[
https://issues.apache.org/jira/browse/AXIOM-282?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15188259#comment-15188259
]
Hudson commented on AXIOM-282:
------------------------------
SUCCESS: Integrated in axiom-trunk #2596 (See
[https://builds.apache.org/job/axiom-trunk/2596/])
Use a different approach to implement AXIOM-282: instead of using lookahead,
allow newly registered custom builders to be applied to the last created node.
Since we don't allow direct access to the underlying parser anymore, this
approach basically has the same effect but allows to simplify the design of the
builder. (veithen: rev 1734317)
*
axiom/aspects/om-aspects/src/main/java/org/apache/axiom/om/impl/common/builder/CustomBuilderManager.java
*
axiom/aspects/om-aspects/src/main/java/org/apache/axiom/om/impl/common/builder/StAXHelper.java
*
axiom/aspects/om-aspects/src/main/java/org/apache/axiom/om/impl/common/builder/StAXOMBuilder.java
*
axiom/aspects/om-aspects/src/main/java/org/apache/axiom/soap/impl/mixin/AxiomSOAPBodySupport.aj
> Get Payload QName information without building OMElement
> --------------------------------------------------------
>
> Key: AXIOM-282
> URL: https://issues.apache.org/jira/browse/AXIOM-282
> Project: Axiom
> Issue Type: Improvement
> Reporter: Rich Scheuerle
> Assignee: Rich Scheuerle
>
> Problem:
> The Axis2 code invokes MessageContext.isFault() before starting the phases.
> This causes an Axiom tree to be built up to and including the
> first child element (payload) in the soap body. The problem is that the
> phases might have special CustomBuilders. Building the Axiom tree past the
> payload
> can defeat these kinds of optimizations.
> Solution1:
> The first solution is tactical and will work with all parsers. The proposal
> is to add some simple, protected "look ahead" code in the StaXOMBuilder.
> The Axiom SOAPBody can use this information to determine whether the payload
> is a fault, without actually building the object. I am working on the
> tests and code for this first solution, and will have it ready in the next
> day or two.
> Solution 2:
> The second solution is support a custom property on the parser. The
> SOAPEnvelope can then call the custom property on the backing parser to get
> the soap body payload qname. The idea is that the parser (woodstox or ibm,
> etc.) may have already looked ahead in the text before the StAX events are
> fully pulled.
> The benefit of this solution is that calls, like MessageContext.isFault(),
> will not cause the expansion of the header elements. I will work on this
> solution
> after solution 1 is done. I don't have a proposed name for the property yet.
> (Note that both of these solutions will require some minor changes to the
> Axis2 code.)
> P.S. I am working with Takahide Nogayama on these changes.
> Thanks,
> Rich Scheuerle
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
