fix: sync update-graph.sh hook extension allowlist with EXTENSIONS

.claude/hooks/update-graph.sh

@@ -24,23 +24,42 @@ if [ -z "$FILE_PATH" ]; then
   exit 0
 fi
 
-# Only rebuild for source files codegraph tracks
-# Skip docs, configs, test fixtures, and non-code files
-case "$FILE_PATH" in
-  *.js|*.ts|*.tsx|*.jsx|*.py|*.go|*.rs|*.java|*.cs|*.php|*.rb|*.tf|*.hcl)
-    ;;
-  *)
-    exit 0
-    ;;
-esac
+PROJECT_DIR=$(git rev-parse --show-toplevel 2>/dev/null) || PROJECT_DIR="${CLAUDE_PROJECT_DIR:-.}"
+
+# Only rebuild for source files codegraph tracks.
+# Skip docs, configs, test fixtures, and non-code files.
+#
+# The real allowlist is EXTENSIONS (src/shared/constants.ts), derived from
+# LANGUAGE_REGISTRY (src/domain/parser.ts) — the single source of truth for
+# every language codegraph parses. `npm run build` snapshots it to
+# dist/hook-extensions.txt (see scripts/gen-hook-extensions.mjs) so this
+# hook can do a fast native bash/grep check on every Edit/Write instead of
+# spawning a second Node process, or hand-copying the list, on every edit.
+#
+# The case statement below is only a fallback for before the first build
+# (no dist/hook-extensions.txt yet). tests/unit/hook-extensions.test.ts
+# fails if it ever drifts behind EXTENSIONS — keep it updated when
+# LANGUAGE_REGISTRY gains a new extension.
+EXT=".${FILE_PATH##*.}"
+GENERATED_EXT_LIST="$PROJECT_DIR/dist/hook-extensions.txt"
+if [ -f "$GENERATED_EXT_LIST" ]; then
+  grep -qxF "$EXT" "$GENERATED_EXT_LIST" || exit 0
+else
+  case "$EXT" in
+    .R|.bash|.c|.cc|.cjs|.clj|.cljc|.cljs|.cpp|.cs|.cu|.cuh|.cxx|.dart|.erl|.ex|.exs|.fs|.fsi|.fsx|.gemspec|.gleam|.go|.groovy|.gvy|.h|.hcl|.hpp|.hrl|.hs|.java|.jl|.js|.jsx|.kt|.kts|.lua|.m|.mjs|.ml|.mli|.php|.phtml|.py|.pyi|.r|.rake|.rb|.rs|.scala|.sh|.sol|.sv|.swift|.tf|.ts|.tsx|.v|.zig)
+      ;;
+    *)
+      exit 0
+      ;;
+  esac
+fi
 
 # Skip test fixtures — they're copied to tmp dirs anyway
 if echo "$FILE_PATH" | grep -qE '(fixtures|__fixtures__|testdata)/'; then
   exit 0
 fi
 
 # Guard: codegraph DB must exist (project has been built at least once)
-PROJECT_DIR=$(git rev-parse --show-toplevel 2>/dev/null) || PROJECT_DIR="${CLAUDE_PROJECT_DIR:-.}"
 DB_PATH="$PROJECT_DIR/.codegraph/graph.db"
 if [ ! -f "$DB_PATH" ]; then
   exit 0

package.json

@@ -79,7 +79,7 @@
     "node": ">=22.12.0"
   },
   "scripts": {
-    "build": "tsc && node -e \"require('fs').writeFileSync('dist/index.cjs',require('fs').readFileSync('src/index.cjs','utf8').replaceAll('./index.ts','./index.js'))\"",
+    "build": "tsc && node -e \"require('fs').writeFileSync('dist/index.cjs',require('fs').readFileSync('src/index.cjs','utf8').replaceAll('./index.ts','./index.js'))\" && node scripts/gen-hook-extensions.mjs",
     "build:wasm": "node scripts/node-ts.js scripts/build-wasm.ts",
     "typecheck": "tsc --noEmit",
     "verify-imports": "node scripts/node-ts.js scripts/verify-imports.ts",

scripts/gen-hook-extensions.mjs

@@ -0,0 +1,30 @@
+#!/usr/bin/env node
+/**
+ * Generates dist/hook-extensions.txt: a plain-text, one-extension-per-line
+ * snapshot of EXTENSIONS (src/shared/constants.ts), which is itself derived
+ * from LANGUAGE_REGISTRY (src/domain/parser.ts) — the single source of
+ * truth for every language codegraph parses.
+ *
+ * Consumed by .claude/hooks/update-graph.sh, a PostToolUse hook that fires
+ * on every Edit/Write in this repo. Reading this pre-generated list with a
+ * native `grep -qxF` lets the hook decide in ~1ms whether an edited file's
+ * extension is one codegraph tracks, without spawning a second Node process
+ * (tens of ms of startup cost) just to check a file extension, and without
+ * hand-copying the extension list a second time where it can silently drift
+ * out of sync (see issue #1736 — `.mjs`/`.cjs` were missing from the old
+ * hardcoded copy).
+ *
+ * Runs as part of `npm run build`, right after `tsc`, so the snapshot is
+ * regenerated automatically whenever LANGUAGE_REGISTRY changes.
+ */
+import { writeFileSync } from 'node:fs';
+import { dirname, join } from 'node:path';
+import { fileURLToPath } from 'node:url';
+import { EXTENSIONS } from '../dist/shared/constants.js';
+
+const root = dirname(dirname(fileURLToPath(import.meta.url)));
+const outFile = join(root, 'dist', 'hook-extensions.txt');
+const sorted = [...EXTENSIONS].sort();
+
+writeFileSync(outFile, `${sorted.join('\n')}\n`);
+console.log(`[gen-hook-extensions] wrote ${outFile} (${sorted.length} extensions)`);

