graphmark 0.3.0__tar.gz → 0.3.1__tar.gz

{graphmark-0.3.0 → graphmark-0.3.1}/CHANGELOG.md

@@ -1,6 +1,32 @@
 # CHANGELOG
 
 
+## v0.3.1 (2026-07-25)
+
+### Bug Fixes
+
+- **graph**: Stop counting same-note anchor links as unresolved
+  ([#99](https://github.com/cdcoonce/graphmark/pull/99),
+  [`a1d03f3`](https://github.com/cdcoonce/graphmark/commit/a1d03f3eb0e7b36b6d44d2a4eecfe8bb70da8d11))
+
+unresolved counted anchor-only wikilinks — [[#Heading]], [[#Heading|alias]], [[#^blockref]] — as
+  broken. Those are Obsidian same-note references: they target no note at all, so they are neither
+  an edge nor a broken link. Counting them corrupted the very metric Track D was built around,
+  max_unresolved_links.
+
+Measured on the owner's live 521-note vault: 40 of 213 reported unresolved links (19%) were
+  anchor-only, and the note reported as the worst offender earned that spot purely by navigating
+  itself heavily.
+
+build() now skips a display whose note part (before | and #) is empty. [[Note#Heading]] is
+  unaffected: it still resolves via Note, and still counts as unresolved when Note is missing.
+
+This refines the #75 semantics, which defined ambiguous-counts and self-link-doesn't but never
+  considered anchor-only links.
+
+Closes #98
+
+
 ## v0.3.0 (2026-07-25)
 
 ### Bug Fixes

{graphmark-0.3.0 → graphmark-0.3.1}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: graphmark
-Version: 0.3.0
+Version: 0.3.1
 Summary: Deterministic knowledge-graph analysis for markdown / [[wikilink]] vaults.
 Project-URL: Homepage, https://github.com/cdcoonce/graphmark
 Project-URL: Repository, https://github.com/cdcoonce/graphmark

{graphmark-0.3.0 → graphmark-0.3.1}/pyproject.toml

@@ -1,6 +1,6 @@
 [project]
 name = "graphmark"
-version = "0.3.0"
+version = "0.3.1"
 description = "Deterministic knowledge-graph analysis for markdown / [[wikilink]] vaults."
 readme = "README.md"
 requires-python = ">=3.11"

{graphmark-0.3.0 → graphmark-0.3.1}/src/graphmark/graph.py

@@ -18,6 +18,16 @@ def _normalize(text: str) -> str:
     return " ".join(text.lower().translate(_PUNCT_TABLE).split())
 
 
+def _is_intra_note_reference(display: str) -> bool:
+    """True for a link that targets no note, e.g. ``[[#Heading]]`` or ``[[#^block]]``.
+
+    Obsidian uses an empty note part to mean "somewhere in this same note". Such a link is
+    neither an edge nor a broken link, so it must not be recorded as unresolved — otherwise
+    a note that navigates itself heavily looks like the vault's worst offender.
+    """
+    return not display.split("|")[0].split("#")[0].strip()
+
+
 def build_catalog(docs: list[Document]) -> dict[str, list[str]]:
     """Map normalized stem → list of rel_paths (len > 1 means ambiguous)."""
     catalog: dict[str, list[str]] = {}
@@ -126,6 +136,8 @@ class VaultGraph:
 
         for doc in docs:
             for display in extractor.extract(doc.text):
+                if _is_intra_note_reference(display):
+                    continue
                 target = resolver.resolve(display, catalog)
                 if target is None:
                     # Unresolvable OR ambiguous — the Resolver protocol conflates the two, and

{graphmark-0.3.0 → graphmark-0.3.1}/tests/test_unresolved.py

@@ -130,3 +130,48 @@ class TestUnresolvedIntegration:
         assert isinstance(graph.unresolved, dict)
         # stats is untouched by this addition — the oracle still governs it.
         assert set(stats(graph)) == {"notes", "edges", "orphans", "clusters", "density"}
+
+
+class TestIntraNoteAnchorLinks:
+    """[[#Heading]] is a same-note reference, not a link to another note.
+
+    It is neither an edge nor a broken link, so counting it as unresolved inflates the
+    check gate's max_unresolved_links threshold (measured: 19% of a real vault's total).
+    """
+
+    def test_anchor_only_link_is_not_unresolved(self, tmp_path):
+        _write(tmp_path, "a.md", "# Top\n\nJump to [[#Section]].\n\n## Section\n")
+        graph = _build(tmp_path)
+        assert graph.unresolved == {}
+
+    def test_anchor_only_link_with_alias_is_not_unresolved(self, tmp_path):
+        _write(tmp_path, "a.md", "See [[#Design cruxes|crux 2]].\n")
+        assert _build(tmp_path).unresolved == {}
+
+    def test_block_reference_is_not_unresolved(self, tmp_path):
+        _write(tmp_path, "a.md", "See [[#^abc123]].\n")
+        assert _build(tmp_path).unresolved == {}
+
+    def test_anchor_only_link_creates_no_edge(self, tmp_path):
+        _write(tmp_path, "a.md", "Jump to [[#Section]].\n")
+        graph = _build(tmp_path)
+        assert graph.out_links["a.md"] == set()
+
+    def test_whitespace_only_display_is_ignored(self, tmp_path):
+        _write(tmp_path, "a.md", "Odd [[   ]] link.\n")
+        assert _build(tmp_path).unresolved == {}
+
+    def test_cross_note_anchor_still_resolves(self, tmp_path):
+        _write(tmp_path, "a.md", "See [[b#Section]].\n")
+        _write(tmp_path, "b.md", "# B\n\n## Section\n")
+        graph = _build(tmp_path)
+        assert graph.unresolved == {}
+        assert graph.out_links["a.md"] == {"b.md"}
+
+    def test_cross_note_anchor_to_a_missing_note_is_still_unresolved(self, tmp_path):
+        _write(tmp_path, "a.md", "See [[Nowhere#Section]].\n")
+        assert _build(tmp_path).unresolved == {"a.md": ["Nowhere#Section"]}
+
+    def test_mixed_note_keeps_only_the_real_break(self, tmp_path):
+        _write(tmp_path, "a.md", "[[#Local]] and [[Missing]] and [[#Other|x]].\n")
+        assert _build(tmp_path).unresolved == {"a.md": ["Missing"]}