Implements SideEffectsOnly checking.

[t-rasmud]

This PR adds an annotation @SideEffectsOnly to the set of dataflow qualifiers. For a method with this annotation, dataflow analysis unrefines just those expressions that are specified as annotation values of @SideEffectsOnly instead of the more conservative alternative of unrefining all expressions.

[mernst]

I'm not sure what this pull request is doing. Could you please clarify?

[t-rasmud]

The @SideEffectsOnly makes a claim that the body of the annotated method only mutates certain values. The checker ought to issue a warning if @SideEffectsOnly is written on a method that makes more changes than the annotation permits.

Added this check in the latest commit.

[mernst]

Thanks for this code. I have a few questions and suggestions.

[mernst]

I'm curious why this is a brand-new class rather than a modification to the existing mutation checker.

[t-rasmud]

I am not aware of a mutation checker and couldn’t find it in the manual either. Could you please point me to that class?
Although I think the PurityChecker can be modified to add all the functionality in the SideEffectsOnlyAnnoChecker. Do you think that's more appropriate?

[mernst]

By "mutation checker" I meant PurityChecker. I'm sorry for being sloppy in my terminology. Thank you for setting me straight. Yes, I was thinking that merging the functionality in one place (maybe controlled by a switch) would be shorter, easier to understand, and less prone to code drift, compared to the two separate implementations. At least, I think that is worth trying to see whether it works out.

Thank you!

[t-rasmud]

Thank you for the suggestions. I tried merging the two functionalities into PurityChecker but wasn’t successful because of the following reason:
SideEffectsOnlyAnnoChecker (a part of the Checker Framework repository) needs to import org.checkerframework.common.basetype.BaseTypeChecker, which is not visible in the dataflow project (dataflow project contains the PurityChecker). Is there any other way around this?

[t-rasmud]

Thanks for the feedback Mike. I just have one question regarding your comment about the mutation checker.