Interested in resolving issues from the SMILES benchmark?

I've put together a benchmark to compare the ability of different tools to read SMILES. The goal is to identify corner cases and improve SMILES reading across the board. Initial results were reported at the recent ICCS:
https://nextmovesoftware.com/blog/2018/06/06/can-we-agree-on-the-structure-represented-by-a-smiles-string/
...and the benchmark itself is available at:
https://github.com/nextmovesoftware/smilesreading

In the benchmark, each toolkit reports the number of hydrogens on each atom in the molecule represented by an aromatic SMILES string, or whether there was an error reading it. If you are interested in working through some of the issues raised by the benchmark, let me know and I am happy to describe them and how they might be solved. I have already done this with a number of toolkits.

Review the SMILES benchmark at github.com/nextmovesoftware/smilesreading and the linked ICCS report to understand the reported corner cases. Ask the maintainer which benchmark discrepancies apply to Indigo and how they should be addressed; done means improving SMILES reading so the relevant benchmark cases report the expected hydrogens or errors.

Written by the indexing model from the issue text.