slab sub-selecting not working in UVCDAT GUI.

[cameronsmith1]

I wanted to load a variable and plot various slices (eg, surface, zonal-mean), but when I tried to sub-select slabs, I got strange behavior that stopped me doing what I needed to do.

This capability is really critical.

I was running on Rhea at OLCF, and got the same result with both 'latest' (ie 2015-06-10) and 'devel' (ie 2015-06-22)

Here is my sequence of operations:
(1) I clicked on the 'green plus' to load a variable (SO2). It had 4 dimensions: time, altitude, latitude, longitude (specifically 2,30,96,144). But I only wanted the last time record, so I slid the slider to the last time record, and clicked 'Load'. So far, so good.
(2) I selected the SO2 variable and clicked the pencil to edit it. I wanted to start with the surface, so I used the slide to get just the bottom level. I clicked 'save as' and named it SO2_sfc. As can be seen in the screenshot below, the time dimension suddenly reappeared.
(3) I selected the new SO2_sfc variable and selected just the last time record, and saved as SO2_sfc2. The vertical levels then came back again.

[durack1]

@cameronsmith1 just FYI there's an inability in cdms2 currently to arbitrarily index variables (I'm referring to command-line execution here), which caused me to submit #1288 which I hope will be addressed along with the planned upgrades coming to a UV-CDAT near you fairly soon (#479, #481, #537, #576, #885, #1125 and #1059)

[cameronsmith1]

@durack1, thanks for pointing out the possible links between these issues. As I understand it, your issue is a feature request to the routines, where as mine is for a bug fix in the GUI, but that some of the underlying changes will probably overlap. Is that correct?

[durack1]

@cameronsmith1 exactly.. I have just pinged a couple of the other "interacting" issues above, apologies I wasn't trying to suggest this is a duplicate, just that with so many moving parts and changes planned and being executed by numerous parties it can't hurt to attempt to link these efforts together.

@doutriaux1 @dnadeau4 and @chaosphere2112 can speak in more detail about current development plans and can chime in about the issue you've highlighted above..

[doutriaux1]

@cameronsmith1 I can reproduce it on my end.
Good news: Vistrails is going away so this bug will disappear with it
Bad news: No ETA yet (probably a couple of months) in the mean time you can apply your double sub-sampling at once on the original file. Or use the dimension browsing in the cell.

Both ways are not great... Sorry...

[doutriaux1]

Reopening in case someone feels like fixing it before we take vistrails out for good.

[potter2]

I encountered the same problem using the sliders to average a dimension. The process seems to work but the new variable cannot be plotted (nothing happens when the variable is dragged to the panel). Is there a quick fix for this until vistrails is out?

[cameronsmith1]

The solution by @doutriaux1, of going back to the load window and doing all of the selecting and averaging at that time, works for me. It is a bit of a nuisance, but workable for me at the moment.

[doutriaux1]

@rexissimus any progress on this?