Dom\HTMLDocument: Impossible to add text node on document level

Description

The following code...

$doc = \Dom\HTMLDocument::createFromString('Hello <b>world</b>!',LIBXML_HTML_NOIMPLIED);
$greeting = $doc->firstChild;
echo get_class($greeting)."<br />";
$new_greeting = $doc->createTextNode('Welcome');
$greeting->replaceWith($new_greeting);
echo htmlentities($doc->saveHtml());

results in this error:

DOMException: Cannot insert text as a child of a document in Dom\CharacterData->replaceWith()

This is different behavior from the old DOM classes which allowed this. I understand, this spec-compliance is supposed to be a feature. If I understand correctly, the new Lexbor-based parser offers validity checks for some methods and not for others:

https://lexbor.com/modules/dom/

E.g. lxb_dom_node_insert_before_spec vs lxb_dom_node_insert_before etc.

I'm unaware of the current PHP implementation but I assume it is using the _spec versions throughout without any option to turn it off.

I think this is problematic. The NOIMPLIED option is very useful indeed if you want to handle document fragments etc. without unnecessary wrapper tags.

The above code leads to the peculiar situation that you can indeed create a Document with textNodes on the document level, but if you just want to replace that textNode with a different textNode it won't work. This complicates things a lot unnecessarily. Other methods like insertAdjacentHTML, appendChild etc are equally stubborn.

There are different solutions for this: IMHO the easiest one would be to use the non-spec versions of these functions if NOIMPLIED was specified. NOIMPLIED explicitly allows a non-spec document to be imported. Any manipulation should also be allowed then.

Another alternative would be to create a new general flag LIBXML_HTML_RELAX_SPEC.

A third alternative would be to specifically add option flags to the individual methods to allow this for a single call.

I'm unsure what's really the best solution but at the moment, porting over existing code is unnecessarily difficult due to this restriction and I really think this feature would make a lot of sense anyway. Thanks for considering.

Empieza reproduciendo el fragmento de PHP con Dom\HTMLDocument::createFromString, LIBXML_HTML_NOIMPLIED y CharacterData::replaceWith(). Traza cómo replaceWith(), insertAdjacentHTML() y appendChild() gestionan los nodos de texto a nivel de documento y, después, compara los puntos de entrada de Lexbor conformes y no conformes con la especificación mencionados en el issue. Se considera terminado cuando se haya llegado a una decisión clara y se haya implementado o documentado un comportamiento coherente para los documentos NOIMPLIED, con pruebas para el caso mostrado.