Status: To Do
Fix Version/s: None
Sprint:SUIT Sprint 2017-01, SUIT Sprint 2017-2, SUIT Sprint 2017-3, SUIT Sprint 2017-4
Team:Science User Interface
There was a brief discussion today of the desirability of supporting multiple-pane displays of (small) plots (i.e., x-y plots and histograms) in the same way that we do for images, allowing the creation of grids of plots, as well as the use of some of the other concepts we have for multi-image viewing such as blinking and manual or auto-play stepping through a series of plots.
There's no doubt that this would be useful!
However, we have to think through what the use cases for this are, because this will affect the way that we allow users to create such displays and how we label them.
Converting this ticket to a story would be a request that we collect a set of use cases.
Here is an example of a very powerful grid-form multi-plot that is used quite a bit in exploratory data analysis:
(Not all packages make it quite so easy to generate!)
Since we were talking about the use of screen real estate it is interesting to note how they have incorporated the axis labels for the variables into the plot.
This one plot actually incorporates elements from several of the improvement tickets I filed today as well: note the display of 1D distributions adjacent to the 2D ones.
A small subset of this functionality has been implemented, both in the standard (e.g., tri-view) query-results viewer, and in the "slate" that's used by the Firefly Python interface in Jupyter.
The remaining ideas are still relevant to future Portal and Python API development. It would be a good idea to issue a new ticket for the remainder and close this one with a little more detail.
Example: a series of plots of a pair of variables x and y, or a series of histograms of a variable x, one per bin of a histogram on another variable z.
"How does the distribution of x and y change as you vary z?"
One can imagine this being interesting both for continuous-valued z variables (e.g., photometric redshift) and for category-valued z variables (e.g., an object classification, or the WISE variability codes). For a category-valued variable, no association with a histogram would be needed.
When coupled with the programmatic (Python) interface to Firefly it would be possible to drive a series of plots with a more complex selection (e.g., non-uniform bins in photo-z) than would be realistic to provide with direct UI control.
In this use case, the labeling of the plots in the grid would need to be able to indicate the values of z used in the selection.