Hi Rich, On 9/20/07, richard apodaca <rich_apod...@yahoo.com> wrote: > How feasible would it be to create a lightweight, > Python-free, GUI-less, barebones RDKit distribution?
what? you want to cut pieces out of my baby? no way! ;-) Just kidding. Pretty easy. You can ignore the Python and GUI stuff (which is also Python) now without difficulties. In theory (minor modifications might be necessary): -- running "make libs" instead of "make all" in the $RDBASE/Code directory will build only the C++ libraries. -- "make regrs" in $RDBASE/Code should also build only tests for c++ stuff. The scripts that run the tests would have to be changed to ignore anything in Wrap directories, but that's also not a big problem. You'd still need a Python distribution and a teeny bit of the stuff in $RDBASE/Python to run the tests themselves though. Alternatively, someone could do shell scripts to run the tests. > > - something that can be compiled (with minimal > dependencies) using the standard unix scripts: > configure, make, make install I would like, very much, to have a configure script for the RDKit. I haven't done it because I don't know how. It's not something I've ever done and anytime I've looked at the documentation it's made my head hurt. > - something that could be used to begin to craft a > SWIG interface > > > My own interest is in creating a Ruby interface to > RDKit. In fact, if recent activity on the Open Babel > project is any indication, there may be a few other > folks interested in a _fill-in-the-blank_ interface to > RDKit. I would very very much like to have a SWIG interface. I will happily provide whatever support I can to anyone who wants to do this. > To the extent that all of these interfaces and the > underlying C++ library were distributed under > MIT/BSD-style licenses, we would truly have something > new under the sun. As I said, I'm happy to help however I can. Particularly if whoever is doing this is willing to contribute the stuff back to the project (under the BSD license, of course). > > And if it were feasible to create this bare-bones > RDKit distribution, what features would be lacking? A lot, but you would still be left with something useful. You'd lose most of the descriptors, the machine-learning code (though some of the utility function for it would still be there), the virtual library functionality, the pharmacophore stuff, and others. You'd still have everything listed on the "general molecular functionality" section (http://rdkit.sourceforge.net/wiki/index.php/Feature_Overview) but: MACCS keys, some of the similarity/diversity picking functionality, and subshape alignment. I might have missed one or two small pieces, but I think that's more or less it. -greg
