fix(indexing): avoid sqlite symbol id collisions

Author: johnhuang316

src/code_index_mcp/indexing/strategies/base_strategy.py

@@ -49,18 +49,22 @@ def _create_symbol_id(self, file_path: str, symbol_name: str) -> str:
         return f"{relative_path}::{symbol_name}"
 
     def _get_relative_path(self, file_path: str) -> str:
-        """Convert absolute file path to relative path."""
-        parts = file_path.replace('\\', '/').split('/')
+        """Normalize path for symbol identifiers relative to project root."""
+        if not file_path:
+            return ""
 
-        # Priority order: test > src (outermost project roots first)
-        for root_dir in ['test', 'src']:
-            if root_dir in parts:
-                root_index = parts.index(root_dir)
-                relative_parts = parts[root_index:]
-                return '/'.join(relative_parts)
+        normalized = os.path.normpath(file_path)
+        if normalized == ".":
+            return ""
 
-        # Fallback: use just filename
-        return os.path.basename(file_path)
+        normalized = normalized.replace("\\", "/")
+        if normalized.startswith("./"):
+            normalized = normalized[2:]
+
+        if not os.path.isabs(file_path):
+            normalized = normalized.lstrip("/")
+
+        return normalized or os.path.basename(file_path)
 
     def _extract_line_number(self, content: str, symbol_position: int) -> int:
         """

tests/indexing/test_symbol_ids.py

@@ -0,0 +1,30 @@
+"""Tests for symbol identifier generation."""
+
+from code_index_mcp.indexing.json_index_builder import JSONIndexBuilder
+
+
+def test_symbol_ids_use_relative_paths(tmp_path):
+    project_dir = tmp_path / "project"
+    scripts_dir = project_dir / "scripts"
+    examples_dir = project_dir / "examples"
+    scripts_dir.mkdir(parents=True)
+    examples_dir.mkdir(parents=True)
+
+    (scripts_dir / "foo.py").write_text(
+        "def foo():\n"
+        "    return 1\n",
+        encoding="utf-8",
+    )
+    (examples_dir / "foo.py").write_text(
+        "def foo():\n"
+        "    return 2\n",
+        encoding="utf-8",
+    )
+
+    builder = JSONIndexBuilder(str(project_dir))
+    index = builder.build_index(parallel=False)
+    symbols = index["symbols"]
+
+    assert "scripts/foo.py::foo" in symbols
+    assert "examples/foo.py::foo" in symbols
+    assert len({sid for sid in symbols if sid.endswith("::foo")}) == 2