider area.
---
But I just discovered that PR can jump into action during a Projections -->
Use Displayed Area change. I just meant to redefine the "Home"
icon to be my zoomed-in area, as the last finishing touch on a bundle.
[cid:221090A0-5D17-4319-BA3A-44CECB9006FD@gateway.2wire.net]
Instead, PR actually overrode spatial subsetting limits I had put on my
datasets,
causing a too-large data read (much larger than the displayed area at that
moment, actually... I was trying to redefine Home view as a zoomed-in area,
but first the zoom jumped out wide, then the data reading over that wide area
began).
Match Display Area was unchecked in my (GFS and point data)
dataset's Spatial Subset -- I had typed in lat/lon bounds manually.
It crashed my session. (grr, sigh, again, I forgot to save often enough,
again...
(p.s. where is autosave?? Even a user-beware, yes conceivably corrupt,
.unidata/... hidden file one_minute_ago.xidv would be a treasure sometimes.
I always know within a couple seconds when a click is regrettable.)
Is it the intention that PR will be active in this way? Or is it a bug?
This sounds more like an "adaptive resampling" and less safe.
At least it should be unbundled from "Projection" since it is more
and different and costly (although I find too often that Projection
is a dangerous game... even Auto-set Projection is a killer when
one dataset is on 0-360 and another on -180 to 180. Support tickets
will be forthcoming...).
We decided that PR mustn't reread data automatically upon every
zoom/unzoom, since that would be a performance killer. Right?
So what list of actions are meant to trigger a re-reading operation?
Only new Create Display, or a well-defined list of other actions?
This bears strongly on what the "PR" function should be named
(and whether it should be on or off by default in Preferences, for
the unwary user just updating to 5.0 because it is available).
Brian
On Feb 18, 2014, at 3:54 PM, Yuan Ho wrote:
Hi All,
I would like to have another round of testing and feedback for the
progressive resolution (PR) and new image chooser. Hopefully, this will be the
final round and I can do a formal release very soon. Please download from the
following link:
http://www.unidata.ucar.edu/downloads/idv/nightly/index.jsp
A few key points since the last communication:
1) PR checkbox has moved to the User Preferences > View. The main concern is
that we don't want to mix variables with PR and without PR and create
complicated derived variable. It would be a better design for PR to be a global
setting.
2) Match Display Area is now a property in the gridded dataset. If this is
checked in the properties window, it would be a default property for the whole
dataset. It can be overwritten in the data subset panel. This change addresses
Jim' request in the last meeting.
3) PR has been implemented in the ADDE point data server.
Another issue is the decision for the Naming of PR. I like to use Progressive
Resolution, and Don's pick is Progressive Disclosure. In most situations of
rubber band box (RBB), there is no feature hiding being disclosed. In the case
of point data, RBB may bring out more stations in the view window. I would like
to know your opinion here.
Thanks,
Yuan
Brian Mapes
bmapes@xxxxxxxxxxxxxxx<mailto:bmapes@xxxxxxxxxxxxxxx>
