Re: Multiple molecules: was [Open Babel] CML bugs

[Michael B. <mb...@gm...>]

On Sun, Sep 21, 2003 at 09:31:43PM -0500, Geoff Hutchison wrote:
> I'm not sure the best and most general way to handle this, since you 
> don't know you have a mult-molecule file until you get into it. One 
> easy solution is to change all of the ReadFormat() and WriteFormat() 
> codes to take OBMolVector() objects instead of OBMol() objects.

I thought about this problem a bit over the last weeks. I didn't come up
with a clear design yet, mostly because I shouldn't think about it in
the first place and rather concentrate on my studies :)

Just some random thoughts:

 * A couple of classes could be derived from OBMolVector for each
   possible (and chemically different) list of molecules. It includes
   an OBMolVector plus additional specific information. OBConformer,
   OBGeomOptStep, OBMoleculeList, OBReaction{Reactants,Products} come to
   mind.

 * Especially for reactions, have a OBReaction class which includes the 
   two vectors OBReactionReactants and OBReactionProducts, plus possible
   information like reaction conditions, catalysts, whether it's an
   reaction equilibrium etc.

 * File formats should be able to declare whether they are able to
   read/write multiple conformers, geometry optimization steps or
   whether they are able to handle reactions. (This goes a bit into the
   "Let each file format be a class of its own, with their own
   attributes and methods (i.e. read/writing molecules) implementation"
   direction)

 * One superclass should be taken by the read/write routines. This
   superclass includes the above MolVectors. The routines then decide
   which part of the information they are able to handle and will write.

 * Alternatively, we could split up the read/write routines to really
   just write one molecule at a time, while the header (and possible glue
   stuff and footer) will be written by a helper routine.

Of course, this is quite different from the way it works right now, but
perhaps basis for discussion. Also, as I said above, this is not thought
out very well, just random thinking.


Michael