Lua filters in R Markdown
In rmarkdown: Dynamic Documents for R

knitr::opts_chunk$set(
  collapse = TRUE,
  comment = "#>"
)

This vignettes gives some detailed information regarding the Lua filters including in the rmarkdown package. You don't have to do anything to use them usually as they just works usaully. This document presents the features they power.

To know more about Lua filters in general, you can jump directly to last section

Pagebreaks

This filter is available since rmarkdown 1.15 and works with Pandoc 2.1+ (shipped in RStudio >= 1.2).

Adding a pagebreak in document was always possible using custom output specific syntax in a rmarkdown file but one drawback was the compatibility with several output format.

With this Lua filter, it is possible to add a \newpage or \pagebreak command in a new line to include a pagebreak in any of these formats: pdf_document(), html_document(), word_document() and odt_document().

# Header 1

Some text

\newpage

# Header 2 on a new page

Some other text

\pagebreak

# Header 3 on a third page

rmarkdown will convert those commands in the correct output format syntax using a Lua filter during pandoc conversion.

Using with PDF/ latex documents {#pdf}

As the commands are the ones already used in latex syntax, this works as expected in a tex output document, and thus with pdf. Adding a pagebreak was already possible with rmarkdown when output is pdf_document() or latex_document(), without any restriction about the version of pandoc.

Using with HTML documents {#html}

A \newpage or \pagebreak command in a rmarkdown document with output as HTML will be converted by default in this html code with inline style using CSS rule page-break-after

<div style="page-break-after: always;"></div>

This will always insert a pagebreak after this div.

To get more flexibility, you can use a HTML class and some custom CSS instead of an inline style. You need to add a metadata field newpage_html_class in your yaml header to set the class.

Then you can control the behavior using custom CSS as in this example

---
output: 
  html_document: default
newpage_html_class: page-break
---

```{css, echo = FALSE}`r ''`
// display the pagebreak only when printing the html page
@media all {
    .page-break { display: none; }
}
@media print {
    .page-break { display: block; break-after: page; }
}
```

# Header 1

Some text

\newpage

# Header 2 on a new page

Some other text

\newpage will be converted here to

<div class="page-break"></div>

and the style will be applied to this class from the CSS included in the chunk.

This customisation can also be achieved by setting the environnement variable PANDOC_NEWPAGE_HTML_CLASS in the R session that will render the document (or in .Renviron file for example)

Let's note that in this example we use break-after property instead of page-break-after as it is recommended now to use the former which is the replacement. The latter is kept around for compatibility reason with browsers.

Using with Word documents {#word}

A \newpage or \pagebreak command in a rmarkdown document with output as Word document will be converted in a pagebreak for word document. Manually, this would mean adding this in your rmarkdown

```{=openxml}
<w:p><w:r><w:br w:type="page"/></w:r></w:p>
```

For example, using the pagebreak feature, this will add the first header in the second page of the work document

---
title: My main title
output: word_document
---

\newpage

# First Header

Using with ODT documents {#odt}

To use the pagebreak feature with odt_document(), you need to provide a reference document that includes a paragraph style with, by default, the name Pagebreak. This named paragraph style should have no extra space before or after and have a pagebreak after it. (see libre office documentation on how to create a style).

The name of the named paragrah style could be customized using newpage_odt_style metadata in yaml header or PANDOC_NEWPAGE_ODT_STYLE environment variable (as in html document).

As the previous one, this example will lead to a two pages document, with first header on the second page.

---
title: My main title
output: 
  odt_document:
    reference_odt: reference.odt
---

\newpage

# First Header

Number sections

This filter is available since rmarkdown 2.4 and works with Pandoc 2.1+ (shipped in RStudio >= 1.2).

Numbering sections are supported by Pandoc for limited formats (e.g., html and pdf). The rmarkdown package adds number_sections.lua to support this feature in other formats (e.g., docx, odt, and so on). Users do not have to know which formats use the Pandoc's feature or the Lua filter.

---
title: My main title
output: 
  md_document:
    number_sections: true # implemented by Lua filter
  html_document
    number_sections: true # implemented by Pandoc
---

# First Header

## A Header Belonging the First Header

# Second Header

In general, numbers and titles of sections are separated by a space. An exception is the word_document function, which separates them by a tab in order to be consistent with Pandoc's number sections for docx format in Pandoc >= 2.10.1. If one wants to have fine controls on the format of section numbers, prepare customized docx file and specify it to the reference_docx argument of the word_document function.

Custom environment for LaTeX

This filter is available since rmarkdown 1.15 and works with Pandoc 2.1+ (shipped in RStudio >= 1.2).

Pandoc supports a syntax called Fenced Divs that allows to create <div> in HTML. For example, the syntax is

::: {#special .sidebar}
Here is a paragraph.

And another.
:::

By default, this special syntax will be ignored in LaTeX conversion for PDF output and the content of the block will be used as-is. However, this syntax could be used to create LaTeX environment.

That is why rmarkdown uses a custom Lua filter to allow any Fenced Divs to be converted as a LaTeX environment (\begin{env} ... \end{env}) when desired by adding a special attribute.

::: {.verbatim data-latex=""}
We show some _verbatim_ text here.
:::

Its LaTeX output will be:

\begin{verbatim}
We show some \emph{verbatim} text here.
\end{verbatim}

This feature we call Custom blocks is documented in details in the R Markdown Cookbook

About Lua filters {#lua-filter}

Since pandoc 2.0, it is possible to use Lua filters to add some extra functionality to pandoc document conversion. Adding a pagebreak command in markdown to be compatible with several output documents is one of them. You can find some more informations about Lua filters in pandoc's documentation and also some examples in a collection of Lua filters for pandoc. These examples, and any other Lua filters, can be use in your Rmarkdown document directly by adding a pandoc argument in yaml header

---
output:
  html_document:
    pandoc_args: ["--lua-filter=filter.lua"]
---

You can also use a special helper using a !expr syntax for yaml in your header

The package rmdfiltr provides a collection of Lua filters and helpers functions to use them.

Before pandoc 2.0, using filter with pandoc was already available through programs that modifies the AST. pandoc-citeproc is an example used to deal with citations. The package pandocfilter is useful to create filters using R.

Using filters with output format

If you want to create a format that uses a special filter, you can add it as a pandoc option if the output format.

Either by modifying an existing format:

custom_format <- function(...) {
  base_format <- rmarkdown::html_document(...)
  # prepending a new Lua filter to html_document() ones
  base_format$pandoc$lua_filters <- c(
    rmarkdown::pandoc_path_arg("new.lua"), 
    base_format$pandoc$lua_filters)
  base_format
}

basename(custom_format()$pandoc$lua_filters)

Or by creating a new format :

custom_format <- function(toc = TRUE, ...) {
  rmarkdown::output_format(
    knitr = rmarkdown::knitr_options(),
    # a new filter will be appended to base_format ones
    pandoc = rmarkdown::pandoc_options(to = "html", lua_filters = "new.lua"),
    base_format = rmarkdown::html_document(toc = toc, ...)
  )
}

basename(custom_format()$pandoc$lua_filters)

As this is an advanced feature, if you want to go further, we encourage you to look at the examples of formats in packages such as rmarkdown, bookdown or pagedown.