fix: surface silently-skipped dirs in enumeration + dedup Pascal edges

Two correctness fixes found while analysing the reported 'graphify update
occasionally writes a partial graph.json' bug.

Enumeration (P0): detect()'s os.walk had no onerror handler, so any os.scandir
failure -- a transient PermissionError, or a directory created/deleted mid-walk
by concurrent writes (e.g. benchmarking racing the scan) -- was silently
swallowed and that entire subtree dropped out of the file list with no log, no
error. Downstream that becomes a silently partial graph.json. The walk now
records each skipped directory (surfaced as walk_errors in detect()'s result)
and warns to stderr, while still enumerating the rest of the tree. This stays
visible even when a --force/GRAPHIFY_FORCE rebuild bypasses the shrink guards.
Relatedly, to_json's #479 anti-shrink guard was fail-OPEN: a non-empty but
unreadable existing graph.json (corrupt or mid-write) proceeded with the
overwrite. It now fails SAFE -- refuse and point at force=True -- while an
empty/whitespace existing file (no nodes to lose) still proceeds. The size-cap
check keeps running before any read, so an oversized existing file is not
loaded into memory.

Pascal edges (P1): a class method declared in the interface section and defined
in the implementation section each emitted a "method" edge to the same node id,
and the edge helpers (unlike the node helpers) did not dedup, so ~half of a
Pascal/Delphi graph's method edges were doubled -- inflating degree/centrality
and tripping the #1739 cross-file resolver's single-owner god-node guard. Both
extractors now dedup edges on (source, target, relation).

Adds regression tests for all three behaviours.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

CHANGELOG.md

@@ -4,6 +4,8 @@ Full release notes with details on each version: [GitHub Releases](https://githu
 
 ## 0.9.11 (2026-07-08)
 
+- Fix: file enumeration no longer silently drops a directory subtree. `detect()`'s `os.walk` had no `onerror` handler, so an `os.scandir` failure (a permission error, or a directory created/deleted mid-walk by concurrent writes) was swallowed and that whole subtree vanished from the scan with no log, yielding a silently partial `graph.json`. The walk now records every skipped directory (surfaced in the result's `walk_errors`) and warns to stderr, while still enumerating the rest. Relatedly, `to_json`'s anti-shrink guard (#479) now fails safe: a non-empty but unreadable existing `graph.json` refuses the overwrite (pass `force=True` to override) instead of silently clobbering a good graph; an empty file still proceeds.
+- Fix: Pascal/Delphi extractors no longer emit duplicate `method`/`contains`/`inherits` edges. A class method declared in the interface section and defined in the implementation section each emitted an edge to the same node, so ~half of a Pascal graph's method edges were doubled (skewing degree/centrality and tripping the new cross-file resolver's god-node guard). Both extractors now dedup edges on (source, target, relation), mirroring the existing node dedup.
 - Fix: Pascal/Delphi call resolution is scoped to the caller's class + inherits chain, and calls to methods inherited across file boundaries now resolve (#1739, thanks @richtext). Both extractors previously resolved every call via a single file-wide `{name: node_id}` dict, so two unrelated classes with a same-named method (property accessors, generated COM/TLB wrappers) collapsed onto whichever was inserted last, producing wrong cross-class `calls` edges. Resolution now walks own-class then ancestor chain then file-level free functions, emitting no edge when ambiguous (same god-node guard as the Ruby resolver). A new corpus-wide resolver (`graphify/pascal_resolution.py`) resolves calls from a descendant to a base-class method declared in a different file (the common generated-base/manual-descendant split). Also stops emitting a duplicate cross-file base-class stub carrying the wrong `source_file`.
 - Fix: query ranking no longer lets a lone generic term that exact-matches a short leaf label hijack seed selection in multi-term queries (#1602/#1724, thanks @fkhawajagh). `_score_nodes` scales the per-term exact/prefix tiers by squared term coverage; single-term and full-coverage queries are unchanged.
 - Fix: Kotlin enum entries are extracted as nodes with `case_of` edges to their enum (#1700, thanks @ivanzhilovich). Closes the Kotlin half of #1700 (the Java half shipped in 0.9.10 via #1719); `enum class ChatType { NORMAL, GROUP, SYSTEM }` now yields NORMAL/GROUP/SYSTEM nodes and "where is ChatType.X used" works for Kotlin.

