This page contains things which would be great to have them in CDK. Everybody is invited to work on them, but they will require some research and might (in the extreme case) even turn out to be infeasible.
Representation-independent STEREO coding
Right now, all wedge bonds from e. g. mol files are put into the CDK molecule objects as a such. This is not too good, since several graphical representations can actually mean the same. The SMILES generator works out these representations and unifies them. The SMILES parser ignores stereochemistry (we would need a routine to transform these to wedges).
It would be better to have a coding of stereo configurations which is independent from graphics/wedges. The SMILES parser could read directly to it, the SMILES generator would generate directly from it. The MDL reader (and others, like CML reader) would use the code which is now in SMILES parser to transform to this layer. New code would be needed to transform from this layer to wedge representation.
RG - If we're going to put in stereochemistry, we should support other types as well such as cis/trans. One possibility is implementing the Dietz representation to allow full generality
There is IAtomParity which is wedge bond independent already; it's just that no IO classes are using it http://chem-bla-ics.blogspot.com/ 06:33, 10 April 2009 (UTC)
Excluded volumes and directionality in pharmacophores
Pretty much self explanatory, though how exactly it would be implemented and how it would be used in pharmacophore definitions and searching remains to be considered. Requires refactoring of the pharmacophore classes
JJTree or JavaCC implemention of the OpenSMILES grammar
Detection of (potential) stereo centres
In a molecule without stereo specification (ie. without wedges), this should detect stereo centres, probably telling the type of stereochemistry.