(OK, KMail + Gmail is not a friendly combination... the slighest mistype results in an undesired submit :( )
On 9/25/07, Egon Willighagen <[EMAIL PROTECTED]> wrote: > in this email I would like to bring forward a request for comments > (RFC) about the SMILES parser. > > One thing that CDK 1.0 has clarified is the difficulties it has with > deducing bond orders, for which it was tried to write a > one-size-fits-all algorithm using available atom type lists. This > failed. One workarou ... One workaround is the DeduceBondOrderTool which takes a rule based approach, which will never provide a full coverage, but at least succeeds to do the most common things right. Getting to the SMILES parser, I would like to request comments on what functionality is to be expected from a SMILES parser. To make the SMILES parser in the CDK back into a consistent working state again, I would like to suggest it will no longer try to resolve bond orders and leave that to the user as post-processing. This will fix all standing open bugs in the CDK SmilesParser, which are all related to failing bond order resolution. The SMILES parser should: 1. parse the SMILES into a connectivity table 2. mark atoms which lower case symbols for steps 2 and 3 3. try to resolve bond orders for atom groups which do not include ring atoms 4. try to deduce bond orders for atom groups which include ring atoms only for very well defined groups using a rule-based approach 5. detect aromaticity for single, isolated ring systems only Not really too different from what is happening right now, but cleaner and better defined. Is the above all we should expect the SMILES parser to do? Egon -- ---- http://chem-bla-ics.blogspot.com/ ------------------------------------------------------------------------- This SF.net email is sponsored by: Microsoft Defy all challenges. Microsoft(R) Visual Studio 2005. http://clk.atdmt.com/MRT/go/vse0120000070mrt/direct/01/ _______________________________________________ Cdk-user mailing list [email protected] https://lists.sourceforge.net/lists/listinfo/cdk-user
