Check lists of nice to haves for CDK 3.0 some of these are large sweeping API changes hence v3.0.

• Switch IElement.H etc to int rather than byte (pre-v3.0 possible)
• Move to Java 11, 17, 21 for JENA dependency (v3.0 only)
• JUnit 5
• API improvments
  • Stereo element enums (v3.0 remove the int fields masks)
  • Stream API on AtomContainer Set etc, should it’s self be iterable
  • Stereo map
  • IAtom.newBond(IAtom, IBond.ORDER.SINGLE) method? - forcing the atoms to exist before they are bonded
  • Update query atom container for better iteration

v3.0 branch

• Java 17, 21
• Remove clone
• Move classes to different packages for CDK modules as Java modules? #626, may not be possible overall but some should be (v3.0 only - will be painful upgrade downstream but good long run, less annoying than IMolecule -> IAtomContainer)
• Use primitive types in IAtom, i.e. atomic number and mass=0 rather than null (pre-v3.0 maybe)
• first-class Sgroup support (need interfaces) (pre-v3.0 possible)
  • update manipulator.copy()
• Remove old AtomContainer implementation (see plan in https://github.com/cdk/cdk/wiki/AtomContainer2) (v3.0 only). Should check some other downstream tools (i.e. AMBIT) to see what is breaking and if we can tighten up the API
  • DepthFirst default substructure matcher
  • return the ref from the container IAtom addAtom(atom) and IBond addBond(IBond bond)
  • Return the AtomRef from getAtom etc
• Remove JNI InChI portability options
• replace VecMath with our own Point2d/Point3d, the cdk "core" should have no 3rd party dependencies
• fix typo: AtomContainerManipulator::percieveAtomTypesAndConfigureAtoms to AtomContainerManipulator::perceiveAtomTypesAndConfigureAtoms
• SMILES Canonicalisation with cdk-inchi
• Merge MDL readers see merge MDL V2000/V3000 readers #952
• Stabalise atom types Unstable CDK atom types #1090
• Merge CML code into a single module merge CML code into a single module #965