tests/unit/hook-extensions.test.ts

@@ -0,0 +1,86 @@
+/**
+ * Drift guard for .claude/hooks/update-graph.sh's extension allowlist.
+ *
+ * That PostToolUse hook fires on every Edit/Write and needs a fast
+ * (no extra Node-startup) way to decide whether an edited file's extension
+ * is one codegraph tracks. It prefers dist/hook-extensions.txt — generated
+ * from EXTENSIONS by scripts/gen-hook-extensions.mjs as part of
+ * `npm run build` — and falls back to a hardcoded `case "$EXT" in` list for
+ * before the first build.
+ *
+ * That fallback list is a hand-maintained second copy of EXTENSIONS, so it
+ * can silently drift out of sync (this is exactly what happened in issue
+ * #1736: `.mjs`/`.cjs` were missing, so editing those files never
+ * triggered a rebuild). This test fails the moment EXTENSIONS and the
+ * hook's fallback list disagree, in either direction.
+ */
+import { execFileSync } from 'node:child_process';
+import fs from 'node:fs';
+import path from 'node:path';
+import { fileURLToPath } from 'node:url';
+import { describe, expect, it } from 'vitest';
+import { EXTENSIONS } from '../../src/shared/constants.js';
+
+const __dirname = path.dirname(fileURLToPath(import.meta.url));
+const REPO_ROOT = path.resolve(__dirname, '..', '..');
+const HOOK_PATH = path.join(REPO_ROOT, '.claude', 'hooks', 'update-graph.sh');
+const DIST_CONSTANTS_PATH = path.join(REPO_ROOT, 'dist', 'shared', 'constants.js');
+
+/** Extracts the `.ext|.ext|...` pattern from the hook's `case "$EXT" in` fallback arm. */
+function parseFallbackAllowlist(script: string): string[] {
+  const match = script.match(/case "\$EXT" in\s*\n\s*([.\w|-]+)\)/);
+  if (!match) {
+    throw new Error(
+      'Could not find `case "$EXT" in <list>)` fallback allowlist in update-graph.sh',
+    );
+  }
+  return match[1].split('|');
+}
+
+describe('update-graph.sh hook extension allowlist', () => {
+  const script = fs.readFileSync(HOOK_PATH, 'utf8');
+
+  it('fallback allowlist matches EXTENSIONS exactly (no drift)', () => {
+    const fallback = new Set(parseFallbackAllowlist(script));
+
+    const missingFromHook = [...EXTENSIONS].filter((ext) => !fallback.has(ext)).sort();
+    const staleInHook = [...fallback].filter((ext) => !EXTENSIONS.has(ext)).sort();
+
+    expect(
+      missingFromHook,
+      `EXTENSIONS has extensions the update-graph.sh fallback allowlist doesn't know ` +
+        `about: ${missingFromHook.join(', ')}. Sync the \`case "$EXT" in\` list in ` +
+        '.claude/hooks/update-graph.sh with EXTENSIONS (src/shared/constants.ts).',
+    ).toEqual([]);
+    expect(
+      staleInHook,
+      `update-graph.sh fallback allowlist has extensions no longer in EXTENSIONS: ` +
+        `${staleInHook.join(', ')}. Remove them from the \`case "$EXT" in\` list in ` +
+        '.claude/hooks/update-graph.sh.',
+    ).toEqual([]);
+  });
+
+  it('reads the generated dist/hook-extensions.txt snapshot as its primary source', () => {
+    expect(script).toMatch(/GENERATED_EXT_LIST="\$PROJECT_DIR\/dist\/hook-extensions\.txt"/);
+    expect(script).toMatch(/grep -qxF "\$EXT" "\$GENERATED_EXT_LIST"/);
+  });
+
+  // Only meaningful once `npm run build` has produced dist/shared/constants.js.
+  // Skipped (not failed) otherwise so this test doesn't require a build step
+  // that isn't a normal prerequisite of `npm test` itself.
+  it.skipIf(!fs.existsSync(DIST_CONSTANTS_PATH))(
+    'scripts/gen-hook-extensions.mjs generates a snapshot covering every EXTENSIONS entry',
+    () => {
+      const generatedPath = path.join(REPO_ROOT, 'dist', 'hook-extensions.txt');
+      execFileSync('node', [path.join(REPO_ROOT, 'scripts', 'gen-hook-extensions.mjs')], {
+        cwd: REPO_ROOT,
+      });
+
+      expect(fs.existsSync(generatedPath)).toBe(true);
+      const generated = new Set(fs.readFileSync(generatedPath, 'utf8').split('\n').filter(Boolean));
+
+      expect([...EXTENSIONS].filter((ext) => !generated.has(ext))).toEqual([]);
+      expect([...generated].filter((ext) => !EXTENSIONS.has(ext))).toEqual([]);
+    },
+  );
+});