Whitespace HTML entities break DOM parsing in Drupal 10.2

Further investigation is that it seems to be related to
Html::normalize()
https://www.drupal.org/node/2441811 →

https://stackoverflow.com/a/9760247/80281

Use DOMDocument::loadHTMLFile() instead of load(). That's what it has been made for. HTML is not XML.

XML does not know the named entity  . However if you use loadHTML, the XML parser will get the HTML named entities loaded so the error goes away.

See as well: XML parser error: entity not defined.

I think this makes sense, that masterminds/html5 has been in there I believe since Drupal 8.0 release because I remember solving problems with it in the tests prior to twig being included in core.

I'm not a fan of the DOMDocument API, so my apologies for being verbose. If only there was an appendHTML() this would have been a 1 liner.

I do believe it's ready to review though, this seems to work and will work in both D10.2 and earlier

Tested but didn't work. It fails silently, headers on nodes text are missing.

My workaround to avoid filling logs, is decoding   to " ".

The same as #9 - my headers (h2 tags) in text are lost.

Szy.

I am curious about the reported failures - I tested and have not be able to reproduce the headings getting lost. Is anybody able to share sample HTML they used when the headers go lost & config?

It looks good to me, resolves the errors reported and I can still see the headings in the text.

@ericgsmith, are you leaving at least one nbsp at the end of any h2? :)

This is a condition, iirc.

Szy.

Can confirm that we encounterd this issue (in our case, an   at the start of a header) and the commit in the MR (1c95fbbb) resolved the issue.

We didn't experience the problem Matthew reported in the MR comment.

I had same result as Szy.

After applying patch, h2 were vanishing from text

Liam Morland → made their first commit to this issue’s fork.

I have put on the merge request a fix that uses html_entity_decode(). This is fixing the error messages and not removing the h2 elements.

This is the changed mentioned in #16 as a patch.

In further testing, I found this approach does not work because it outputs characters instead of entities for something that need to be entities in XML.

@ericgsmith, are you leaving at least one nbsp at the end of any h2? :)

Yes - hence saying that the patch resolved this issue.

I am curious - people reporting the headers are missing - do you have a custom toc-header.html.twig file that adds additional elements to the header? There is a flaw in the original patch in that it expects only 1 root level element in the header, I have been able to reproduce HTML being removed with a custom template that was adding additional elements in the header element.

Ok, I have opened MR12 for reasons described above (a change in approach IMO should be done in a new MR so I didn't want to revert the approach introduced MR11)

This goes back to the original approach from @joelpittet and includes a minor change to account for instances where the toc header may return more than 1 child element.

I believe the issue some described with missing text was caused by either theme debug being enabled, or by having a template what introduced additional elements prior to the heading.

#21 works for me on 1.4.0

for us the patch#21 seems to solve the problem too.
would be nice, to have an update of the module.

I can also confirm that patch#21 seems to solve the problem in Drupal 10.2.5

Patch from #21 fixes the issue for us with Drupal 10.2.5 as well.

I can confirm #21 works here too with Drupal 10.2.4

#21 also works for me.

Automatically closed - issue fixed for 2 weeks with no activity.