Poor behavior for plot yields no data cases

[markcmiller86]

Describe the bug

1. Open globe.silo
2. Put up mesh plot of mesh
3. Open rect3d.silo
4. Put up Pseudocolor plot of d
5. Draw

Note that you get the rect3d box down inside the center of globe.silo. That is fine. With the globe mesh plot opaque mode in auto, it will display wireframe to reveal the pc plot.

Ok, now lets use threshold to create a plot yielded no data condition. Add threshold operator to just the PC plot of d. Set min of operator above the max value in d, say 1. Hit apply.

You get the plot yielded no data message...which is fine. But, you also get the legend and color bar for the PC plot and you get the wireframe mesh plot of globe.silo. In other words, even though VisIt has told you that the plot has no data, it nonetheless is treating the plot like it does have data.

What it should do is NOT display the legend and color bar (it might be ok to show something like Pseudocolor: no data as a legend) and the mesh plot should not go into wireframe mode. In fact, I think having the legend display Pseudocolor: no data and NOT put up the warning about plot yielded no data is probably a much better behavior IN GENERAL.

[markcmiller86]

This originates from @nDimensionalSpace

[markcmiller86]

@nDimensionalSpace, a clumsy work-around in the meantime is to manually set the mesh plot's opaque mode to on and hide the no-data Pseudcolor plot.

If its a big enough issue, we can probably craft some Python CLI code that might provide a 1-button click solution for this as a macro in the GUI. Call it Patchup No Data Plots. A user could just hit that macro button once. The CLI code would introspect the plots present and adjust settings to have the intended effect. I think that is possible.

[cyrush]

Even with only one plot that has been thresholded away, you still see the legend and even the spatial dataset bounds.
PC legend will show the full data extents, and the axes will show the full dataset bounds.

Can be helpful when folks make a mistake with filter parameters.
If the plot info reports they plot is empty, then use CLI to hide?

[nDimensionalSpace]

@nDimensionalSpace, a clumsy work-around in the meantime is to manually set the mesh plot's opaque mode to on and hide the no-data Pseudcolor plot.

If its a big enough issue, we can probably craft some Python CLI code that might provide a 1-button click solution for this as a macro in the GUI. Call it Patchup No Data Plots. A user could just hit that macro button once. The CLI code would introspect the plots present and adjust settings to have the intended effect. I think that is possible.

@markcmiller86: Clearly, there are ways to manage this manually . . . I will note that the use case for this observation seemed to be animations/movies.

I don't think the person who noticed this is in a huge hurry, so no heroic efforts are required.

I will post some screenshots shortly.

[nDimensionalSpace]

Even with only one plot that has been thresholded away, you still see the legend and even the spatial dataset bounds. PC legend will show the full data extents, and the axes will show the full dataset bounds.

Can be helpful when folks make a mistake with filter parameters. If the plot info reports they plot is empty, then use CLI to hide?

@cyrush: Oh interesting . . . If I understand correctly, these behaviors are meant to address another use case, where a state of zero length data is induced accidentally. Maybe there should exist a flag to indicate whether the data writer is intentionally writing zero length data; if intentional, don't do all of the screen modifications, otherwise, yes do screen modifications.

[nDimensionalSpace]

As follows:

• fig1: some quantity of interest, looking as intended,
• fig2: the same quantity of interest in a "disabled" state, and the way that that is currently depicted in visit (with colorbar, and with wireframe mesh), and
• fig3: the same quantity of interest when hidden by the user.

I guess the whole point here is that the state of the data that currently visualizes like "fig2" should look more like that of "fig3", at least for cases where the field values of a variable have been intentionally omitted.