Re: [Cdk-devel] reading the qsar-dict entries

[Egon W. <e.w...@sc...>]

On Wednesday 09 February 2005 08:06 pm, Peter Murray-Rust wrote:
> At 11:06 09/02/2005 +0100, Egon Willighagen wrote:
> >On Tuesday 08 February 2005 08:57 pm, Rajarshi Guha wrote:
> > > For what I'm doing now, it seems the answer is yes. But in the long run
> > > will we want to access the BibTeX information from the dictionaries? I
> > > think yes, otherwise why would it be there :)
> >
> >In the long run, we'll use CMLDOM for this. But rumours go that even
> > dictRef like things are going to be replaced by RDF...
>
> I think we'll see that other disciplines (like bioscience) make rapid
> progress with ontologies and we should narrow where appropriate. 

Is there an overview of ontologies people are working on?

> But for 
> now continue to use dictRef with the semantics:
> <blurf xmlns:foo="http://www.zzz.org/dict/foo.xml">
> <xyzzy dictRef="foo:bar" >abc</xyzzy>
> </blurf>
>
> The namespace foo must now have a namespace URI/URL (or Java 1.5 will fail
> to parse). This URL should be an absolute or relative address. "bar" should
> point to the entry with id="bar".

Rajarshi, did you read the QSAR dictionary into CDK with Java 1.5? Did you 
have problems with the default JAXP XML parser of 1.5?

> The main problem is the addressing. I am still struggling with this:
> - if you use relative URLs and relocate the file, the addresses break
> - if you use absolute ones you rely on a web presence or a duplicate of the
> local file system

Maybe this is of help. I use jEdit as XML editor which uses XML Schema 
Instance (or so, 'xsi'):

<dictionary
  xmlns:xsi="http://www.w3.org/2001/XMLSchema-instance" 
  xsi:schemaLocation="http://www.xml-cml.org/schema/cml2/core cmlAll.xsd
                      http://www.w3.org/1998/Math/MathML      
mathml2/mathml2.xsd
                      http://bibtexml.sf.net/                 bibtexml.xsd"
  
  xmlns="http://www.xml-cml.org/schema/cml2/core"
  xmlns:bibtex="http://bibtexml.sf.net/"
  xmlns:cvs="https://www.cvshome.org/"
  xmlns:dc="http://dublincore.org/"
  
  xmlns:qsar-descriptors="http://qsar.sourceforge.net/dicts/qsar-descriptors"
  xmlns:qsar-descriptors-metadata="http://qsar.sourceforge.net/dicts/qsar-descriptors-metadata"
  id="qsar-descriptors" title="QSAR.sf.net Descriptor Dictionary">

where xsi:schemaLocation gives hints to XML editors where schema's can be 
found. Also note the definition of the namespace definitions of stuff in 
dictRef and metadata@name, e.g. qsar-descriptors-metadata, cvs, dc and 
bibtex.

> XML./SGML addresses this sort of thing with a catalog. It isn't perfect and
> is non-trivial to implement. So we have to have a local dictionary catalog
> of some sort. I think we need to use the same $user.home mechanism as for
> .jmol. .jchempaint, .jumbo  where each directory has a config.properties
> file. In whatever case we have to have:
>   - the user maintains their dictionaries
>   - we provide a dictionary installer
>   - the dictionaries are on the web

Indeed. In JChemPaint, STMML based dictionaries can manually be copied into 
the $HOME/.jchempaint/dicts/ directly...

Yes, an installer would be valuable. It would 
detect .jumbo, .jmol, .jchempaint directory structures, download an 
dictionary index from, say, the WWMM server, and propose to install the 
dictionary for the available programs.

> It is here where I think RDF will help as the browsers become RDF-aware

Egon