Re: [Cdk-devel] MoleculeSet accepting atomcontainers

[Nina J. <ni...@ac...>]

Hi All,

Just few thoughts, in support to Stefan suggestions:

A possible solution toward more consistent design (though requiring some
API changes) would be to make use of generics, by
defining base class e.g. IChemObjectSet<T extends IAtomContainer> with
methods add(T) instead of addAtomContainer(),
iterator<T>  instead of iterator<IAtomContainer> ,Iterable<T>  etc.

and then
 IAtomContainerSer extends IChemObjectSet<IAtomContainer>
and
IMoleculeSet extends IChemObjectSet<IMolecule>

which will naturally follow to iterable<IAtomcontainer> and
iterable<IMolecule>


Best regards,
Nina



Stefan Kuhn wrote:
> Hi all,
> I fixed today the problem with JCP throwing ClassCastExceptions when saving. 
> The background was that the ControllerHub actually added IAtomContainers to 
> the MoleculeSet in the ChemModel. This is possible since the IMolecule set, 
> apart from addMolecule(IMolecule) etc., also has 
> addAtomContainer(IAtomContainer) etc. from the IAtomContainer it extends. 
> Then there is the method Iterable<IAtomContainer> molecules() which returns 
> IAtomContainers since it uses atomContainers() in the superclass.
> I think this is inconsistent design and is prone to cause trouble for other 
> people as well. I am not quite sure what to do best, I must admit. The only 
> idea I have right now is to overwrite the addAtomContainer etc. methods in 
> MoleculeSet and have them throw an exception. The molecules() method would 
> then be guaranteed to return IMolecules, but it might still be good to change 
> its return type to IMolecule and do the cast in there.
> Stefan
>