Varying results with pagedown::chrome_print()

[snhansen]

Description

I'm using RevealJS html-slides created with Quarto more and more, and to create pdf handouts I've been using pagedown::chrome_print() which works very well. I have however noticed a very random bug when converting my slides: the pdf produced by pagedown::chrome_print() is not always exactly the same. Most of the times, the conversion is flawless but some times all LaTeX formulas are missing while other times some but not all formulas are missing. It appears more frequently with large decks, so it's taken me some time to create a minimal reproducible example. I'm not sure whether this is a Quarto or pagedown/chromote thing, but I'd be happy to see if anyone other than me can recreate the problem and to get some input on what might be causing this.

Consider this slides.qmd:

---
format: 
  revealjs:
    slide-number: true
    smaller: true
---

## The average

Let $y_1,\ldots,y_n$ be a sample of size $n$. The average is defined as

$$
\bar{y} = \frac{1}{n}\sum_{i=1}^n y_i = (y_1+y_2+\cdots + y_n)/n
$$


```{r}
#| echo: false
#| out-width: 600px
ggplot2::ggplot(data = mtcars, ggplot2::aes(x = mpg, y = disp)) + ggplot2::geom_point()
```

I've used the following script to create 350 pdf's of the html slides

convert <- function(i) {
  pagedown::chrome_print(input = "slides.html", output = paste0("slides-", stringr::str_pad(i, width = 4, pad = "0"), ".pdf"))
}

quarto::quarto_render("slides.qmd", cache_refresh = TRUE)
purrr::map(1:350, convert)

When I look at their sizes I see this pattern:

pdf_sizes
63604 68842 69374 
    6   340     4 

The majority of the slides look like this (which matches the html):

The larger files are because the plot is pushed to its own slide which is weird because everything else is unchanged. The smaller files are because some of the LaTeX is stripped like this:

The above has been created using

• Windows 11
• Quarto 1.6.28
• R 4.4.1
• pagedown 0.21

but the issue has been bugging me for quite a while, so it's definitely not something new. Any ideas what's at stake here?

[mcanouil]

Did you try setting some wait time before the screenshots happen? It might simply be because MathJax was not fully loaded which means the math layout is not finished.

[snhansen]

Did you try setting some wait time before the screenshots happen? It might simply be because MathJax was not fully loaded which means the math layout is not finished.

Yes, I have experimented with that argument. I've just tried the example above with wait = 200 and can still reproduce the issue.

[mcanouil]

What's the unit?

(I'm not sure this is a Quarto issue here)

[snhansen]

What's the unit?

(I'm not sure this is a Quarto issue here)

It's in seconds. The default is 2. Not sure it's a Quarto issue either, but maybe someone else using the Quarto + pagedown combo has experienced similar issues which is why I've posted here.

[cderv]

For revealjs printing to PDF, I found I had the best results with using decktape to make the PDF printing.
This is discussed in

• Support other PDF export tools for revealjs (e.g decktape) #4677

and linked issues

I don't have ideas on what could cause the Math to not be fully rendered 🤔 chrome_print() has a specific JS script to try to sync things up before making the printing and maybe it needs adjustement to latest reveal version. Script is here: https://github.com/rstudio/pagedown/blob/main/inst/resources/js/chrome_print.js
There is a wait for Mathjax there.

You could also check what Revealjs gives you before printing in browser by looking at the print-pdf version of the slide. This would make sure there is nothing weird already there before triggering a chrome print.

All this PDF printing is touchy, with lots of element. I would really try out other printing tools so that we can know if this is an issue with pagedown itself, or other layers.

Hope it helps

[mcanouil]

This seems to be a Reveal.js issue.

[cderv]

One thing I did notice however is that for some slides pressing "E" yields slides.html?print-pdf=#/... while for others it yields slides.html?view=print#/... Is there a difference between the two?

The second URL is from Revealjs v5 update that we included in Quarto 1.6. So you may not be rendering the presentation with the same version. Though both url are fine and supported by revealjs. They introduced the second one because of new view=scroll feature.

And by the way, I tested this html-math-method: katex and can't reproduce the issue with that, so it seems connected to MathJax.
(...)
This seems to be a Reveal.js issue.

That is really good to know. Could be a Mathjax issue or Revealjs issue integration with Mathjax 🤔

To be sure, it would be great to reproduce without all Quarto doing. One easy way would be to take the .qmd, remove everything that makes is a .qmd to keep only what is supported by pandoc, then make it a .md. From there, call pandoc --to reveal --mathjax ... with all command line that are needed.

If the issue reproduce, then this is definitely not Quarto !

[snhansen]

To be sure, it would be great to reproduce without all Quarto doing. One easy way would be to take the .qmd, remove everything that makes is a .qmd to keep only what is supported by pandoc, then make it a .md. From there, call pandoc --to reveal --mathjax ... with all command line that are needed.

If the issue reproduce, then this is definitely not Quarto !

I'll give it a try, thanks.

Edit: I can't seem to reproduce the error this way.. weird!

[cderv]

I can't seem to reproduce the error this way.. weird!

Not so weird. It could mean this is one of the configuration that Quarto is doing.

Either configuration we set as default, which is not the default for revealjs in Pandoc, or it could be one of the specific processing Quarto is doing as we tweak the format. So that is really helpful.

We need to dig a little more to find the problem though 😞

[mcanouil]

Quarto also adds plugins which might not interact well with MathJax.
It feels more and more as a JavaScript (conflict) issue.

Is there a way to opt-out Reveal.js plugins set by Quarto that I am not aware of?