Deserialization loses data when HTML looking data is passed

[keenan]

I'm seeing data loss when deserializing a document that has HTML data in a string typed element. I'm using Document Literal, JWSDP 1.6 and have tested with JBoss 4.0.2 and 4.0.3RC2.

I can reproduce the problem with a slight modification to the JBossWS test suite. In the test client named /org/jboss/test/webservice/marshalltest/MarshallDocLiteralTestCase.java, if you change the test data in the first part of the testEchoString method from "Hello" to

"Hello <em> Mom </em>"

in both the call and the assert, you'll see the effects.

I don't think the problem is simply that this results in an invalid XML document, intercepting the SOAP reveals that the special characters were serialized into < notation. Instead, looking at what happens in a debugger, it seems that the value object's set method is being called repeatedly with different parts of the string, so that the last part of the string overwrites all the previous parts.

Can someone confirm the problem, or tell me what I'm doing wrong. Thanks,
--keenan

[anil.saldhana]

http://jira.jboss.com/jira/browse/JBWS-362

We had an issue to investigate the return of "carriage return". I have updated that issue to investigate the retutn of html elements in a string.

[anil.saldhana]

http://jira.jboss.com/jira/browse/JBWS-362

lists guidelines on passing HTML tags within a string. Basically xsd:String cannot be used. You will have to use xsd:any for this. As the String basic schema type forbids nested tags. Look at the aforementioned JIRA issue.

[keenan]

Anil,
Thanks for your attention.

However, the instance document does not contain HTML because the serializer on the client side (generated by JWSDP) converted

<em>

to

<em>

(The second block is supposed to contain the characters "& l t ; e m & g t ;" in case they get changed back to the symbols).

So the problem isn't that I'm passing HTML (which I agree is invalid unless the schema uses xsd:any), instead I think the string parsing routine is splitting the string up on "special character" boundaries, and only storing the last chunk into the value object artifact.

For example, although I'm not in a position to test it right now, I believe changing the test case to attempt to pass the string

Joan & Ed

will result in just the " Ed" part being stored into the value object.
--keenan