Improve handling of "mixed content" with XML module

[cowtowncoder]

This is a placeholder issue for improving how Jackson handles XML content like:

<root>
   <value>123</value>
some text
   <value>345</value>
</root>

and other cases where a nesting level has both element(s) and at least one non-blank text value (content other than whitespace).

Currently such content can not be supported at all, and it is difficult to really find a general way of mapping this into POJOs.

However, it should be possible to have some level of support, for example maybe allow reading into JsonNode in some form -- considering that such text content could perhaps be mapped into empty key (and assuming handling of repeated Object keys can be resolved).
They may be other approaches too; to be discussed here.

[cowtowncoder]

Looks like JsonNode only builds mixed content that comes right after start element, but not one(s) after end-element. Seems like token-stream level problem, now what JsonNode itself can express multiple values as array.

[cowtowncoder]

Exposing of mixed content through XmlTokenStream (low-level) and FromXmlParser completed, which allowed fixing #405 so now mixed content is available via:

1. JsonNode, with logical key of "" (empty String) -- and as per Make JsonNode implicitly create ArrayNodes for repeated XML Elements #403 value becomes ArrayNode if multiple mixed content found
2. "Untyped" case (java.lang.Object) as well: Map key of "" (empty String), with either single String value or List<String> for multiple mixed content segments

Note that due to limitation in logic blank mixed content can not be included, so all-whitespace segments are not retained. This is unfortunately necessary to support use of indentation for non-mixed-content cases.

I will leave this issue open for possible future improvements regarding POJO (typed) use cases: these will not yet allow more than one @XmlText annotated segment.

[Ace2Pic]

Could you please detail how is it supposed to be used with POJOs?

I have this problem in 2.11.2, tried your solution with the 2.12.0-snapshot, created a similar case with a root element containing a value field with @JacksonXMLText annotation.
I can confirm that trying to deserialize the element does not throw error anymore, but I can only get the raw text, but not any inner element ("some text" in your example). Am I missing something?

[cowtowncoder]

@Ace2Pic As things are, this would not help use with POJOs, unfortunately. It does help with JsonNode (which will retain duplicates) and, to a degree, "untyped" (read as Object) including case of binding into Map.

[Ace2Pic]

@cowtowncoder Thank you for your answer. It helped!
Using a simple custom deserializer and with JsonNode I am now able to get the whole "mixed content".

[cowtowncoder]

@Ace2Pic ah yes, that makes sense!

[nrodriguezmore]

@cowtowncoder Thank you for your answer. It helped!
Using a simple custom deserializer and with JsonNode I am now able to get the whole "mixed content".

can you share the custom deserializer you wrote? thanks!

[cowtowncoder]

Closing this due to following improvements that will be in 2.12:

• XmlMapper/UntypedObjectDeserializer swallows duplicated elements in XML documents #205: when binding to Object, will now create Lists for duplicate elements
• Make JsonNode implicitly create ArrayNodes for repeated XML Elements #403, Mixed content not exposed through FromXmlParser, lost by JsonNode #405: when binding to JsonNode will similarly create ArrayNode for duplicate.

Further improvements would be useful too but require specific ideas, and will likely need to go in 2.13 or later.