2010/11/28 Chris Selwyn <chris@selwyn-family.me.uk>

Hi Wolfgang,

Thanks for the helpful response.

I do find myself using "contains" and "memberof" quite a lot depending on whether I am going "up" or "down" the tree structure. Is this a performance problem? Is it better to use "from"?

contains and memberof are OK to use; they are 1st class "citizen" operators.

Thanks for the hint about adding the "parent" pointer in the child. I'll have to think about that one because the schema is actually shared with external trading partners.

The addition I propose has no influence on the XML data; it'll just let you generate what you need in you JAXB generated classes. Or a simple XSLT transformation might create the xjc input from the "official" XML schema.

I guess I could have an "external" version and and "internal" version but then I would have 2 versions to maintain and that is not very desirable.

Consider the XSLT proposal.

As someone else has asked, I do actually make modifications to the model that I want to return to the client. So it is not read-only. That means I would then have to do a post-rule processing pass to null out the parent pointers (otherwise I would have cycles in the graph!) before returning the answer. Actually... @XmlTransient will probably have the desired effect.

This is an annotation you can add when you start from hand-coded Java. But it shouldn't be a problem to write rules (or Java code) setting all the parent pointer to null before you marshal everything back.

(Thinks: I wonder whether it would be possible to write a JAXB plugin to do this stuff and make the generated classes more "rules-friendly")

There's no generally applicable procedure, I'd say.

Thanks also for the hint about the code injection plugin. I was not aware of that. I have just done some investigation into that and I find that the code-injection plugin is present in my Glassfish server but is not named in the META-INF/services/com.sun.tools.xjc.Plugin file. So I'm not sure that I can use it without getting a separate installation of JAXB.

This is a deficiency in some JAXB version that (I think) is distributed with Java 1.6.x. You can add the missing line to the file, and it'll work; or you can get the latest JAXB release and use this, superceding what comes with JDK.

-W

Chris

On 28/11/2010 13:12, Wolfgang Laun wrote:
An XML schema just fitting your XML data will frequently not result in a class
hierarchy that's well suited for writing rules.

One point deals with representing the parent-child relationship, especially
when there are multiple children of a kind, i.e., the parent class contains a
List<?>. You have this in XML the natural way; you lose it when you just
insert parents and children. If extending the XML schema is possible, you
can add another element with minOccurs="0" to the Child complexType of
type Parent and set the reference to the current parent when you walk the
object tree. (You can erase the List<?> in the parent.) This avoids the
frequent use of the "from" clause in conditions.

You may also have to handle repeated occurrences of elements that are
"equal" as objects but appear in full XML text; this is usually simple to handle
with a temporary Set, but you'll have to add hashCode and equals to the
JAXB generated classes using the code injection plugin.

Inserting just the root element is not advisable, as you are well aware of.

-W
On 28 November 2010 12:38, Chris Selwyn <chris@selwyn-family.me.uk> wrote:

I am working on a project that is using Drools to perform validation of
hierarchical XML messages.

So I have passed the XSDs through JAXB and have a set of interrelated
Java objects.
One of these objects (naturally) represents the root of the messages and
the others represent the intermediate and leaf nodes of the message.

My question is: Should I traverse the tree and insert each object into
the working memory or should I just insert the root object into the
working memory?

If I insert each object into the memory then I have a lot of flexibility
about how I write my rules... I can start by focussing on the particular
part of the message that I want to detect an error in.
However, I then have to perform a whole bunch of "joining" clauses to
correctly "connect up" to the other objects in the working memory.
(This is how I am doing it at the moment... I have a mechanism that uses
the Java Introspector. It identifies by package name which parts of the
data model get inserted into the WM)

On the other hand, I could just insert the root object into the WM. This
means that I would have to write really complicated where clauses on the
root object but very much simplifies the WM loading process.

All of the examples that I have seen have a very simple flat
(non-hierarchical) structure that don't really give a hint as to what a
"best practice" might be.

Any suggestions/insights welcome :-)

Chris Selwyn
_______________________________________________
rules-users mailing list
rules-users@lists.jboss.org
https://lists.jboss.org/mailman/listinfo/rules-users
_______________________________________________
rules-users mailing list
rules-users@lists.jboss.org
https://lists.jboss.org/mailman/listinfo/rules-users
No virus found in this message.
Checked by AVG - www.avg.com
Version: 10.0.1170 / Virus Database: 426/3284 - Release Date: 11/27/10
_______________________________________________
rules-users mailing list
rules-users@lists.jboss.org
https://lists.jboss.org/mailman/listinfo/rules-users