Added SDF Export to PandasTools and adjusted SDF Import

[nhfechner]

The pull request contains the commit from @samoturk in the pull request #401. In addition to these changes, I made the generation of a smiles column optional and defaulting to not generate it. Adding a smiles column later on is simple, but automatically doing it silently changes the matching of SD tags to dataframe columns.

The export is fairly self contained and is nearly identically reproducing the SDF input. The deviations are mainly occuring when some SD tags are not present in all molecules in the input SDF. Such cells are filled in the dataframe with "nans" which consequently end up in the written SDF.
Another deviation is that the generated SD file contains the valences for some atoms (e.g. Cu,Si,Se) in the atoms block even if they were not present in the input.

Are these deviations acceptable?

@samoturk : Do you think the method signatures and the changes to the input are reasonable or would you suggest doing things differently?

[samoturk]

I agree with the changes. I guess round trip works in most cases which is good enough. We could remove SD tags with nan from mols in cases when some SD tags are not in all molecules but it would complicate the code and probably still not be 100% fool proof.

Nice effort! I think it can be merged.

[greglandrum]

Great. I will merge next time I am in front of my laptop

On Tuesday, December 23, 2014, Samo Turk notifications@github.com wrote:

I agree with the changes. I guess round trip works in most cases which is
good enough. We could remove SD tags with nan from mols in cases when some
SD tags are not in all molecules but it would complicate the code and
probably still not be 100% fool proof.

Nice effort! I think it can be merged.

—
Reply to this email directly or view it on GitHub
#404 (comment).

[greglandrum]

@nhfechner and @samoturk : thanks for this!