How to parse CDATA text in HTML

[JanPretzel]

In 1.23.1 the CDATA tokenization was aligned with the HTML spec, so CDATA syntax in HTML content is now parsed as a bogus comment rather than a CDataNode.
Is there a recommended migration path for a case wher I would want CDATA as text in HTML still?

Context

I process markup from a third-party system that mixes HTML with custom namespaced elements. Some of these elements carry text bodies wrapped in CDATA, e.g.:

xml

<ns:foo><![CDATA[some text
with newlines and    spacing]]></ns:foo>

I've been reading the body with element.wholeText(). Under 1.23.1 this now returns an empty string, because the CDATA is tokenized as a bogus comment under the HTML parser and no longer appears as a CDataNode.

Should I manually pares the Comment nodes, switch to XML parsing or is there another option?

[jhy]

Hi there,

CDATA sections are XML syntax, so if the source format uses them inside custom namespaced elements, I’d parse the document as XML:

Document doc = Jsoup.parse(input, Parser.xmlParser());
String text = doc.expectFirst("ns|foo").wholeText();

The HTML syntax only recognizes CDATA inside SVG or MathML foreign content. A tag such as <ns:foo> does not itself move the HTML parser into an XML namespace, so treating the CDATA opener as a bogus comment is the expected HTML behavior.

I wouldn’t recommend extracting it from Comment nodes. A HTML bogus comment ends at the first >, so the original CDATA body may already have been split across multiple nodes.

If the surrounding document genuinely requires HTML parsing, the producer should instead emit ordinary text with < and & escaped.

Otherwise, preprocessing the source before HTML parsing would be the remaining workaround.