From 48a8da896be0158fa89f5e4314b950e0d8f85822 Mon Sep 17 00:00:00 2001
From: abhirupone <abhirup.karmakar00@gmail.com>
Date: Fri, 25 Sep 2026 15:27:43 +0530
Subject: [PATCH] fix(markdown): respect ignore rules when building wikilink
 index (#3822)

---
 graphify/extractors/markdown.py | 19 +++++++++++-----
 tests/test_languages.py         | 40 +++++++++++++++++++++++++++++++++
 2 files changed, 53 insertions(+), 6 deletions(-)

diff --git a/graphify/extractors/markdown.py b/graphify/extractors/markdown.py
index 4dc2b8feb..d2d28f30d 100644
--- a/graphify/extractors/markdown.py
+++ b/graphify/extractors/markdown.py
@@ -147,20 +147,27 @@ def _build_link_index(root: Path) -> "dict[str, list[tuple[int, str, Path]]]":
     """Index every linkable document under *root* by NFC-normalized basename.
 
     Each entry maps basename -> [(depth, root-relative posix path, absolute
-    path)]. Directories in detect._SKIP_DIRS and dot-directories are pruned —
-    the same corpus boundary the scanner draws, and Obsidian itself does not
-    index dot-folders.
+    path)]. Directories in detect._SKIP_DIRS, dot-directories, and paths
+    excluded by .graphifyignore/.gitignore are pruned — the same corpus
+    boundary the scanner draws, and Obsidian itself does not index dot-folders
+    or ignored paths (#3822).
     """
-    from graphify.detect import _SKIP_DIRS
+    from graphify.detect import _SKIP_DIRS, ignored_predicate
+    root = Path(root)
+    ignored = ignored_predicate(root)
     index: dict[str, list[tuple[int, str, Path]]] = {}
     for dirpath, dirnames, filenames in os.walk(root):
+        dp = Path(dirpath)
         dirnames[:] = sorted(
-            d for d in dirnames if not d.startswith(".") and d not in _SKIP_DIRS
+            d for d in dirnames
+            if not d.startswith(".") and d not in _SKIP_DIRS and not ignored(dp / d)
         )
         for fname in filenames:
             if Path(fname).suffix.lower() not in _MD_LINKABLE_EXTS:
                 continue
-            abs_path = Path(dirpath) / fname
+            abs_path = dp / fname
+            if ignored(abs_path):
+                continue
             rel = os.path.relpath(str(abs_path), str(root)).replace("\\", "/")
             index.setdefault(_nfc(fname), []).append(
                 (rel.count("/"), _nfc(rel), abs_path)
diff --git a/tests/test_languages.py b/tests/test_languages.py
index 4508bbfdd..7416425bf 100644
--- a/tests/test_languages.py
+++ b/tests/test_languages.py
@@ -3119,6 +3119,46 @@ def test_markdown_wikilink_fallback_unicode_normalization(tmp_path):
                for e in refs), f"NFD wikilink missed the NFC file: {refs}"
 
 
+def test_markdown_wikilink_index_prunes_ignored_directories(tmp_path, monkeypatch):
+    """#3822: _build_link_index must prune directories and files matched by
+    .graphifyignore/.gitignore, preventing both 50+ min directory walks on
+    large ignored trees and wikilinks erroneously resolving into ignored files."""
+    vault = tmp_path / "vault"
+    vault.mkdir()
+    (vault / ".graphifyignore").write_text("bigdata/\nsecret.md\n", encoding="utf-8")
+
+    (vault / "notes").mkdir()
+    (vault / "notes" / "hub.md").write_text("# Hub\nSee [[target]] and [[secret]].\n", encoding="utf-8")
+    (vault / "notes" / "target.md").write_text("# Target\n", encoding="utf-8")
+
+    (vault / "bigdata").mkdir()
+    (vault / "bigdata" / "sub").mkdir()
+    (vault / "bigdata" / "sub" / "secret.md").write_text("# Secret in bigdata\n", encoding="utf-8")
+
+    import os
+    visited = []
+    real_walk = os.walk
+
+    def spy_walk(top, *args, **kwargs):
+        for dirpath, dirnames, filenames in real_walk(top, *args, **kwargs):
+            visited.append(os.path.relpath(dirpath, str(vault)))
+            yield dirpath, dirnames, filenames
+
+    monkeypatch.setattr(os, "walk", spy_walk)
+
+    node_ids, refs, page_id = _vault_extract(
+        vault, [vault / "notes" / "hub.md", vault / "notes" / "target.md"]
+    )
+
+    bigdata_walked = [v for v in visited if v.startswith("bigdata")]
+    assert not bigdata_walked, f"ignored directory descended during index walk: {bigdata_walked}"
+
+    hub_id = page_id(vault / "notes" / "hub.md")
+    target_id = page_id(vault / "notes" / "target.md")
+    assert any(e["source"] == hub_id and e["target"] == target_id for e in refs), f"valid target link lost: {refs}"
+    assert not any("bigdata" in e["target"] for e in refs), f"wikilink resolved into ignored path: {refs}"
+
+
 # ── Groovy ───────────────────────────────────────────────────────────────────