We should find a more concrete way to test registry behavior in general, since the inconsistent stereochemistry error currently used to see whether a call hits OE vs. RDK relies on a lot of subtlety and could change in a subsequent OE or RDKit release. My rough idea would be some new tests that try to request an OE-only method (eg. to_openeye) from a registry that shouldn't have the NAGLOpenEyeToolkitWrapper, but I'd need to look more into the structure to figure out whether this is feasible.
Originally posted by @j-wags in #242 (comment)
We should find a more concrete way to test registry behavior in general, since the inconsistent stereochemistry error currently used to see whether a call hits OE vs. RDK relies on a lot of subtlety and could change in a subsequent OE or RDKit release. My rough idea would be some new tests that try to request an OE-only method (eg.
to_openeye) from a registry that shouldn't have the NAGLOpenEyeToolkitWrapper, but I'd need to look more into the structure to figure out whether this is feasible.Originally posted by @j-wags in #242 (comment)