Linking with static plot seems to have some errors

[dicook]

When linking between a static plot and the tour, using code from your upcoming paper, the points that have been brushed don't necessarily match the corresponding points in the tour view.

See that brushing the small cluster, shows as this cluster and the smile in the tour view. Is there a problem linking the row to row?

Here is the code to reproduce the problem.

# remotes::install_github("dicook/mulgar")
library(mulgar)
library(uwot)
library(detourr)
library(dplyr)
library(crosstalk)
library(plotly)

set.seed(42)
cnl_umap <- umap(clusters_nonlin)
umap_df <- data.frame(umapX = cnl_umap[, 1],
                      umapY = cnl_umap[, 2])

cnl_df <- bind_cols(clusters_nonlin, umap_df)
shared_cnl <- SharedData$new(cnl_df)

detour_plot <- detour(shared_cnl, tour_aes(
  projection = starts_with("x"))) |>
    tour_path(grand_tour(2), fps = 60) |>
       show_scatter(alpha = 0.7, axes = FALSE,
                    width = "100%", height = "450px")

tsne_plot <- plot_ly(shared_cnl,
                    x = ~umapX, 
                    y = ~umapY,
                    color = I("black"),
                    height = 450) %>%
    highlight(on = "plotly_selected", 
              off = "plotly_doubleclick") %>%
    add_trace(type = "scatter", 
              mode = "markers")

[casperhart]

In the example, I don't think the smile is highlighted, it's just the points are packed so densely that they appear darker due to the overlap, which makes it harder to distinguish the highlighted vs. non-highlighted points.

For example here, I have only the small cluster selected:

And it isn't clear that it's only the small cluster that's selected until I change the colour:

To show this more clearly, If I select both the small cluster along with some of the smile, the selected points in the smile do become darker:

So this could definitely be made better, but I'm not sure the best approach that will retain the alpha transparency in both the selected and non-selected points. What do you think?

[dicook]

Oh, I'm sure you're right. I kept thinking going back and forth on whether it was correct or not.What about automatically colouring the selection with the current colour? This would then make it clear what is selected or not, and allow the alpha transparency to stay as it is.

[casperhart]

The problem with applying the current brush colour immediately is if some brushing / selection has already been done, the same issue will crop up again.

I think having the selection highlighted with some colour could work, but it should work separately from the brush and be cleared when the selection is cleared. There would still be some risk of the highlight colour clashing with a brush colour, but this shouldn't be much of an issue.

Alternatively, the highlighted points could be given a red border or something. Maybe a red and white border in case the points have already been brushed red.

I'm not sure what these would take to implement, but I might have time to look in to it in a few weeks.