Emit SCIP shards directly from plugins

build.sbt

@@ -71,18 +71,16 @@ commands +=
       "scalafixAll --check" :: "publishLocal" :: s
   }
 
-// Shared module that owns the canonical SemanticDB proto schema and the
-// associated symbol/builder utilities. Both the Java compiler plugin
-// (semanticdb-javac) and the Kotlin compiler plugin (semanticdb-kotlinc)
-// depend on it instead of carrying their own divergent copies of the proto.
+// Shared module with the SCIP shard utilities (symbol encoder, document
+// builder, on-disk writer) consumed by both the Java compiler plugin
+// (semanticdb-javac) and the Kotlin compiler plugin (semanticdb-kotlinc).
 lazy val semanticdbShared = project
   .in(file("semanticdb-shared"))
   .settings(
     moduleName := "semanticdb-shared",
     javaOnlySettings,
-    (Compile / PB.targets) :=
-      Seq(PB.gens.java(V.protobuf) -> (Compile / sourceManaged).value),
-    libraryDependencies += "com.google.protobuf" % "protobuf-java" % V.protobuf
+    libraryDependencies +=
+      "org.scip-code" % "scip-java-bindings" % V.scipBindings
   )
 
 lazy val gradlePlugin = project
@@ -329,8 +327,10 @@ lazy val semanticdbKotlinc = project
     // classpath via Provided so the assembled fat-jar does not bundle it.
     libraryDependencies +=
       "org.jetbrains.kotlin" % "kotlin-stdlib" % V.kotlinVersion % Provided,
-    // The SemanticDB proto schema and the generated Java classes live in
-    // semanticdbShared; we get them transitively via .dependsOn below.
+    // SCIP message classes come from semanticdbShared (which depends on
+    // scip-java-bindings); this adds the Kotlin DSL extensions on top.
+    libraryDependencies +=
+      "org.scip-code" % "scip-kotlin-bindings" % V.scipBindings,
     // kotlin-compiler-embeddable is supplied by kotlinc at runtime
     libraryDependencies +=
       "org.jetbrains.kotlin" % "kotlin-compiler-embeddable" % V.kotlinVersion %

scip-java/src/main/scala/com/sourcegraph/scip_java/buildtools/GradleBuildTool.scala

@@ -35,13 +35,13 @@ class GradleBuildTool(index: IndexCommand) extends BuildTool("Gradle", index) {
   }
 
   /**
-   * Diagnose the case where Gradle finished successfully but our SemanticDB
-   * compiler plugin never produced any `.semanticdb` files. This used to be
-   * silently rescued by a `-javaagent` fallback; now it surfaces as a clear
-   * error pointing at the two known causes.
+   * Diagnose the case where Gradle finished successfully but our SCIP compiler
+   * plugin never produced any `.scip` shards. This used to be silently rescued
+   * by a `-javaagent` fallback; now it surfaces as a clear error pointing at
+   * the two known causes.
    */
   private def reportMissingSemanticdbOutput(): Unit = {
-    if (containsFileWithSuffix(targetroot, ".semanticdb"))
+    if (containsFileWithSuffix(targetroot, ".scip"))
       return
     if (!containsFileWithSuffix(index.workingDirectory, ".class"))
       // Project produced no compiled JVM output — nothing to index, stay quiet.
@@ -50,9 +50,9 @@ class GradleBuildTool(index: IndexCommand) extends BuildTool("Gradle", index) {
       .app
       .reporter
       .error(
-        s"""scip-java: Gradle finished successfully but produced no SemanticDB output in $targetroot.
+        s"""scip-java: Gradle finished successfully but produced no SCIP shards in $targetroot.
            |
-           |This means our SemanticDB compiler plugin was not attached to one or more JavaCompile tasks. Two known causes:
+           |This means our SCIP compiler plugin was not attached to one or more JavaCompile tasks. Two known causes:
            |
            |  1. The 'compileOnly' configuration was already resolved before our init script ran.
            |     Check the Gradle output above for warnings of the form:

scip-java/src/main/scala/com/sourcegraph/scip_java/commands/IndexSemanticdbCommand.scala

@@ -17,7 +17,7 @@ import moped.cli.Command
 import moped.cli.CommandParser
 import org.scip_code.scip.ToolInfo
 
-@Description("Converts SemanticDB files into a single SCIP index file.")
+@Description("Aggregates per-source SCIP shards into a single SCIP index file.")
 @Usage("scip-java index-semanticdb [OPTIONS ...] [POSITIONAL ARGUMENTS ...]")
 @ExampleUsage(
   "scip-java index-semanticdb --out=myindex.scip my/targetroot1 my/targetroot2"
@@ -26,14 +26,14 @@ import org.scip_code.scip.ToolInfo
 final case class IndexSemanticdbCommand(
     @Description("The name of the output file.")
     output: Path = Paths.get("index.scip"),
-    @Description("Whether to process the SemanticDB files in parallel")
+    @Description("Whether to process the SCIP shards in parallel")
     parallel: Boolean = true,
     @Description(
       "Whether to emit parent->child relationships for 'Find references' and 'Find implementations'. " +
         "This flag exists as a workaround for the issue https://github.com/sourcegraph/sourcegraph/issues/50927"
     )
     emitInverseRelationships: Boolean = true,
-    @Description("Directories that contain SemanticDB files.")
+    @Description("Directories that contain SCIP shards.")
     @PositionalArguments()
     targetroot: List[Path] = Nil,
     @Description(

scip-java/src/main/scala/com/sourcegraph/scip_java/commands/SnapshotCommand.scala

@@ -51,27 +51,35 @@ case class SnapshotCommand(
               attrs: BasicFileAttributes
           ): FileVisitResult = {
             if (scipPattern.matches(file)) {
-              foundScipFile = true
               val index = Index.parseFrom(Files.readAllBytes(file))
-              val root = URI.create(index.getMetadata.getProjectRoot)
-              index
-                .getDocumentsList
-                .asScala
-                .foreach { doc =>
-                  val sourcepath = Paths.get(root.resolve(doc.getRelativePath))
-                  val source =
-                    new String(
-                      Files.readAllBytes(sourcepath),
-                      StandardCharsets.UTF_8
+              // Per-source SCIP shards under META-INF/scip/ carry no Metadata;
+              // only the aggregated index does. Skip shards so `scip-java
+              // snapshot <targetroot>` doesn't trip over them.
+              val projectRoot = index.getMetadata.getProjectRoot
+              if (!projectRoot.isEmpty) {
+                foundScipFile = true
+                val root = URI.create(projectRoot)
+                index
+                  .getDocumentsList
+                  .asScala
+                  .foreach { doc =>
+                    val sourcepath = Paths.get(
+                      root.resolve(doc.getRelativePath)
                     )
-                  val document = ScipPrinters.printTextDocument(doc, source)
-                  val snapshotOutput = output.resolve(doc.getRelativePath)
-                  Files.createDirectories(snapshotOutput.getParent)
-                  Files.write(
-                    snapshotOutput,
-                    document.getBytes(StandardCharsets.UTF_8)
-                  )
-                }
+                    val source =
+                      new String(
+                        Files.readAllBytes(sourcepath),
+                        StandardCharsets.UTF_8
+                      )
+                    val document = ScipPrinters.printTextDocument(doc, source)
+                    val snapshotOutput = output.resolve(doc.getRelativePath)
+                    Files.createDirectories(snapshotOutput.getParent)
+                    Files.write(
+                      snapshotOutput,
+                      document.getBytes(StandardCharsets.UTF_8)
+                    )
+                  }
+              }
             }
             super.visitFile(file, attrs)
           }