I think a key part of releasing 5.0 with PR enabled is to have some clear use
cases for its use included in the documentation.
_____________________________________________
Kevin Tyle, Systems Administrator
Dept. of Atmospheric & Environmental Sciences
University at Albany
Earth Science 235, 1400 Washington Avenue
Albany, NY 12222
Email: ktyle@xxxxxxxxxx<mailto:ktyle@xxxxxxxxxx>
Phone: 518-442-4578
_____________________________________________
From: Brian Mapes [mailto:bmapes@xxxxxxxxxxxxxxx]
Sent: Tuesday, March 11, 2014 4:46 AM
To: Tyle, Kevin R
Cc: Jim Steenburgh; Yuan Ho; <idv-steering@xxxxxxxxxxxxxxxx>;
idvdevelopers@xxxxxxxxxxxxxxxx
Subject: Re: New test release of Image Chooser and PR
On Mar 10, 2014, at 8:13 PM, Tyle, Kevin R wrote:
I too echo Jim's thoughts ... great job to Yuan and all the developers!
This shames me that I forgot to lead with the praise: Great job!
I have learned to use RBB and really enjoy UsePD for exploring.
How it beats the old imagery bounding!
User Preferences allows user to enable Progressive Disclosure, but it also
appears in the Projections Menu in Map View windows. If User Prefs has it set
to no, then setting it to yes in Projections has no effect.
I don't think this is correct -- I have set it to default no in my Prefs, but
am able to switch it on as desired from the Projections menu.
Projections --> Use Displayed Area is non-intuitive ... zooms out and reloads
data, and in my case, did not re-draw contours. What is the purpose of this
option?
It does redraw the contours --- but 360 degrees of longitude east!
You have to zoom out 2 planets worth to see them - screenshot. (Bug!)
[cid:image001.png@01CF3D16.896A8CB0]
Kevin, It sets a (square) wireframe box covering the Displayed Area, which you
can use to clip displays.
I find that useful for clipping 3D isosurface displays to clear cluttered sight
lines for oblique views.
(although RBB may be better for that clipping some day!)
Also it redefines the height for 3D displays if they are too tall or short.
Not clear to me how one can take Don's Local Weather (Boulder) bundle and
change projection so data
loads over another city (e.g. Miami) ... when I try this, map view changes
projection but no new
data loads. APPARENTLY THIS IS ONLY AN ISSUE FOR OLD BUNDLES
*
http://motherlode.ucar.edu/repository/entry/get/BoulderWx.xidv?entryid=b3c53509-4db9-45de-b2ae-800d2a833ecc<qobject:history:0>
I loaded this, then turned on UsePD for the session, and also for each display.
Then I RBB an area. The contours and stations move upon RBB, but the imagery
fields do not.
(the +360 longitude bug upon Projections --> Use Displayed Area can be seen
using this bundle).
-------------main issue: there are now two kinds of displays, old and new
Starting with other old bundles, I find I am able to convert most existing
displays to UsePD displays by turning the UsePD checkbox on in the View menu of
that display's control window. (not 3D displays yet, though, says Yuan)
But there is no going back: I find that once an old-style display has ever
touched UsePD,
* it will never again respect the Dataset spatial subset box.
* it will always reopen in a .xidv to cover the entire and only the initial
displayed region.
Even if its UsePD check is turned off before saving, and/or UsePD is off in the
session when you open the bundle -- in short, no matter what you do. Currently.
(is this structural, or strategic?)
So UsePD is a change in usage culture... one that many users will find purely a
100% improvement, and one I could come around to if I rethink my strategy and
approaches. But I would like some cues about future intentions. Before a big
release, it would be nice to see some draft of the release notes and.or
documentation text about what UsePD was designed to do, what it was expanded to
also do (and why), what the future intention is, what aspects are irreversible
or hard to change structurally, pitfalls or surprises, etc .
For example, dataset-level subsetting would have only one useful effect in a
UsePD-driven new IDV, if I understand right: It clips the data going into a
.zidv file. That can be confusing, as a .zidv bundle may come up blank because
you subset the data for a different region than the one you were happily
examining at the moment you saved it. (I tried this).
I'm looking forward to hearing about the Miami IDV workshop ... I am sure Brian
will have lots of good ideas to keep Yuan and company busy!
I hope there could be some strategic guidance, not just training on individual
capabilities and features and letting users string them together in an ad hoc
way. Maybe that is too much a matter of personal style? But there are getting
to be so many features ... it's not easy to find a satisfying, reliable
workflow that doesn't lead to surprises or dead ends or clutter. I have been
down many such dead ends in trial and error, and now may be missing some great
features for fear of deviating from safe pathways.
Examples of strategic-use questions I think even 2-day beginners should hear
about at training--
* Sphere view or flat view? Choose wisely, you can never convert later. Should
we all be moving to sphere? These were developed in some order, far apart in
time, presumably. What are the issues? (seams, size, transects, etc)
* UsePD or not to UsePD? Will one ever, down the line, want to save this
display over a fixed area? Then you have to refrain from ever touching or even
having checked the UsePD (or Match Display Area) feature during its creation,
however tempting that may be.
* Cross sections vs transects? Cross sections are easy, they are in the
displays menu. But down the line you discover that sharing several cross
sections shares the color bar, not just the location. Surprise! and there is
no back button or autosave recovery. If transects are clearly better, let's say
so right up front and train on them instead of cross sections.
* How to meaningfully reuse the considerable work one puts into a nice IDV
session? Save whole bundles (with or without datasets?), data favorites,
displays, display settings, colorbars, all of the above? Can saved bundles be
partly-reused, or is the Display the most efficient and meaningful unit of
recycling? What's the strategy? Down the line it may become clear what you wish
you saved, but can best practices be trained at the outset?
* Does it matter if you save a bundle without datasets, vs. opening it later
with Change Data Path? (YES, I think... In the latter case, you have to go off
and find a Data Sources file, which may have to match the format of the saved
file exactly, while in the former case you can select fields in sources you may
already have attached...) But unchecking Data Sets in the Save dialog is
dangerous: it's easy to forget you have done it, and it applies as default to
all future saves. Surprise! Your bundles all have no datasets. Argh.
* Imagery choices are so many and inscrutable. Is there a reliable subset, or a
guide page? Can some dead ones be pruned?
* Dealing with errors. What are the most common ones, what do they mean, when
should you try to parse them vs. just click "OK".
Well something to think about.
Sleepless in Miami
Brian
p.s. anyone else have this happen? Dead views piling up? Maybe it's harmless.
[cid:image002.png@01CF3D16.896A8CB0]
