Quick start guide

Ecological Metadata Language (EML) is a “comprehensive vocabulary and a readable XML markup syntax for documenting research data” (Jones et al. 2019). Supplying metadata in this format is a requirement of several data-sharing institutions, notably including the Global Biodiversity Information Facility (GBIF) and its partner nodes. While the EML package already exists to support building and manipulating EML within the workspace, there is no comparable system for writing these documents in a more human-readable format, such as R Markdown or Quarto markdown. delma provides this capability.

EML is quite stringent in the types of data that are allowed, as well as their order and placement in the hierarchy. Getting these points right can be challenging, and we suggest using check_metadata() in combination with the EML schema documentation to get it right.

Formatting your metadata statement

There are several points that are unusual about Rmarkdown documents formatted using delma, which we discuss below

Document structure

Create a metadata statement template using use_metadata_template()

use_metadata_template("my-metadata-statement.Rmd")

Header levels in markdown-formatted metadata statements determine the nested structure of the resulting xml. For example, this text in markdown…

# EML

## Dataset

### Title
My title

…parses to this in xml.

<EML>
  <dataset>
    <title>My title</title>
  </dataset>
<EML>

Attributes

Attributes can be added to a particular EML tag by including them in a list within a code block, the label of which is used by delma to link tags to their attributes. To add attributes to the userId field, for example, you would add the following code under the ## userId heading:

```{r}
#| label: 'userId'
#| include: false

list(directory = "https://orcid.org")
```

The include: false tag is added so that this content isn’t displayed when the document is knit.

Setting a unique ID

Every EML document must open with the tag eml, and the attributes of that element must contain a unique identifier in the packageId field, as well as a link to the system within which that key is unique. A logical example might be a DOI:

```{r}
#| label: 'eml'
#| include: false

list(packageId = "https://doi.org/10.32614/CRAN.package.galah",
     system = "https://doi.org")
```

A valid alternative might be a GitHub release:

```{r}
#| label: 'eml'
#| include: false

list(packageId = "https://github.com/AtlasOfLivingAustralia/galah-R/releases/download/v2.1.0/galah_2.1.0.tar.gz",
     system = "https://github.com")
```

Note that the eml tag is unusual in delma in that it is added automatically if not supplied. Where this occurs, all tag levels are also incremented by one to account for this change.

Dynamic content

delma will call rmarkdown::render() internally whenever read_md() or render_metadata() is used, meaning that it is possible to add dynamic content to your metadata statements. The boilerplate statement that ships with delma uses this feature to automatically populate the Title and Pubdate fields from the YAML section, for example:

```{r}
#| echo: false
#| results: 'asis'

# NOTE: This is set using the yaml above; do not edit by hand
cat(rmarkdown::metadata$date)
```

You could also implement dynamic content using inline code. For example:

This data contains `r readr::read_csv("my_data.csv") |> nrow()` rows.

Reading, writing, and format conversion

Internally, delma uses the lightparser package to convert markdown files to tibbles, and the xml2 package to convert lists to xml and write them to file. Between these two packages, we have written functions to convert between tibble and list versions of EML-formatted markdown.

Under the hood, read_md(), which does a few things:

calls rmarkdown::render() on your chosen file, meaning any code blocks or inline code is executed properly.
appends code blocks whose label matches an EML tag to the attributes of that EML tag; this allows quite complex attribute addition without affecting rendered text.
‘cleans’ the imported tibble to a small number of required columns.

This tibble can then be rendered to EML using write_eml(). The inverse operation is accomplished by calling read_eml() followed by write_md().