[Cdk-devel] Processing CML Conventions

[rich a. <che...@ya...>]

Hi all,

It was great meeting some of you in San Diego. It's
inspired me to get to work on CML support in Octet
(http://octet.sf.net).

I've been looking at how CDK implements support for
CML and I had some questions about the
org.openscience.cdk.io.cml package, although some of
these questions may be better addressed at CML. (feel
free to correct any of my misinterpretations below)

I was surprised to see that common file
formats/applications (i.e. molfile, pdb) and others
that are more specialized (jmol) each have their own
unique CML conventions. My impression was that CML was
a universal molecular information exchange format -
write one DocumentHandler in SAX and be done, but it
looks like the picture is not that simple?

From Egon's article at
http://www.openscience.org/~egonw/cml/cml_conventions.html
 I gather that there are a number of ways to encode
molecular information in CML.

For example, the meaning of bond order = 4 will vary.
In some ontologies it will mean an aromatic bond and
in others it will mean a quadruple bond - other
ontologies may have still another interpretation. To
be able to assign meaning to bond order = 4, a CML
convention needs to be specified in the CML document
itself as an aid to the parser.

It looks like CDK's approach is to use the Strategy
Pattern. The core strategy is CMLCoreModule which
dynamically delegates to one of a few possible
concrete ConventionInterface implementations as the
need arises.

This approach, if I am understanding it correctly,
makes sense from the perspective of software
development.

But I'm concerned about a larger issue: doesn't this
essentially leave us in the same position we were in
before CML? Sure, we can take advantage of the
semi-automated file parsing available through SAX or
DOM's, but the core issue of multiple, often
incompatible interpretations of the data in these
files remains, doesn't it?

What happens when CDK encounters a CML processing
convention for which no ConventionInterface
implementation exists? Isn't this the same as having a
file format without a parser - or a file format with a
partial parser?

thanks for your help,
rich


		
__________________________________ 
Do you Yahoo!? 
Yahoo! Small Business - Try our new resources site!
http://smallbusiness.yahoo.com/resources/