Deserialization of `null` String values in Arrays / `Collection`s not working as expected

[mbellade]

Environment: Jackson Dataformat XML 2.14.2

We are trying to serialize and deserialize arrays or collections of Strings containing both empty and null values.
We noticed that nulls are handled correctly when serializing:

• null values are rendered as <item />, while empty strings as <item></item>
• enabling ToXmlGenerator.Feature.WRITE_NULLS_AS_XSI_NIL works as expected

When parsing this values back null elements have become empty ("") values.
We tried:

• explicitly enabling FromXmlParser.Feature.PROCESS_XSI_NIL, even though it should be enabled by default
• enabling FromXmlParser.Feature.EMPTY_ELEMENT_AS_NULL
  But we still got empty String values as a result.

Testing showed that with simple String attributes inside a class both null and empty values are correctly handled during deserialization, as expected (see also #354).

The behavior of String values inside collections should be aligned to how String attributes inside objects are handled.

Here's an example:

XmlMapper xmlMapper = new XmlMapper();
xmlMapper.enable( ToXmlGenerator.Feature.WRITE_NULLS_AS_XSI_NIL );
xmlMapper.enable( FromXmlParser.Feature.PROCESS_XSI_NIL );

String string = xmlMapper.writeValueAsString( new String[] { "", "test", null, "test2" } );
// string: <Strings><item></item><item>test</item><item xmlns:xsi="http://www.w3.org/2001/XMLSchema-instance" xsi:nil="true"/><item>test2</item></Strings>
String[] parsed =  xmlMapper.readValue( string, String[].class );
// parsed: ["", "test", "", "test2"]

[cowtowncoder]

Quick note(s):

<item /> and <item></item> are semantically equivalent in XML and although (some) parsers can indicate difference (Woodstox does) -- and most can produce one or the other explicitly -- it's not a good idea to try to use difference for semantics, due to interoperability concerns.

Having said that, yes, handling of String values in Arrays or typed Collections should work the same way as individual properties. Handling of difference between null and empty String are tricky, and I think Array/Collection deserializers probably have some shortcuts that would need to be disabled for XML.

[mbellade]

Thanks for the feedback.

<item /> and <item></item> are semantically equivalent in XML

I agree that's working as expected, that is why I went for the xsi:nil approach in the example - just wanted to exhaust all possibilities with the different configurations/features and let you know what I found from my testing.

[cowtowncoder]

@mbladel yes good. I did realize it is/was not the main point. Mostly mentioned since at some point I was trying to consider/expose distinction and spent some time thinking about it.