fix: complete monthly DOI conflict repair

citeforge/merge_utils.py

@@ -1400,8 +1400,18 @@ def _replace_existing() -> None:
                 if prefer_entry:
                     pf_fields = sum(1 for v in prefer_entry.get("fields", {}).values() if v)
                     nf_fields = sum(1 for v in entry.get("fields", {}).values() if v)
-                    pf_doi = (prefer_entry.get("fields", {}).get("doi") or "").strip()
-                    if pf_fields > nf_fields or (pf_fields == nf_fields and pf_doi):
+                    pf_doi = _norm_doi(prefer_entry.get("fields", {}).get("doi")) or ""
+                    same_doi_work = bool(
+                        pf_doi
+                        and new_doi
+                        and pf_doi == new_doi
+                        and evaluate_identity(
+                            prefer_entry,
+                            entry,
+                            context=IdentityContext.DISK_SURVIVOR,
+                        ).verdict
+                    )
+                    if not same_doi_work and (pf_fields > nf_fields or (pf_fields == nf_fields and pf_doi)):
                         logger.debug(
                             f"FILE_CLEANUP_BLOCKED | prefer={os.path.basename(prefer_path)} "
                             f"({pf_fields} fields, doi) > new ({nf_fields} fields) | keeping enriched",

citeforge/text_utils.py

@@ -256,7 +256,7 @@ def normalize_person_name(n: Any | None) -> str:
     if not n:
         return ""
     n_str = to_text(n)
-    n2 = strip_accents(n_str).lower()
+    n2 = latex_to_ascii(n_str, math_mode="remove").lower()
     n2 = n2.replace("'", "").replace("\u2019", "").replace("\u02bc", "")
     n2 = _PERSON_PUNCT_RE.sub(" ", n2)
     return " ".join(n2.split())

tests/test_save_entry.py

@@ -405,3 +405,51 @@ def test_year_change_renames_the_file_to_match_the_new_year(tmp_path: Path) -> N
     assert len(survivors) == 1, f"the superseded name must not linger: {survivors}"
     assert "2023" in survivors[0], f"filename must carry the new year: {survivors[0]}"
     assert "2020" not in survivors[0]
+
+
+def test_validated_same_doi_metadata_replaces_equal_field_baseline(tmp_path: Path) -> None:
+    """A DOI-bearing baseline cannot block corrected metadata solely because field counts tie."""
+    author_dir = tmp_path / format_author_dirname("Haque, Israat", "x1")
+    author_dir.mkdir(parents=True)
+    doi = "10.1109/tmlcn.2025.3575368"
+    old_path = factories.write_bib(
+        author_dir,
+        factories.article(
+            key="Hasan2024",
+            title="A Generalized Transformer-based Radio Link Failure Prediction Framework in 5G RANs",
+            author="Kazi Hasan and Thomas Trappenberg and Israat Haque",
+            year="2024",
+            journal="IEEE Transactions on Machine Learning in Communications and Networking",
+            doi=doi,
+            url=f"https://doi.org/{doi}",
+            volume="3",
+            pages="1-12",
+        ),
+        "Hasan2024-GeneralizedTransformer.bib",
+    )
+    incoming = factories.article(
+        key="Hasan2025",
+        title="A Generalized GNN-Transformer-Based Radio Link Failure Prediction Framework in 5G RAN",
+        author="Kazi Hasan and Khaleda Papry and Thomas Trappenberg and Israat Haque",
+        year="2025",
+        journal="IEEE Transactions on Machine Learning in Communications and Networking",
+        doi=doi,
+        url=f"https://doi.org/{doi}",
+        volume="3",
+        pages="1-12",
+    )
+
+    path, written = save_entry_to_file(
+        str(tmp_path),
+        "x1",
+        incoming,
+        prefer_path=str(old_path),
+        author_name="Haque, Israat",
+    )
+
+    assert written is True
+    assert not old_path.exists()
+    assert "2025" in Path(path).name
+    saved = _read(author_dir, Path(path).name)
+    assert "Khaleda Papry" in saved
+    assert "year = {2025}" in saved

tests/test_text_utils.py

@@ -142,6 +142,10 @@ def test_author_overlap_preserves_apostrophes_inside_surnames() -> None:
     assert author_overlap_ratio("Sageev Oore and Jason D'eon", "Oore, Sageev and D'eon, Jason") == 1.0
 
 
+def test_author_overlap_decodes_bibtex_accent_macros() -> None:
+    assert author_overlap_ratio(r"Anne Bergh{\"o}fer", "Anne Berghofer") == 1.0
+
+
 def _preprint_side() -> dict[str, str]:
     return {"title": "T", "author": "Smith, John", "year": "2021", "journal": "arXiv"}