Normalize relative source_file to POSIX separators

graphify/extract.py

@@ -6519,6 +6519,21 @@ def _sf_entry(sf: str, sf_path: Path) -> tuple[str, str, tuple[str, ...]]:
             continue
         sf_path = Path(sf)
         if not sf_path.is_absolute():
+            # No id-remap needed here (an id is minted independently of this
+            # string), but the raw source_file itself can still carry a
+            # native `\` separator on Windows — every extractor sets it via
+            # plain str(path), and only the absolute-input branch below ever
+            # ran it through as_posix(). A relative-path caller (the
+            # documented `extract(paths)` entry point with no `root`, or
+            # `python -m graphify.extract <file>...`) got whichever
+            # separator convention that one extractor happened to use,
+            # diverging from the canonical POSIX form every other path
+            # (explicit `root`, the CLI, which always passes one) produces.
+            # That fragments string-equality lookups keyed on source_file
+            # (build._norm_source_file, analyze.find_import_cycles, the
+            # semantic-cache key) into two spellings of one file (#2625).
+            if "\\" in sf:
+                item["source_file"] = sf.replace("\\", "/")
             continue
         new_sf, canonical_id, keys = _sf_entry(str(sf), sf_path)
         if "id" in item:

tests/test_extract.py

@@ -3812,3 +3812,62 @@ def test_inferred_uses_edge_dropped_for_module_top_level_reference(tmp_path):
     uses = _inferred_uses(result)
 
     assert not any(tgt == "helpers_helper" for _, tgt in uses)
+
+
+def test_relative_input_source_file_is_normalized_to_posix(tmp_path, monkeypatch):
+    """#2625: extract()'s canonicalization pass only ran a relative source_file
+    through as_posix() when the caller passed an explicit `root` — the
+    documented `extract(paths)` entry point with no root (and
+    `python -m graphify.extract <file>...`) left it exactly as the extractor
+    set it. Every extractor sets source_file via plain str(path), which on
+    Windows renders with a native `\\` separator, diverging from the POSIX
+    form every other path (explicit root, the CLI) produces — fragmenting
+    the string-equality lookups keyed on source_file (build._norm_source_file,
+    analyze.find_import_cycles, the semantic-cache key) into two spellings of
+    one file.
+
+    A real WindowsPath can't be constructed on a POSIX test runner, so this
+    monkeypatches one extension's dispatch to return a node/edge whose
+    source_file is a literal backslash-separated string — exactly the shape
+    str(WindowsPath("sub/file.py")) produces on a real Windows machine —
+    and asserts extract()'s post-pass normalizes it, exercising the actual
+    fixed code path rather than the OS-dependent symptom.
+    """
+    import graphify.extract as extract_mod
+
+    def _fake_windows_extractor(path):
+        return {
+            "nodes": [{
+                "id": "fake_node", "label": "fake", "file_type": "code",
+                "source_file": "sub\\file.py", "source_location": "L1",
+            }],
+            "edges": [{
+                "source": "fake_node", "target": "other_node", "relation": "calls",
+                "confidence": "EXTRACTED", "source_file": "sub\\file.py",
+            }],
+        }
+
+    real_get_extractor = extract_mod._get_extractor
+
+    def _patched_get_extractor(path):
+        if path.suffix == ".fakewin":
+            return _fake_windows_extractor
+        return real_get_extractor(path)
+
+    monkeypatch.setattr(extract_mod, "_get_extractor", _patched_get_extractor)
+
+    (tmp_path / "sub").mkdir()
+    target = tmp_path / "sub" / "file.fakewin"
+    target.write_text("irrelevant", encoding="utf-8")
+
+    old_cwd = Path.cwd()
+    try:
+        os.chdir(tmp_path)
+        result = extract([Path("sub/file.fakewin")])
+    finally:
+        os.chdir(old_cwd)
+
+    node = next(n for n in result["nodes"] if n["id"] == "fake_node")
+    edge = next(e for e in result["edges"] if e["source"] == "fake_node")
+    assert node["source_file"] == "sub/file.py", node["source_file"]
+    assert edge["source_file"] == "sub/file.py", edge["source_file"]