"data-" prefix is added to custom attributes, how to prevent this?

Question

I have upgraded Pandoc to v2, and some of my regression tests fail now.

It's the ones where I add custom attributes to elements like so:

# Test {role="heading" aria-level="7"}

In earlier versions of Pandoc, this resulted in

<h1 role="heading" aria-level="7">Test</h1>

While in Pandoc 2, it's

<h1 data-role="heading" data-aria-level="7">Test</h1>

How can I change it back? Both role and aria- attributes are 100% valid HTML, so no data- prefix is needed.

Answer 1

It seems to be a difference with Pandoc in the conversion from Markdown to HTML5 vs HTML4.

Using their test editor the HTML4 conversion works perfectly fine.

Pandoc HTML4 Conversion Example

However, when we switch to HTML5 the data- portion of the attribute gets injected.

Pandoc HTML5 Conversion Example

This makes me think that this is either intentionally done, for differences between HTML4 and HTML5, or that it is potentially a Pandoc bug, which you should probably raise an issue with them in their github issue tracker: https://github.com/jgm/pandoc/issues

From the Pandoc docs - html defaults to html5 as follows:

• html or html5 ([HTML], i.e. [HTML5]/XHTML [polyglot markup])
• html4 ([XHTML] 1.0 Transitional)

---

Update:

It looks like this IS because of HTML5 spec - the "role" and "aria" attributes themselves are no longer in the attribute list. Reference: https://developer.mozilla.org/en-US/docs/Web/HTML/Attributes

So this NOT a bug with Pandoc - this is expected behavior. My suggestion, use the html4 conversion is your answer.

Thanks to @Caramiriel in the comments for additional reference in the Pandoc code.