Selection displayed on incorrect subplot

[paulsputer]

When using subplots and performing selections if the selection is not removed by double clicking prior to performing a select on a different subplot the selection is drawn on the previous subplot rather than the current subplot. However, when selecting only horizontal or vertical the handles are shown on the correct subplot. The data returned from the event however is correct.

[alexcjohnson]

Good catch @paulsputer ! Looks though like the problem is a bit deeper than that - points selected in one subplot keep their selected/deselected styling until you explicitly double click in that subplot. I don't think that's what we want, unless you multi-select, right?

cc @etpinard @dfcreative

[paulsputer]

Thanks @alexcjohnson - Ahh I hadn't noticed that bit, good point. Seems to get a bit funky with multi select too.

Along with deselecting points on a normal select, perhaps we need to keep the previous selection visible when performing a multi-select?

Would you recommend that multi-select should also be capable of selecting traces over multiple subplots?

[alexcjohnson]

I'm not 100% sure what the right behavior is here, it may be that we need several different ways for this to work. For example, if you have independent data in each subplot, it would be nice to be able to select points from different subplots together. But if the subplots are coupled (like a SPLOM showing slices of higher-dimensionality data eg https://rreusser.github.io/plotly-mock-viewer/#18) it can be nice to have one subplot in control of the others - whether we do it automatically or through some layered-on crossfilter functionality. Though even in that case being able to multiselect across subplots could be cool, as a way to select a sub-population that might not be disjoint within any single subplot.

Anyway, it's clear that this is going to be a bit of a complex issue if we want to fully exploit the possibilities. We could certainly start with a bugfix that doesn't allow multi-subplot select, and just clears any selection in other subplots when you start a new selection, along with fixing the transform issue that kicked this off.

[paulsputer]

Looks like some delayed push it's already been fixed in ebaf705 and appears to support multiselect across multi-subplot.

[alexcjohnson]

Very impressive @dfcreative - that's a lot of issues closed in one PR #2258!

I still think there's a discussion to be had about whether we want selection in one subplot to clear selections in another, but that's for another time.