Apply style to specific figures when rendering Markdown using Cmarker

Questions
1

I have a large Markdown project which I’m rendering in Typst. There are a lot of tables in the Markdown file that I like to be figures and have some tags for each for custom styling. In my research I found out that Cmarker does pass the html classes down to the typst renderer. Here is an example:

#let wide-table = "
<figure class=\"wide\">

| Column 1 | Column 2 |
| -------- | -------- |
| Text 1   | Text 2   | 
<figcaption>Example Caption</figcaption>
</figure>
"

#let normal-table = "
<figure class=\"normal\">

| Column 1 | Column 2 |
| -------- | -------- |
| Text 1   | Text 2   | 
<figcaption>Example Caption</figcaption>
</figure>
"

#import "@preview/cmarker:0.1.8"

#cmarker.render(wide-table + normal-table, show-source: true)

This, with the show-source option on, produces this:

image
image991×377 33.3 KB

The issue is I don’t know how to query and style it. Here is what I tried: Using #show figure but the class is not present in the fields when converted to native figure (I think). Using #show html.figure tells me “only element functions can be used as selectors”. There’s also html.elem which is an element function but it’s a different thing from html.figure.

2

Regular expressions can be used with the match method of string to extract the class:

#import "@preview/cmarker:0.1.8"

#let get-class(markdown-table) = {
  let table-raw = cmarker.render(markdown-table, show-source: true)
  let pattern = `"class":\W(\w*)\W`.text
  let matches = table-raw.text.match(regex(pattern))
  assert(matches.captures.len() == 1, message: "Number of matches must be 1")

  matches.captures.first()
}

The wide table has a class of: #get-class(wide-table)

The normal table has a class of: #get-class(normal-table)

image

Once the class is known you can apply a style this way:

#let style-table(markdown-table, class) = {
  set table(
    stroke: if class == "wide" {blue} else {1pt}
  )

  cmarker.render(markdown-table)
}

#style-table(wide-table, get-class(wide-table))

#style-table(normal-table, get-class(normal-table))

image
image206×200 13.1 KB

Full Code 
#let wide-table = "
<figure class=\"wide\">

| Column 1 | Column 2 |
| -------- | -------- |
| Text 1   | Text 2   | 
<figcaption>Example Caption</figcaption>
</figure>
"

#let normal-table = "
<figure class=\"normal\">

| Column 1 | Column 2 |
| -------- | -------- |
| Text 1   | Text 2   | 
<figcaption>Example Caption</figcaption>
</figure>
"

#import "@preview/cmarker:0.1.8"

#let get-class(markdown-table) = {
  let table-raw = cmarker.render(markdown-table, show-source: true)
  let pattern = `"class":\W(\w*)\W`.text
  let matches = table-raw.text.match(regex(pattern))
  assert(matches.captures.len() == 1, message: "Number of matches must be 1")

  matches.captures.first()
}

The wide table has a class of: #get-class(wide-table)

The normal table has a class of: #get-class(normal-table)

#pagebreak()
#let style-table(markdown-table, class) = {
  set table(
    stroke: if class == "wide" {blue} else {1pt}
  )

  cmarker.render(markdown-table)
}

#style-table(wide-table, get-class(wide-table))

#style-table(normal-table, get-class(normal-table))
3

This makes sense, but there’s one issue. All the tables are packed in a single markdown file with no way to separate them, and as you have put it “Number of matches must be 1”.

4

Tables can be split with (another) function:

#let split-markdown-tables(markdown-tables) = {
  let tables-text = markdown-tables

  let regex-options = `(?s)`.text//Allow . to match newlines
  let pattern-start = `<figure`.text
  let pattern-middle = `.*?`.text
  let pattern-end = `</figure>`.text
  
  let individual-tables = tables-text.matches(
    regex(regex-options + pattern-start + pattern-middle + pattern-end)
  )
  
  individual-tables.map(match => "\n" + match.text + "\n")
}
Full Code 
#let wide-table = "
<figure class=\"wide\">

| Column 1 | Column 2 |
| -------- | -------- |
| Text 1   | Text 2   | 
<figcaption>Example Caption</figcaption>
</figure>
"

#let normal-table = "
<figure class=\"normal\">

| Column 1 | Column 2 |
| -------- | -------- |
| Text 1   | Text 2   | 
<figcaption>Example Caption</figcaption>
</figure>
"

#let all-tables = wide-table + normal-table

#import "@preview/cmarker:0.1.8"

#let split-markdown-tables(markdown-tables) = {
  let tables-text = markdown-tables

  let regex-options = `(?s)`.text//Allow . to match newlines
  let pattern-start = `<figure`.text
  let pattern-middle = `.*?`.text
  let pattern-end = `</figure>`.text
  
  let individual-tables = tables-text.matches(
    regex(regex-options + pattern-start + pattern-middle + pattern-end)
  )
  
  individual-tables.map(match => "\n" + match.text + "\n")
}
#let get-class(markdown-table) = {
  let table-raw = cmarker.render(markdown-table, show-source: true)
  let pattern = `"class":\W(\w*)\W`.text
  let matches = table-raw.text.match(regex(pattern))
  assert(matches.captures.len() == 1, message: "Number of matches must be 1")

  matches.captures.first()
}
#let style-table(markdown-table, class) = {
  set table(
    stroke: if class == "wide" {blue} else {1pt}
  )

  cmarker.render(markdown-table)
}

*All together:*\
#for found-table in split-markdown-tables(all-tables) {
  let table-class = get-class(found-table)

  style-table(found-table, table-class)
}

Note the use of (?s) to match across more than one line of text. Typst’s regex spec comes from Rusts spec (linked to on Typst’s regex page).

5

Even ignoring the fact that this solution doesn’t account for elements other than tables (like texts, headers, blockquotes, etc) but surely there has to be a way less convoluted way to mark a table for styling?