oplot - Crude Observable Plot htmlwidget

Currently oplot is in a very alpha experimental (slightly embarrassing) form where we pass JavaScript through the htmlwidgets wrapper and then just eval on the JavaScript side. While this works I know we can do so much better, but the design of Observable Plot with functions at the core poses some difficult challenges. This all started with this issue requesting a Plot wrapper.

Installation

You can install the development version of oplot like so:

remotes::install_github("timelyportfolio/oplot")

Example

htmlwidget

Below you will see how ugly this looks in R, but we do get the plot that we expect as output.

library(oplot)
library(htmltools)

oplot('Plot.plot({
  marks: [
    Plot.frame(),
    Plot.text(["Hello, world!"], {frameAnchor: "middle"})
  ]
})')

# using sprintf with jsonlite we can provide data from R to JS
#  but of course this gets messy
#  be careful with % since they should be %% inside sprintf
oplot(HTML(
  sprintf(
'{ //# enclose in {} to prevent mtcars leakage into global/window
  let mtcars = %s
  Plot.dot(mtcars, {x: "hp", y: "mpg", fill: "cyl"}).plot()
}',
    jsonlite::toJSON(mtcars, auto_unbox = TRUE, dataframe = "rows")
  )
))

dependencies

If you are just looking for the convenience of locally installed JS dependencies for Rmd/Md/HTML usage, then this package provides a helper function oplot_dependencies().

md/Rmd

Add this bit of code anywhere in the md/Rmd.

htmltools::attachDependencies(htmltools::tagList(), oplot::oplot_dependencies())

HTML

tagList(
  oplot::oplot_dependencies()
  ... html ...
)