Re: [Cdk-devel] Pubchem Fingerprinter Test Failures

[Egon W. <ego...@gm...>]

Hi Nina,

On Fri, Dec 14, 2012 at 10:01 AM, Nina Jeliazkova
<jel...@gm...> wrote:
> If this is the expected behaviour, it at least should be documented
> why it is so (may be some preprocessing)?  Or the code above is wrong?

Two initial thoughts:

1. algorithms should not change the data model, so that rerunning it
results in the same result.
2. the JavaDoc should list that requirement of atom typing, but the
algorithm is fine to give different results if not done

Ad 1. some QSAR descriptors cache input fields, in order to undo
changes it made during the calculate() and there is a unit test for
that, IIRC
Ad 2. the immutable and caching after CDK 1.6.0 should help here

Egon

--
Dr E.L. Willighagen
Postdoctoral Researcher
Department of Bioinformatics - BiGCaT
Maastricht University (http://www.bigcat.unimaas.nl/)
Homepage: http://egonw.github.com/
LinkedIn: http://se.linkedin.com/in/egonw
Blog: http://chem-bla-ics.blogspot.com/
PubList: http://www.citeulike.org/user/egonw/tag/papers