This article shows how to parse MDEx-generated HTML into a nested data structure that is used to build a table of contents component for a Phoenix blog.
Introduction
In this article, we are going to build a table of contents component that can be used in blogs or other markdown based pages. The table of contents component will display a link to each section of your content that allows users to quickly navigate through the content.
We assume that you have set up a Phoenix project and are using the MDEx markdown parser to render markdown content. We will configure MDEx to include IDs and anchor links for headings in the generated HTML that we can use to build the table of contents component. If you are not using MDEx, you can still follow along, but you will need to make some adjustments to the code to make it work with your markdown parser.
Steps
Configure MDEx
The first step is to configure the MDEx markdown parser to include IDs and anchor links for headings in the generated HTML. By default, MDEx does not include any IDs and or additional anchor links. To enable this feature, we need to set the header_ids option in the extensions keyword list when calling the MDEx.to_html function. This tells the MDEx markdown parser to include IDs and anchor links for headings in the generated HTML.
MDEx.to_html(body, extension: [header_ids: ""])
In the code snippet above, we pass an empty string to the header_ids option. The string will be used as a prefix for the generated IDs. We do not necessarily need a prefix, but it can be useful to avoid conflicts with other IDs on the page.
You can see a list of all available extension in the comrak documentation as MDEx uses Rust’s comrak crate under the hood.
We can write a simple test to verify that the generated HTML includes the IDs and anchor links for headings:
use ExUnit.Case
test "MDEx includes IDs and anchor links" do
assert MDEx.to_html("## Introduction\n", extension: [header_ids: ""]) ==
"<h2><a href=\"#introduction\" aria-hidden=\"true\" class=\"anchor\" id=\"introduction\"></a>Introduction</h2>\n"
end
As you can see, the generated HTML includes an anchor link with the ID introduction for the heading ## Introduction.
Parse and convert headings
The next step is to parse the generated HTML, extract the headings from it and transform them into a suitable data structure that we can use to build the table of contents component. We will use the Floki library for this. From the README: “Floki is a simple HTML parser that enables search for nodes using CSS selectors.”
We will write a function that takes the generated HTML content as input, parses the headings, and returns a list of headings with labels, hrefs and subheadings.
defp parse_headings(content) do
content
|> Floki.parse_fragment!()
|> Enum.reduce([], fn
{"h2", _class, child} = el, acc ->
acc ++ [%{label: Floki.text(el), href: get_href(child), childs: []}]
{"h3", _class, child} = el, acc ->
List.update_at(acc, -1, fn %{childs: subs} = h2 ->
%{h2 | childs: subs ++ [%{label: Floki.text(el), href: get_href(child), childs: []}]}
end)
_other, acc ->
acc
end)
end
def get_href(heading_element) do
attr = heading_element |> Floki.find("a") |> Floki.attribute("href")
case attr do
[] -> nil
[href | _] -> href
end
end
The parse_headings/1 function takes the generated HTML content as input. Since we are dealing with a string we first use the Floki.parse_fragment! function to parse the string into a Floki html_tree() data structure. We then use Enum.reduce/3 to iterate over the html tree. We check if the element is an h2 or h3 element and extract the label and href. The label is the text content of the heading element and the href is generated using the get_href/1 helper function (the get_href/1 function searches for the anchor link in the heading element and returns the corresponding href attribute). If the element is an h3 element, we need to add it as a child of the last h2 element. This results in a nested data structure.
We can validate the function by writing a test that also demonstrates the resulting data structure:
use ExUnit.Case
test "parse_headings" do
headings =
"""
## Section 1
### Subsection 1.1
### Subsection 1.2
## Section 2
"""
|> MDEx.to_html(extension: [header_ids: ""])
|> parse_headings()
expected = [
%{
label: "Section 1",
childs: [
%{label: "Subsection 1.1", childs: [], href: "#subsection-11"},
%{label: "Subsection 1.2", childs: [], href: "#subsection-12"}
],
href: "#section-1"
},
%{label: "Section 2", childs: [], href: "#section-2"}
]
assert headings = expected
end
As you can see, the parse_headings/1 function returns a list of headings with labels, hrefs and subheadings. This data structure is suitable for building the table of contents component because it represents the hierarchy of the headings in the content.
Note that we only parse h2 and h3 elements in this example. You can extend the function to include deeper heading levels if needed.
Build the table of contents component
The last step is to build the actual table of contents component. It takes the list of parsed headings from the previous steps and an optional list of classes. Since we are using daisyUI in our Phoenix project, we can use the menu class to make the table of contents look like a sidebar menu by default (see daisyUI menu component). We also add some other default classes. You can customize the component to fit your design.
The toc/1 component renders an unordered list with list items and a link for each heading. If the current heading has child headings, the toc component is called recursively with them as input. This creates a nested HTML list structure that represents the headings data structure in the UI.
@doc """
Renders a table of contents from a list of headings.
"""
attr :headings, :list, required: true
attr :class, :string, default: "menu w-56 p-0 opacity-60"
def toc(assigns) do
~H"""
<ul class={@class}>
<li :for={%{label: label, href: href, childs: childs} <- @headings}>
<.link href={href}>
<%= label %>
</.link>
<.toc :if={childs != []} headings={childs} class={nil} />
</li>
</ul>
"""
end
How to use the component
We are now ready to use the component in our layout. On my blog, I only show the table of contents on larger screens, so I can place it next to the main content. I also made it sticky so that it stays visible while scrolling. Here is an example of how I placed the component in my layout:
<div class="relative">
<div class="absolute top-16 bottom-16 -left-14 w-10 text-xs">
<div class="sticky top-6 hidden xl:block">
<.toc headings={@article.heading_links} />
</div>
</div>
<!-- Blog page content -->
</div>
The above code uses absolute positioning to place the table of contents on the left side of the main content. It is only visible on larger screens (xl:block) and is sticky so that it stays visible while scrolling. To not stick to the top of the page, we use the top-6 class to save some space.
Conclusion
In this article, we configured the Elixir MDEx markdown parser to include IDs and anchor links for headings in the generated HTML. We then parsed the generated HTML into a nested data structure we can pass to a table of contents component we built. This component renders a list of links to the headings in the content, creating a table of contents that allows users to quickly navigate through the content.