graphify/detect.py

@@ -1107,9 +1107,29 @@ def detect(root: Path, *, follow_symlinks: bool | None = None, google_workspace:
     seen: set[Path] = set()
     all_files: list[Path] = []
 
+    # os.walk swallows os.scandir errors by default (no onerror -> the failing
+    # directory subtree is silently skipped). That turns a transient
+    # PermissionError, or a directory created/deleted mid-walk (e.g. concurrent
+    # writes racing the scan), into a partial file list and, downstream, a
+    # silently partial graph.json. Record and surface every skipped directory
+    # so an incomplete enumeration is visible rather than silent.
+    walk_errors: list[str] = []
+
+    def _on_walk_error(err: OSError) -> None:
+        import sys as _sys
+        target = getattr(err, "filename", None) or "<unknown>"
+        walk_errors.append(f"{target}: {err}")
+        print(
+            f"[graphify] WARNING: could not scan {target} ({err}); "
+            f"its files are missing from this run's enumeration.",
+            file=_sys.stderr,
+        )
+
     for scan_root in scan_paths:
         in_memory_tree = memory_dir.exists() and str(scan_root).startswith(str(memory_dir))
-        for dirpath, dirnames, filenames in os.walk(scan_root, followlinks=follow_symlinks):
+        for dirpath, dirnames, filenames in os.walk(
+            scan_root, followlinks=follow_symlinks, onerror=_on_walk_error
+        ):
             dp = Path(dirpath)
             if follow_symlinks and os.path.islink(dirpath):
                 real = os.path.realpath(dirpath)
@@ -1245,6 +1265,7 @@ def detect(root: Path, *, follow_symlinks: bool | None = None, google_workspace:
         "warning": warning,
         "skipped_sensitive": skipped_sensitive,
         "unclassified": sorted(unclassified),
+        "walk_errors": walk_errors,
         "graphifyignore_patterns": len(ignore_patterns),
         "scan_root": str(root.resolve()),
     }

graphify/export.py

@@ -184,11 +184,44 @@ def to_json(G: nx.Graph, communities: dict[int, list[str]], output_path: str, *,
     # Safety check: refuse to silently shrink an existing graph (#479)
     existing_path = Path(output_path)
     if not force and existing_path.exists():
+        from graphify.security import check_graph_file_size_cap
         try:
-            from graphify.security import check_graph_file_size_cap
             check_graph_file_size_cap(existing_path)
-            existing_data = json.loads(existing_path.read_text(encoding="utf-8"))
-            existing_n = len(existing_data.get("nodes", []))
+        except Exception:
+            # Existing graph.json trips the size cap; reading it to compare would
+            # be the very DoS the cap guards against. Can't verify — let the new
+            # graph replace the oversized file.
+            oversized = True
+        else:
+            oversized = False
+        if not oversized:
+            try:
+                raw = existing_path.read_text(encoding="utf-8")
+            except Exception:
+                raw = ""
+            if not raw.strip():
+                # Empty/whitespace existing file (e.g. a freshly touched path):
+                # no nodes to lose, so any new graph is a growth — proceed.
+                existing_n = 0
+            else:
+                try:
+                    existing_data = json.loads(raw)
+                    existing_n = len(existing_data.get("nodes", []))
+                except Exception as exc:
+                    # Non-empty but unparseable existing graph (corrupt or a
+                    # mid-write): we cannot verify the new graph is not a silent
+                    # shrink. Fail SAFE — refuse rather than overwrite. A
+                    # fail-OPEN here (the prior behavior) is the silent data-loss
+                    # path #479 exists to prevent: a transiently unreadable
+                    # graph.json would let a partial rebuild clobber a good one.
+                    import sys as _sys
+                    print(
+                        f"[graphify] WARNING: existing {existing_path} could not be "
+                        f"read to verify the new graph is not smaller ({exc}). "
+                        f"Refusing to overwrite; pass force=True to override.",
+                        file=_sys.stderr,
+                    )
+                    return False
             new_n = G.number_of_nodes()
             if new_n < existing_n:
                 import sys as _sys
@@ -203,8 +236,6 @@ def to_json(G: nx.Graph, communities: dict[int, list[str]], output_path: str, *,
                     file=_sys.stderr,
                 )
                 return False
-        except Exception:
-            pass  # unreadable existing file — proceed with write
 
