[colbymchenry]

Summary

A Scala call like RemoteNode.Obfuscated("x"), Uci.Move(from, to) or Header.Raw(name, value), a case-class or companion apply on a nested type, now links to the Scala case class or object it builds, and a Scala call no longer takes an unrelated Java constructor through a return-type annotation.

Root cause

Scala qualified names carry no package (Outer::Inner), so a call written Outer.Inner(...) found no type by its written name and fell through to the name strategies. Under def f(): WebSocket = WebSocket.Accepted(flow), the local-declaration reader took WebSocket for the receiver's type and handed over an unrelated Java constructor (Other::Accepted::Accepted).

From the contributor, and what we changed

From @danusha2345 (#1951), unchanged: the Scala Outer.Inner type lookup before the receiver-word strategy (one file, or the call's own file; a class or trait before an object), the guard that keeps the typed-receiver path from handing a Scala call another language's constructor, and the five resolution.test.ts cases.

Our changes: the CHANGELOG bullet rewritten in the house format (without the "never guessed" claim: only the typed path is guarded, though the receiver-word strategy never produced such an edge in our fixtures or repos) and moved next to the JVM overload bullet. Rebased onto main 81559dcc, which has #2875 (#1950), #2874 and #2923 (#1936).

Validation

Tests. 5 cases in resolution.test.ts: 4 red on main, all green here; the fifth, a Scala call to a real Java static method, passes on both and guards that link.

A/B against main 81559dcc (with #1950 and #1936), fresh indexes, edges compared by site (file, kind, position, ref name) and target identity (name, file, line):

repo	files	edge sites added	removed	retargeted
lichess-org/scalachess a498819	245	77	0	0
http4s/http4s 53b931f	765	371	1	0
playframework f3eb322 (Java + Scala)	2,727	291	1	0
playframework/play-samples 19eb65b	867	2	0	0
yahoo/CMAK 30abfde	190	identical
pallets/click, gin-gonic/gin (controls)		identical

Triage:

• Added: case-class and object applies, all sampled right: scalachess Uci.Move (17), Uci.Drop (13), Position.AndFullMoveNumber (11), Clock.Config (7), Eval.Mate (an object); http4s Header.Raw (66), H2Frame.Data (36), Uri.Ipv6Address (20), WebSocketFrame.Text (16), Uri.RegName (16), Request.Connection over the same-named headers.Connection. playframework: all 291 are Scala applies of nested case classes, recorded as instantiations at 0.85: RemoteNode.Ip (54), RemoteNode.Obfuscated (53), RemoteNode.Unknown (33), NodePort.Numeric (31), HttpEntity.Strict (20), HttpEntity.Streamed (19), WebSocket.Accepted (15), ApplicationLoader.Context under import play.api._ beside the renamed Java JApplicationLoader; these include the HttpEntity and WebSocket applies that fix(play): look up Class.method among the class's own members, not the whole file #1950 unlinked from Java constructors.
• Removed: both are retargets (a call becomes an instantiation, so the site reads as removed and added). http4s: Origin.Host(...) moves from a 0.7 guess on defaults::Host to the Origin.Host case class. playframework: Writeable.scala's HttpEntity.Strict(...), in package play.api.http, moves from the Java play.http.HttpEntity.Strict constructor to the Scala case class.
• With fix(play): look up Class.method among the class's own members, not the whole file #1950: on play-samples, fix(play): look up Class.method among the class's own members, not the whole file #1950 removed two wrong edges from UserActor.WatchStocks(...) and UserParentActor.Create(...) to Java constructors of another sample; this PR links both to their Scala case classes. On scalachess and http4s the two are independent: fix(play): look up Class.method among the class's own members, not the whole file #1950 gave +6 on scalachess and nothing on http4s, and this PR gives the same +77 and +371/-1 with or without it. The two merged cleanly in either order.

Cost (http4s, median of 3 interleaved fresh indexes, measured on main 69af19ac, where the A/B gave the same edges as on 81559dcc): wall 8.6 s on main vs 7.8 s with this PR, process CPU 27.5 s vs 26.2 s: within noise.

Suite: CODEGRAPH_KERNEL_EXPECT=1 npx vitest run --maxWorkers=4 --minWorkers=1 on c3937888 (Windows 11, kernel built): 760 files passed, 6 skipped; 9,037 tests passed, 83 skipped, 0 failed.

Lands #1951 by @danusha2345.

🤖 Generated with Claude Code