Case Class References Missing Instantiations

[wchau]

lsif-java version: 0.7.6

Type of code: case classes in Scala

Expected behavior: Case class "Find Reference" should contain instantiations.

Actual behavior: Case class "Find References" does not contain instantiations. You can find other instantiations if you "Find Reference" any one instantiation.

If the companion object is explicitly defined, you can find instantiations there by "Find Reference".

Note that "Go to Definition" from instantiations does go to the case class if the companion object does not exist (otherwise, it would go to the companion object), but the hover references the companion object(???). Ideally, all the synthetic functions should reference the case class (and vice versa) similar to IntelliJ's code reference.

case class TestA(a: String) {}  // "Find References" missing the below instantation

...

TestA("blah")  // Hover references object TestA but "Go to Definition" jumps to case class definition

[wchau]

It is already a lot better than before, though, thanks! :)

[olafurpg]

Thank you for reporting! I am able to reproduce and I opened a PR fixing this issue in #404

Please keep the bug reports coming, I appreciate your patience with all of the different releases. It's tricky to get this right for case classes since they generate so many different kinds of symbols (some which we want to include with "find ref" results while not others)

[olafurpg]

Reopening this issue since "Find references" on the case class does not show usages of the synthetic companion object. Here's a minimized test case that reproduces the problem

scip-java/tests/snapshots/src/main/generated/tests/minimized-scala/src/main/scala/minimized/Issue396.scala

Line 16 in bf4ca0c

// relationship is_definition semanticdb maven . . minimized/Issue396#

To fix this problem, we should add an is_reference relationship alongside the is_definition relationship. I estimate this is an easy issue to fix and it does not require changes in the Sourcegraph backend.

[olafurpg]

The root cause of this issue is here where we intentionally return false for objects

scip-java/scip-semanticdb/src/main/java/com/sourcegraph/scip_semanticdb/ScipSemanticdb.java

Line 350 in bf4ca0c

return false;

I'm wondering if the fix is to remove the supportsReferenceRelationship condition where we set is_definition. As long as the names of the symbols are the same then it's OK for them to share a reference relationship.

scip-java/scip-semanticdb/src/main/java/com/sourcegraph/scip_semanticdb/ScipSemanticdb.java

Line 147 in bf4ca0c

&& supportsReferenceRelationship(info)));

[olafurpg]

Another possible fix, which I think might be a better solution, is to add an is_reference relationship on the synthetic apply(). on the companion object. This solution is better because it works the same regardless if there is an explicit companion or not.

[olafurpg]

Concretely, the proposal is to update this line here in the snapshots to include is_reference

scip-java/tests/snapshots/src/main/generated/tests/minimized-scala/src/main/scala/minimized/Issue403.scala

Line 13 in 8964623

// relationship is_definition semanticdb maven . . minimized/Issue403#

[olafurpg]

Fixed in #561

Link to reproduce https://sourcegraph.com/github.com/sourcegraph/scip-java/-/blob/tests/minimized-scala/src/main/scala/minimized/Issue403.scala?L3:12-3:20#tab=references

Note that the empty line is refers to a zero-length range for a usage of the apply() symbol. I have opened a separate issue to track that #564