     node_community = _node_community_map(communities)
     _labels: dict[int, str] = {int(k): v for k, v in (community_labels or {}).items()}

graphify/extractors/pascal.py

@@ -243,6 +243,7 @@ def _extract_pascal_regex(path: Path) -> dict:
     edges: list[dict] = []
     seen_ids: set[str] = set()
     seen_call_pairs: set[tuple[str, str]] = set()
+    seen_edges: set[tuple[str, str, str]] = set()
 
     def _add_node(nid: str, label: str, line: int) -> None:
         if nid not in seen_ids:
@@ -256,6 +257,14 @@ def _extract_pascal_regex(path: Path) -> dict:
             })
 
     def _add_edge(src: str, tgt: str, relation: str, line: int, context: str | None = None) -> None:
+        # A class method declared in the interface section and defined in the
+        # implementation section both emit a `method` edge to the same node, so
+        # dedup on (src, tgt, relation) to keep the graph from carrying doubled
+        # method/contains/inherits edges (mirrors _add_node's seen_ids guard).
+        key = (src, tgt, relation)
+        if key in seen_edges:
+            return
+        seen_edges.add(key)
         edge: dict = {
             "source": src,
             "target": tgt,
@@ -459,6 +468,7 @@ def extract_pascal(path: Path) -> dict:
     nodes: list[dict] = []
     edges: list[dict] = []
     seen_ids: set[str] = set()
+    seen_edges: set[tuple[str, str, str]] = set()
     proc_bodies: list[tuple[str, Any, str, str]] = []
     # (proc_nid, body_node, container, name_lower)
 
@@ -478,6 +488,14 @@ def extract_pascal(path: Path) -> dict:
         confidence: str = "EXTRACTED", weight: float = 1.0,
         context: str | None = None,
     ) -> None:
+        # A class method declared in the interface section and defined in the
+        # implementation section both emit a `method` edge to the same node, so
+        # dedup on (src, tgt, relation) to keep the graph from carrying doubled
+        # method/contains/inherits edges (mirrors add_node's seen_ids guard).
+        key = (src, tgt, relation)
+        if key in seen_edges:
+            return
+        seen_edges.add(key)
         edge: dict[str, Any] = {
             "source": src, "target": tgt, "relation": relation,
             "confidence": confidence, "source_file": str_path,

tests/test_detect.py

@@ -1597,3 +1597,38 @@ def test_detect_unclassified_empty_when_all_supported(tmp_path):
     (tmp_path / "README.md").write_text("# hi\n")
     res = detect(tmp_path)
     assert res.get("unclassified", []) == []
+
+
+def test_detect_reports_walk_errors_key():
+    """detect() always surfaces a walk_errors list so callers can tell whether
+    enumeration was complete."""
+    import tempfile
+    d = Path(tempfile.mkdtemp())
+    (d / "a.py").write_text("def f(): pass\n")
+    res = detect(d)
+    assert "walk_errors" in res
+    assert res["walk_errors"] == []
+
+
+def test_detect_surfaces_unreadable_dir_instead_of_silent_skip(tmp_path, capsys):
+    """os.walk silently skips a subtree whose scandir raises (permissions, or a
+    dir deleted mid-walk); that under-enumeration used to be invisible and could
+    yield a silently partial graph. detect() now records it in walk_errors and
+    warns, while still enumerating the rest of the tree."""
+    import os
+    if os.geteuid() == 0:
+        import pytest
+        pytest.skip("running as root: chmod 000 does not block scandir")
+    (tmp_path / "a.py").write_text("def f(): pass\n")
+    locked = tmp_path / "locked"
+    locked.mkdir()
+    (locked / "b.py").write_text("def g(): pass\n")
+    os.chmod(locked, 0o000)
+    try:
+        res = detect(tmp_path)
+    finally:
+        os.chmod(locked, 0o755)  # restore for cleanup
+    code = res["files"]["code"]
+    assert any(f.endswith("a.py") for f in code)  # rest of tree still enumerated
+    assert len(res["walk_errors"]) >= 1
+    assert "could not scan" in capsys.readouterr().err

