[chemfp] chemfp 3.0 and 1.3

[Andrew Dalke]

Hi all,

  Greetings from a very inactive list. I thought that people here might be interested in knowing what's going on with chemfp.

The main development branch is now at chemfp 3.0. This is the version for people who have purchased a copy of chemfp. The 2.x series added support for web service development, including the FPB binary fingerprint format for fast loading, and an API for working with structures and fingerprints through strings. It also added Tversky search.

The big change in version 3.0 is Python 3 support. Chemfp 3.0 runs under both Python 2.7 and 3.5+. The API changed so that identifiers are now Unicode strings. I've also improved the similarity performance by 5-20%, depending on the fingerprint size, and (for 2048 bit fingerprints) if you've compiled for AVX2 support.

Now I'm going to update chemfp 1.1. The new version will be called chemfp 1.3. (The code which would have been chemfp 1.2 became chemfp 2.0.) It will still be released at no cost under the MIT license.

The major change will be to update the fingerprint types to recognize the newest versions of the respective toolkit.

Also, as part of a discussion thread on the RDKit mailing list, I extracted the Taylor-Butina algorithm from the documentation and made it a stand-alone program, at:

  http://dalkescientific.com/writings/taylor_butina.py

I'm thinking to include a set of chemfp-based tools in the release. I would like it people could try out the program and give feedback, especially in what would be a better output format.

I would also like to hear any other suggestions for light-weight programs that I might include.

Finally, I've started writing up a chemfp paper, for J. Chem. Inf. It'll be a while though - I would rather write code than papers!

Cheers,

				Andrew
				dalke at dalkescientific.com