style(hoveron="fills") doesn't work with ggplotly() when there are multiple polygons with the same fill color

[homerhanumat]

The following code based on the standard North Carolina shape-file fills counties by the values of BIR74, except that one county has been given an NA value for the variable.

library(ggplot2)
library(plotly)
library(sf)

nc_1 <-
  st_read(system.file("shape/nc.shp", package = "sf"), 
              quiet = TRUE) %>% 
  mutate(BIR74 = ifelse(NAME == "Ashe", NA, BIR74)) %>% 
  mutate(tooltip_text = ifelse(is.na(BIR74), 
                               paste0(NAME, " not reporting"), 
                               paste0(NAME, " reports ", BIR74)))
p_1 <-
  nc_1 %>% 
  ggplot() +
  geom_sf(aes(fill = BIR74, text = tooltip_text))

ggplotly(p_1, tooltip = "text") %>% 
  style(hoveron = "fills")

The custom tooltip indicates the missing value (hover over Ashe County) as expected.

The following code sets BIR74 to NA for two counties:

nc_2 <-
  st_read(system.file("shape/nc.shp", package = "sf"), 
          quiet = TRUE) %>% 
  mutate(BIR74 = ifelse(NAME %in% c("Ashe", "Alleghany"), NA, BIR74)) %>% 
  mutate(tooltip_text = ifelse(is.na(BIR74), 
                               paste0(NAME, " not reporting"), 
                               paste0(NAME, " reports ", BIR74)))
p_2 <-
  nc_2 %>% 
  ggplot() +
  geom_sf(aes(fill = BIR74, text = tooltip_text))

ggplotly(p_2, tooltip = "text") %>% 
  style(hoveron = "fills")

Hover over either Ashe or Alleghany and you get identical very long tooltips.

The output of plotly_json(p_2) indicates that Ashe and Alleghany have somehow been combined. See, e.g.:

json_2 <- plotly_json(p_2)
json_2$x$data %>%
  as.character() %>% 
  RJSONIO::fromJSON() %>% 
  .$data %>% 
  .[[100]] %>% 
  .[1:3]

[cpsievert]

Ahh, for performance reasons, ggplotly() uses one plotly.js trace to draw multiple polygons with the same fill color, which leads to issues with hoveron="fills". I'm not sure if I'll be able to provide a fix for this anytime in the near future as there's no easy way to let ggplotly() know about hoveron="fills" before it has mapped ggplot2 to plotly.

Anyway, in this case, there is a workaround: you can trick ggplotly() into creating one trace for each simple feature by mapping a unique identifier to color like this:

p_2 <- ggplot(nc_2) +
  geom_sf(aes(fill = BIR74, text = tooltip_text, color = NAME))

gg_2 <- ggplotly(p_2)

gg_2 %>%
  style(
    hoveron = "fills",
    # override the color mapping
    line.color = toRGB("gray40"),
    # don't apply these style rules to the first trace, which is the background graticule/grid
    traces = seq.int(2, length(gg_2$x$data))
  ) %>%
  hide_legend()