[Cdk-devel] Re: [Cdk-commits] CVS: cdk/src/org/openscience/cdk/qsar

[E.L. W. <e.w...@sc...>]

=2D----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1


Matteo,

thanx for starting QSAR descriptor calculation in CDK. Below are some JavaD=
oc=20
comments, but some more important things first...

Christoph might have told you about the qsar.sf.net project where we will t=
ry=20
to get opensource QSAR programs to interoperate... there already is an=20
interface for data classes, but an interface to descriptor calculation is=20
important too...

Therefore, I think it's very important that we have an interface so that=20
descriptor calculators can be pluged into any program...

At this moment the classes have a straightforward JavaBean like interface=20
which is good, but maybe we can think about a more general API.
I'm thinking about one class for each descriptor type...

Each type has one or more parameters that should be set before calculating =
the=20
descriptor. For example this example:

public class AtomCountDescriptor implements cdk.qsar.Descriptor {

    /**
     * Sets the element name of the atom type to count.
     */
    public void setParameters(Object[] params) throws CDKExceptiion {
        if (params.length > 1)
           throw new CDKException("AtomCount only expects one parameter");
        if (!(params[0] instanceof String))=20
           throw new CDKException("The parameter must be of type String");
        // process parameter
    }

    public Object calculate(AtomContainer container) {
         return new Integer(1); // well, I hope CDK can do better :)
    }=20

}

with this interface:

public interface org.openscience.cdk.qsar.Descriptor {

    // an exception would be thrown when the parameters are not
    // appropriate for this descriptor
    public void setParameters(Object[] params) throws CDKException;
    public Object[] getParameters();

    public Object calculate(AtomContainer container);

    // we should also have something like these. The order of the String[]
    // return should match the expected order in setParameters;
    public String[] getParameterNames();
    public Object getParameterType(String name);

}

Parameter processing is specific, we cannot work around that, but the=20
interface should have mechanism of introspecting that... An object is=20
returned so that we can return any type of object. Some descriptors are not=
 a=20
single value but e.g. matrix instead.

Joerg, does this match a bit the stuff in JOELib?
Rich, what about Octet?

Egon



The JavaDoc comments...

On Monday 22 November 2004 13:48, Matteo Floris wrote:
> /**
>  *  Description of the mfe4

Please do not add such comments, but either leave them out, or actually wri=
te=20
it.

>  *
>  *@author     cubic

Add your initials/name there.

>  *@created    November 13, 2004

That should be @cdk.created

>  */
> public class QsarDescriptors {
>
> 	/**
> 	 *  Constructor for the QsarDescriptors object
> 	 */

Same as above. Either write, or leave out.

> 	public QsarDescriptors() { }
>
>
> 	/**
> 	 *  Gets the atomCount attribute of the QsarDescriptors object

Ok, good. Missing dot at the end of sentence.

> 	 *
> 	 *@param  ac      Description of the Parameter
> 	 *@param  symbol  Description of the Parameter
> 	 *@return         The atomCount value
> 	 */

Remove or write, again.


Rest of that file similar comments.

> --- NEW FILE: QsarDescriptors2D.java ---
> /**
>  *  Description of the Class

Same as above.

>  *@author     mfe4
>  *@created    November 13, 2004

Please use #cdk.created and use the format YYYY-MM-DD.

>  */
> public class QsarDescriptors2D {
>
> 	/**
> 	 *  Constructor for the QsarDescriptors2D object
> 	 */
> 	public QsarDescriptors2D() { }
>
>
> 	/**
> 	 *  Gets the mW attribute of the QsarDescriptors2D object

Well... the rest is just the same.

So:
=2D - write JavaDoc or remove your editor's default JavaDoc.
=2D - use @cdk.created YYYY-MM-DD instead of @created
=2D - dot at end of first JavaDoc description sentence is *very* important



=2D --=20
e.w...@sc...
PhD-student on Molecular Representation in Chemometrics
Radboud University Nijmegen
http://www.cac.science.ru.nl/people/egonw/
GPG: 1024D/D6336BA6
=2D----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.0.7 (SunOS)

iD8DBQFBoemAd9R8I9Yza6YRArGSAKCB7T4krbW4PSIsPuQSg+LSgri8DACgnkdg
hFZn6CufJFDa0+QQbnv5cf8=3D
=3DAKZT
=2D----END PGP SIGNATURE-----