Closed Aariq closed 4 months ago
It seems like the built-in version of OpenBabel included in the windows version of ChemmineOB might behave differently when it comes to creating the formula and InChI that we are using to diagnose parsing issues. Need to add some tests and investigate.
It looks like this strategy doesn't currently work on Windows because the windows version of ChemmineR::propOB()
doesn't generate InChI. I've opened an issue: https://github.com/girke-lab/ChemmineOB/issues/37.
ChemmineOB
can't generate InChI on Windows https://github.com/girke-lab/ChemmineOB/issues/37#issuecomment-2008364855
For now I may just only allow the validate
argument to work on non-windows operating systems.
ChemmineOB currently is encountering errors when being built from source on macOS: https://github.com/girke-lab/ChemmineOB/issues/35.
Ok, going to go ahead and merge despite checks on macOS failing since I've confirmed checks work locally on macOS with binary version of ChemmineOB
. Might do a separate PR to force GH actions to install ChemmineOB
from binary rather than source if I can figure out how to do that easily.
Addresses #85 (and also closes #56) by checking the output of
ChemmineR::propOB()
for the symptoms of OpenBabel issues parsing SDFs. WhypropOB()
? Simply because that is the onlyChemmineR
function we have found that actually complains about problems in .mol files.Currently this is tested with two cases:
I'm happy to have this PR merged as-is or add additional tests—I'm sure there are other ways that inputs (.mol files or SMILES) can be just broken enough to trigger OpenBabel errors, but not R errors, and return incorrect results.
Unfortunately, since InChI can't be generated with the windows version of
ChemmineOB
, this validation option can't be used on Windows. This is documented in the help file forget_fx_groups()
and a warning is printed whenvalidate = TRUE
is used on Windows.Still to do?:
validate
to Getting Started vignette