Drop semanticdb step when indexing Java code

build.sbt

@@ -154,7 +154,7 @@ lazy val javacPlugin = project
           .inAll
       )
   )
-  .dependsOn(semanticdb)
+  .dependsOn(semanticdb, scipProto)
 
 lazy val scipProto = project
   .in(file("scip-java-proto"))
@@ -693,6 +693,11 @@ lazy val fatjarPackageSettings = List[Def.Setting[_]](
       MergeStrategy.discard
     case PathList("META-INF", "versions", "9", "module-info.class") =>
       MergeStrategy.discard
+    // Bazel BUILD files live next to *.proto sources in our subprojects; they are
+    // not needed at runtime and would conflict when multiple proto modules are
+    // merged into the same fat jar.
+    case PathList("BUILD") =>
+      MergeStrategy.discard
     case x =>
       val oldStrategy = (assembly / assemblyMergeStrategy).value
       oldStrategy(x)

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

@@ -14,6 +14,7 @@ import com.sourcegraph.scip_semanticdb.ConsoleScipSemanticdbReporter
 import com.sourcegraph.scip_semanticdb.ScipOutputFormat
 import com.sourcegraph.scip_semanticdb.ScipSemanticdb
 import com.sourcegraph.scip_semanticdb.ScipSemanticdbOptions
+import com.sourcegraph.scip_semanticdb.ScipShardAggregator
 import moped.annotations._
 import moped.cli.Application
 import moped.cli.Command
@@ -60,6 +61,13 @@ final case class IndexSemanticdbCommand(
         "Maven->Maven or Gradle->Gradle projects because those build tools compile sources to classfiles inside directories."
     )
     allowExportingGlobalSymbolsFromDirectoryEntries: Boolean = true,
+    @Description(
+      "If true, walk targetroots for *.scip shards (META-INF/scip/...) emitted by the " +
+        "compiler plugin instead of *.semanticdb files. The aggregator rewrites placeholder " +
+        "symbols into the final 'scip-java' scheme and merges per-source shards into the " +
+        "output index."
+    )
+    useScipShards: Boolean = false,
     @Inline()
     app: Application = Application.default
 ) extends Command {
@@ -108,7 +116,10 @@ final case class IndexSemanticdbCommand(
         allowEmptyIndex,
         allowExportingGlobalSymbolsFromDirectoryEntries
       )
-    ScipSemanticdb.run(options)
+    if (useScipShards)
+      ScipShardAggregator.run(options)
+    else
+      ScipSemanticdb.run(options)
     postPackages(packages)
     if (!app.reporter.hasErrors()) {
       app.info(options.output.toString)

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

@@ -51,31 +51,37 @@ case class SnapshotCommand(
               attrs: BasicFileAttributes
           ): FileVisitResult = {
             if (scipPattern.matches(file)) {
-              foundScipFile = true
               val index = Scip.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
+              // Skip per-source shards emitted by the compiler plugin (those don't have a
+              // project_root). The aggregator produces a single top-level index file that
+              // carries the project_root and is the canonical input for snapshot rendering.
+              val rawProjectRoot = index.getMetadata.getProjectRoot
+              if (rawProjectRoot.nonEmpty) {
+                foundScipFile = true
+                val projectRoot = URI.create(rawProjectRoot)
+                index
+                  .getDocumentsList
+                  .asScala
+                  .foreach { doc =>
+                    val sourcepath = Paths.get(projectRoot.resolve(doc.getRelativePath))
+                    val source =
+                      new String(
+                        Files.readAllBytes(sourcepath),
+                        StandardCharsets.UTF_8
+                      )
+                    val document = ScipPrinters.printTextDocument(
+                      doc,
+                      source,
+                      CommentSyntax.default
                     )
-                  val document = ScipPrinters.printTextDocument(
-                    doc,
-                    source,
-                    CommentSyntax.default
-                  )
-                  val snapshotOutput = output.resolve(doc.getRelativePath)
-                  Files.createDirectories(snapshotOutput.getParent)
-                  Files.write(
-                    snapshotOutput,
-                    document.getBytes(StandardCharsets.UTF_8)
-                  )
-                }
+                    val snapshotOutput = output.resolve(doc.getRelativePath)
+                    Files.createDirectories(snapshotOutput.getParent)
+                    Files.write(
+                      snapshotOutput,
+                      document.getBytes(StandardCharsets.UTF_8)
+                    )
+                  }
+              }
             }
             super.visitFile(file, attrs)
           }