XOM API for Processing XML with Java

[peter murray-rust <pm286 AT cam.ac.uk>]

At 19:32 27/12/2005, Elliotte Harold wrote:
> Kevin Y. Kim (Lists) wrote: > Hey all, > > Is there someone who's 
> willing to share/provide a good example of > Element subclassing > 
> and a custom node factory?  Conceptually, I think it's what I want 
> to > do based on the > tutorial, but I'm not sure how to begin. >

I have done this to support Chemical Markup Language (CML) - you are 
welcome to take any ideas from the latest release:
http://sourceforge.net/project/showfiles.php?group_id=51361&package_id=47718&release_id=375325

It's quite large but has  a very regular format. I'd also be grateful 
for comments on the overall approach because it tackles a general 
problem that may have wider applicability. My strategy is:
- create an XML Schema to describe the chemical domain using XSD 
(this has many limitations but it seems the best approach at 
present). There are over 100 elements and 100 attributes. Many 
attributes and other simpleContent have restrictions (e.g. numeric, 
lists, minimum and maximum values). For example the elementType 
attributes must conform to the 118 atoms in the current periodic 
table for chemical elements. hydrogenCount must be a non-negative integer.
- generate code to support the functionality in the schema. There is 
an custom subclass for each XOM element (e.g. <atom> creates 
AbstractAtom. This has all the getters and setters defined by the 
schema. Thus atom has AbstractAtom.setElementType(String) which sets 
the chemical element. This automatically validates whether String is 
an allowed chemicalElement. Most XOM Elements have between 5 and 15 
such attributes. One advantage of hardcoding these is that the 
application programmer gets guidance as to what attributes are 
allowed rather than having this fail at runtime.
- generate a custom CMLNodeFactory that creates objects of the 
correct type during the parse process. This delegates the 
domain-specific stuff to finishMakingElement (see below);
- subclass the Attribute class to support a range of simpleContent. 
There are about 6 hardcoded subclassed attributes. Thus 
getHydrogenCount() returns an instance of IntegerAttribute, whose 
getters and setters deal with ints rather than Strings. Further 
restriction is done though the setCMLValue() method on the attribute 
which can check for values, cardinality, etc.

Apart from the subclassed attributes, all code is now autogenerated 
from the schema. If a document is parsed it will automatically 
generate elements of the correct class, attributes of the correct 
subclass and will schema-validate all the attribute values on input. 
(Note :I do not validate content models as they are almost useless 
for what I want to do and I would really like to see a more flexible 
approach to content models).

So far the system is automatic and knows nothing about chemistry. But 
every XOM element can be subclassed by hand, so we have CMLAtom 
extends AbstractAtom. A key feature is that finishMakingElement is 
overridden so chemical logic can be added. For example a chemical 
bond has to refer to existing atoms and this requirement can be 
validated on input. This makes XOM an extremely powerful system - it 
can carry out schema- and non-schema validation on input and also 
carry out complex transformations of the data.

I am now starting to add handcrafted subclassed Attributes, so that 
there is effectively an attributeFactory. For example many of my 
elements can have a UnitsAttribute which points to a range of 
dictionaries supporting scientific units of measurement. This 
requires considerable validation and logic and again all this can be 
done on input when the attributes are created as XOM Nodes.

Because the schema continues to evolve it would be almost impossible 
to keep a handwritten set of subclasses in sync. Now every time I 
modify the schema I regenerate the abstract subclasses in less than a 
minute. I'd be very interested in hearing from those who have the 
same requirements and are tackling them in a XOM context.

P.


Peter Murray-Rust
Unilever Centre for Molecular Sciences Informatics
University of Cambridge,
Lensfield Road,  Cambridge CB2 1EW, UK
+44-1223-763069