Andreas Veithen created AXIOM-420:
-------------------------------------
Summary: Support push-only OMDataSources
Key: AXIOM-420
URL: https://issues.apache.org/jira/browse/AXIOM-420
Project: Axiom
Issue Type: Improvement
Components: API, LLOM
Affects Versions: 1.2.13
Reporter: Andreas Veithen
Assignee: Andreas Veithen
Fix For: 1.2.14
The OMSourcedElement/OMDataSource feature assumes that a data source is equally
well capable of providing an XMLStreamReader and of serializing the data to an
XMLStreamWriter. However, this is often not the case: serializing to an
XMLStreamWriter is in generally easy, but constructing an XMLStreamReader may
be highly non trivial. The consequence is that many OMDataSource
implementations implement getReader by first serializing the data to an
intermediary representation and than constructing an XMLStreamReader from that
representation. The following two approaches are used most often:
1. getReader serializes the data as XML to a byte array (using one of the
serialize methods defined by OMDataSource) and then returns an XMLStreamReader
pull parser constructed from that byte array. Note that this is highly
inefficient (and even gives unexpected results) if the serialization produces
optimizable base64 binary data.
2. getReader first creates an Axiom tree and then uses
OMContainer#getXMLStreamReader(). For example, ADB does this by using an
XMLStreamWriter implementation (called MTOMAwareOMBuilder) that constructs the
Axiom tree. That XMLStreamWriter is passed to the serialize(XMLStreamWriter)
method. That approach is inefficient because in the worst case, the Axiom tree
is built twice: once by getReader and then again by OMSourcedElement. This
should be considered as an anti-pattern (see
http://ws.apache.org/axiom/userguide/ch05.html#d5e679).
The issue can be avoided if the implementation of OMSourcedElement is changed
such that for this type of OMDataSource it uses a different approach to expand
the element: instead of using a standard OM builder to expand the element, it
would use the OMDataSource#serialize(XMLStreamWriter) method together with an
XMLStreamWriter that builds the nodes and adds them to the OMSourcedElement.
This is similar to what ADB already does, except that the tree would be built
only once because the constructed nodes are directly added to the
OMSourcedElement. Of course, since this approach is push based, deferred
building will not work and the OMSourcedElement is always built completely if
it gets expanded. It is nevertheless better than the other approaches currently
in use.
--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators:
https://issues.apache.org/jira/secure/ContactAdministrators!default.jspa
For more information on JIRA, see: http://www.atlassian.com/software/jira
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
