Select the active array to visualize in a data set

[cryos]

Multidimensional data sets have more than one scalar array, at present Tomviz can only support s single array per data set. I think at a minimum we should support selecting the active data set from a list of available arrays read in from a file. We need to explore lazily loading the active array(s) due to memory constraints, and offering a separate array per module.

If we embrace the use of multiple scalar arrays per data set, we probably want to consider folding label arrays from segmentation into the main data set, and improving the interface to expose multiple arrays. This bug should extend the data panel once #1322 has moved forward to a place where it can be built upon. This is part of the core of supporting multidimensional/multimodal reconstructions (in addition to the RGB support being explored for three channel data sets).

[cquammen]

#1380 works towards this.

[cquammen]

@cryos Things mentioned on this issue that haven't been implemented:

• Lazy loading of data arrays
• Selecting an alternative data array to color by in modules
• Folding label arrays into transformed data results/child data sets

Should we break those out into separate issues to track them more easily?

[cryos]

I would say the 2nd and 3rd bullet are highly related, if you change the code to select alternate arrays to color by to look at all in a data source, then will it be easier to drop the support for selecting the alternative array from others? I would focus on 2 and 3 in the short term, if you would rather use this issue as a tracker, and break out a few smaller issues that is fine by me. I know this and a few other issues are on the larger side.

[cquammen]

The 2nd bullet makes doing the 3rd useful, it's true, but they are orthogonal in terms of implementation.

if you change the code to select alternate arrays to color by to look at all in a data source, then will it be easier to drop the support for selecting the alternative array from others?

I think you may be referring to the Contour module, which lets you color by a child data set which is usually assumed to be a label map produced by a segmentation algorithm. Saving the label map data into the data source would indeed simplify this color choice.

I'll focus on 2 and 3 in the short term.

[cjh1]

@cquammen Did you make any progress on 2 and 3? May be this is something @psavery could take over.

[cryos]

This won't make 1.4, we can revisit.

[cquammen]

@cjh1 The most recent progress was in #1411, but that was based on the old pipeline. It would be better off to start fresh IMHO.