feat(normalizer): add PARENT_OF Bemerkung extraction to persons_tree

tools/import-normalizer/persons_tree.py

@@ -113,6 +113,7 @@ def _build_index(persons: list[dict]) -> dict[str, list[str]]:
         if maiden:
             _add(_norm_tree(f"{first} {maiden}"), row_id)
         _add(_norm_tree(last), row_id)
+        _add(_norm_tree(first), row_id)
 
     return index
 
@@ -242,3 +243,66 @@ def _resolve_spouses(
             })
 
     return relationships, unresolved
+
+
+_CHILD_RE = re.compile(r"^(?:Sohn|Tochter)\s+v(?:on)?\s+(.+)", re.I)
+_PARENT_RE = re.compile(r"^(?:Vater|Mutter)\s+v(?:on)?\s+(.+)", re.I)
+_AND_RE = re.compile(r"\s+u(?:nd)?\s+", re.I)
+
+
+def _parse_bemerkung(
+    row_id: str, bemerkung: str, index: dict[str, list[str]]
+) -> tuple[list[dict], list[dict], str]:
+    """Extract PARENT_OF edges from a Bemerkung cell.
+
+    Returns (relationships, unresolved, remaining_notes).
+    Text that doesn't match a parent pattern goes to remaining_notes unchanged.
+    """
+    if not bemerkung or not bemerkung.strip():
+        return [], [], ""
+
+    s = bemerkung.strip()
+
+    for pattern, direction in ((_CHILD_RE, "child"), (_PARENT_RE, "parent")):
+        m = pattern.match(s)
+        if not m:
+            continue
+
+        name_part = m.group(1).strip().rstrip("!., ")
+        parts = [p.strip() for p in _AND_RE.split(name_part) if p.strip()]
+        rels: list[dict] = []
+        unres: list[dict] = []
+
+        for part in parts:
+            part = part.rstrip("!., ")
+            matched_id, reason = _resolve_one(part, index)
+            if matched_id:
+                if direction == "child":
+                    rels.append({
+                        "personId": matched_id,
+                        "relatedPersonId": row_id,
+                        "type": "PARENT_OF",
+                        "source": "bemerkung",
+                        "rawBemerkung": bemerkung,
+                    })
+                else:
+                    rels.append({
+                        "personId": row_id,
+                        "relatedPersonId": matched_id,
+                        "type": "PARENT_OF",
+                        "source": "bemerkung",
+                        "rawBemerkung": bemerkung,
+                    })
+            else:
+                unres.append({
+                    "rowId": row_id,
+                    "field": "bemerkung",
+                    "raw": bemerkung,
+                    "reason": reason,
+                })
+
+        remainder = s[m.end():].strip().lstrip(".,! ")
+        return rels, unres, remainder
+
+    # No pattern matched — full text goes to notes, nothing to unresolved
+    return [], [], s

tools/import-normalizer/tests/test_persons_tree.py

@@ -318,3 +318,103 @@ def test_resolve_spouses_empty_spouse_field():
     idx = persons_tree._build_index(persons)
     rels, unres = persons_tree._resolve_spouses(persons, idx)
     assert rels == [] and unres == []
+
+
+def _register(*args):
+    """Build index from (rowId, first, last, maiden) tuples."""
+    persons = [
+        {"rowId": a[0], "firstName": a[1], "lastName": a[2], "maidenName": a[3]}
+        for a in args
+    ]
+    return persons, persons_tree._build_index(persons)
+
+
+def test_parse_bemerkung_sohn_two_parents():
+    _, idx = _register(
+        ("row_019", "Clara", "Cram", "de Gruyter"),
+        ("row_028", "Herbert", "Cram", None),
+    )
+    rels, unres, notes = persons_tree._parse_bemerkung(
+        "row_021", "Sohn v Clara u Herbert", idx
+    )
+    assert len(rels) == 2
+    assert all(r["type"] == "PARENT_OF" for r in rels)
+    child_ids = {r["relatedPersonId"] for r in rels}
+    parent_ids = {r["personId"] for r in rels}
+    assert child_ids == {"row_021"}
+    assert "row_019" in parent_ids and "row_028" in parent_ids
+    assert unres == []
+    assert notes == ""
+
+
+def test_parse_bemerkung_tochter_von():
+    _, idx = _register(("row_019", "Clara", "Cram", None))
+    rels, unres, notes = persons_tree._parse_bemerkung(
+        "row_036", "Tochter von Clara Cram", idx
+    )
+    assert len(rels) == 1
+    assert rels[0] == {
+        "personId": "row_019",
+        "relatedPersonId": "row_036",
+        "type": "PARENT_OF",
+        "source": "bemerkung",
+        "rawBemerkung": "Tochter von Clara Cram",
+    }
+    assert notes == ""
+
+
+def test_parse_bemerkung_vater():
+    _, idx = _register(("row_028", "Herbert", "Cram", None))
+    rels, unres, notes = persons_tree._parse_bemerkung(
+        "row_031", "Vater v Herbert", idx
+    )
+    assert len(rels) == 1
+    assert rels[0]["personId"] == "row_031"
+    assert rels[0]["relatedPersonId"] == "row_028"
+    assert rels[0]["type"] == "PARENT_OF"
+
+
+def test_parse_bemerkung_unmatched_parent_name():
+    _, idx = _register()   # empty index
+    rels, unres, notes = persons_tree._parse_bemerkung(
+        "row_004", "Sohn v Elsgard A.", idx
+    )
+    assert rels == []
+    assert len(unres) == 1
+    assert unres[0]["reason"] == "not_found"
+    assert notes == ""
+
+
+def test_parse_bemerkung_skip_nichte():
+    _, idx = _register(("row_028", "Herbert", "Cram", None))
+    rels, unres, notes = persons_tree._parse_bemerkung(
+        "row_002", "Nichte von Herbert", idx
+    )
+    assert rels == []
+    assert unres == []
+    assert notes == "Nichte von Herbert"
+
+
+def test_parse_bemerkung_skip_bruder():
+    _, idx = _register(("row_028", "Herbert", "Cram", None))
+    rels, unres, notes = persons_tree._parse_bemerkung(
+        "row_033", "Bruder v Herbert", idx
+    )
+    assert rels == []
+    assert unres == []
+    assert notes == "Bruder v Herbert"
+
+
+def test_parse_bemerkung_empty():
+    _, idx = _register()
+    rels, unres, notes = persons_tree._parse_bemerkung("row_004", "", idx)
+    assert rels == [] and unres == [] and notes == ""
+
+
+def test_parse_bemerkung_plain_remark():
+    _, idx = _register()
+    rels, unres, notes = persons_tree._parse_bemerkung(
+        "row_029", "Verfasserin der Cram-Chronik !!", idx
+    )
+    assert rels == [] and unres == []
+    assert notes == "Verfasserin der Cram-Chronik !!"