tests/test_export.py

@@ -603,3 +603,39 @@ def test_backup_env_disable(tmp_path, monkeypatch):
     (tmp_path / "graph.json").write_text('{"nodes":[],"links":[]}')
     (tmp_path / ".graphify_semantic_marker").write_text("{}")
     assert backup_if_protected(tmp_path) is None
+
+
+def _mkG(n):
+    import networkx as nx
+    G = nx.Graph()
+    for i in range(n):
+        G.add_node(f"n{i}", label=f"n{i}", community=0)
+    return G
+
+
+def test_to_json_refuses_shrink(tmp_path):
+    """#479: refuse to silently overwrite an existing graph with fewer nodes."""
+    p = tmp_path / "graph.json"
+    json.dump({"nodes": [{"id": f"n{i}"} for i in range(5)]}, p.open("w"))
+    assert to_json(_mkG(2), {}, str(p), force=False) is False
+    assert to_json(_mkG(2), {}, str(p), force=True) is True  # force overrides
+
+
+def test_to_json_fails_safe_on_corrupt_existing(tmp_path):
+    """A non-empty but unparseable existing graph.json (corrupt or mid-write)
+    must NOT be silently overwritten — we can't verify the new graph isn't a
+    partial shrink, so fail safe (refuse) unless force is given."""
+    p = tmp_path / "graph.json"
+    p.write_text("{ this has content but is not valid json")
+    assert to_json(_mkG(10), {}, str(p), force=False) is False
+    assert to_json(_mkG(10), {}, str(p), force=True) is True
+
+
+def test_to_json_proceeds_on_empty_existing(tmp_path):
+    """An empty/whitespace existing file has no nodes to lose, so it is not a
+    shrink risk — the write proceeds."""
+    p = tmp_path / "graph.json"
+    p.write_text("")
+    assert to_json(_mkG(3), {}, str(p), force=False) is True
+    data = json.loads(p.read_text())
+    assert len(data["nodes"]) == 3

tests/test_pascal.py

@@ -320,3 +320,26 @@ def test_dfm_dispatch_registered():
 def test_dfm_detect_extension_registered():
     from graphify.detect import CODE_EXTENSIONS
     assert ".dfm" in CODE_EXTENSIONS
+
+
+def _dup_edges(r):
+    from collections import Counter
+    triples = Counter((e["source"], e["target"], e["relation"]) for e in r["edges"])
+    return {k: v for k, v in triples.items() if v > 1}
+
+
+def test_pascal_no_duplicate_method_edges_tree_sitter():
+    """A class method appears in both the interface declaration and the
+    implementation; each used to emit a `method` edge to the same node, so the
+    graph carried doubled method/contains/inherits edges (skewing degree and
+    breaking the cross-file inherited-call resolver's god-node guard). Edges are
+    now deduped on (source, target, relation)."""
+    from graphify.extract import extract_pascal
+    r = extract_pascal(FIXTURES / "sample.pas")
+    assert _dup_edges(r) == {}, f"duplicate edges: {_dup_edges(r)}"
+
+
+def test_pascal_no_duplicate_method_edges_regex():
+    from graphify.extract import _extract_pascal_regex
+    r = _extract_pascal_regex(FIXTURES / "sample.pas")
+    assert _dup_edges(r) == {}, f"duplicate edges: {_dup_edges(r)}"