Hugo-ify R Markdown documents

Hugo is a static site builder that is fast and lightweight, it is the preferred engine for building {blogdown} sites.

govukhugo is a theme for Hugo that applies the GOV.UK Design System. The theme is designed with dashboard-/book-style output common for publishing data products, and has been strongly influenced by the application of the Design System’s principle and components in Public Health England’s Coronavirus dashboard.

{govukhugo} is an R package that enables the use of R Markdown documents in Hugo static sites using the govukhugo theme, it provides an easy way for R users to initialise a Hugo site using the govukhugo theme. The govukhugo theme is not compatible with {blogdown}, therefore the package also provides a means to render and build Rmarkdown documents.

Examples of how {govukhugo} converts a static R Markdown page and an interactive R Markdown page are provided in the tests section of this documentation.

govdown or govukhugo

The work of govukhugo is inspired by the earlier efforts of the {govdown} package that worked entirely within the Rmarkdown ecosystem, govukhugo builds on the early lessons learnt by the {govdown} project and aims to overcome some of its limitations.

The design of {govdown} is itself inspired by the {bookdown} package which is designed to enable R users to write books using Rmarkdown, including rendering them as HTML. Unfortunately {bookdown} rendered websites do not easily support sub-directories and all content is assumed to sit at the same level.

{govdown} renders content to look similar to that for standard GOV.UK service pages, this is a layout optimised for prose (guidance and reports) and transactional services (such as applying for a passport). However, this layout is not necessarily optimised for communicating data, statistics and analysis, govukhugo seeks to overcome this by adopting a ‘dashboard’ layout. In simple terms the standard layout supported by {govdown} can be considered a ‘portrait’ orientation, whereas the dashboard layout of govukhugo as having a ‘landscape’ orientation by default.

As a complete Rmarkdown package, {govdown} uses Pandoc and its lua filters to render content into the GOV.UK design system. This is particularly complicated to maintain and update, especially the support of design system components. The Hugo templating system makes it much easier to maintain the base scaffold of the govukhugo theme, and to customise individual sites as necessary. Furthermore, the Hugo shortcode system makes it easy to extend support for different design system components and patterns.

govukhugo and its R package are not a complete Rmarkdown package, as discussed below you cannot “knit” an Rmarkdown document into a govukhugo page, rather Rmarkdown is used to create interactive pages within the framework of a govukhugo site. The added benefit here is that the govukhugo can be used with other data science technologies and languages that can generate markdown or HTML content.

Installing and using {govukhugo}

The {govukhugo} package is not currently on CRAN, you can install it from GitHub:

1
remotes::install_github("co-analysis/govuk-hugo-r")

Within RStudio you can use init_govuk_hugo() to create a Hugo site with the govukhugo theme. Otherwise follow the general getting started info.

1
2
library(govukhugo)
init_govuk_hugo()

The package includes an R Markdown template that provides an outline skeleton which you can access from the R Markdown menu in R Studio.

It is recommended that you save R Markdown documents in the R/Rmd folder of your project. In the R Markdown YAML header you should specify the section (i.e. sub-folder of the content directory) where you want the Rmd file to be placed after it has been converted.

Once you have written and saved your R Markdown document(s), you can then build your Hugo site by calling build_hugo_site().

1
build_hugo_site()

By default build_hugo() will look for and process Rmd files in the R/Rmd folder, you can specify a different folder by the altering the rmd_folder argument, setting with_rmd to false skips the building of Rmd files.

1
2
3
4
5
# specify a different folder
build_hugo(rmd_folder = "code/Rmarkdown")

# skip Rmd processing
build_hugo(with_rmd = FALSE)

In more advanced site building you might not want to build the Hugo site, just convert your R Markdown files. You can do this by calling build_hugo_rmd(). As with build_hugo() the default assumption is that R Markdown is stored in the R/Rmd folder.

1
2
3
4
build_hugo_rmd()

# specify a different folder
build_hugo_rmd(rmd_folder = "code/Rmarkdown")

You can render a single R Markdown file by calling render_rmd(). This will use the section as defined in the YAML header, however for testing purposes you may want to output file somewhere else, which you can do by specifying a different folder using the out_dir argument.

1
2
3
render_rmd("R/Rmd/my_file.Rmd")

render_rmd("R/Rmd/my_file.Rmd", out_dir = "R/test_output")

Interactive elements from the {DT}, {plotly} and {leaflet} packages are supported (see below).

