Custom deserialization logic not reporting CDATA events

[mmkathurima]

Using the following code:

@Override
public XmlNode deserialize(JsonParser p, DeserializationContext ctxt) {
    if (p instanceof FromXmlParser) {
        XMLStreamReader reader = ((FromXmlParser) p).getStaxReader();
        while (reader.hasNext()) {
             switch (reader.next()) {
                 case XMLStreamConstants.CDATA:
                     System.out.println("Found CDATA");
                     break;
             }
         }
     }
}

The CDATA branch is never reached.
Even after enabling the P_REPORT_CDATA and disabling the IS_COALESCING flags:

XMLInputFactory xmlInputFactory = new WstxInputFactory();
xmlInputFactory.setProperty(XMLInputFactory2.P_REPORT_CDATA, Boolean.TRUE);
xmlInputFactory.setProperty(XMLInputFactory2.IS_COALESCING, Boolean.FALSE);
XmlMapper xmlMapper = new XmlMapper(xmlInputFactory);

[cowtowncoder]

What kind of XML document are trying to read? Is content being reported as XMLStreamConstants.CHARACTERS?

[mmkathurima]

Any kind of document. An example is as below

<data><![CDATA[Some <CDATA> content]]></data>

This causes the content to be reported as characters. I looked into the Woodstox source and found that if the flags P_REPORT_CDATA is not enabled and IS_COALESCING is not disabled then CDATA will be reported as characters.

[mmkathurima]

Is it feasible to make the Woodstox XMLInputFactory accessible and enable changing flags?

[cowtowncoder]

@mmkathurima I don't think XML module should allow that -- but you can pass XMLInputFactory like you are doing.

Without looking at the code (don't have time do dig deeper right now), it is possible/likely that XmlMapper may force some settings.

... Oh, yes: it's almost certainly due to P_COALESCING which is needed to avoid splitting of CHARACTERS segments; segmentation would not play nicely with data-binding (because one property would get multiple values). Otherwise settings are left as-is.

One possibility would be to allow a new FromXmlParser.Feature that would prevent applying of this setting. I don't think the default settings can be changed (coalescing is pretty much required for most regular usage), but non-default features are ok for changing them.
Setting could either affect P_COALESCING, or enable/disable XmlInputFactory settings overall.

But overall I suspect it's going to be challenging to detect "CDATA-ness" of text content.

[mmkathurima]

One thing to note is that when CDATA is reported as characters, the CDATA tags are obviously skipped and text is parsed as Some <CDATA> content.

Additionally, passing XMLInputFactory to XmlMapper doesn't work. You're right by adding a new FromXmlParser.Feature just for reporting CDATA.

[cowtowncoder]

Additionally, passing XMLInputFactory to XmlMapper doesn't work

Do you mean it does not lead to CDATA events being reported or something else?

[mmkathurima]

Do you mean it does not lead to CDATA events being reported or something else?

Yes, CDATA events are not being reported.