Modifying a molecule should not automatically clear SubstanceGroups

[greglandrum]

The current behavior - clearing all SubstanceGroups whenever a molecule is modified - was a sensible initial implementation strategy given that we were still gaining experience with the use cases for SubstanceGroups. Taking a "better safe than sorry" approach made sense.

Unfortunately, this decision has some fairly unfortunate side effects: if a molecule includes any Hs, all SubstanceGroups will be removed when the Hs are removed, which is the default behavior. This happens regardless of whether or not the Hs are involved in the SubstanceGroups.

I think this has a large enough impact on the usefulness of SubstanceGroups that it should be fixed.

We can argue aboutdiscuss whether or not this is actually a bug since the current behavior is intentional. I think that from the user perspective it feels like a bug when all SubstanceGroups vanish just because a molecule contained one or more H atoms that weren't involved in the SubstanceGroups.

The next argumentdiscussion is whether or not this should be fixed in a patch release. I would like to do so because I need to be able to use this fix in a project I'm working on. That's not a good argument if it introduces potential problems for the broader community. However, I would guess that the SubstanceGroup functionality is currently not heavily used by the community. It's something of an advanced user thing and is hardly documented. Still, if this makes the other maintainers uneasy, we can shift this out to the 2020.09 release and I will set up a fork for my other project.

[greglandrum]

@ricrogz, @d-b-w : you guys may have some thoughts/opinions on this one.

[greglandrum]

This is related to #2716

[ricrogz]

I think it makes sense to do this sooner than later.

As you say, I don't think stripping the Substance Groups on modification of the molecule is a bug, since we decided it to be that way, although them getting stripped from molecules with H atoms is, since that is an unplanned side effect of that decision.

I think the best way to fix the H atoms problem is to implement the code that would allow modification of the molecule without losing the Substance Groups, so I think it makes sense to include it in a patch release.

I don't think that would impact the user community in a bad way, because, as you said, Substance Groups are mostly undocumented, and not widely used. And, even for the few people that use them, this change would be beneficial, since I guess the bug with the H atoms is annoying for them.