On Wed, Sep 22, 2010 at 10:00 AM, Noel O'Boyle <baoille...@gmail.com> wrote: > Looks very good now. Nice work.
Yes, I enabled the python bindings again here. I still have swig 1.3 though... I reimplemented rule 1 for the rule based stereogenic units and the results are now the same as for the automorphisms based version. The automorphism based version was not usable for general use since there were memory issues. The automorphism search is actually memory efficient but simply storing all automorphisms for many cases can require GBs of memory. As a result, the automorphisms are not used for anything right now. The canonical coding still needs to be optimized but since we only need to keep track of the best labeling/canonical code, the same problem is not seen here. Tim > On 21 September 2010 11:38, Noel O'Boyle <baoille...@gmail.com> wrote: >> I'm afraid I'm seeing problems with roundtripping from SMILES through >> 0D, 2D or 3D MDL. >> >> If you run pytest_sym (e.g. "ctest -R pytest_sym -VV"), you should see >> several errors >> (http://my.cdash.org/testDetails.php?test=3025177&build=99128). >> >> As an example, one of the errors can be shown as follows: >> >> $ ./obabel -:"Cl[C@@](Br)(F)I" -osdf -Otmp.sdf >> ============================== >> *** Open Babel Warning in WriteMolecule >> No 2D or 3D coordinates exist. Any stereochemical information will >> be lost. To generate 2D or 3D coordinates use --gen2D or --gen3d. >> 1 molecule converted >> >> (tmp.sdf is correct) >> >> $ ./obabel tmp.sdf -osmi >> ClC(Br)(F)I # should be the starting SMILES with stereo >> >> Whatever you think of the merits of preserving chiral stereochemistry >> in a 0D mol file, this was working until the 31 file update on Sunday >> (http://my.cdash.org/index.php?project=Open+Babel&date=2010-09-19). >> >> - Noel >> >> On 20 September 2010 21:23, Tim Vandermeersch >> <tim.vandermeer...@gmail.com> wrote: >>> Hi, >>> >>> Here are some comments regarding my commits from yesterday. >>> >>> The type used to represent automorphisms has been changed from >>> std::vector< std::map<unsigned int, unsigned int> > to std::vector< >>> std::pair<unsigned int, unsigned int> >. This makes a huge difference >>> in performance but might require small changes in code using it (code >>> in svn is updated of course). When doing an automorphisms search for a >>> multiple fragment molecule, each fragment is treated separately and >>> the resulting set of automorphisms is simply the union of the >>> individual results. In other words, not all automorphisms will have >>> the same size and cover the whole molecule. Substructure search should >>> now also be much faster using OBIsomorphismMapper compared to smarts >>> although I haven't tested this yet. >>> >>> The canonical labeling code now uses memory more efficiently and also >>> treats connected fragments individually. Performance analysis shows >>> that precomputing and caching stereo descriptors would be the next >>> part to optimize and I'll do this tonight together with some other >>> smaller improvements. If anyone still encounters problematic cases >>> (now or later), they can always send me the cases. >>> >>> The canonical order is now "stable" and I'll add a unit test to check >>> this tonight. >>> >>> Still todo: Add square-planar stereochemistry + isotopes >>> >>> I've added a DataCache class to help with storing symmetry classes, >>> automorphisms and stereogenic units as generic data. The class uses >>> graph invariants and bitvec masks to check if the data needs to be >>> recalculated. I still need to document this although we might want to >>> keep this as private API? >>> >>> That's it for now, feel free to ask more questions... >>> >>> Tim >>> >>> ------------------------------------------------------------------------------ >>> Start uncovering the many advantages of virtual appliances >>> and start using them to simplify application deployment and >>> accelerate your shift to cloud computing. >>> http://p.sf.net/sfu/novell-sfdev2dev >>> _______________________________________________ >>> OpenBabel-Devel mailing list >>> OpenBabel-Devel@lists.sourceforge.net >>> https://lists.sourceforge.net/lists/listinfo/openbabel-devel >>> >> > ------------------------------------------------------------------------------ Start uncovering the many advantages of virtual appliances and start using them to simplify application deployment and accelerate your shift to cloud computing. http://p.sf.net/sfu/novell-sfdev2dev _______________________________________________ OpenBabel-Devel mailing list OpenBabel-Devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/openbabel-devel
