From 444d73e1c893de30f741ae0fc71a83f6f57a0383 Mon Sep 17 00:00:00 2001
From: Timo Derstappen <teemow@gmail.com>
Date: Thu, 11 Jun 2026 09:36:38 +0200
Subject: [PATCH] fix(global-graph): rewire edges to deduplicated external
 nodes (#1250)

global_add deduplicates external-library nodes (no source_file) by label
against externals already in the global graph, but dropped every edge
incident to a skipped node. Each repo after the first lost its edges to
shared externals, so cross-repo "what uses library X" queries only saw
the first-added repo.

Build a remap from each deduplicated external to the surviving global
node and rewrite edge endpoints through it before add_edge, skipping
self-loops introduced by the remapping.

Co-authored-by: Cursor <cursoragent@cursor.com>
---
 graphify/global_graph.py   | 14 ++++++++-----
 tests/test_global_graph.py | 41 ++++++++++++++++++++++++++++++++++++++
 2 files changed, 50 insertions(+), 5 deletions(-)

diff --git a/graphify/global_graph.py b/graphify/global_graph.py
index c6310f94..4c4817fa 100644
--- a/graphify/global_graph.py
+++ b/graphify/global_graph.py
@@ -105,20 +105,24 @@ def global_add(source_path: Path, repo_tag: str) -> dict:
         for n, d in G.nodes(data=True)
         if not d.get("source_file") and d.get("label")
     }
-    nodes_to_skip = set()
+    # Map each deduplicated external onto the existing global node so that
+    # edges incident to it can be rewired instead of dropped.
+    remap = {}
     for node, data in prefixed.nodes(data=True):
         if not data.get("source_file") and data.get("label") in external_labels:
-            nodes_to_skip.add(node)
+            remap[node] = external_labels[data["label"]]
 
     # Compose: add prefixed nodes (except deduplicated externals) into global graph
     for node, data in prefixed.nodes(data=True):
-        if node not in nodes_to_skip:
+        if node not in remap:
             G.add_node(node, **data)
     for u, v, data in prefixed.edges(data=True):
-        if u not in nodes_to_skip and v not in nodes_to_skip:
+        u = remap.get(u, u)
+        v = remap.get(v, v)
+        if u != v:  # don't introduce self-loops via remapping
             G.add_edge(u, v, **data)
 
-    added = prefixed.number_of_nodes() - len(nodes_to_skip)
+    added = prefixed.number_of_nodes() - len(remap)
     _save_global_graph(G)
 
     manifest["repos"][repo_tag] = {
diff --git a/tests/test_global_graph.py b/tests/test_global_graph.py
index f40d9c6d..b91a1ebe 100644
--- a/tests/test_global_graph.py
+++ b/tests/test_global_graph.py
@@ -279,6 +279,47 @@ def test_merge_graphs_prefixes_ids(tmp_path):
     assert merged.number_of_nodes() == 2  # no silent collapse
 
 
+def test_global_add_rewires_edges_to_deduplicated_externals(tmp_path):
+    """Edges incident to an external node that gets deduplicated against an
+    already-present external must be rewired to the existing node, not dropped."""
+    g1 = tmp_path / "graph1.json"
+    g2 = tmp_path / "graph2.json"
+    GA = _make_graph(
+        [
+            {"id": "moda", "label": "ModA", "source_file": "src/a.py"},
+            {"id": "requests", "label": "requests"},
+        ],
+        [{"source": "moda", "target": "requests", "relation": "imports"}],
+    )
+    GB = _make_graph(
+        [
+            {"id": "modb", "label": "ModB", "source_file": "src/b.py"},
+            {"id": "requests", "label": "requests"},
+        ],
+        [{"source": "modb", "target": "requests", "relation": "imports"}],
+    )
+    _graph_to_json(GA, g1)
+    _graph_to_json(GB, g2)
+
+    global_dir = tmp_path / ".graphify"
+    with patch("graphify.global_graph._GLOBAL_DIR", global_dir), \
+         patch("graphify.global_graph._GLOBAL_GRAPH", global_dir / "global-graph.json"), \
+         patch("graphify.global_graph._GLOBAL_MANIFEST", global_dir / "global-manifest.json"):
+        from graphify.global_graph import global_add, _load_global_graph
+        global_add(g1, "repoA")
+        global_add(g2, "repoB")
+        G = _load_global_graph()
+
+    # repoB's external "requests" was deduplicated against repoA's
+    assert "repoA::requests" in G.nodes
+    assert "repoB::requests" not in G.nodes
+    # repoA's edge is untouched
+    assert G.has_edge("repoA::moda", "repoA::requests")
+    # repoB's edge must be rewired to the existing external node, not dropped
+    assert G.has_edge("repoB::modb", "repoA::requests")
+    assert G.edges["repoB::modb", "repoA::requests"]["relation"] == "imports"
+
+
 def test_global_add_rejects_oversized_source_graph(monkeypatch, tmp_path):
     """#F4: global_add must refuse to read a source graph.json that
     exceeds the size cap, rather than json.loads-ing it into memory."""