Headless browser PDF rendering pipeline as Unicode-friendly alternative to pdflatex

[VivaldoMendes]

Description

Hi, I am using Quarto to produce a document that must be compiled into both PDF and HTML. One of its sections contains Julia code chunks with Unicode characters. Julia allows this handy functionality, and for teaching, it works like a charm.

My problem is that when I compile the document into HTML, no problem at all, Quarto and VSCode do a wonderful job. However, when I compile the exact same document to PDF, I get a nightmare: it fails to compile because it contains Unicode characters.

Well, most people may find this pretty acceptable, given that compiling Unicode characters with pdflatex is an old, well-known problem that we can do very little about.

However, I have a doubt. I have slides that I produced with Marp for VSCode, and they have the same Julia code chunks with plenty of Unicode characters. When I compile the same (Marp) markdown file with pdflatex and with html, I get exactly the same NICE, clean output. Can this also be done with Quarto?

I am not a developer, but I am a regular user of Quarto and its nice functionalities. However, I find this problem with Unicode characters very annoying. Why should a document display the output properly when compiled to HTML, but blow up completely when compiled to PDF? I checked the new functionalities of AI, and I got the following advice:

" When you raise the issue, it might be worth framing it as a feature request rather than a bug — something like: "provide a modern PDF rendering pipeline via headless browser (similar to Marp) as an alternative to pdflatex/lualatex, to handle Unicode natively without workarounds." That framing is more likely to get traction than reporting it as a broken feature.

You could also mention weasyprint and pagedjs-cli as existing PDF engines that Quarto already supports in principle — it would just be a matter of making them work smoothly within the existing workflow, including with extensions like aea."

Thanks.

[mcanouil]

• Would you mind sharing a small reproducible example?
• What version of Quarto are you using? In recent versions, the default engine is lualatex not pdflatex because of the issue you mentioned and others.
• Have you tried typst format for the PDF?

Regarding Marp, have you looked for other discussions first?

• https://github.com/orgs/quarto-dev/discussions/7377

Can this also be done with Quarto?

Extensions are there for that.
Quarto CLI cannot implement all possible tools that exist.

You could also mention weasyprint and pagedjs-cli as existing PDF engines that Quarto already supports in principle — it would just be a matter of making them work smoothly within the existing workflow, including with extensions like aea."

See the open issues and discussions such as:

• Other pdf engines like wkhtmltopdf and weasyprint? #222

---

Finally, don't follow AI guidelines, follow repository guidelines, see a subset of the guidelines given for "Feature Requests" discussion:

If you want to ask for help, please use the Q&A GitHub Discussions.
First, please check Quarto Issues and Feature Requests GitHub Discussions to see if your feature request has already been discussed.
If it has, please consider adding a comment to the existing discussion instead of creating a new one.

[VivaldoMendes]

Hi, thanks for your reply.

Yes, I can explain in some details waht I do using Marp and VScode.

• I open a pretty regular MD file in VSCode
• The file has many ingredients that we may find in slides: regular text, mathematics, tables, code (with and without Unicode), etc.
• The same file exports to PDF and HTML
• When I want to export to PDF, this is the file's frontmatter (see first code chunk below). And this is an example of a PDF file with clean Julia code (with plenty of unicode characters): Slide 19, in a PDF doc, produced by me
• When I want to export to HTML, the frontmatter becomes ridiculously simple (see second code chunk below), and this is the same output, but now as an HTML file: Slide 19, in an HTML document, produced by me
• The Quarto version I am currently using is: v1.8.27
• Typst is not a solution either: it does not compile.

As for using AI guidelines, I used them as a last resort. I am not an eager consumer of AI technologies. I like to know what I do. However, in the current case, I don't have a clue how to solve it. I have been dealing with this problem for two days already with no success, and I have gone through a lot of your documentation (as I have in the past, where you helped by pointing out some Quarto extensions), but I have not found a solution. As I am aware that less sophisticated software (Marp) allows me to avoid that same problem out of the blue, maybe I was doing something wrong, and this is where AI assistance came into play.

Edited. One of the links above was duplicated. Problem now corrected.

Exporting to PDF (frontmatter)

---

marp: true
html: true
author: Vivaldo Mendes
paginate: true
math: true
style: |
  .columns {
    display: grid;
    grid-template-columns: minmax(0, 80fr) minmax(0, 20fr);
  }
  h2 {
    position: absolute;
    top: 50px;
    left: 75px;
    right: 75px;
    color: #34429a;
    font-size: 46px;
  }

  h1 {
    font-size: 60px
  }
  .red  { color: #e53935; }
  .blue { color: #1e88e5; }
  .cyan { color: #00bcd4; }
  .purple { color: #8e24aa; }
  .teal { color: #00897b; }
  .orange { color: #fb8c00; }

---

%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

Exporting to HTML (frontamatter)

---
toc: true 

---

[mcanouil]

Important

Accessibility: For better accessibility, please ensure link text is descriptive and meaningful rather than using generic terms such as "here". This helps all users understand the link's destination. See Writing meaningful link text (WCAG).

[mcanouil]

Thanks for the additional information.

Note that you can use Quarto profiles to have conditional metadata.
Regarding Marp, it would be better to have a proof of concept via a Quarto extension first as it seems to be a bit niche in Quarto's context (and there is the other discussion about it).

[VivaldoMendes]

Thank you for the guidelines about better accessibility. I have been using these forums for more than a decade now and was not aware that I was doing anything wrong. But it's never too late to learn. I have changed the text, but I am not sure it has improved the reading or accessibility.