[java] Make LawOfDemeter results deterministic

[adangel]

Describe the PR

I believe, this should fix the spurious differing results in the regression tester. At least, I could reproduce on first try the problem when reversing the order of the reaching assignments.

That should then make the test-case-3 of pmd/pmd-regression-tester#119 happy.

Ready?

• Added unit tests for fixed bug/feature
• Passing all unit tests
• Complete build ./mvnw clean verify passes (checked automatically by github actions)
• Added (in-code) documentation (if needed)

[pmd-test]

	1 Message
📖	Compared to master: This changeset changes 11 violations, introduces 2 new violations, 0 new errors and 0 new configuration errors, removes 2 violations, 0 errors and 0 configuration errors. Full report
✅	Compared to master: This changeset changes 11 violations, introduces 2 new violations, 0 new errors and 0 new configuration errors, removes 2 violations, 0 errors and 0 configuration errors. Full report
✅	Compared to master: This changeset changes 4 violations, introduces 0 new violations, 0 new errors and 0 new configuration errors, removes 0 violations, 12 errors and 0 configuration errors. Full report
✅	Compared to master: This changeset changes 4 violations, introduces 0 new violations, 0 new errors and 0 new configuration errors, removes 0 violations, 12 errors and 0 configuration errors. Full report
✅	Compared to master: This changeset changes 13 violations, introduces 2 new violations, 0 new errors and 0 new configuration errors, removes 2 violations, 12 errors and 0 configuration errors. Full report

Generated by 🚫 Danger

[adangel]

Ok, the baseline seems to have been created with a different order of the reaching assignments. I'll give it a try and remove the reversed() to see, whether we match the baseline then. Ideally we wouldn't see any changed violations.

[oowekyala]

Maybe a better fix would be to replace all new HashSet and new HashMap in DataflowPass.java with their LinkedHashX version... Then we wouldn't need the sorting at all, and we ensure other rules won't have this problem.

[adangel]

Maybe a better fix would be to replace all new HashSet and new HashMap in DataflowPass.java with their LinkedHashX version... Then we wouldn't need the sorting at all, and we ensure other rules won't have this problem.

I wanted to make it minimally invasive. I can give it a shot, though...

[adangel]

Maybe a better fix would be to replace all new HashSet and new HashMap in DataflowPass.java with their LinkedHashX version... Then we wouldn't need the sorting at all, and we ensure other rules won't have this problem.

I wanted to make it minimally invasive. I can give it a shot, though...

I've updated the Dataflow code now. I needed to swap the test case again - that means, if this was really the reason for the random results, then we should get the numbers from the first run again:

This changeset changes 13 violations,
introduces 2 new violations, 0 new errors and 0 new configuration errors,
removes 2 violations, 12 errors and 0 configuration errors.

But it should be stable now...