Converting RMarkdown documents

While Hugo is intended for converting plain markdown (.md) files, {blogdown} websites rely on the fact that Hugo also processes HTML files (given certain parameters). However, as outlined in the {blogdown} documentation, Hugo themes are not always compatible and “it can be time-consuming” to adapt them.

This is the case for govukhugo. Trying to solve this problem, from investigating the inner workings of {blogdown} one learns that its html_page() format is a wrapper around the bookdown::html_document2() format, which itself is a wrapper around rmarkdown::html_document().

Rather than invest time in making the govukhugo theme work with {blogdown} the {govukhugo} package goes back to first principles to create HTML that Hugo will process.

For end users the build_hugo() and build_hugo_rmd() functions are all that are required.

Under the hood: the “Hugo-ification” process

The site building functions call render_rmd() which processes an Rmd file and converts it to an HTML file that Hugo will process. The Rmd is minimally rendered into HTML by rmarkdown::html_document(): without a theme, without MathJax, data frames are printed via knitr::kable() and as a self-contained document. A self-contained document is chosen so that static plots are stored as base64 content within the file rather than having to deal with separate images.

Having rendered the HTML file, it is then passed to a back-end function (hugo_html()), which “Hugo-ifies” the HTML. It replaces the HTML entities &lt; and &gt; with < and > so that code fragments will be processed by Hugo properly. It replaces the HTML tags for code blocks <pre><code> with Hugo’s shortcode for code block syntax highlighting. Finally, it extracts the HTML content between the <body> tags, excluding the lines where the title and date have been written by html_document().

The Hugo-ified HTML is then returned to render_rmd() which finishes the process by prepending the YAML front-matter of the Rmd file to the HTML. A rmarkdown: true parameter is also added to the YAML. Finally, the YAML is scanned for the section parameter, if present this is used to determine the sub-folder of the site’s content directory that the HTML will be placed in.

Supporting htmlwidgets

R Markdown documents that do not include interactive content, e.g. that print a data frame or include a plot image, will now be converted when Hugo builds the site. However, a common reason for writing R Markdown documents is to include interactive components. When rendering an R Markdown document that include an {htmlwidgets} component the rendering process will insert the relevant dependencies into the <head> component of the resulting HTML document.

As just described, the “Hugo-ified” HTML process of {govukhugo} removes the head component, thus removing these dependencies. This is rectified by the setting of the rmarkdown: true parameter in the YAML front matter of the HTML page. When Hugo processes the HTML it reads this parameter in adds back in the dependencies to the head component of the relevant page. At present it adds in the dependencies for the {DT}, {plotly} and {leaflet}.

The {crosstalk} package’s filtering controls that allow dynamic filtering of the data used by these three widgets are also supported. However, users should wrap any calls to these controls in the package’s unstrap() function which removes the dependencies associated with the Bootstrap framework, if you do not remove these the page may render incorrectly. Note that these controls make use of the selectize.js library and may not work properly for users of assistive technologies.

govuk_datatable()

The {DT} package provides an opinionated implementation of the DataTables jQuery plugin, which does not easily conform to the GOV.UK Design System. As a result {govukhugo} includes a govuk_datatable() function that acts as a wrapper around DT::datatable(). Note that govuk_datatable() only provides access to some aspects of the datatable() arguments, and does not support passing of additional arguments via a ... argument, this is to minimise the potential for disrupting the application of GOV.UK style.

1
2
3
4
5
6
7
8
govuk_datatable <- function(data,
                            element_id = NULL,
                            col_names = NULL,
                            page_length = 10,
                            search = TRUE,
                            small_text = FALSE,
                            buttons = TRUE,
                            col_defs = NULL)

govuk_datatable() requires only a data argument, this should be either a data frame or a matrix, it provides the following optional arguments:

  • element_id: an id for the HTML chunk
  • col_names: a vector of column names, if not supplied the column names of the data object will be used
  • page_length: the number of rows to show in the table, note that end-users cannot change this setting
  • search: whether to include the table search functionality
  • small_text: whether the table should render with a smaller font-size (only recommended for very large tables)
  • buttons: whether to include a copy and csv download button with the table
  • col_defs: a list of column definitions (see the {DT} and DataTables documentation for more details)

govuk_datatable() returns a DT::datatable object, meaning that {DT} formatting functions, such as DT::formatPercentage(), should work on the returned object. Styling functions should be used with caution.

An example of govuk_datatable() can be seen on the interactive R Markdown test page