Drop SemanticDB from Kotlin compiler plug-in; emit SCIP shards only

Stacked on the Java SCIP-shard drop. semanticdb-kotlinc now bypasses the
SemanticDB protobuf step entirely: the FIR visitor feeds
ScipTextDocumentBuilder directly, the plug-in writes per-source *.scip
shards under META-INF/scip/, and the legacy SemanticDB writers, schema,
shared SemanticdbDocumentBuilder, and protobuf codegen are removed.

The aggregator's SemanticDB fallback and the --use-scip-shards opt-in
flag are also dropped — SCIP-shard aggregation is now the only path.

Author: jupblb

CONTRIBUTING.md

@@ -31,8 +31,8 @@ If you'd rather install tools manually, you'll need at least:
 
 These are the main components of the project.
 
-- `semanticdb-javac/src/main/java`: the Java compiler plugin that creates
-  SemanticDB files.
+- `semanticdb-javac/src/main/java`: the Java compiler plugin that emits
+  SCIP shard files.
 - `tests/minimized`: minimized Java source files that reproduce interesting test
   cases.
 - `tests/unit`: fast running unit tests that are helpful for local edit-and-test

build.sbt

@@ -531,7 +531,7 @@ lazy val semanticdbKotlincMinimized = project
           Def.sequential(
             Compile / compile,
             (cli / Compile / runMain).toTask(
-              s" $mainCls index-semanticdb --no-emit-inverse-relationships --use-scip-shards --cwd $srcRoot --output $scipOut $tgtRoot"
+              s" $mainCls index-semanticdb --no-emit-inverse-relationships --cwd $srcRoot --output $scipOut $tgtRoot"
             ),
             (cli / Compile / runMain).toTask(
               s" $mainCls snapshot --cwd $srcRoot --output $snapDir ${file(

docs/design.md

@@ -5,10 +5,9 @@ title: Design
 
 This project is implemented as a
 [Java compiler plugin](https://docs.oracle.com/en/java/javase/11/docs/api/jdk.compiler/com/sun/source/util/Plugin.html)
-that generates one
-[SemanticDB](https://scalameta.org/docs/semanticdb/specification.html) file for
-every `*.java` source file. After compilation completes, the SemanticDB files
-are processed to produce SCIP.
+that emits one [SCIP](https://github.com/sourcegraph/scip) shard file for every
+`*.java` source file. After compilation completes, the per-file SCIP shards are
+aggregated into a single SCIP index.
 
 ### Why Java compiler plugin?
 
@@ -24,24 +23,3 @@ There are several benefits to implementing scip-java as a compiler plugin:
   tool, we minimize the risk of diverging from the CI build environment such as
   installed system dependencies, custom compiler options and custom annotation
   processors.
-
-### Why SemanticDB?
-
-SemanticDB is Protobuf schema for information about symbols and types in Java
-programs and other languages. There are several benefits to using SemanticDB as
-an intermediary representation for SCIP:
-
-- **Simplicity**: It's easy to translate a single Java source file into a single
-  SemanticDB file inside a compiler plugin. It's more complicated to produce
-  SCIP because compiler plugins does not have access to a project-wide context,
-  which is necessary to produce accurate definitions and hovers in multi-module
-  projects with external library dependencies.
-- **Performance**: SemanticDB is fast to write and read. Each compilation unit
-  can be processed independently to keep memory usage low. The final conversion
-  from SemanticDB to SCIP can be safely parallelized.
-- **Cross-repository**: Compiler plugins have access to both source code and the
-  classpath (compiled bytecode of upstream dependencies). SemanticDB has been
-  designed so that it's also possible to generate spec-compliant symbols from
-  the classpath alone (no source code) and from the syntax tree of an individual
-  source file (no classpath). This flexibility will be helpful for scip-java in
-  the future to unblock cross-repository navigation.

docs/getting-started.md

@@ -351,10 +351,10 @@ Next, run the following command to generate the SCIP index (`index.scip`).
 ```
 bazel run @scip_java//scip-semanticdb:bazel -- --sourceroot $PWD
 
-# (optional) Validate that SemanticDB files were generated.
+# (optional) Validate that SCIP shard files were generated.
 # The command below works for the `examples/bazel-example` directory in the sourcegraph/scip-java repository.
-❯ jar tf bazel-bin/src/main/java/example/libexample.jar | grep semanticdb$
-META-INF/semanticdb/src/main/java/example/Example.java.semanticdb
+❯ jar tf bazel-bin/src/main/java/example/libexample.jar | grep scip$
+META-INF/scip/src/main/java/example/Example.java.scip
 ```
 
 Finally, run the following commands to upload the SCIP index to Sourcegraph.

docs/manual-configuration.md

@@ -12,10 +12,9 @@ fails.
 
 Indexing a codebase consists of two independent phases:
 
-- Compile the codebase with the SemanticDB compiler plugin.
-- Generate SCIP index from SemanticDB files.
-
-![A three stage pipeline that starts with a list of Java sources, creates a list of SemanticDB files that then become a single SCIP index.](assets/semanticdb-javac-pipeline.svg)
+- Compile the codebase with the SemanticDB compiler plugin, which writes one
+  SCIP shard per Java source file.
+- Aggregate the SCIP shards into a single SCIP index.
 
 The first phase can be complicated to configure and it can take a while to run.
 The second phase is quite simple to configure and it usually runs very fast.
@@ -63,7 +62,7 @@ compiler plugin. To do this you need to explicitly configure two directories:
   It's important that all of the source files that should be index live under
   this directory.
 - `-targetroot:PATH`: the absolute path to the directory where to generate
-  SemanticDB file. This directory can be anywhere on your file system.  
+  SCIP shard files. This directory can be anywhere on your file system.  
   Alternatively, pass in `-targetroot:javac-classes-directory` for the plugin to
   automatically use the `javac` output directory.
 
@@ -112,13 +111,13 @@ examples:
 - Maven: `mvn clean verify -DskipTests`
 - Bazel: `bazel build //...`
 
-If everything went well, you should have a lot of `*.semanticdb` files in the
+If everything went well, you should have a lot of `*.scip` shard files in the
 targetroot directory.
 
 ```
 ❯ find $TARGETROOT -type f
-build/semanticdb-targetroot/META-INF/semanticdb/j11/src/test/java/example/ExampleTest.java.semanticdb
-build/semanticdb-targetroot/META-INF/semanticdb/j11/src/main/java/example/Example.java.semanticdb
+build/semanticdb-targetroot/META-INF/scip/j11/src/test/java/example/ExampleTest.java.scip
+build/semanticdb-targetroot/META-INF/scip/j11/src/main/java/example/Example.java.scip
 ...
 ```
 
@@ -198,13 +197,13 @@ Which allows you to invoke it by simply running `mvn sourcegraph:sourcegraphDepe
 Cross-repository navigation is a feature that allows "goto definition" and "find
 references" to show results from multiple repositories.
 
-## Step 5: Generate SCIP index from SemanticDB files
+## Step 5: Aggregate SCIP shards into a single SCIP index
 
 First, install the `scip-java` command-line tool according to the instructions
 in the [getting started guide](getting-started.md).
 
-Next, run the `scip-java index-semanticdb` command to convert SemanticDB files
-into SCIP.
+Next, run the `scip-java index-semanticdb` command to aggregate the per-file
+SCIP shards into a single SCIP index.
 
 ```sh
 ❯ scip-java index-semanticdb $TARGETROOT

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

@@ -206,7 +206,7 @@ class ScipBuildTool(index: IndexCommand) extends BuildTool("SCIP", index) {
           .app
           .reporter
           .info(
-            "Some SemanticDB files got generated even if there were compile errors. " +
+            "Some SCIP shard files got generated even if there were compile errors. " +
               "In most cases, this means that scip-java managed to index everything " +
               "except the locations that had compile errors and you can ignore the compile errors."
           )

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

@@ -29,7 +29,7 @@ case class IndexCommand(
     @Description("The path where to generate the SCIP index.")
     output: Path = Paths.get("index.scip"),
     @Description(
-      "The directory where to generate SemanticDB files. " +
+      "The directory where to generate SCIP shard files. " +
         "Defaults to a build-specific path. " +
         "For example, the default value for Gradle is 'build/semanticdb-targetroot' and for Maven it's 'target/semanticdb-targetroot'"
     )

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

@@ -10,7 +10,6 @@ import com.sourcegraph.io.AbsolutePath
 import com.sourcegraph.scip_java.BuildInfo
 import com.sourcegraph.scip_java.buildtools.ClasspathEntry
 import com.sourcegraph.scip_semanticdb.ConsoleScipSemanticdbReporter
-import com.sourcegraph.scip_semanticdb.ScipSemanticdb
 import com.sourcegraph.scip_semanticdb.ScipSemanticdbOptions
 import com.sourcegraph.scip_semanticdb.ScipShardAggregator
 import moped.annotations._
@@ -21,7 +20,7 @@ import org.scip_code.scip.ToolInfo
 import ujson.Arr
 import ujson.Obj
 
-@Description("Converts SemanticDB files into a single SCIP index file.")
+@Description("Aggregates SCIP shard files 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"
@@ -30,10 +29,10 @@ import ujson.Obj
 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 shard files in parallel")
     parallel: Boolean = true,
     @Description(
-      "Whether to infer the location of SemanticDB files based as produced by Bazel"
+      "Whether to infer the location of SCIP shard files based as produced by Bazel"
     )
     bazel: Boolean = true,
     @Description(
@@ -44,7 +43,7 @@ final case class IndexSemanticdbCommand(
     @Description("URL to a PackageHub instance")
     @Hidden
     packagehub: Option[String] = None,
-    @Description("Directories that contain SemanticDB files.")
+    @Description("Directories that contain SCIP shard files.")
     @PositionalArguments()
     targetroot: List[Path] = Nil,
     @Description(
@@ -60,11 +59,6 @@ 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, aggregate *.scip shards under META-INF/scip/ instead of *.semanticdb files. " +
-        "Pass --use-scip-shards=false to fall back to the SemanticDB-based aggregator."
-    )
-    useScipShards: Boolean = true,
     @Inline()
     app: Application = Application.default
 ) extends Command {
@@ -102,10 +96,7 @@ final case class IndexSemanticdbCommand(
         allowEmptyIndex,
         allowExportingGlobalSymbolsFromDirectoryEntries
       )
-    if (useScipShards)
-      ScipShardAggregator.run(options)
-    else
-      ScipSemanticdb.run(options)
+    ScipShardAggregator.run(options)
     postPackages(packages)
     if (!app.reporter.hasErrors()) {
       app.info(options.output.toString)

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

@@ -65,7 +65,7 @@ public boolean hasErrors() {
             /* allowEmptyIndex */ true,
             /* indexDirectoryEntries */ false // because Bazel only compiles to jar files.
             );
-    ScipSemanticdb.run(scipOptions);
+    ScipShardAggregator.run(scipOptions);
 
     if (!scipOptions.reporter.hasErrors()) {
       System.out.println("done: " + scipOptions.output);