core: reduce amount of scalac dependencies #2048
Conversation
| @@ -0,0 +1,10 @@ | |||
| package scala.meta.internal.symtab | |||
There was a problem hiding this comment.
I am not super happy about this hack, since eventual drift between this version and the upstream one will cause runtime errors in reflect and cli, but for the foreseeable future tests should pick that up and I think it's worth it to keep core clean until we contractualize with scalameta what could be exposed as a public API.
| @@ -41,6 +41,7 @@ object Dependencies { | |||
| val scalatest = "org.scalatest" %% "scalatest" % scalatestV | |||
| val munit = "org.scalameta" %% "munit" % munitV | |||
| val semanticdbScalacCore = "org.scalameta" % "semanticdb-scalac-core" % scalametaV cross CrossVersion.full | |||
| val semanticdbSharedFor3Use2_13 = "org.scalameta" % "semanticdb-shared" % scalametaV cross CrossVersion.for3Use2_13 | |||
There was a problem hiding this comment.
leverage scalameta/scalameta#3887
| @@ -149,6 +148,7 @@ lazy val reflect = projectMatrix | |||
| moduleName := "scalafix-reflect", | |||
| isFullCrossVersion, | |||
| libraryDependencies ++= Seq( | |||
| semanticdbScalacCore, | |||
There was a problem hiding this comment.
Unrelated to this PR, but this change highlights a limitation: I wonder how much of the current implementation of GlobalSymbolTable works against class files produced by dotty.... We haven't got any issue but I guess it's just because the Symbol lookup API is not used much on symbols that are not in the document, or maybe it's mostly on Java and Scala standard library classes.
bjaglin
changed the title
Follows #2047 / scalameta/scalameta#3747
See initial discussion in #2024 (comment)
I initially tried to inline everything that was needed from semanticdb-scalac-core, but I ended up pulling pretty much everything.... It's not a big difference compared to before anyway, as the Scala 2 compiler is brought by scalameta (the main artifact). As discussed in one of the comment, that logic will have to be revisited anyway to better support Scala 3 (#2049), so I am not sure it's worth investing time on that now.