Markdown2Html-Transformer with turned off LinkValidation causes error 'unresolved link id reference'

[sven42]

Hello,

I have the following Markdown (generated by LLM):

[p][Att Doc 30636](https://Universe.edu)

Using the following transformer with explicitly turned off Link-Validation causes the following error:

Transformer
    .from(Markdown)
    .to(HTML)
    .using(GitHubFlavor)
    .withConfigValue(LinkValidation.Off)
    .build

[1]: unresolved link id reference: att doc 30636

   [p][Att Doc 30636](https://Universe.edu)
   ^

Is this a bug or what can I do to make the transformer work?

Best regards,
Sven

[valencik]

I don't think this is a bug with LinkValidation.
It's possible it's a parsing bug though.
Either way, I admit it can be surprising behaviour.

It's worth noting that [p][Att Doc 30636](https://Universe.edu) is a tricky input.
It wasn't obvious to me how that would parse.
Is it a reference-style link ([click here to read more][1]) followed by some text in parenthesis?
Or is it some bracketed texted followed immediately by a link?

It turns out Laika parses this as the first, a reference-style link followed by some text in parens.
It then tries to resolve (not validate) this link reference, basically looking further in the file for the real link value.

But we don't have a link reference definition anywhere else in our input.
In fact, if we add a link reference definition on another line, then we don't get an error:

//> using scala 3.3.7
//> using dep org.typelevel::laika-core:1.3.2

import laika.api._
import laika.format._
import laika.config.LinkValidation

@main def repro(): Unit = {
  val markdown = """[p][Att Doc 30636](https://Universe.edu)
  
  [Att Doc 30636]: https://example.com"""

  val transformer = Transformer
    .from(Markdown)
    .to(HTML)
    .using(Markdown.GitHubFlavor)
    .withConfigValue(LinkValidation.Off)
    .build

  val result = transformer.transform(markdown)
  println(result)
}

prints:

Right(
  <p><a href="https://example.com">p</a>(<a href="https://Universe.edu">https://Universe.edu</a>)</p>
)

The error we're getting on the original input is from LinkResolver.
We are encountering the error before we hit the link validation.
Which is ultimately why setting withConfigValue(LinkValidation.Off) does not affect the results for this particular input.

Ok, so is this a parsing bug though? It's possible it might be.
At the very least, Laika is parsing this differently from how the spec.commonmark.org tool does: