feat(codeintel): Add precise auto-inferencing for dotnet (#63348)

Adds support for dotnet inference by adding the lua script and wiring it
up in the inference engine.

Author: mmanela

internal/codeintel/autoindexing/internal/inference/BUILD.bazel

@@ -50,6 +50,7 @@ go_test(
     timeout = "short",
     srcs = [
         "infer_test.go",
+        "lang_dotnet_test.go",
         "lang_go_test.go",
         "lang_java_test.go",
         "lang_python_test.go",

internal/codeintel/autoindexing/internal/inference/lang_dotnet_test.go

@@ -0,0 +1,26 @@
+package inference
+
+import (
+	"testing"
+)
+
+func TestDotNetGenerator(t *testing.T) {
+	testGenerators(t,
+		generatorTestCase{
+			description: "dotnet sln files exist",
+			repositoryContents: map[string]string{
+				"one.sln":            "",
+				"one.csproj":         "",
+				"foo/baz/two.sln":    "",
+				"foo/baz/two.vbproj": "",
+			},
+		},
+		generatorTestCase{
+			description: "dotnet sln files do not exist",
+			repositoryContents: map[string]string{
+				"one.csproj":         "",
+				"foo/baz/two.vbproj": "",
+			},
+		},
+	)
+}

internal/codeintel/autoindexing/internal/inference/libs/indexes.go

@@ -22,6 +22,7 @@ var defaultIndexers = map[string]string{
 	"rust":       "sourcegraph/scip-rust",
 	"typescript": "sourcegraph/scip-typescript",
 	"ruby":       "sourcegraph/scip-ruby",
+	"dotnet":     "sourcegraph/scip-dotnet",
 }
 
 // To update, run `DOCKER_USER=... DOCKER_PASS=... ./update-shas.sh`
@@ -32,6 +33,7 @@ var defaultIndexerSHAs = map[string]string{
 	"sourcegraph/scip-python":     "sha256:e3c13f0cadca78098439c541d19a72c21672a3263e22aa706760d941581e068d",
 	"sourcegraph/scip-typescript": "sha256:3df8b36a2ad4e073415bfbeaedf38b3cfff3e697614c8f578299f470d140c2c8",
 	"sourcegraph/scip-ruby":       "sha256:ef53e5f1450330ddb4a3edce963b7e10d900d44ff1e7de4960680289ac25f319",
+	"sourcegraph/scip-dotnet":     "sha256:1d8a590edfb3834020fceedacac6608811dd31fcba9092426140093876d8d52e",
 }
 
 func DefaultIndexerForLang(language string) (string, bool) {

internal/codeintel/autoindexing/internal/inference/libs/update-shas.sh

@@ -16,7 +16,7 @@ fi
 SCRIPT_DIR="$(dirname "${BASH_SOURCE[0]}")"
 
 # No scip-clang as that doesn't have a Docker image
-for indexer in scip-go scip-rust scip-java scip-python scip-typescript scip-ruby; do
+for indexer in scip-go scip-rust scip-java scip-python scip-typescript scip-ruby scip-dotnet; do
   tag="latest"
   if [[ "${indexer}" = "scip-python" ]] || [[ "${indexer}" = "scip-typescript" || "${indexer}" = "scip-ruby" ]]; then
     tag="autoindex"

internal/codeintel/autoindexing/internal/inference/lua/BUILD.bazel

@@ -8,6 +8,7 @@ go_library(
         "README.md",
         "config.lua",
         "embed.go",
+        "dotnet.lua",
         "go.lua",
         "indexes.lua",
         "java.lua",

