XML Empty tag to Empty string in the object during xml deserialization

[thefalconfeat]

I am using jackson-dataformat-xml:2.6.2

If an empty tag is available in the xml and the respective field in the object is a string, it would be nice to have jackson populate the field with empty string rather than with null value.

Ex:

@XmlAccessorType(XmlAccessType.FIELD)
@XmlType(
        name = "",
        propOrder = {"first", "last"}
)
@XmlRootElement(
        name = "name"
)
public class Name {
private String first;
private String last;
}

xml:

<name>
<first>Ryan</first>
<last></last>
</name>

mapper creation:

new XmlMapper(XMLInputFactory2.newFactory(), XMLOutputFactory2.newFactory());

I would expect

Name name = mapper.readValue(xml, Name.class)

name.getLast().equals("") to return true but name.getLast() ends up being null.

Any help would be appreciated.

[cowtowncoder]

Yes I can reproduce this. It is a tricky problem to solve but I added a failing unit test so that hopefully it may be resolved in future.

In the meantime a work-around would be to add explicit setLast() that can convert null into empty String (and perhaps also pre-initialize last as empty string in constructor.

[cowtowncoder]

I have some ideas as to how this could be implemented; but actual fix will need to wait until 2.8 unfortunately, as some API changes will be needed.

[RobinAllen]

I'm also facing what I think is the same issue.
In my case (simplified here to just the basic problem), I'm reading xml which defines a 'Thing's state as A or B:

State A : "<Thing><A /></Thing>"
State B : "<Thing><B /></Thing>"

Both A and B are defined in XSD as being 'xsd:string' with max length 0 characters. 'Thing' is a complextype, defined as being xsd:choice between element A or element B. They are intended to have no content; it is purely the presence or absence of the empty element which declares which state we are in. Sadly I don't have control over the interface to be able to change the logic. I have to live with this XML interface.

In Java, the POJO generated from XSD has String fields 'a' and 'b'.

If I serialize from POJO to XML I have no problem, but the trouble comes when I try to deserialize again.

If I deserialize "<Thing><A /></Thing>", then I want to end up with [ a="", b=null ]
If I deserialize "<Thing><B /></Thing>", then I want to end up with [ a=null, b="" ]
But whichever input I deserialize, I get the same results: [ a=null, b=null ]

So you see, I'm unable to detect WHICH empty element was present, which leaves me unable to know which state it was in.

If I could submit some content inside the A / B, even just a space character, then it would deserialize properly. But I can't validly submit content inside the A / B element, because that would not be acceptable according to the XSD (max length 0 characters).

I see this as a fault, really, because dataformat-xml cannot read back in the xml which it itself has outputted. If I serialize a Thing from POJO to XML string and then deserialize back to POJO again, I don't get back in the same object I wrote out. (Quoting the readme page: "What should be guaranteed is that any XML written using this module must be readable using module as well")

[cowtowncoder]

@RobinAllen This is indeed problematic, but consider that without mandating use of XML Schema's nillable, there really is nothing to definitely differentiate empty String from null.

Anyway, I do hope that with 2.8 (to be worked on next now that 2.7.0 is ready to release) we can at least ensure that empty element, for String valued property, will become "". And total absence of element would in that case remain null. And finally, maybe nullability with XML schema can be eventually supported as well, for explicit nulls.

[manueljordan]

Hello

what is the status of this situation?

I am working with Jackson 2.8.2 and seems this situation is reflected until now in:
Spring Rest & Jackson: Empty values are transformed in Null values in XML transformation

Thanks

[cowtowncoder]

@manueljordan If there is progress, there will be update here. I have not worked on this, and am unlikely to have much time in near future.

[manueljordan]

@cowtowncoder thanks by the quick reply. I understand. The important is it was reported. Yes, take your time, I know you are busy. Meanwhile I have turned around this situation in other way.

[jlous]

Eagerly awaiting this! Would a pull-request make any difference in ETA in an official release?

[jlous]

Any idea when the PR might be merged if approved?

[manueljordan]

What is needed to approve the PR?

[oistein]

Yeah, what is stopping this?

[zhoulei-patsnap]

@cowtowncoder
Hello,what is the status of this situation? Which is version that the bug is fiexed?

[jhaber]

This is a pretty painful issue to run into, but after a lot of not-so-fun experimentation I came up with two different workarounds that I think should handle most use-cases. You can check them out here

[cowtowncoder]

Quick note: related to #209, Jackson 2.9 SHOULD allow use of:

@JsonSetter(nulls=Nulls.AS_EMPTY)
public String value;

to basically coerce null value into empty String (or one of alternatives; skip assignment to leave as default, throw exception). It's via databind feature:

FasterXML/jackson-databind#1402

I will test that this works, and if so, mark this as closed.

[cowtowncoder]

Ok. Hmmh. So, good and bad news.

Good news: woth empty tags, works as expected.

Bad news: open+start, without even intervening space (so simple <value></value>) does NOT work as expected. Can not remember if (... and, how?) there was intentional checking to effect different handling. But XML semantics should not differ between two representations.

[cowtowncoder]

So. Works as suggested with empty tag, in 2.9. And with ACCEPT_EMPTY_STRING_AS_NULL_OBJECT can make other cases work too.

In 3.0.0 will further work the same way for "split" empty tag (start+end).

[cowtowncoder]

Note: some of the fixes to jackson-databind will go in 2.9.6. So above comments reflect to that upcoming version.

[itisvamsi]

Is there any fix for this which is available?