fix: disambiguate colliding node IDs from same-name files

tests/fixtures/collision/app1/Program.cs

@@ -0,0 +1,9 @@
+namespace App1;
+
+public class Program
+{
+    public static void Main()
+    {
+        var svc = new MyService();
+    }
+}

tests/fixtures/collision/app2/Program.cs

@@ -0,0 +1,9 @@
+namespace App2;
+
+public class Program
+{
+    public static void Run()
+    {
+        var helper = new Helper();
+    }
+}

tests/fixtures/collision/preexist/proj1/Startup.cs

@@ -0,0 +1,2 @@
+namespace Proj1;
+public class Startup { public void Run() { } }

tests/fixtures/collision/preexist/proj1_startup.cs

@@ -0,0 +1,4 @@
+namespace Preexist;
+// This file's stem is "proj1_startup", so _make_id("proj1_startup", "Preexist")
+// produces an ID that could collide with the renamed "startup" from proj1/Startup.cs
+public class Preexist { public void Work() { } }

tests/fixtures/collision/preexist/proj2/Startup.cs

@@ -0,0 +1,2 @@
+namespace Proj2;
+public class Startup { public void Init() { } }

tests/fixtures/collision/src/app/Startup.cs

@@ -0,0 +1,6 @@
+namespace SrcApp;
+
+public class Startup
+{
+    public void Configure() { }
+}

tests/fixtures/collision/tests/app/Startup.cs

@@ -0,0 +1,6 @@
+namespace TestApp;
+
+public class Startup
+{
+    public void ConfigureTest() { }
+}

tests/test_node_id_collision.py

@@ -0,0 +1,83 @@
+"""Tests for node ID collision disambiguation in extract()."""
+
+from pathlib import Path
+
+from graphify.extract import extract, collect_files
+
+COLLISION_DIR = Path(__file__).parent / "fixtures" / "collision"
+
+
+def _extract_collision_files(subdirs: list[str]) -> dict:
+    """Run extract() on files from the given subdirectories."""
+    files = []
+    for subdir in subdirs:
+        d = COLLISION_DIR / subdir
+        files.extend(collect_files(d))
+    return extract(files)
+
+
+def test_same_stem_different_dirs_get_unique_ids():
+    """Two Program.cs in different dirs must produce distinct node IDs."""
+    result = _extract_collision_files(["app1", "app2"])
+    ids = [n["id"] for n in result["nodes"]]
+    program_ids = [i for i in ids if "program" in i and "program_cs" not in i.replace("program_cs", "")]
+    # There must be at least two distinct Program-class nodes
+    class_ids = [i for i in ids if i.endswith("_program") or "_program_" in i]
+    assert len(set(class_ids)) >= 2, f"Expected distinct IDs, got: {class_ids}"
+
+
+def test_no_node_ids_collide_after_disambiguation():
+    """After disambiguation, all node IDs must be unique."""
+    result = _extract_collision_files(["app1", "app2"])
+    ids = [n["id"] for n in result["nodes"]]
+    assert len(ids) == len(set(ids)), f"Duplicate IDs found: {[i for i in ids if ids.count(i) > 1]}"
+
+
+def test_edges_reference_valid_nodes():
+    """All edge sources and targets must exist in the node set."""
+    result = _extract_collision_files(["app1", "app2"])
+    node_ids = {n["id"] for n in result["nodes"]}
+    for e in result["edges"]:
+        assert e["source"] in node_ids, f"Dangling source: {e['source']}"
+        # targets may reference external (unresolved) nodes, but sources must exist
+
+
+def test_same_parent_dir_uses_deeper_path():
+    """src/app/Startup.cs and tests/app/Startup.cs share parent 'app',
+    so disambiguation must use deeper path components."""
+    result = _extract_collision_files(["src/app", "tests/app"])
+    ids = [n["id"] for n in result["nodes"]]
+    startup_ids = [i for i in ids if "startup" in i and not i.endswith("_cs")]
+    assert len(set(startup_ids)) >= 2, (
+        f"Expected distinct IDs even with same parent dir, got: {startup_ids}"
+    )
+
+
+def test_cross_file_edges_remapped():
+    """Edges crossing file boundaries must have their targets remapped."""
+    result = _extract_collision_files(["app1", "app2"])
+    node_ids = {n["id"] for n in result["nodes"]}
+    for e in result["edges"]:
+        if e["source"] in node_ids:
+            # If we can verify the target, it must not be an old pre-rename ID
+            if e["target"] in node_ids:
+                continue
+            # Target may be an external ref (MyService, Helper) — that's fine
+
+
+def test_rename_does_not_collide_with_preexisting_id():
+    """Regression: a renamed ID must not collide with an existing node ID
+    from a non-colliding group. proj1/Startup.cs and proj2/Startup.cs collide
+    on 'startup_startup'; the rename of proj1's node must not produce an ID
+    that clashes with the node from proj1_startup.cs."""
+    result = _extract_collision_files(["preexist/proj1", "preexist/proj2", "preexist"])
+    ids = [n["id"] for n in result["nodes"]]
+    assert len(ids) == len(set(ids)), (
+        f"Duplicate IDs after disambiguation (preexisting collision): "
+        f"{[i for i in ids if ids.count(i) > 1]}"
+    )
+    # All file nodes and class nodes must survive — none dropped
+    node_files = {n.get("source_file", "") for n in result["nodes"]}
+    assert any("proj1/Startup.cs" in f or "proj1\\Startup.cs" in f for f in node_files)
+    assert any("proj2/Startup.cs" in f or "proj2\\Startup.cs" in f for f in node_files)
+    assert any("proj1_startup.cs" in f for f in node_files)