Re: [Open Babel] cansmi of molfile vs cansmi of cansmi

[Geoffrey H. <ge...@ge...>]

> I've attached an example sdf file with 20 structures, all of which show this behaviour.
> I hope someone can take a look and see if I'm doing something odd or how I
> might correct this.

As Craig said, you're not doing anything odd. Breaking ties is tough, and this type of testing really taxes the aromaticity detection as well.

I just tried your file (as well as several others) with the 2.3 trunk, and found that 44/48 cases passed. That still leaves 4 failures:
* One funny canonicalization bug with Cu phthalocyanine.
* Two aromaticity failures with fused rings containing multiple nitrogens
* One silly case with un-needed '[n]' sequences.

The good news is this -- it's been added as a unit test, so if you find more failures, just send me more files.

I can definitely tackle case #2 and case #3 in the near future. I'd need Craig to look at #1, but if we can reduce his test cases, we all profit. :-)
  [Cu].c1ccc2C3=NC(=NC4=NC(=NC5=NC(=NC6=NC(=N3)c3ccccc63)c3ccccc53)c3ccccc43)c2c1
  [Cu].c1ccc2C3=NC4=NC(=NC5=NC(=NC6=NC(=NC(=N3)c2c1)c1ccccc61)c1ccccc51)c1ccccc41

Thanks for the bug report TJ -- please send us any other failures you encounter.
-Geoff