Printing xaringan slides with chromote

R Markdown Source

There are a number of options for producing PDF versions of xaringan slides, provided you use the standard xaringan features.

If you add interactive elements, like panelsets from xaringanExtra, printing your slides to convert them to PDF may not capture everything in your slides.

This post demonstrates a function that uses chromote to print xaringan slides to PDF files that should give better results, in particular when using panelsets.

The Problem

Typically, it’s fairly easy to convert xaringan slides to PDF. There are three methods that I’ve used that each work well and produce relatively similar results:

  1. Print the slides from a browser. Typically this works best in Chrome.

  2. Use xaringan::decktape() to virtually print the slides to PDF. This requires docker or an installed version of the decktape.js utility.

  3. Use pagedown::chrome_print(). This is similar to the first option, but uses a headless version of Chrome to do the printing behind the scenes.

These methods all work well but have one significant drawback: they don’t work well with xaringanExtra’s panelsets. The problem with the panelsets is that they essentially add “within-slide” slides. All of the panels are contained in a single slide, so when printed, only the first panel in the panelset is shown.

The Idea

The solution is easy but took me a bit of fiddling to figure out: we use chromote to control our own headless version of Chrome. Then we ask a programmatic monkey to push the button repeatedly, once per second(-ish), to advance through the slides, printing each slide to its own PDF.

Once all the slides are printed, we ask our monkey assistant to please staple the slides together into one (big) PDF file. All of this happens inside a headless Chrome browser controlled by chromote.

That’s it! And okay, the monkey assistant is actually a little bit of JavaScript that mashes a virtual right arrow key. And the stapler is actually the fantastic pdftools package.

Oh, and headless isn’t as spooky as it sounds. Basically it’s Chrome without the chrome. In other words, it’s a version of Chrome that runs as a command line utility and doesn’t have a user interface that you can click around in. Instead, you communicate with the browser by sending it commands, all of which are made easier by chromote.

The Solution

I’ve written a function, xaringan_to_pdf(), that you can point either at your slides online or your at local rendered slides (which may require file:// before the full path to the .html file).

You can copy the code below, or you can source it from this gist using the code and shortlink below1. Please check the source before you run it!

source("https://git.io/xaringan2pdf")

Running xaringan_to_pdf() will walk through your slides, printing them one-by-one, and returning one big PDF file. I’m not linking one here because they can be big.

The Code

Before you can print your slides with xaringan_to_pdf(), you’ll need a few things:

  1. Google Chrome

  2. The chromote package, which isn’t on CRAN yet:

    remotes::install_github("rstudio/chromote")
  3. A few other packages that you can get from CRAN:

    install.packages(c("progress", "jsonlite", "pdftools", "digest"))

Once you have those things installed and ready to go, copy the source code below and then run xaringan_to_pdf() to print your slides!

If you have any problems with the code, feel free to leave a comment on the gist.

#' Print xaringan slides to PDF
#'
#' Prints xaringan slides to a PDF file, even complicated slides
#' with panelsets or other html widgets or advanced features.
#' Requires a local installation of Chrome.
#'
#' @param url The url to the xaringan slides, either online or local.
#'   If the slides are stored locally, you may need to add `file://`
#'   before the complete file path.
#' @param path The path where the PDF file should be saved
#' @param delay Seconds of delay between advancing to and printing
#'   a new slide.
xaringan_to_pdf <- function(
  url, 
  path = paste0(basename(dirname(url)), ".pdf"), 
  delay = 0.5
) {
  if (!requireNamespace("chromote", quietly = TRUE)) {
    stop("`chromote` is required: devtools::install_github('rstudio/chromote')")
  }
  required_packages <- c("progress", "jsonlite", "pdftools", "digest")
  for (pkg in required_packages) {
    if (!requireNamespace(pkg, quietly = TRUE)) {
      stop("`", pkg, "` is required: install.packages('", pkg, "')")
    }
  }

  b <- chromote::ChromoteSession$new()
  on.exit(b$close(), add = TRUE)

  b$Page$navigate(url, wait_ = TRUE)
  b$Page$loadEventFired()

  current_slide <- function() {
    x <- b$Runtime$evaluate("slideshow.getCurrentSlideIndex()")$result$value
    as.integer(x) + 1L
  }

  hash_current_slide <- function() {
    digest::digest(b$Runtime$evaluate(
      "document.querySelector('.remark-visible').innerHTML"
    )$result$value)
  }

  get_ratio <- function() {
    r <- b$Runtime$evaluate('slideshow.getRatio()')$result$value
    r <- lapply(strsplit(r, ":"), as.integer)
    width <- r[[1]][1]
    height <- r[[1]][2]
    page_width <- 8/width * width
    list(
      width = as.integer(908 * width / height),
      height = 681L,
      page = list(width = page_width, height = page_width * height / width)
    )
  }

  slide_size <- get_ratio()

  expected_slides <- as.integer(
    b$Runtime$evaluate("slideshow.getSlideCount()")$result$value
  )

  max_slides <- expected_slides * 4

  b$Browser$setWindowBounds(1, bounds = list(
    width = slide_size$width, 
    height = slide_size$height
  ))

  b$Emulation$setEmulatedMedia("print")
  b$Runtime$evaluate(paste0(
    "let style = document.createElement('style')\n",
    "style.innerText = '@media print { ",
    ".remark-slide-container:not(.remark-visible){ display:none; ",
    "}}'\n",
    "document.head.appendChild(style)"
  ))

  pb <- progress::progress_bar$new(
    format = "Slide :slide (:part) [:bar] Eta: :eta", 
    total = expected_slides
  )

  idx_slide <- current_slide()
  last_hash <- ""
  idx_part <- 0L
  pdf_files <- c()
  for (i in seq_len(max_slides)) {
    if (i > 1) {
      b$Input$dispatchKeyEvent(
        "rawKeyDown", 
        windowsVirtualKeyCode = 39, 
        code = "ArrowRight", 
        key = "ArrowRight", 
        wait_ = TRUE
      )
      Sys.sleep(delay)
    }

    if (current_slide() == idx_slide) {
      step <- 0L
      idx_part <- idx_part + 1L
    } else {
      step <- 1L
      idx_part <- 1L
    }
    idx_slide <- current_slide()
    pb$tick(step, tokens = list(slide = idx_slide, part = idx_part))

    this_hash <- hash_current_slide()
    if (identical(last_hash, this_hash)) break
    last_hash <- this_hash

    pdf_file_promise <- b$Page$printToPDF(
      landscape = TRUE,
      printBackground = TRUE,
      paperWidth = 12,
      paperHeight = 9,
      marginTop = 0,
      marginRight = 0,
      marginBottom = 0,
      marginLeft = 0,
      pageRanges = "1",
      preferCSSPageSize = TRUE,
      wait_ = FALSE
    )$then(function(value) {
      filename <- tempfile(fileext = ".pdf")
      writeBin(jsonlite::base64_dec(value$data), filename)
      filename
    })
    pdf_files <- c(pdf_files, b$wait_for(pdf_file_promise))
  }

  pdftools::pdf_combine(pdf_files, output = path)
  file.remove(pdf_files)

  invisible(path)
}

  1. I wasn’t really sure where to put this function. Maybe I’ll eventually add it to xaringanExtra, but for now it’ll live here where hopefully it can still be useful to you!↩︎