internal/codeintel/autoindexing/internal/inference/lua/dotnet.lua

@@ -0,0 +1,65 @@
+local path = require("path")
+local pattern = require("sg.autoindex.patterns")
+local recognizer = require("sg.autoindex.recognizer")
+
+local indexer = require("sg.autoindex.indexes").get("dotnet")
+
+local env_steps = {
+
+  -- macOS enables W^X (https://en.wikipedia.org/wiki/W%5EX) which means
+  -- that when running in development, we need to set the following environment
+  -- variable for `dotnet restore` to work correctly. It is technically not needed on
+  -- Linux in production, and removing this might improve performance, but we
+  -- currently do not have a way of conditionally passing in OS-level configuration
+  -- to the Lua script, and doing so would create a divergence between dev vs prod,
+  -- so leave this as-is for now.
+  --
+  -- See also: https://github.com/dotnet/runtime/issues/97828
+  "export DOTNET_EnableWriteXorExecute=0",
+}
+
+local generate_dotnet_jobs = function(_, paths)
+  local jobs = {}
+  for i = 1, #paths do
+    table.insert(jobs, {
+      indexer = indexer,
+      root = path.dirname(paths[i]),
+      local_steps = env_steps,
+      indexer_args = { "scip-dotnet", "index", paths[i], "--output", "index.scip" },
+      outfile = "index.scip",
+    })
+  end
+
+  return jobs
+end
+
+local dotnet_proj_recognizer = recognizer.new_path_recognizer({
+  patterns = {
+    pattern.new_path_extension("csproj"),
+    pattern.new_path_extension("vbproj"),
+  },
+
+  generate = generate_dotnet_jobs,
+})
+
+local dotnet_sln_recognizer = recognizer.new_path_recognizer({
+  patterns = {
+    pattern.new_path_extension("sln"),
+  },
+
+  generate = generate_dotnet_jobs,
+})
+
+-- For .NET projects we are employing a fallback recognizer that will first
+-- look for only solution files in a repo. This is the most common structure for .NET repos
+-- and we will build an indexing job for each solution file we find.
+-- If we find no solution files, we will then look for any .csproj or.vbproj files
+-- and build an indexing job for each of them. This structure can happen since solution files are not required
+-- and sometimes people auto-generate and don't version control them just for IDE support.
+-- A repo with a more exotic structure, mixing projects that are part of solutions with
+-- projects that are outside of solutions, would need to be indexed by the user manually
+-- or have the user customize the inference.
+return recognizer.new_fallback_recognizer({
+  dotnet_sln_recognizer,
+  dotnet_proj_recognizer,
+})

internal/codeintel/autoindexing/internal/inference/lua/recognizers.lua

@@ -1,14 +1,15 @@
-local config = require("sg.autoindex.config").new {}
+local config = require("sg.autoindex.config").new({})
 
-for _, name in ipairs {
+for _, name in ipairs({
   "go",
   "java",
   "python",
   "ruby",
   "rust",
   "test",
   "typescript",
-} do
+  "dotnet",
+}) do
   -- Backdoor set `sg.`-prefixed recognizers
   rawset(config, "sg." .. name, require("sg.autoindex." .. name))
 end

internal/codeintel/autoindexing/internal/inference/testdata/dotnet_sln_files_do_not_exist.yaml

@@ -0,0 +1,26 @@
+- steps: []
+  local_steps:
+    - export DOTNET_EnableWriteXorExecute=0
+  root: ""
+  indexer: sourcegraph/scip-dotnet@sha256:1d8a590edfb3834020fceedacac6608811dd31fcba9092426140093876d8d52e
+  indexer_args:
+    - scip-dotnet
+    - index
+    - one.csproj
+    - --output
+    - index.scip
+  outfile: index.scip
+  requestedEnvVars: []
+- steps: []
+  local_steps:
+    - export DOTNET_EnableWriteXorExecute=0
+  root: foo/baz
+  indexer: sourcegraph/scip-dotnet@sha256:1d8a590edfb3834020fceedacac6608811dd31fcba9092426140093876d8d52e
+  indexer_args:
+    - scip-dotnet
+    - index
+    - foo/baz/two.vbproj
+    - --output
+    - index.scip
+  outfile: index.scip
+  requestedEnvVars: []

internal/codeintel/autoindexing/internal/inference/testdata/dotnet_sln_files_exist.yaml

@@ -0,0 +1,26 @@
+- steps: []
+  local_steps:
+    - export DOTNET_EnableWriteXorExecute=0
+  root: ""
+  indexer: sourcegraph/scip-dotnet@sha256:1d8a590edfb3834020fceedacac6608811dd31fcba9092426140093876d8d52e
+  indexer_args:
+    - scip-dotnet
+    - index
+    - one.sln
+    - --output
+    - index.scip
+  outfile: index.scip
+  requestedEnvVars: []
+- steps: []
+  local_steps:
+    - export DOTNET_EnableWriteXorExecute=0
+  root: foo/baz
+  indexer: sourcegraph/scip-dotnet@sha256:1d8a590edfb3834020fceedacac6608811dd31fcba9092426140093876d8d52e
+  indexer_args:
+    - scip-dotnet
+    - index
+    - foo/baz/two.sln
+    - --output
+    - index.scip
+  outfile: index.scip
+  requestedEnvVars: []

internal/codeintel/autoindexing/internal/store/config_inference.go

@@ -92,6 +92,7 @@ return require("sg.autoindex.config").new({
 	-- ["sg.ruby"] = false,
 	-- ["sg.rust"] = false,
 	-- ["sg.typescript"] = false,
+	-- ["sg.dotnet"] = false,
 	["acme.custom"] = custom_recognizer,
 })
 `