Animated plot with a moving facet zoom via gganimate and ggforce?

Question

Goal

I would like to zoom in on the GDP of Europe throughout the years. The phantastic ggforce::facet_zoom allows this for static plots (i.e., for one specific year) very easily.

Moving scales, however, prove harder than expected. gganimate seems to take the x-axis limits from the first frame (year == 1952) and continute until the end of the animation. This related, but code-wise outdated question did not yield an answer, unfortunately. Neither + coord_cartesian(xlim = c(from, to)), nor facet_zoom(xlim = c(from, to)) seems to be able to influence the facet_zoom window beyond static limits.

• Is there any way to make gganimate 'recalculate' the facet_zoom scales for every frame?

Ideal result

First frame

Last frame

Current code

library(gapminder)
library(ggplot2)
library(gganimate)
library(ggforce)
p <- ggplot(gapminder, aes(gdpPercap, lifeExp, size = pop, color = continent)) +
    geom_point() + scale_x_log10() +
    facet_zoom(x = continent == "Europe") +
    labs(title = "{frame_time}") +
    transition_time(year) 

animate(p, nframes = 30)

Answer 1

I don't think it's possible quite yet with the current dev version of gganimate as of Dec 2018; there seem to be some bugs which prevent facet_zoom from playing nice with gganimate. Fortunately, I don't think a workaround is too painful.

First, we can tween to fill in the intermediate years:

# Here I tween by fractional years for more smooth movement
years_all <- seq(min(gapminder$year), 
                 max(gapminder$year), 
                 by = 0.5)

gapminder_tweened <- gapminder %>%
  tweenr::tween_components(time = year, 
                           id   = country, 
                           ease = "linear", 
                           nframes = length(years_all))

Then, adopting your code into a function that takes a year as input:

render_frame <- function(yr) {
  p <- gapminder_tweened %>%
    filter(year == yr) %>%
    ggplot(aes(gdpPercap, lifeExp, size = pop, color = continent)) +
    geom_point() +
    scale_x_log10(labels = scales::dollar_format(largest_with_cents = 0)) +
    scale_size_area(breaks = 1E7*10^0:3, labels = scales::comma) +
    facet_zoom(x = continent == "Europe") +
    labs(title = round(yr + 0.01) %>% as.integer) 
    # + 0.01 above is a hack to override R's default "0.5 rounds to the
    #   closest even" behavior, which in this case gives more frames
    #   (5 vs. 3) to the even years than the odd years
  print(p) 
}  

Finally, we can save an animation by looping through through the years (which in this case include fractional years):

library(animation)
oopt = ani.options(interval = 1/10)
saveGIF({for (i in 1:length(years_all)) {
  render_frame(years_all[i])
  print(paste0(i, " out of ",length(years_all)))
  ani.pause()}
},movie.name="facet_zoom.gif",ani.width = 400, ani.height = 300) 

or, alternatively, using gifski for a smaller file <2MB:

gifski::save_gif({ for (i in 1:length(years_all) {
  render_frame(years_all[i])
  print(paste0(i, " out of ",length(years_all)))
}
},gif_file ="facet_zoom.gif", width = 400, height = 300, delay = 1/10, progress = TRUE) 

(When I have more time, I'll try to remove the distracting changes in the legends by using manually specified breaks.)