Is it possible to add more complex objects like ggplots?

[Make42]

Is it possible to add more complex objects like ggplots?
For example I might have a data.frame with columns lat, long, x, y:

lat, long, x, y
40, 5, 1, 6
40, 5, 2, 8
40, 5, 5, 3
41, 7, 3, 9
41, 7, 7, 10
41, 7, 9, 13

and I would like to build two ggplots(...) + geom_line(...) - one for 40, 5 and one for 41, 7 - which are printed as embetted graphs on their respective coordinates, namely 40, 5 and 41, 7.

Would that be possible and if so, how?
If not, what would be an option? (E.g. a different package instead of rstudio/leaflet or a different package instead of ggplot2.)

[yihui]

It is certainly possible, although not straightforward at the moment. Hopefully it will become easier after we merge #48.

[Make42]

That'd be great! What is the timeline for that - if there is?

In the meantime: How would I do this until then? An example would be great.

[yihui]

Hopefully by the end of next week. Examples will be given when this feature is ready.

[Make42]

How is the status? Could you provide a link to the examples?

[Make42]

How is this going? How is the status? Link to examples?

[jcheng5]

It's possible now. The trickiest part is encoding the plot as a data URI; I've provided a makePlotURI function to help you with that here.

library(leaflet)
library(ggplot2)

makePlotURI <- function(expr, width, height, ...) {
  pngFile <- plotPNG(function() { expr }, width = width, height = height, ...)
  on.exit(unlink(pngFile))

  base64 <- httpuv::rawToBase64(readBin(pngFile, raw(1), file.size(pngFile)))
  paste0("data:image/png;base64,", base64)
}

plot1 <- makePlotURI({
  print(ggplot(cars, aes(speed, dist)) + geom_point())
}, 200, 200, bg = "transparent")

plot2 <- makePlotURI({
  print(ggplot(cars, aes(speed, dist)) + geom_point())
}, 200, 200, bg = "transparent")

df <- data.frame(
  lat = c(40, 41),
  lng = c(5, 7),
  plots = c(plot1, plot2),
  stringsAsFactors = FALSE
)

leaflet(df) %>% addTiles() %>%
  addMarkers(icon = ~icons(plots))

[timelyportfolio]

very nice example @jcheng5. It might be helpful to add library(shiny) for plotPNG. Thanks for posting this.